Anda di halaman 1dari 537

Prevention of Cancer: a Global Perspective

Food, Nutrition, Physical Activity, and the

Food, Nutrition,
Physical Activity,
World American
and the Prevention Cancer Institute for
of Cancer: Research Fund Cancer Research
a Global Perspective

The most definitive review of the science to date,

and the most authoritative basis for action to
prevent cancer worldwide.

Recommendations based on expert

judgements of systematic reviews of the
world literature.
The result of a five-year examination by a
panel of the worlds leading scientists.
Includes new findings on early life, body
fatness, physical activity, and cancer
Recommendations harmonised with
prevention of other diseases and promotion
of well-being.
A vital guide for everybody, and the
indispensable text for policy-makers and

Food, Nutrition,
Physical Activity,
and the Prevention
Research Fund
of Cancer:

a Global Perspective
Cancer Research
Institute for

World Cancer World Cancer Wereld Kanker World Cancer Fonds Mondial
American Institute Research Fund
Research Fund for Cancer Research Research Fund Onderzoek Fonds de Recherche
International Hong Kong contre le Cancer

ER HARD FINAL.indd 1 30/10/07 10:07:25

Food, Nutrition,
P h y s i c a l A c t i v i t y, a n d t h e
P re v e n t i o n o f C a n c e r :
a Global Perspective
Please cite the Report as follows:
World Cancer Research Fund / American Institute for Cancer Research.
Food, Nutrition, Physical Activity, and the Prevention of Cancer: a Global Perspective.
Washington DC: AICR, 2007

First published 2007 by the American Institute for Cancer Research

1759 R St. NW, Washington, DC 20009

2007 World Cancer Research Fund International

All rights reserved

Readers may make use of the text and graphic material in this Report
for teaching and personal purposes, provided they give credit to
World Cancer Research Fund and American Institute for Cancer Research.

ISBN: 978-0-9722522-2-5

CIP data in process

Printed in the United States of America by RR Donnelley

Food, Nutrition,
P h y s i c a l A c t i v i t y, a n d t h e
P re v e n t i o n o f C a n c e r :
a Global Perspective

A project of
World Cancer Research Fund
P re f a c e

I am very grateful to the special group of distinguished scientists who made up the Panel
and Secretariat for this major review of the evidence on food, nutrition, physical activity
and cancer. The vision of WCRF International in convening this Panel and confidence in
letting a strong-willed group of scientists have their way is to be highly commended.
In our view, the evidence reviewed here that led to our recommendations provides a
wonderful opportunity to prevent cancer and improve global health. Individuals and
populations have in their hands the means to lead fuller, healthier lives. Achieving that
will take action, globally, nationally, and locally, by communities, families, and
It is worth pausing to put this Report in context. Public perception is often that experts
disagree. Why should the public or policy-makers heed advice if experts differ in their
views? Experts do disagree. That is the nature of science and a source of its strength.
Should we throw up our hands and say one opinion is as good as another? Of course not.
Evidence matters. But not evidence unguided by human thought. Hence the process that
was set up for this review: use a systematic approach to examine all the relevant evidence
using predetermined criteria, and assemble an international group of experts who, having
brought their own knowledge to bear and having debated their disagreements, arrive at
judgements as to what this evidence means. Both parts of the exercise were crucial: the
systematic review and, dare I say it, the wisdom of the experts.
The elegance of the process was one of the many attractions to me of assuming the role
of chair of the Panel. I could pretend that it was the major reason, and in a way it was, but
the first reason was enjoyment. What a pleasure and a privilege to spend three years in
the company of a remarkable group of scientists, including world leaders in research on
the epidemiology of cancer, as well as leaders in nutrition and public health and the
biology of cancer, to use a relatively new methodology (systematic literature reviews),
supported by a vigorous and highly effective Secretariat, on an issue of profound
importance to global public health: the prevention of cancer by means of healthy patterns
of eating and physical activity. It was quite as enjoyable as anticipated.
Given this heady mix, the reasons why I might have wanted to take on the role of Panel
chair were obvious. I did question the wisdom of WCRF International in inviting me to do
it. Much of my research has been on cardiovascular disease, not cancer. What I described
as my ignorance, WCRF International kindly labelled impartiality.
WCRF also appreciated the parallels between dietary causes of cardiovascular disease
and cancer. There is a great deal of concordance. In general, recommendations in this
Report to prevent cancer will also be of great relevance to cardiovascular disease. The only
significant contradiction is with alcohol. From the point of view of cancer prevention, the
best level of alcohol consumption is zero. This is not the case for cardiovascular disease,
where the evidence suggests that one to two drinks a day are protective. The Panel
therefore framed its recommendation to take this into account.
The fact that the conclusions and recommendations in this Report are the unanimous
view of the Panel does not imply that, miraculously, experts have stopped disagreeing. The
Panel debated the fine detail of every aspect of its conclusions and recommendations with
remarkable vigour and astonishing stamina. In my view, this was deliberation at its best. If
conclusions could simply fall out of systematic literature reviews, we would not have
needed experts to deliberate. Human judgement was vital; and if human, it cannot be
infallible. But I venture to suggest this process has led to as good an example of evidence-
based public health recommendations as one can find.
Throughout the Panels deliberations, it had in mind the global reach of this Report.
Most of the research on diet and cancer comes from high-income countries. But

noncommunicable diseases, including cancer, are now major public health burdens in
every region of the world. An important part of our deliberations was to ensure the
global applicability of our recommendations.
One last point about disagreement among experts: its relevance to the link between
science and policy. A caricature would be to describe science as precise and policy-makers
as indecisive. In a way, the opposite is the case. Science can say: could be, might be,
some of us think this, and some think that. Policy-makers have either to do it or not
do it more often, not. Our effort here was to increase the precision of scientific
judgements. As the Report makes clear, many of our conclusions are in the could be
category. None of our recommendations is based on these could be conclusions. All are
based on judgements that evidence was definite or probable. Our recommendations, we
trust, will serve as guides to the population, to scientists, and to opinion-formers.
But what should policy-makers do with our judgements? A year after publication of
this Report, we will publish a second report on policy for diet, nutrition, physical activity,
and the prevention of cancer. As an exercise developing out of this one, we decided to
apply, as far as possible, the same principles of synthesis of evidence to policy-making.
We enhanced the scientific panel that was responsible for this Report with experts in
nutrition and food policy. This policy panel will oversee systematic literature reviews on
food policy, deliberate, and make recommendations.
The current Report and next years Policy Report have one overriding aim: to reduce
the global burden of cancer by means of healthier living.
Michael Marmot


P re f a c e iv Chapter 7 Cancers 244

Contents vi 7.1 Mouth, pharynx and larynx 245

7.2 Nasopharynx 250
Acknowledgements viii 7.3 Oesophagus 253
7.4 Lung 259
S u m m a ry xiv 7.5 Stomach 265
7.6 Pancreas 271
I n t ro d u c t i o n xxii 7.7 Gallbladder 275
7.8 Liver 277
7.9 Colon and rectum 280
PA RT O N E B A C K G R O U N D 1 7.10 Breast 289
7.11 Ovary 296
Chapter 1 I n t e r n a t i o n a l v a r i a t i o n s a n d t re n d s 4 7.12 Endometrium 299
7.13 Cervix 302
1.1 Food systems and diets throughout history 5 7.14 Prostate 305
1.2 Foods and drinks, physical activity, 7.15 Kidney 310
body composition 11 7.16 Bladder 312
1.3 Migrant and other ecological studies 22 7.17 Skin 315
1.4 Conclusions 25 7.18 Other cancers 318

Chapter 2 T h e c a n c e r p ro c e s s 30 Chapter 8 D e t e rm i n a n t s o f w e i g h t g a i n ,
o v e rw e i g h t , o b e s i t y 322
2.1 Basic concepts and principles 31
2.2 Cellular processes 32 Chapter 9 C a n c e r s u rv i v o r s 342
2.3 Carcinogen metabolism 36
2.4 Causes of cancer 37 C h a p t e r 1 0 F i n d i n g s o f o t h e r re p o r t s 348
2.5 Nutrition and cancer 41
2.6 Conclusions 46 10.1 Method 349
10.2 Interpretation of the data 350
Chapter 3 Judging the evidence 48 10.3 Nutritional deficiencies 350
10.4 Infectious diseases 351
3.1 Epidemiological evidence 49 10.5 Chronic diseases other than cancer 352
3.2 Experimental evidence 52 10.6 Cancer 355
3.3 Methods of assessment 55 10.7 Conclusions 358
3.4 Causation and risk 57
3.5 Coming to judgement 58 C h a p t e r 1 1 R e s e a rc h i s s u e s 360
3.6 Conclusions 62

PA RT T H R E E R E C O M M E N D AT I O N S 365
PA RT T W O E V I D E N C E A N D J U D G E M E N T S 63
Chapter 12 Public health goals and
Chapter 4 Foods and drinks 66 p e r s o n a l re c o m m e n d a t i o n s 368

4.1 Cereals (grains), roots, tubers and plantains 67 12.1 Principles 369
4.2 Vegetables, fruits, pulses (legumes), nuts, 12.2 Goals and recommendations 373
seeds, herbs, spices 75 12.3 Patterns of food, nutrition and
4.3 Meat, poultry, fish and eggs 116 physical activity 391
4.4 Milk, dairy products 129
4.5 Fats and oils 135
4.6 Sugars and salt 141 APPENDICES 395
4.7 Water, fruit juices, soft drinks and hot drinks 148
4.8 Alcoholic drinks 157 A p p e n d i x A P ro j e c t p ro c e s s 396
4.9 Food production, processing, preservation
and preparation 172 A p p e n d i x B T h e f i r s t W C R F / A I C R E x p e rt R e p o rt 398
4.10 Dietary constituents and supplements 179
4.11 Dietary patterns 190 Appendix C WCRF global network 400

G l o s s a ry 402
Chapter 5 Physical activity 198
R e f e re n c e s 410
Chapter 6 G ro w t h , d e v e l o p m e n t , b o d y
composition 210 Index 506

6.1 Body fatness 211

6.2 Growth and development 229
6.3 Lactation 239

CHAPTER BOXES Box 4.6.1 Sugar, sugars, sugary foods and drinks 142
Box 4.6.2 Salt and salty, salted and salt-preserved
PA RT O N E B A C K G R O U N D foods 143
Box 4.6.3 Chemical sweeteners 143
Chapter 1 I n t e r n a t i o n a l v a r i a t i o n s a n d t re n d s Box 4.6.4 Refrigeration 144
Box 4.7.1 High temperature, and irritant drinks
Box Egypt 6 and foods 150
Box South Africa 8 Box 4.7.2 Contamination of water, and of foods
Box China 10 and other drinks 150
Box 1.1 Measurement of food supply Box 4.8.1 Types of alcoholic drink 159
and consumption 13 Box 4.9.1 Food systems 173
Box India 14 Box 4.9.2 Organic farming 174
Box Japan 16 Box 4.9.3 Regulation of additives and
Box UK 18 contaminants 175
Box 1.2 Measurement of cancer incidence Box 4.9.4 Water fluoridation 176
and mortality 18 Box 4.10.1 Food fortification 182
Box Poland 20 Box 4.10.2 Functional foods 182
Box Spain 22 Box 4.10.3 Levels of supplementation 183
Box USA 24
Box Mexico 26 Chapter 5 Physical activity
Box Australia 27
Box Brazil 28 Box 5.1 Energy cost and intensity of activity 200
Box 5.2 Sedentary ways of life 201
Chapter 2 T h e c a n c e r p ro c e s s
Chapter 6 G ro w t h , d e v e l o p m e n t , b o d y c o m p o s i t i o n
Box 2.1 Nutrition over the life course 34
Box 2.2 Oncogenes and tumour suppressor genes 35 Box 6.2.1 Sexual maturity 232
Box 2.3 Mechanisms for DNA repair 37 Box 6.2.2 Age at menarche and risk of
Box 2.4 Body fatness and attained height 39 breast cancer 232
Box 2.5 Energy restriction 46
Chapter 7 Cancers
Chapter 3 Judging the evidence
Box 7.1.1 Cancer incidence and survival 246
Box 3.1 Issues concerning interpretation of Box 7.2.1 Epstein-Barr virus 251
the evidence 50 Box 7.5.1 Helicobacter pylori 266
Box 3.2 Dose-response 52 Box 7.8.1 Hepatitis viruses 278
Box 3.3 Forest plots 53 Box 7.13.1 Human papilloma viruses 303
Box 3.4 Systematic literature reviews 54
Box 3.5 Experimental findings 55 Chapter 8 D e t e rm i n a n t s o f w e i g h t g a i n ,
Box 3.6 Effect modification 56 o v e rw e i g h t , o b e s i t y
Box 3.7 Energy adjustment 57
Box 3.8 Criteria for grading evidence 60 Box 8.1 Energy density 324
Box 8.2 Fast food 325
Box 8.3 Body fatness in childhood 326
PA RT T W O E V I D E N C E A N D J U D G E M E N T S Box 8.4 Television viewing 331

Chapter 4 Foods and drinks Chapter 9 C a n c e r s u rv i v o r s

Box 4.1.1 Wholegrain and refined cereals and Box 9.1 Conventional and unconventional
their products 69 therapies 345
Box 4.1.2 Foods containing dietary fibre 69 Box 9.2 Use of supplements by cancer survivors 346
Box 4.1.3 Glycaemic index and load 69
Box 4.1.4 Aflatoxins 70 Chapter 10 F i n d i n g s o f o t h e r re p o r t s
Box 4.2.1 Micronutrients and other bioactive
compounds and cancer risk 78 Chapter 11 R e s e a rc h i s s u e s
Box 4.2.2 Phytochemicals 79
Box 4.2.3 Preparation of vegetables and nutrient
bioavailability 79 PA RT T H R E E R E C O M M E N D AT I O N S
Box 4.2.4 Foods containing dietary fibre 80
Box 4.3.1 Processed meat 117 Chapter 12 Public health goals and personal
Box 4.3.2 Nitrates, nitrites and N-nitroso re c o m m e n d a t i o n s
compounds 118
Box 4.3.3 Foods containing iron 118 Box 12.1 Quantification 371
Box 4.3.4 Heterocyclic amines and polycyclic Box 12.2 Making gradual changes 372
aromatic hydrocarbons 119 Box 12.3 Height, weight and ranges of BMI 375
Box 4.3.5 Cantonese-style salted fish 120 Box 12.4 When supplements are advisable 387
Box 4.4.1 Foods containing calcium 131 Box 12.5 Regional and special circumstances 392
Box 4.5.1 Hydrogenation and trans-fatty acids 137

F O O D , N U T R I T I O N , P H Y S I C A L A C T I V I T Y, A N D T H E P R E V E N T I O N O F C A N C E R : A G L O B A L P E R S P E C T I V E


Panel Jim Mann DM PhD FFPH Robert Beaglehole ONZM World Health Organization
Sir Michael Marmot University of Otago Chair 2003 Geneva, Switzerland
MB BS MPH PhD FRCP FFPH Dunedin, New Zealand Was at: World Health Denise Coitinho PhD
Chair Organization (WHO) Ruth Bonita MD
University College London Hilary J Powers PhD RNutr Geneva, Switzerland Chizuru Nishida PhD MA
UK University of Sheffield, UK Now at: University of Pirjo Pietinen DSc
Auckland, New Zealand
Tola Atinmo PhD K Srinath Reddy MD DM MSc
University of Ibadan, Nigeria Institute of Medical Sciences Additional
New Delhi, India Panel observers members for policy
Tim Byers MD MPH panel
University of Colorado Health Elio Riboli MD ScM MPH Food and Agriculture
Sciences Center Was at: International Agency Organization of the United Barry Popkin PhD MSc BSc
Denver, CO, USA for Research on Cancer Nations (FAO) Carolina Population Center,
(IARC), Lyon, France Rome, Italy University of North Carolina,
Junshi Chen MD Now at: Imperial College Guy Nantel PhD Chapel Hill, NC, USA
Chinese Centre for Disease London, UK Prakash Shetty MD PhD
Control and Prevention Jane Wardle PhD MPhil
Beijing, Peoples Republic of Juan A Rivera PhD International Food Policy University College London, UK
China Instituto Nacional de Salud Research Institute (IFPRI)
Publica Washington DC, USA Nick Cavill MPH
Tomio Hirohata MD Cuernavaca, Mexico Lawrence Haddad PhD British Heart Foundation
DrScHyg PhD Marie Ruel PhD Health Promotion
Kyushu University Arthur Schatzkin MD DrPH Research Group
Fukuoka City, Japan National Cancer Institute International Union of University of Oxford, UK
Rockville, MD, USA Nutritional Sciences (IUNS)
Alan Jackson CBE MD FRCP Mark Wahlqvist MD AO
FRCPCH FRCPath Jacob C Seidell PhD
University of Southampton Free University Amsterdam Mechanisms Working Group
UK The Netherlands John Milner PhD

W Philip T James CBE MD David E G Shuker PhD FRSC Methodology Task Force
DSc FRSE FRCP The Open University Jos Kleijnen MD PhD
International Obesity Task Milton Keynes, UK Gillian Reeves PhD
London, UK Ricardo Uauy MD PhD Union Internationale Contre
Instituto de Nutricion y le Cancer (UICC)
Laurence N Kolonel MD PhD Tecnologia de los Alimentos Geneva, Switzerland
University of Hawaii Santiago, Chile Annie Anderson PhD
Honolulu, HI, USA Curtis Mettlin PhD
Walter C Willett MD DrPH Harald zur Hausen MD DSc
Shiriki Kumanyika PhD MPH Harvard School of Public
University of Pennsylvania Health United Nations Childrens
Philadelphia, PA, USA Boston, MA, USA Fund (UNICEF)
New York, NY, USA
Claus Leitzmann PhD Steven H Zeisel MD PhD Ian Darnton-Hill MD MPH
Justus Liebig University University of North Carolina Rainer Gross Dr Agr
Giessen, Germany Chapel Hill, NC, USA


Systematic Rajendra Persad ChM FEBU Carlos A Gonzalez PhD MPH Karen Robinson MSc
Literature Review FRCS MD Johns Hopkins University
Centres United Bristol Healthcare Catalan Institute of Oncology Baltimore, MD, USA
Trust & Bristol Urological Barcelona, Spain
University of Bristol, UK Institute, UK Xuguang (Grant) Tao MD
Vittorio Krogh MD MSc PhD
George Davey Smith Massimo Pignatelli MD PhD Istituto Nazionale Tumori Johns Hopkins University
FMedSci FRCP DSc FRCPath Milan, Italy Baltimore, MD, USA
University of Bristol , UK University of Bristol, UK
Sylvie Menard ScD
Jonathan Sterne PhD MSc Jelena Savovic PhD Istituto Nazionale Tumori University of Leeds, UK
MA University of Bristol, UK Milan, Italy
University of Bristol, UK David Forman PhD FFPH
Steve Thomas MB BS PhD Eugenio Mugno ScD University of Leeds, UK
Chris Bain MB BS MS MPH FRCS Istituto Nazionale Tumori
University of Queensland University of Bristol, UK Milan, Italy Victoria J Burley PhD MSc
Brisbane, Australia RPHNutr
Tim Whittlestone MA MD Valeria Pala ScD University of Leeds, UK
Nahida Banu MB BS FRCS Istituto Nazionale Tumori
University of Bristol, UK United Bristol Healthcare Milan, Italy Janet E Cade PhD BSc
Trust, UK RPHNutr
Trudy Bekkering PhD Sabina Sieri ScD University of Leeds, UK
University of Bristol, UK Luisa Zuccolo MSc Istituto Nazionale Tumori
University of Bristol, UK Milan, Italy Darren C Greenwood MSc
Rebecca Beynon MA BSc University of Leeds, UK
University of Bristol, UK
Istituto Nazionale Johns Hopkins Doris S M Chan MSc
Margaret Burke MSc Tu m o r i M i l a n , I t a l y U n i v e r s i t y, B a l t i m o re , University of Leeds, UK
University of Bristol, UK MD, USA
Franco Berrino MD Jennifer A Moreton PhD MSc
David de Berker MB BS MRCP Istituto Nazionale Tumori Anthony J Alberg PhD MPH University of Leeds, UK
United Bristol Healthcare Milan, Italy University of South Carolina
Trust, UK Columbia, SC, USA James D Thomas
Patrizia Pasanisi MD MSc University of Leeds, UK
Anna A Davies MSc BSc Istituto Nazionale Tumori Kristina Boyd MS
University of Bristol, UK Milan, Italy Johns Hopkins University Yu-Kang Tu PhD MSc DDS
Baltimore, MD, USA University of Leeds, UK
Roger Harbord MSc Claudia Agnoli ScD
University of Bristol, UK Istituto Nazionale Tumori Laura Caulfield PhD Iris Gordon MSc
Milan, Italy Johns Hopkins University University of Leeds, UK
Ross Harris MSc Baltimore, MD, USA
University of Bristol, UK Silvana Canevari ScD Kenneth E L McColl FRSE
Istituto Nazionale Tumori Eliseo Guallar MD DrPH FMedSci FRCP
Lee Hooper PhD SRD Milan, Italy Johns Hopkins University Western Infirmary
University of East Anglia Baltimore, MD, USA Glasgow, UK
Norwich, UK Giovanni Casazza ScD
Istituto Nazionale Tumori James Herman MD Lisa Dyson MSc
Anne-Marie Mayer PhD MSc Milan, Italy Johns Hopkins University University of Leeds, UK
University of Bristol, UK Baltimore, MD, USA
Elisabetta Fusconi ScD
Andy Ness PhD FFPHM MRCP Istituto Nazionale Tumori Genevieve Matanoski MD
University of Bristol, UK Milan, Italy DrPH
Johns Hopkins University
Baltimore, MD, USA

F O O D , N U T R I T I O N , P H Y S I C A L A C T I V I T Y, A N D T H E P R E V E N T I O N O F C A N C E R : A G L O B A L P E R S P E C T I V E

London School of P e n n S t a t e U n i v e r s i t y, Dina M Gifkins MPH Anouk Geelen PhD

H y g i e n e & Tr o p i c a l University Park, The Cancer Institute of New Wageningen University
Medicine, UK PA , U S A Jersey The Netherlands
New Brunswick, NJ, USA
Alan D Dangour PhD MSc Terryl J Hartman PhD MPH Evelien Smit MSc
London School of Hygiene & RD Marjorie L McCullough RD Wageningen University
Tropical Medicine, UK Penn State University, ScD The Netherlands
University Park, PA, USA American Cancer Society
Shefalee Mehta MSc New York, NY, USA Salome Scholtens MSc
London School of Hygiene & David Mauger PhD Wageningen University
Tropical Medicine, UK Penn State College of The Netherlands
Medicine, Wa g e n i n g e n U n i v e r s i t y,
Abigail Perry MSc University Park, PA, USA The Netherlands Evert-Jan Bakker PhD
London School of Hygiene & Wageningen University
Tropical Medicine, UK Lindsay Camera MS Pieter van t Veer PhD The Netherlands
Penn State College of Wageningen University
Sakhi Kiran Dodhia MSc Medicine, The Netherlands Jan Burema MSc
London School of Hygiene & University Park, PA, USA Wageningen University
Tropical Medicine, UK Ellen Kampman PhD The Netherlands
M Jenny Harris Ledikwe PhD Wageningen University
Vicki Pyne MSc Penn State University, The Netherlands Marianne Renkema PhD
London School of Hygiene & University Park, PA, USA Wageningen University
Tropical Medicine, UK Marije Schouten PhD The Netherlands
Linda Kronheim MS Wageningen University
Penn State University, The Netherlands Henk van Kranen PhD
U n i v e r s i t y o f Te e s s i d e , University Park, PA, USA National Institute for Health
UK Bianca Stam MSc and the Environment (RIVM)
Keith R Martin PhD MTox Wageningen University Bilthoven, the Netherlands
Carolyn Summerbell PhD Penn State University, The Netherlands
SRD University Park, PA, USA
University of Teesside Claudia Kamphuis MSc Narrative review
Middlesbrough, UK Tara Murray Wageningen University authors
Penn State University, The Netherlands
Sarah Kelly PhD University Park, PA, USA Liju Fan PhD
University of Teesside Maureen van den Donk PhD Ontology Workshop
Middlesbrough, UK Michele L Shaffer PhD Wageningen University Columbia, MD, USA
Penn State College of The Netherlands
Louisa Ells PhD Medicine, Luigino Dal Maso ScD
University of Teesside University Park, Marian Bos MSc Aviano Cancer Center
Middlesbrough, UK PA, USA Wageningen University Italy
The Netherlands
Frances Hillier MSc Kim Spaccarotella PhD Michael Garner MSc
University of Teesside Rutgers, The State University Akke Botma MSc University of Ottawa
Middlesbrough, UK of New Jersey, New Wageningen University Ontario, Canada
Brunswick, NJ, USA The Netherlands
Sarah Smith MSc Frank M Torti MD MPH
University of Teesside Simone Croezen MSc Wake Forest University,
Middlesbrough, UK K a i s e r P e rm a n e n t e , Wageningen University Comprehensive Cancer Unit
Oakland, California, USA The Netherlands Winston-Salem, NC, USA
Alan Batterham PhD
University of Teesside Elisa V Bandera MD PhD Mirjam Meltzer MSc Christine F Skibola PhD
Middlesbrough, UK The Cancer Institute of New Wageningen University University of California,
Jersey The Netherlands Berkeley, CA, USA
Laurel Edmunds PhD New Brunswick, NJ, USA
University of Teesside Fleur Schouten MSc
Middlesbrough, UK Lawrence H Kushi ScD Wageningen University
Kaiser Permanente The Netherlands
Vicki Whittaker MSc Oakland, California, USA
University of Teesside Janneke Ploemacher MSc
Middlesbrough, UK Dirk F Moore PhD Wageningen University
The Cancer Institute of New The Netherlands
Ian Macdonald PhD Jersey
University of Nottingham, UK New Brunswick, NJ, USA Khahn Le MSc
Wageningen University
The Netherlands


Methodology Gillian Reeves PhD Mechanisms Peer reviewers and

Task Force Cancer Research UK Working Group other contributors
Epidemiology Unit,
Martin Wiseman FRCP University of Oxford John Milner PhD David S Alberts MD
FRCPath Oxford, UK Chair Arizona Cancer Center
Chair National Cancer Institute Tucson, AZ, USA
Project Director Elio Riboli MD ScM MPH Rockville, MD, USA
WCRF International Was at: International Agency Chris Bain MBBS MPH
for Research on Cancer Nahida Banu MBBS University of Queensland
Sheila A Bingham PhD (IARC), Lyon, France University of Bristol, UK Brisbane, Australia
FMedSci Now at: Imperial College
MRC Dunn Human Nutrition London, UK Xavier Castellsagu Pique Amy Berrington de Gonzalez
Unit PhD MD MPH DPhil MSc
Cambridge, UK Arthur Schatzkin MD DrPH Catalan Institute of Oncology Johns Hopkins University
National Cancer Institute Barcelona, Spain Baltimore, MD, USA
Heiner Boeing PhD Rockville, MD, USA
German Institution of Human Sanford M Dawsey MD Sheila A Bingham PhD
Nutrition David E G Shuker PhD National Cancer Institute FMedSci
Berlin, Germany The Open University Rockville, MD, USA MRC Dunn Human Nutrition
Milton Keynes, UK Unit
Eric Brunner PhD FFPH Carlos A Gonzalez PhD MPH Cambridge, UK
University College London, Michael Sjstrm MD PhD MD
UK Karolinska Institute Catalan Institute of Oncology Diane Birt PhD
Stockholm, Sweden Barcelona, Spain Iowa State University
H Bas Bueno de Mesquita MD Ames, IA, USA
MPH PhD Pieter van t Veer PhD James Herman MD
National Institute of Public Wageningen University Johns Hopkins University Steven Blair PED
Health and the Environment The Netherlands Baltimore, MD, USA University of South Carolina
(RIVM) Columbia, SC, USA
Bilthoven, the Netherlands Chris Williams MD Stephen Hursting PhD
Cochrane Cancer Network University of North Carolina Judith Bliss MSc
David Forman PhD FFPH Oxford, UK Chapel Hill, NC, USA The Institute of Cancer
University of Leeds, UK Research
Henry Kitchener MD Sutton, UK
Ian Frayling PhD MRCPath University of Manchester, UK
Addenbrookes Hospital Cristina Bosetti ScD
Cambridge, UK Keith R Martin PhD MTox Istituto di Recherche
Penn State University Farmacologiche Mario
Andreas J Gescher DSc University Park, PA, USA Negri
University of Leicester, UK Milan, Italy
Kenneth E L McColl FRSE
Tim Key PhD FMedSci FRCP Paul Brennan PhD MSc
Cancer Research UK Western Infirmary International Agency for
Epidemiology Unit, Glasgow, UK Research on Cancer (IARC)
University of Oxford Lyon, France
Oxford, UK Sylvie Menard ScD
Istituto Nazionale Tumori Johannes Brug PhD FFPH
Jos Kleijnen MD PhD Milan, Italy Institute for Research in
Was at: University of York, UK Extramural Medicine
Now at: Kleijnen Systematic Massimo Pignatelli MD PhD (EMGO),
Reviews, York, UK MRCPath VU University Medical Centre
University of Bristol, UK Amsterdam, the Netherlands
Barrie Margetts MSc PhD
MFPH Henk van Kranen PhD Eric Brunner PhD FFPH
University of Southampton, National Institute of Public University College London,
UK Health and the Environment UK
Robert Owen PhD Bilthoven, the Netherlands H Bas Bueno de Mesquita MD
German Cancer Research MPH PhD
Centre National Institute of Public
Heidelberg, Germany Health and the Environment
Bilthoven, the Netherlands

F O O D , N U T R I T I O N , P H Y S I C A L A C T I V I T Y, A N D T H E P R E V E N T I O N O F C A N C E R : A G L O B A L P E R S P E C T I V E

Noel Cameron BEd MSc Elizabeth TH Fontham DrPH Victor Kipnis PhD Tony McMichael MB BS PhD
Loughborough University, UK Louisiana State University of National Cancer Institute FAFPHM
Moira Chan-Yeung MBBS Public Health Rockville, MD, USA The Australian National
FRCP FACP New Orleans, LA, USA University
University of Hong Kong Paul Knekt PhD Canberra, Australia
China Terrence Forrester MB BS DM National Public Health
FRCP Institute Dominique Michaud ScD
Robert Clarke DSc PhD University of the West Indies Helsinki, Finland Harvard School of Public
Lombardi Comprehensive Kingston, Jamaica Health
Cancer Center, Georgetown Thilo Kober PhD Boston, MA, USA
University Teresa Fung ScD RD MSc Cochrane Haematological
Washington DC, USA Simmons College and Malignancies Group Anthony B Miller MD FRCP
Harvard School of Public Cologne, Germany FACE
Steven K Clinton MD PhD Health University of Toronto
The Ohio State University Boston, MA, USA Suminori Kono PhD MD MSc Canada
Columbus, OH, USA Kyushu University
John Garrow MD PhD FRCP Fukuoka, Japan Sidney Mirvish PhD
Karen Collins MS RD University of London, UK University of Nebraska
Nutrition Advisor Nancy Kreiger PhD MPH Omaha, NE, USA
AICR Glenn Gibson PhD Cancer Care Ontario and
University of Reading, UK University of Toronto Max Parkin MD
Brian Cox MBChB PhD Canada International Agency for
FAFPHM Ian Gilmore MD PRCP Research on Cancer (IARC)
University of Otago Royal College of Physicians Petra Lahmann PhD Lyon, France
Dunedin, New Zealand London, UK University of Queensland
Brisbane, Australia Charlotte Paul MB ChB PhD
Cindy Davis PhD Vay Liang W Go MD University of Otago
National Cancer Institute University of California Fabio Levi MD MSc Dunedin, New Zealand
Rockville, MD, USA Los Angeles, CA, USA Institut Universitaire de
Mdecine Sociale et John Reilly PhD
Diana Dyer MS RD Per Hall MD PhD Prventive University of Glasgow, UK
Ann Arbor, MI, USA Karolinska Institutet Lausanne, Switzerland
Stockholm, Sweden Richard Rivlin MD
Jonathan Earle MB BCh FCAP Ruth Lewis MSc Strang Cancer Research
Memorial Sloan Kettering Laura Hardie PhD Cardiff University, UK Laboratory
Cancer Center University of Leeds, UK New York, NY, USA
New York, NY, USA Albert B Lowenfels MD
Peter Herbison MSc New York Medical College Andrew Roddam DPhil
Alison Eastwood MSc University of Otago New York, NY, USA Cancer Research UK
University of York, UK Dunedin, New Zealand Epidemiology Unit
Graham A MacGregor FRCP University of Oxford
Ibrahim Elmadfa PhD Melvyn Hillsdon PhD St Georges University of Oxford, UK
University of Vienna University of Bristol, UK London, UK
Austria Leo Schouten MD PhD
Edward Hurwitz DC PhD Geoffrey Marks PhD MS Nutrition and Toxicology
Dallas English PhD MSc University of Hawaii University of Queensland Research Institute Maastricht
University of Melbourne Honolulu, HI, USA Brisbane, Australia The Netherlands
Victoria, Australia
Susan Jebb PhD John Mathers PhD DipNutr Jackilen Shannon PhD MPH
Michael Fenech PhD MSc BSc MRC Human Nutrition University of Newcastle, UK RD
Commonwealth Scientific and Research Oregon Health and Science
Industrial Research Cambridge, UK Sam McClinton MD FRCS University
Organization (CSIRO) NHS Grampian Portland, OR, USA
Adelaide, Australia Stanley B Kaye MD FRCP Aberdeen, UK
FMedSci Keith Singletary PhD
Justin Fenty MSc The Institute of Cancer Fiona Mensah University of Illinois
University of Nottingham, UK Research University of York, UK Urbana, IL, USA
Sutton, UK
Lynn Ferguson DSc DPhil MSc Margaret McCredie PhD Rashmi Sinha PhD
Univerity of Auckland Tim Key PhD University of Otago National Cancer Institute
New Zealand Cancer Research UK Dunedin, New Zealand Rockville, MD, USA
Epidemiology Unit,
University of Oxford
Oxford, UK


Rachael Stolzenberg-Solomon WCRF/AICR Report Kate Coughlin BSc Ivana Vucenik PhD
PhD MPH RD Executive Team Science Programme Manager Associate Director for
National Cancer Institute WCRF International Research
Baltimore, MD, USA Marilyn Gentry AICR
President Cara James
Boyd Swinburn MB ChB MD WCRF Global Network Associate Director for Joan Ward
Deakin University Research Research Administration
Melbourne, Australia Kelly Browning AICR Assistant
Chief Financial Officer From 2003 to 2005 WCRF International
Peter Szlosarek MRCP PhD WCRF Global Network From 2001 to 2003
St Bartholomews Hospital Jennifer Kirkwood
London, UK Kate Allen PhD Research Administration Julia Wilson PhD
Director Assistant Science Programme Manager
Paul Talalay MD WCRF International WCRF International WCRF International
Johns Hopkins University From 2003 to 2004
Baltimore, MD, USA Kathryn L Ward
Senior Vice-President Anja Kroke MD PhD MPH Art & production
Margaret Thorogood PhD AICR Consultant
University of Warwick, UK University of Applied Sciences Chris Jones
Deirdre McGinley-Gieser Fulda, Germany Design and Art Director
Stewart Truswell MD DSc WCRF International 2002 Design4Science Ltd
FRCP London, UK
University of Sydney Jeffrey R Prince PhD Kayte Lawton
Australia Vice-President for Education Research Administration Emma Copeland PhD
and Communications Assistant Text Editor
Paolo Vineis MD MPH AICR WCRF International Brighton, UK
Imperial College From 2006 to 2007
London, UK Rosalind Holmes
Secretariat Lisa Miles MSc Production Manager
Steven Waggoner MD Science Programme Manager London, UK
Case Comprehensive Cancer Martin Wiseman FRCP WCRF International
Center FRCPath From 2002 to 2006 Mark Fletcher
Cleveland, OH, USA Project Director Graphics
WCRF International Sarah Nalty MSc Fletcher Ward Design
Christopher P Wild PhD Science Programme Manager London, UK
University of Leeds, UK Geoffrey Cannon WCRF International
Chief Editor Ann OMalley
Anthony Williams DPhil FRCP WCRF International Edmund Peston Print Manager
FRCPCH Research Administration AICR
St Georges University of Ritva R Butrum PhD Assistant
London, UK Senior Science Advisor WCRF International Geoff Simmons
AICR From 2004 to 2006 Design & Production Manager
Frederic M Wolf PhD MEd WCRF UK
University of Washington Greg Martin MB BCh MPH Serena Prince
Seattle, WA, USA Project Manager Research Administration
WCRF International Assistant
Jian-Min Yuan MD PhD WCRF International
University of Minnesota, Susan Higginbotham PhD From 2004 to 2005
Minneappolis, MN, USA Director for Research
AICR Melissa Samaroo
Maurice Zeegers PhD MSc Research Administration
University of Birmingham, UK Steven Heggie PhD Assistant
Project Manager WCRF International
WCRF International From 2006 to 2007
From 2002 to 2006
Elaine Stone PhD
Alison Bailey Science Programme Manager
Science Writer WCRF International
Redhill, UK From 2001 to 2006

Poling Chow BSc Rachel Thompson PhD

Research Administration RPHNutr
Assistant Review Coordinator
WCRF International

F O O D , N U T R I T I O N , P H Y S I C A L A C T I V I T Y, A N D T H E P R E V E N T I O N O F C A N C E R : A G L O B A L P E R S P E C T I V E

Introduction summarised here. A more detailed explanation of this
process is given in Chapter 3 and the research teams and
This summary provides an abbreviated version of the full investigators involved are listed on pages viiixi.
Report. It highlights the wealth of information and data This Report is a guide to future scientific research, cancer
studied by the Panel and is designed to give readers an prevention education programmes, and health policy
overview of the key issues contained within the Report, around the world. It provides a solid evidence base for
notably the process, the synthesis of the scientific evidence, policy-makers, health professionals, and informed and
and the resulting judgements and recommendations. interested people to draw on and work with.

T h e f i r s t a n d s e c o n d R e p o rt s
Food, Nutrition and the Prevention of Cancer: a global per- Overview of the second expert Report
spective, produced by the World Cancer Research Fund
together with the American Institute for Cancer Research, This Report has a number of inter-related general purposes.
has been the most authoritative source on food, nutrition, One is to explore the extent to which food, nutrition, phys-
and cancer prevention for 10 years. On publication in 1997, ical activity, and body composition modify the risk of can-
it immediately became recognised as the most authoritative cer, and to specify which factors are most important. To the
and influential report in its field and helped to highlight the extent that environmental factors such as food, nutrition,
importance of research in this crucial area. It became the and physical activity influence the risk of cancer, it is a pre-
standard text worldwide for policy-makers in government ventable disease. The Report specifies recommendations
at all levels, for civil society and health professional organ- based on solid evidence which, when followed, will be
isations, and in teaching and research centres of academic expected to reduce the incidence of cancer.
Since the mid-1990s the amount of scientific literature on P a r t 1 B a c k g ro u n d
this subject has dramatically increased. New methods of Chapter 1 shows that patterns of production and con-
analysing and assessing evidence have been developed, sumption of food and drink, of physical activity, and of body
facilitated by advances in electronic technology. There is composition have changed greatly throughout human
more evidence, in particular on overweight and obesity and history. Remarkable changes have taken place as a result
on physical activity; food, nutrition, physical activity, and of urbanisation and industrialisation, at first in Europe,
cancer survivors is a new field. The need for a new report North America, and other economically advanced coun-
was obvious; and in 2001 WCRF International in collabora- tries, and increasingly in most countries in the world.
tion with AICR began to put in place a global process in Notable variations have been identified in patterns of can-
order to produce and publish the Report in November 2007. cer throughout the world. Significantly, studies consistently
show that patterns of cancer change as populations migrate
H o w t h i s R e p o rt h a s b e e n a c h i e v e d from one part of the world to another and as countries
The goal of this Report is to review all the relevant research, become increasingly urbanised and industrialised. Pro-
using the most meticulous methods, in order to generate a jections indicate that rates of cancer in general are liable
comprehensive series of recommendations on food, nutri- to increase.
tion, and physical activity, designed to reduce the risk of Chapter 2 outlines current understanding of the biology
cancer and suitable for all societies. This process is also the of the cancer process, with special attention to the ways in
basis for a continuous review of the evidence. which food and nutrition, physical activity, and body com-
Organised into overlapping stages, the process has been position may modify the risk of cancer. Cancer is a disease
designed to maximise objectivity and transparency, separat- of genes, which are vulnerable to mutation, especially over
ing the collection of evidence from its assessment and the long human lifespan. However, evidence shows that
judgement. First, an expert task force developed a method only a small proportion of cancers are inherited.
for systematic review of the voluminous scientific literature. Environmental factors are most important and can be mod-
Second, research teams collected and reviewed the litera- ified. These include smoking and other use of tobacco;
ture based upon this methodology. Third, an expert Panel infectious agents; radiation; industrial chemicals and pollu-
has assessed and judged this evidence and agreed recom- tion; medication; and also many aspects of food, nutrition,
mendations. The results are published in this Report and physical activity, and body composition.


Chapter 3 summarises the types of evidence that the physical activity themselves affect the risk of obesity and
Panel has agreed are relevant to its work. No single study associated factors. The judgements, which concern the bio-
or study type can prove that any factor definitely is a cause logical and associated determinants of weight gain, over-
of, or is protective against, any disease. In this chapter, weight, and obesity, are based on a further systematic
building on the work of the first report, the Panel shows literature review, amplified by knowledge of physiological
that reliable judgements on causation of disease are based processes.
on assessment of a variety of well-designed epidemiologi- The relevance of food, nutrition, physical activity, and
cal and experimental studies. body composition to people living with cancer, and to the
The prevention of cancer worldwide is one of the most prevention of recurrent cancer, is summarised in Chapter 9.
pressing challenges facing scientists and public health Improved cancer screening, diagnosis, and medical services
policy-makers, among others. These introductory chapters are, in many countries, improving survival rates. So the
show that the challenge can be effectively addressed and number of cancer survivors people living after diagnosis
suggest that food, nutrition, physical activity, and body of cancer is increasing.
composition play a central part in the prevention of cancer. The Panel agreed that its recommendations should also
take into account findings on the prevention of other chron-
P a rt 2 E v i d e n c e a n d J u d g e m e n t s ic diseases, and of nutritional deficiencies and nutrition-
The judgements made by the Panel in Part 2 are based on related infectious diseases, especially of childhood. Chapter
independently conducted systematic reviews of the litera- 10, also based on a systematic literature review, is a sum-
ture commissioned from academic institutions in the USA, mary of the findings of expert reports in these areas.
UK, and continental Europe. The evidence has been metic- The research issues identified in Chapter 11 are, in the
ulously assembled and, crucially, the display of the evi- view of the Panel, the most promising avenues to explore in
dence was separated from assessments derived from that order to refine understanding of the links between food,
evidence. Seven chapters present the findings of these nutrition, physical activity, and cancer, and so improve the
reviews. The Panels judgements are displayed in the form prevention of cancer, worldwide.
of matrices that introduce five of these chapters, and in the
summary matrix on the fold-out page inside the back cover. P a rt 3 R e c o m m e n d a t i o n s
Chapter 4, the first and longest chapter in Part 2, is con- Chapter 12, the culmination of the five-year process, pre-
cerned with types of food and drink. The judgements of the sents the Panels public health goals and personal recom-
Panel are, whenever possible, food- and drink-based, mendations. These are preceded by a statement of the
reflecting the most impressive evidence. Findings on principles that have guided the Panel in its thinking.
dietary constituents and micronutrients (for example foods The goals and recommendations are based on convinc-
containing dietary fibre) are identified where appropriate. ing or probable judgements made by the Panel in the chap-
Evidence on dietary supplements, and on patterns of diet, ters in Part 2. These are proposed as the basis for public
is included in the two final sections of this chapter. policies and for personal choices that, if effectively imple-
Chapters 5 and 6 are concerned with physical activity mented, will be expected to reduce the incidence of cancer
and with body composition, growth, and development. for people, families, and communities.
Evidence in these areas is more impressive than was the Eight general and two special goals and recommenda-
case up to the mid-1990s; the evidence on growth and tions are detailed. In each case a general recommendation
development indicates the importance of an approach to is followed by public health goals and/or personal recom-
the prevention of cancer that includes the whole life mendations, together with further explanation or clarifica-
course. tion as required. Chapter 12 also includes a summary of the
Chapter 7 summarises and judges the evidence as evidence, justification of the goals and recommendations,
applied to 17 cancer sites, with additional briefer sum- and guidance on how to achieve them.
maries based on narrative reviews of five further body sys- The process of moving from evidence to judgements and
tems and cancer sites. The judgements shown in the to recommendations has been one of the Panels main
matrices in this chapter correspond with the judgements responsibilities, and has involved discussion and debate
shown in the matrices in the previous chapters. until final agreement has been reached. The goals and rec-
Obesity is or may be a cause of a number of cancers. ommendations here have been unanimously agreed.
Chapter 8 identifies what aspects of food, nutrition, and The goals and recommendations are followed by the

F O O D , N U T R I T I O N , P H Y S I C A L A C T I V I T Y, A N D T H E P R E V E N T I O N O F C A N C E R : A G L O B A L P E R S P E C T I V E

Panels conclusions on the dietary patterns most likely to The Panels recommendations
protect against cancer. In order to discern the big picture of
healthy and protective diets, it is necessary to integrate a The Panels goals and recommendations that follow are
vast amount of detailed information. The Panel used a guided by several principles, the details of which can be
broad, integrative approach that, while largely derived from found in Chapter 12. The public health goals are for
conventional reductionist research, has sought to find pat- populations, and therefore for health professionals; the
terns of food and drink consumption, of physical activity, recommendations are for people, as communities, families,
and of body fatness, that enable recommendations designed and individuals.
to prevent cancer at personal and population levels. The Panel also emphasises the importance of not smoking
The goals and recommendations are designed to be gen- and avoiding exposure to tobacco smoke.
erally relevant worldwide and the Panel recognises that in
national settings, the recommendations of this Report will Format
be best used in combination with recommendations, issued The goals and recommendations begin with a general state-
by governments or on behalf of nations, designed to prevent ment. This is followed by the population goal and the per-
chronic and other diseases. In addition, the Panel cited sonal recommendation, together with any necessary
three specific cases where the evidence is strong enough to footnotes. These footnotes are an integral part of the
be the basis for goals and recommendations, but which cur- recommendations. The full recommendations, including
rently are relevant only in discrete geographical regions: further clarification and qualification, can be found in
mat in Latin America, Cantonese-style salted fish in the Chapter 12.
Pearl River Delta in Southern China, and arsenic contami-
nating water supplies in several locations. Further details on
nutritional patterns and regional and special circumstances
can be found in section 12.3.
The main focus of this Report is on nutritional and other
biological and associated factors that modify the risk of can-
cer. The Panel is aware that as with other diseases, the risk
of cancer is also modified by social, cultural, economic, and
ecological factors. Thus the foods and drinks that people
consume are not purely because of personal choice; likewise
opportunities for physical activity can be constrained.
Identifying the deeper factors that affect cancer risk enables
a wider range of policy recommendations and options to be
identified. This is the subject of a separate report to be pub-
lished in late 2008.
The public health goals and personal recommendations of
the Panel that follow are offered as a significant contribu-
tion towards the prevention and control of cancer through-
out the world.




Be as lean as possible within B e p h y s i c a l l y a c t i v e a s p a rt o f e v e ry d a y l i f e

t h e n o rm a l r a n g e1 o f b o d y w e i g h t
The proportion of the population that is sedentary1
Median adult body mass index (BMI) to be to be halved every 10 years
between 21 and 23, depending on the
normal range for different populations2 Average physical activity levels (PALs)1 to be above 1.6

The proportion of the population that is overweight

or obese to be no more than the current level, PERSONAL RECOMMENDATIONS
or preferably lower, in 10 years
Be moderately physically active, equivalent
to brisk walking,2 for at least 30 minutes every day
As fitness improves, aim for 60 minutes or more
Ensure that body weight through of moderate, or for 30 minutes or more of
childhood and adolescent growth projects3 towards the vigorous, physical activity every day2 3
lower end of the normal BMI range at age 21
Limit sedentary habits such as watching television
Maintain body weight within
the normal range from age 21
The term sedentary refers to a PAL of 1.4 or less. PAL is a way of representing
the average intensity of daily physical activity. PAL is calculated as total energy
Avoid weight gain and increases in expenditure as a multiple of basal metabolic rate
waist circumference throughout adulthood 2
Can be incorporated in occupational, transport, household, or leisure activities
This is because physical activity of longer duration or greater intensity is more
Normal range refers to appropriate ranges issued by national governments or
the World Health Organization
To minimise the proportion of the population outside the normal range
Projects in this context means following a pattern of growth (weight and
height) throughout childhood that leads to adult BMI at the lower end of the Justification
normal range. Such patterns of growth are specified in International Obesity
Task Force and WHO growth reference charts Most populations, and people living in industrialised and
urban settings, have habitual levels of activity below levels
to which humans are adapted.
Maintenance of a healthy weight throughout life may be With industrialisation, urbanisation, and mechanisation,
one of the most important ways to protect against cancer. populations and people become more sedentary. As with
This will also protect against a number of other common overweight and obesity, sedentary ways of life have been
chronic diseases. usual in high-income countries since the second half of the
20th century. They are now common if not usual in most
Weight gain, overweight, and obesity are now generally countries.
much more common than in the 1980s and 1990s. Rates of All forms of physical activity protect against some can-
overweight and obesity doubled in many high-income coun- cers, as well as against weight gain, overweight, and obesi-
tries between 1990 and 2005. In most countries in Asia and ty; correspondingly, sedentary ways of life are a cause of
Latin America, and some in Africa, chronic diseases includ- these cancers and of weight gain, overweight, and obesity.
ing obesity are now more prevalent than nutritional defi- Weight gain, overweight, and obesity are also causes of
ciencies and infectious diseases. some cancers independently of the level of physical activity.
Being overweight or obese increases the risk of some can- Further details of evidence and judgements can be found in
cers. Overweight and obesity also increase the risk of condi- Chapters 5, 6, and 8.
tions including dyslipidaemia, hypertension and stroke, type The evidence summarised in Chapter 10 also shows that
2 diabetes, and coronary heart disease. Overweight in child- physical activity protects against other diseases and that
hood and early life is liable to be followed by overweight sedentary ways of life are causes of these diseases.
and obesity in adulthood. Further details of evidence and
judgements can be found in Chapters 6 and 8. Maintenance
of a healthy weight throughout life may be one of the most
important ways to protect against cancer.

F O O D , N U T R I T I O N , P H Y S I C A L A C T I V I T Y, A N D T H E P R E V E N T I O N O F C A N C E R : A G L O B A L P E R S P E C T I V E


PROMOTE WEIGHT GAIN Eat mostly foods of plant origin

L i m i t c o n s u m p t i o n o f e n e rg y - d e n s e f o o d s1 PUBLIC HEALTH GOALS

Av o i d s u g a r y d r i n k s 2
Population average consumption of non-starchy1
PUBLIC HEALTH GOALS vegetables and of fruits to be at least 600 g (21 oz) daily2

Average energy density of diets3 to be lowered Relatively unprocessed cereals (grains) and/or pulses
towards 125 kcal per 100 g (legumes), and other foods that are a natural source of
dietary fibre, to contribute to a population average
Population average consumption of sugary drinks2 of at least 25 g non-starch polysaccharide daily
to be halved every 10 years

Eat at least five portions/servings
Consume energy-dense foods1 4 sparingly (at least 400 g or 14 oz) of a variety2 of
non-starchy vegetables and of fruits every day
Avoid sugary drinks2
Eat relatively unprocessed cereals (grains)
Consume fast foods5 sparingly, if at all and/or pulses (legumes) with every meal3
Energy-dense foods are here defined as those with an energy content of more
than about 225275 kcal per 100 g Limit refined starchy foods
This principally refers to drinks with added sugars. Fruit juices should also be
limited People who consume starchy roots or tubers4
This does not include drinks
Limit processed energy-dense foods (also see recommendation 4). Relatively as staples also to ensure intake of sufficient
unprocessed energy-dense foods, such as nuts and seeds, have not been shown non-starchy vegetables, fruits, and pulses (legumes)
to contribute to weight gain when consumed as part of typical diets, and these
and many vegetable oils are valuable sources of nutrients 1
This is best made up from a range of various amounts of non-starchy vegetables
The term fast foods refers to readily available convenience foods that tend to
and fruits of different colours including red, green, yellow, white, purple, and
be energy-dense and consumed frequently and in large portions
orange, including tomato-based products and allium vegetables such as garlic
Relatively unprocessed cereals (grains) and/or pulses (legumes) to contribute
to an average of at least 25 g non-starch polysaccharide daily
These foods are low in energy density and so promote healthy weight
For example, populations in Africa, Latin America, and the Asia-Pacific region
Consumption of energy-dense foods and sugary drinks is
increasing worldwide and is probably contributing to the
global increase in obesity. Justification
An integrated approach to the evidence shows that most
This overall recommendation is mainly designed to prevent diets that are protective against cancer are mainly made up
and to control weight gain, overweight, and obesity. from foods of plant origin.
Further details of evidence and judgements can be found in
Chapter 8. Higher consumption of several plant foods probably protects
Energy density measures the amount of energy (in kcal against cancers of various sites. What is meant by plant-
or kJ) per weight (usually 100 g) of food. Food supplies that based is diets that give more emphasis to those plant foods
are mainly made up of processed foods, which often contain that are high in nutrients, high in dietary fibre (and so in non-
substantial amounts of fat or sugar, tend to be more energy- starch polysaccharides), and low in energy density. Non-
dense than food supplies that include substantial amounts starchy vegetables, and fruits, probably protect against some
of fresh foods. Taken together, the evidence shows that it is cancers. Being typically low in energy density, they probably
not specific dietary constituents that are problematic, so also protect against weight gain. Further details of evidence
much as the contribution these make to the energy density and judgements can be found in Chapters 4 and 8.
of diets. Non-starchy vegetables include green, leafy vegetables,
Because of their water content, drinks are less energy- broccoli, okra, aubergine (eggplant), and bok choy, but not,
dense than foods. However, sugary drinks provide energy for instance, potato, yam, sweet potato, or cassava. Non-
but do not seem to induce satiety or compensatory reduc- starchy roots and tubers include carrots, Jerusalem arti-
tion in subsequent energy intake, and so promote overcon- chokes, celeriac (celery root), swede (rutabaga), and turnips.
sumption of energy and thus weight gain. Continued on next page




L i m i t i n t a k e o f re d m e a t1 and
L i m i t a l c o h o l i c d r i n k s1
a v o i d p ro c e s s e d m e a t 2


Proportion of the population drinking
Population average consumption of red meat
more than the recommended limits to be
to be no more than 300 g (11 oz) a week,
reduced by one third every 10 years1 2
very little if any of which to be processed

If alcoholic drinks are consumed,
People who eat red meat1
limit consumption to no more than two drinks a day
to consume less than 500 g (18 oz) a week,
for men and one drink a day for women1 2 3
very little if any to be processed2

Red meat refers to beef, pork, lamb, and goat from domesticated animals 1
This recommendation takes into account that there is a likely protective effect
including that contained in processed foods for coronary heart disease
Processed meat refers to meat preserved by smoking, curing or salting, or 2
Children and pregnant women not to consume alcoholic drinks
addition of chemical preservatives, including that contained in processed foods 3
One drink contains about 1015 grams of ethanol

Justification Justification
An integrated approach to the evidence also shows that The evidence on cancer justifies a recommendation not to
many foods of animal origin are nourishing and healthy if drink alcoholic drinks. Other evidence shows that modest
consumed in modest amounts. amounts of alcoholic drinks are likely to reduce the risk of
coronary heart disease.
People who eat various forms of vegetarian diets are at low
risk of some diseases including some cancers, although it is The evidence does not show a clear level of consumption of
not easy to separate out these benefits of the diets from alcoholic drinks below which there is no increase in risk of
other aspects of their ways of life, such as not smoking, the cancers it causes. This means that, based solely on the
drinking little if any alcohol, and so forth. In addition, meat evidence on cancer, even small amounts of alcoholic drinks
can be a valuable source of nutrients, in particular protein, should be avoided. Further details of evidence and judge-
iron, zinc, and vitamin B12. The Panel emphasises that this ments can be found in Chapter 4. In framing the recom-
overall recommendation is not for diets containing no meat mendation here, the Panel has also taken into account the
or diets containing no foods of animal origin. The evidence that modest amounts of alcoholic drinks are likely
amounts are for weight of meat as eaten. As a rough con- to protect against coronary heart disease, as described in
version, 300 g of cooked red meat is equivalent to about Chapter 10.
400450 g raw weight, and 500 g cooked red meat to about The evidence shows that all alcoholic drinks have the
700750 g raw weight. The exact conversion will depend same effect. Data do not suggest any significant difference
on the cut of meat, the proportions of lean and fat, and the depending on the type of drink. This recommendation
method and degree of cooking, so more specific guidance is therefore covers all alcoholic drinks, whether beers, wines,
not possible. Red or processed meats are convincing or spirits (liquors), or other alcoholic drinks. The important
probable causes of some cancers. Diets with high levels of factor is the amount of ethanol consumed.
animal fats are often relatively high in energy, increasing The Panel emphasises that children and pregnant women
the risk of weight gain. Further details of evidence and should not consume alcoholic drinks.
judgements can be found in Chapters 4 and 8.

Recommendation 4, continued from page xviii of relatively unprocessed cereals (grains), non-starchy veg-
The goals and recommendations here are broadly similar etables and fruits, and pulses (legumes), all of which contain
to those that have been issued by other international and substantial amounts of dietary fibre and a variety of
national authoritative organisations (see Chapter 10). They micronutrients, and are low or relatively low in energy den-
derive from the evidence on cancer and are supported by sity. These, and not foods of animal origin, are the recom-
evidence on other diseases. They emphasise the importance mended centre for everyday meals.

F O O D , N U T R I T I O N , P H Y S I C A L A C T I V I T Y, A N D T H E P R E V E N T I O N O F C A N C E R : A G L O B A L P E R S P E C T I V E


P R E S E R VAT I O N , P R O C E S S I N G , D I E TA R Y S U P P L E M E N T S
Aim to meet nutritional needs
Limit consumption of s a l t1 t h ro u g h d i e t a l o n e 1
A v o i d m o u l d y c e re a l s ( g r a i n s ) o r p u l s e s ( l e g u m e s )


Maximise the proportion of the population achieving
Population average consumption of salt from nutritional adequacy without dietary supplements
all sources to be less than 5 g (2 g of sodium) a day

Proportion of the population consuming more than 6 g PERSONAL RECOMMENDATION

of salt (2.4 g of sodium) a day to be halved every 10 years
Dietary supplements are not recommended
Minimise exposure to aflatoxins for cancer prevention
from mouldy cereals (grains) or pulses (legumes) 1
This may not always be feasible. In some situations of illness or dietary
inadequacy, supplements may be valuable


Avoid salt-preserved, salted, or salty foods; Justification

preserve foods without using salt1 The evidence shows that high-dose nutrient supplements
can be protective or can cause cancer. The studies that
Limit consumption of processed foods with added salt demonstrate such effects do not relate to widespread use
to ensure an intake of less than 6 g (2.4 g sodium) a day among the general population, in whom the balance of
risks and benefits cannot confidently be predicted. A gen-
Do not eat mouldy cereals (grains) or pulses (legumes) eral recommendation to consume supplements for cancer
prevention might have unexpected adverse effects.
Methods of preservation that do not or need not use salt include refrigeration, Increasing the consumption of the relevant nutrients
freezing, drying, bottling, canning, and fermentation through the usual diet is preferred.

The recommendations of this Report, in common with its

Justification general approach, are food based. Vitamins, minerals, and
The strongest evidence on methods of food preservation, pro- other nutrients are assessed in the context of the foods and
cessing, and preparation shows that salt and salt-preserved drinks that contain them. The Panel judges that the best
foods are probably a cause of stomach cancer, and that foods source of nourishment is foods and drinks, not dietary sup-
contaminated with aflatoxins are a cause of liver cancer. plements. There is evidence that high-dose dietary supple-
ments can modify the risk of some cancers. Although some
Salt is necessary for human health and life itself, but at lev- studies in specific, usually high-risk, groups have shown evi-
els very much lower than those typically consumed in most dence of cancer prevention from some supplements, this
parts of the world. At the levels found not only in high- finding may not apply to the general population. Their level
income countries but also in those where traditional diets of benefit may be different, and there may be unexpected
are high in salt, consumption of salty foods, salted foods, and uncommon adverse effects. Therefore it is unwise to
and salt itself is too high. The critical factor is the overall recommend widespread supplement use as a means of can-
amount of salt. Microbial contamination of foods and drinks cer prevention. Further details of evidence and judgements
and of water supplies remains a major public health prob- can be found in Chapter 4.
lem worldwide. Specifically, the contamination of cereals In general, for otherwise healthy people, inadequacy of
(grains) and pulses (legumes) with aflatoxins, produced by intake of nutrients is best resolved by nutrient-dense diets
some moulds when such foods are stored for too long in and not by supplements, as these do not increase consump-
warm temperatures, is an important public health problem, tion of other potentially beneficial food constituents. The
and not only in tropical countries. Panel recognises that there are situations when supplements
Salt and salt-preserved foods are a probable cause of are advisable. See box 12.4.
some cancers. Aflatoxins are a convincing cause of liver can-
cer. Further details of evidence and judgements can be
found in Chapter 4.




M o t h e r s t o b re a s t f e e d ; c h i l d re n t o b e b re a s t f e d 1 F o l l o w t h e re c o m m e n d a t i o n s
f o r c a n c e r p re v e n t i o n 2


The majority of mothers to breastfeed
exclusively, for six months2 3 All cancer survivors3 to receive nutritional care
from an appropriately trained professional
If able to do so, and unless otherwise advised,
Aim to breastfeed infants exclusively2 aim to follow the recommendations for
up to six months and continue diet, healthy weight, and physical activity2
with complementary feeding thereafter3
Cancer survivors are people who are living with a diagnosis of cancer, including
those who have recovered from the disease
Breastfeeding protects both mother and child 2
This recommendation does not apply to those who are undergoing active
Exclusively means human milk only, with no other food or drink, including treatment, subject to the qualifications in the text
water 3
This includes all cancer survivors, before, during, and after active treatment
In accordance with the UN Global Strategy on Infant and Young Child Feeding

Justification Justification
The evidence on cancer as well as other diseases shows Subject to the qualifications made here, the Panel has
that sustained, exclusive breastfeeding is protective for the agreed that its recommendations apply also to cancer sur-
mother as well as the child. vivors. There may be specific situations where this advice
may not apply, for instance, where treatment has compro-
This is the first major report concerned with the prevention mised gastrointestinal function.
of cancer to make a recommendation specifically on breast-
feeding, to prevent breast cancer in mothers and to prevent If possible, when appropriate, and unless advised otherwise
overweight and obesity in children. Further details of evi- by a qualified professional, the recommendations of this
dence and judgements can be found in Chapters 6 and 8. Report also apply to cancer survivors. The Panel has made
Other benefits of breastfeeding for mothers and their this judgement based on its examination of the evidence,
children are well known. Breastfeeding protects against including that specifically on cancer survivors, and also on
infections in infancy, protects the development of the its collective knowledge of the pathology of cancer and its
immature immune system, protects against other childhood interactions with food, nutrition, physical activity, and body
diseases, and is vital for the development of the bond composition. In no case is the evidence specifically on can-
between mother and child. It has many other benefits. cer survivors clear enough to make any firm judgements or
Breastfeeding is especially vital in parts of the world where recommendations to cancer survivors. Further details of
water supplies are not safe and where impoverished fami- evidence and judgements can be found in Chapter 9.
lies do not readily have the money to buy infant formula Treatment for many cancers is increasingly successful,
and other infant and young child foods. This recommenda- and so cancer survivors increasingly are living long enough
tion has a special significance. While derived from the evi- to develop new primary cancers or other chronic diseases.
dence on being breastfed, it also indicates that policies and The recommendations in this Report would also be expect-
actions designed to prevent cancer need to be directed ed to reduce the risk of those conditions, and so can also be
throughout the whole life course, from the beginning of recommended on that account.

F O O D , N U T R I T I O N , P H Y S I C A L A C T I V I T Y, A N D T H E P R E V E N T I O N O F C A N C E R : A G L O B A L P E R S P E C T I V E

I n t ro d u c t i o n

The proposals that cancer might be preventable, and that abundantly clear that the incidence of all the common
food, nutrition, physical activity, and body composition cancers in humans is determined by various potentially
might affect the risk of cancer, were first made before controllable external factors. This is surely the most
science emerged in its modern form in the 19th and 20th comforting fact to come out of cancer research, for it
centuries. Throughout recorded history, wise choices of means that cancer is, in large part, a preventable disease.1
food and drink, and of habitual behaviour, have been This is the conclusion of a report on diet and the
recommended to protect against cancer, as well as other prevention of cancer published a quarter of a century
diseases, and to improve well-being. before this Report.
Reports such as this, which incorporate systematic Since the early 1980s, relevant United Nations agencies,
examination of all relevant types of research, differ from national governments, authoritative non-governmental
ancient, historical, and even relatively recent accounts, and organisations, and researchers and other experts in the
descriptive studies of the type detailed in Chapter 1, not field have agreed that food and nutrition, physical activity,
only in the quantity and quality of evidence, but also in the and body composition are individually and collectively
reliability of the judgements and recommendations that important modifiers of the risk of cancer, and taken
derive from it. together may be at least as important as tobacco.
By the mid-1990s the general consensus became more
T h e p u r p o s e o f t h i s R e p o rt solidly based on methodical assessment of the totality of
This Report has been commissioned and resourced by the the relevant literature. Thus: It is now established that
World Cancer Research Fund (WCRF) International and its cancer is principally caused by environmental factors, of
sister organisation the American Institute for Cancer which the most important are tobacco; diet and factors
Research (AICR), who provided the Secretariat that has related to diet, including body mass and physical activity;
supported the Panel responsible for the Report. Panel and exposures in the workplace and elsewhere. This
members, observers, review centres, and other contributors statement introduces the recommendations made in the
are listed on the preceding pages. The five-year project that first WCRF/AICR report.
has resulted in this Report follows a previous five-year Expert reports may be accompanied by guidebooks
project that resulted in the first WCRF/AICR report written for general readers. Thus: A healthy eating
published in 1997, which was the responsibility of the strategy is an important part of protecting yourself
former distinguished international multidisciplinary panel against a long list of diseases. These include heart disease,
chaired by Professor John Potter. stroke, several common cancers, cataract formation, other
This Report has two overall general purposes. The first is age-related diseases, and even some types of birth defects.
to summarise, assess, and judge the most comprehensive When combined with not smoking and regular exercise, this
body of evidence yet collected and displayed on the subject kind of healthy diet can reduce heart disease by 80 per
of food, nutrition, physical activity, body composition, and cent, and stroke and some cancers by 70 percent, compared
the risk of cancer, throughout the life-course. The second with average rates.2 This is a conclusion of a book written
purpose is to transform the evidence-derived judgements by a member of the Panel responsible for this Report.
into goals and personal recommendations that are a Some general judgements are now well known and not a
reliable basis for sound policies and effective actions at matter for serious debate. Cancer in general, and cancers
population, community, family, and individual level, in of different types and sites, are agreed to have various
order to prevent cancer, worldwide. causes, among which are inherited genetic predisposition
and the increasing likelihood that cells will accumulate
W h a t i s a l re a d y k n o w n genetic defects as people age. This is discussed in more
The Panel is aware of the general consensus shared by detail in Chapter 2. Also, people die less frequently from
scientists, health professionals, and policy-makers on the nutritional deficiencies, infectious diseases, predation, and
relationships between food, nutrition, physical activity, accidents, whereas chronic diseases including cancer
body composition, and the risk of cancer. which are more common in older people become more
This consensus, based on the findings of a rapidly common.
growing mass of increasingly well-designed However, cancer is not an inevitable consequence of
epidemiological and experimental studies and other ageing, and peoples susceptibility to it varies. There is
relevant evidence, emerged in the early 1980s. Thus: It is abundant evidence that the main causes of patterns of


cancer around the world are environmental. This does S p e c i a l f e a t u re s o f t h i s R e p o r t

indeed mean that at least in principle, most cancer is This Report in part adapts and builds on the work of the
preventable, though there is still discussion about the previous WCRF/AICR report. It also has central features
relative importance of various environmental factors. that are new. It is not simply an update of the previous
But what are these environmental factors, what is their report. Since the mid-1990s a substantial body of relevant
relative importance, and how may they vary in different literature has been published in peer-reviewed journals.
times in the life-course and in different parts of the world, Further, the executive officers of WCRF/AICR, its
and how might they interact with each other? Many Secretariat, and the Panel responsible, decided at the
thousand epidemiological and experimental studies have outset that developments in scientific method since the
looked for answers. Some answers are now agreed to be mid-1990s, notably in systematic approaches to
unequivocal. Thus, smoking is the chief cause of lung synthesising evidence, and as enabled by the electronic
cancer. Alcohol is also an established carcinogen in revolution, have been so remarkable that a whole new
humans, as are types of radiation such as those used in process was justified.
medical treatments and as released by nuclear weapons
and accidents. Certain infectious agents are undoubtedly a Systematic literature reviews
cause of some cancers. This process (described in Appendix A) has involved
systematic literature reviews (SLRs), which have been used
The need for a new initiative as the main basis for the Panels judgements in this Report.
Many questions, particularly in the field of food, nutrition, These are described in more detail in Chapter 3. They were
and associated factors, remain. Some are fundamental. Do undertaken by independent centres of research and review
statements such as those quoted above remain valid? Do excellence in North America and Europe, to a common
they apply worldwide? Have the reviews and reports so far agreed protocol, itself the product of an expert
published overlooked key findings? How do the large Methodology Task Force. As a result, the judgements of the
prospective studies, meta-analyses, pooling projects, and Panel now are as firmly based as the evidence and the state
randomised controlled trials undertaken and published of the science allow. Some are new. Some are different
since the mid-1990s impact on earlier conclusions and from those previously published. Findings that may at first
recommendations? Are there areas in this field that have reading seem to repeat those of the first report are in fact
been neglected? Is entirely new evidence coming to light? the result of an entirely new process.
Questions such as these led to the commissioning of this
Report by WCRF/AICR in 2001. The Panel responsible for Rigorous criteria to assess evidence
the Report first convened in 2003, and has met twice a The criteria used in this Report to assess the evidence
year until 2007. The terms of reference accepted by the presented in the SLRs and from other sources are more
Panel at its first meeting were to: precise and explicit than, and in some respects different
from and more stringent than, those used in the previous
Judge the reviews of the scientific and other literature report. During its initial meetings, the Panel reviewed and
prepared for the Panel by the assigned review teams agreed these criteria before embarking on the formal
evidence review. More details are given in Chapter 3.
Devise a series of dietary, associated, and other Nevertheless, readers and users of this Report should be
recommendations suitable for all societies, designed to able to see how and why the development of scientific
reduce the risk of cancer method and research since the mid-1990s has resulted in
conclusions and recommendations here that sometimes
Evaluate the consistency between such vary from, sometimes are much the same as, and
recommendations and those designed to prevent other sometimes reinforce those of the previous Report.
food-related diseases.
Graphic display of Panel judgements
The Panel believes that these terms of reference have been The Panel has retained the matrix technique of displaying
fulfilled. The public policy implications of the its judgements, which introduce the chapters and chapter
recommendations made in this Report are the subject of a sections throughout Part 2 of this Report. This technique,
further report, to be published in late 2008. pioneered in the first report, has been adapted by the

F O O D , N U T R I T I O N , P H Y S I C A L A C T I V I T Y, A N D T H E P R E V E N T I O N O F C A N C E R : A G L O B A L P E R S P E C T I V E

World Health Organization in its 2003 report on diet, evidence that greater body fatness (there termed high
nutrition, and the prevention of chronic diseases. Some body mass) is a convincing or probable cause of cancers of
members of the expert consultation responsible for the the endometrium, breast (postmenopausal), and kidney.
WHO report, including its chair and vice-chair, have served For this Report, the commissioned SLRs not only included
as members of the Panel responsible for this Report. the evidence linking body fatness directly with cancer, but
In further adapting the format of the matrices used in a separate review was also commissioned specifically on
the first report, the Panel was careful to distinguish the biological and associated determinants of body fatness
between evidence strong enough to justify judgements of itself. The evidence and the Panels judgements, which
convincing or probable causal relationships, on which include assessment of the physiology of energy
recommendations designed to prevent cancer can be metabolism, are summarised in Chapters 6 and 8.
based, and evidence that is too limited in amount, The Panel is aware that weight gain, overweight, and
consistency, or quality to be a basis for public and personal obesity, and their antecedent behaviours, are critically
health recommendations, but which may nevertheless in determined by social, cultural, and other environmental
some cases be suggestive of causal relationships. factors. This is one topic for the separate report on policy
implications to be published in late 2008.
Food-based approach
Since the 1990s a broad food- and drink-based approach to Cancer survivors
interpreting the evidence on food, nutrition, and the risk of There are increasing numbers of cancer survivors people
cancer has increasingly been used, in contrast to the who have at some time been diagnosed with cancer. What
overwhelming research emphasis on individual food should those people living with cancer do? Particularly
constituents. The previous report included three chapters since the 1990s, this question is being asked increasingly,
showing the findings on dietary constituents (including as more and more people are diagnosed with and treated
energy and related factors, notably physical activity), for cancer, and are seeking ways in which they can add to
foods and drinks, and food processing (meaning their medical or surgical management to help themselves
production, preservation, processing, and preparation), in to remain healthy. Are the circumstances of people who
that order. have recovered from cancer any different from those of
This Report has taken a food-based approach, as shown people who are free from cancer? Questions such as these
throughout Chapter 4, more closely reflecting the nature of are addressed in Chapter 9.
the evidence. Thus many findings on dietary constituents
and micronutrients, when their dietary sources are from Life-course approach
foods rather than supplements, are here identified as, for Unlike this Report, the reviews conducted for the first
example, findings on foods containing dietary fibre or report did not consider the literature on food and nutrition
foods containing folate. Findings on methods of food in the first two years of life. Increasingly, evidence is
processing are, wherever possible, shown as part of the accumulating on the importance of early life-events on
evidence on the associated foods, so that, for example, later health. Evidence and judgements on the impact of
meat processing is integrated with the evidence on meat. birth weight and adult attained height on cancer risk are
The evidence and judgements focused on cancer are presented in Chapter 6, though the detailed processes
summarised and displayed in Chapter 7. underpinning these associations with cancer risk are not
yet clear. Findings on the relationship between not being
Physical activity breastfed and later overweight and obesity in children are
The scope of the work of this Panel is wider than that of reported in Chapter 8, and on lactation and lower breast
the previous panel. The previous report judged that the cancer risk in the mother are reported in Chapter 7. These
evidence that physical activity protects against cancer of findings form part of a general life-course approach
the colon was convincing. Since then evidence on physical summarised in Chapter 2, reflecting an appreciation of the
activity (and physical inactivity, especially when this importance of the accumulation of nutritional and other
amounts to generally sedentary ways of life) has become experiences throughout life, as well as genetic endowment,
more impressive. Correspondingly, the review centres were in influencing susceptibility to disease.
requested specifically to examine the literature on physical
activity (and inactivity) as well as on foods and drinks. The Goals and recommendations
results of this work, and the Panels judgements, are shown The Panels recommendations are set out in Chapter 12
in Chapter 5. and in abbreviated form in the Summary, on the preceding
Body fatness The previous report agreed 14 recommendations. This
As with physical inactivity, the evidence that body fatness Report makes eight general and two special
including degrees of fatness throughout the range of recommendations for specific target groups. These are set
body weight, from underweight and normal to overweight out in more detail than in the previous report. As before,
and obesity, as well as any specific effect of weight gain principles that guide the goals and recommendations are
directly influences risk of some cancers has also become set out. The recommendations themselves are displayed in
more impressive. The previous report judged that the boxes and are accompanied by text that justifies them, and


by practical guidance. The recommendations are addressed scale this represents over 3 to 4 million cases of cancer that
to people, as members of communities and families and can be prevented in these ways, every year.
also as individuals. In many of its forms, cancer is a disease that can cause
Recommendations and options addressed to UN and great suffering and claims many lives. The overall
other international organisations, national governments, commitment of scientists and other professionals
industry, health professional and civil society organisations, committed to disease prevention, as exemplified by this
and the media are set out in the separate report on policy Report, is to reduce the rates not just of cancer, but of all
implications, to be published in late 2008. diseases, so that more people enjoy good health until they
eventually die in old age.
A note of caution
The Panel is confident that its findings are soundly based,
and that its recommendations, when translated into R e f e re n c e s
effective public policy programmes and personal choices, 1.
National Research Council. Diet, Nutrition and Cancer. Washington
will reduce the risk of cancer. That said, the available DC: National Academy of Sciences, 1982
evident is imperfect. The Panels conclusions derive from
Willett W. Summary. In: Eat, Drink, and Be Healthy. The Harvard
Medical School Guide to Healthy Eating. New York: Free Press, 2003
the best evidence now available, which reflects past and 3.
Doll R, Peto R. The causes of cancer: quantitative estimates of
recent research priorities mostly in high-income countries, avoidable risks of cancer in the United States today. J Natl Cancer Inst
though synthesised and judged in as meticulous and 1981;66:1191-308.
rigorous way as possible. What is here is therefore an
incomplete picture.
The tendency of reports such as this is to consider
diseases in isolation. In the case of this Report, the
relationship of weight gain, overweight, and obesity on the
risk of some cancers is so clear that determinants of these
factors have also been considered. But the Panel agrees, as
evident in Chapters 10 and 12, that many chronic diseases,
including type 2 diabetes and its precursors, cardiovascular
diseases and their precursors, and also perhaps other
diseases of the digestive, musculoskeletal, and nervous
systems, are to a large extent caused by environmental
factors, including inappropriate food and nutrition,
physical inactivity, overweight and obesity, and associated
factors. Following from this, future reports should consider
the promotion of health and the prevention of disease as a

H o w m u c h c a n c e r i s p re v e n t a b l e ?
As shown in its title, the purpose of this Report is to
prevent cancer. The term prevention needs definition. It
does not mean the elimination of cancer. It means
reduction in its occurrence, such that at any age fewer
people have cancer than otherwise would be the case.
If all factors are taken into account, cancer is mostly a
preventable disease. The authors of a landmark study
published in the early 1980s concluded: It is highly likely
that the United States will eventually have the option of
adopting a diet that reduces its incidence of cancer by
approximately one third, and it is absolutely certain that
another one third could be prevented by abolishing
smoking.3 Cancers of some sites, notably of the colon, are
generally agreed to be greatly or mostly affected by food
and nutrition.
Since then, authoritative estimates of the preventability
of cancer by means of food and nutrition and associated
factors have been in broad agreement with the around one
third figure. The estimate of the previous WCRF/AICR
Report was that cancer is 30 to 40 per cent preventable
over time, by appropriate food and nutrition, regular
physical activity, and avoidance of obesity. On a global


Part 1

> V } `

Chapter 1
International variations and trends 4

Chapter 2
The cancer process 30

Chapter 3
Judging the evidence 48
P A R T 1

Introduction to Part 1

This Report has a number of inter-related general purposes. One is to explore

the extent to which food, nutrition, physical activity, and body composition
modify the risk of cancer, and to specify as far as possible the importance of
specific factors. To the extent that environmental factors such as food,
nutrition, and physical activity influence risk of cancer, it is a preventable
disease. The Report specifies recommendations based on solid evidence
which, when followed, will be expected to reduce the incidence of cancer.

Part 1 of the Report begins with two chapters summarising the first lines of
evidence from observations of human populations, and from experimental
and basic science, pointing to the conclusion that cancer is preventable. The
third chapter summarises the types of evidence that are relevant in
identifying the causes of cancer, and explains the process used by the Panel
to assess the strength of this evidence and to come to judgement.

Chapter 1 shows that patterns of production and consumption of food and

drink, of physical activity, and of body composition have changed greatly
throughout different periods of human history. Remarkable changes have
taken place as a result of urbanisation and industrialisation, at first in
Europe, North America, and other economically advanced countries, and
increasingly in most countries in the world.

With the establishment of reliable records in the second half of the 20th
century, notable variations have been identified in patterns of cancer
throughout the world. Some cancers, such as those of the upper
aerodigestive tract, stomach, liver, and cervix, are more common in lower
income countries; others, such as those of the colorectum, breast, ovary,
endometrium, prostate, and lung, are more common in higher income

More significant, as shown in Chapter 1, are studies consistently showing

that patterns of cancer change as populations migrate from one part of the
world to another and as countries become increasingly urbanised and
industrialised. Projections indicate that rates of cancer in general are liable
to increase.

Chapter 2 outlines current understanding of the biology of the cancer

process, with special attention given to the ways in which food and
nutrition, physical activity, and body composition may modify it.

Cancer is a disease of genes, which are vulnerable to beneficial or harmful

mutation, especially over the long human lifespan. Nutritional factors are
important in determining the likelihood of some mutations, as well as in
changing the functions of genes even without mutation. However, both
epidemiological and experimental evidence shows that only a small
proportion of cancers are inherited. Environmental factors are most

important and can be modified. These include smoking and other use of
tobacco; infectious agents; radiation; industrial chemicals and pollution;
medication and also many aspects of food, nutrition, physical activity, and
body composition. Essentially this is good news. It means that healthy
environments can stop cancer before it starts. The evidence also indicates that
such environments, including the factors that are the subject of this Report,
may be able to check the cancer process after it has started.

The third chapter summarises the types of evidence that the Panel has agreed
are relevant to its work. No one study can prove that any factor definitely is a
cause of or is protective against any disease. Also while some study designs
are more reliable than others, they often cannot be used to answer many types
of question; so no one kind of study, however careful its methods, can ever
produce definitive results. In this chapter, building on the work of the first
report, the Panel shows that all study designs have strengths and weaknesses,
and that reliable judgements on causation of disease are based on assessment
of a variety of well designed epidemiological and experimental studies.

The judgements made by the Panel in Part 2 of this Report are based on
independently commissioned and conducted systematic reviews of the
literature. This has ensured that the evidence has been assembled using
methods that are as meticulous as possible, and that the display of the
evidence is separated from assessments derived from this evidence, which are
made in Part 2.

The prevention of cancer worldwide is one of the most pressing challenges

facing scientists and public health policy-makers, among others. These
introductory chapters show that the challenge can be effectively addressed.
They also suggest that food and nutrition, physical activity, and body
composition all play a central part in the prevention of cancer.


C H A P T E R 1

International variations
and trends
The first lines of evidence suggesting that cancer is aerodigestive tract (of the mouth, pharynx, larynx,
a largely preventable disease have come from nasopharynx, and oesophagus), and of the stomach,
studies noting variations in cancer incidence across liver (primary), and cervix. Rates of some cancers,
time and place. The most impressive initial evidence especially stomach cancer, are now generally
showing that patterns of cancer are altered by decreasing.
environmental factors, and are not mainly In contrast, high-income countries, and urbanised
genetically determined, comes from studies and industrialised areas of middle- and low-income
describing changes in the rates of different cancers regions and countries, have higher rates of
in genetically identical populations that migrate colorectal cancer and of hormone-related cancers (of
from their native countries to other countries. Such the breast, ovary, endometrium, and prostate). Lung
studies consistently show that changes in the rates cancer is now the most common type in the world
of some of the most common cancers, including because of the increase in tobacco smoking and
those of the stomach, colorectum, breast, and exposure to environmental tobacco smoke. Rates of
prostate, can be remarkable, even over one or two these cancers, some of which may have been
generations. historically rare, are increasing.
This first introductory chapter summarises current Globally, the number of people with cancer is
knowledge of the variations in food, nutrition, projected to double by the year 2030, with most of
physical activity, body composition, and cancer in this increase likely to occur in middle- and low-
different parts of the world. This assessment income countries. Such an increase would only
provides strong circumstantial evidence that partly be accounted for by the projected rise in the
continues to prompt systematic studies including size and average age of the global population. This
interventions of various types, and also reports such makes the task of cancer prevention all the more
as this, which collect and judge the available urgent and important.
evidence. Such systematic work has already led the
United Nations and other international bodies,
national governments, and authoritative
independent organisations to be confident that
most cancers are largely preventable.
Patterns of food and drink, of physical activity,
and of body composition have changed remarkably
throughout human history. With industrialisation
and urbanisation, food supplies usually become
more secure, and more food is available for
consumption. In general, diets become more energy
dense, containing fewer starchy foods, more fats
and oils, sugars, and additives, and often more
alcoholic drinks. At the same time, patterns of
physical activity change: populations become
increasingly sedentary, their need for energy from
food drops, and rates of overweight and obesity
These changes correlate with changes in the
patterns of cancer throughout the world. Middle-
and low-income regions and countries within Africa,
Asia, and Latin America have generally experienced
comparatively high rates of cancers of the upper


This chapter outlines the historic, recent, and current varia- 1.1 Food systems and diets: historical and
tions and trends in food, nutrition, physical activity, over- current
weight and obesity, and in patterns of cancer.
Peoples diets reflect the times and situations in which they Throughout history, food systems, and thus human diets,
live. It is only relatively recently in history that urbanindus- have been and are shaped by climate, terrain, seasons, loca-
trial ways of life have evolved, with many or most people tion, culture, and technology. They can be grouped into three
living in towns and cities rather than in the countryside. In broad types: gathererhunter, peasantagricultural, and
much of Africa and Asia, most people still live in rural com- urbanindustrial. These and other food systems (for exam-
munities, and peasantagricultural and urbanindustrial ple, pastoralist, the semi-mobile farming of herds of large
ways of life still coexist in most countries. Such patterns animals such as sheep and cattle) have their roots in histo-
change very rapidly as countries become increasingly ry. All have coexisted in recent millennia with the exception
urbanised and industrialised. of industrial food systems, which are the consequence of the
The different food systems and diets that are part of these industrial revolution that began in Europe in the late 18th
diverse ways of life affect peoples levels of physical activity, century. These systems still exist in the world today.
their body composition and stature, their life expectancy, and
patterns of disease, including cancer. With the move to 1.1.1 Gathererhunter
urbanindustrial ways of life, populations have become taller Since the emergence of Homo sapiens around 250 000 years
and heavier, their life expectancy has increased, and they are ago, gathererhunter food systems have taken different
usually adequately nourished (although poverty, and even forms, depending on the environments in which people lived.
destitution, remains a major problem in most big cities). On These systems still exist in parts of the world that are remote
the other hand, urban populations are at increased risk of from cities and roads. They supply diets that usually include
chronic diseases such as obesity, type 2 diabetes, coronary moderate amounts of starchy foods, and which are high in
heart disease, and also some cancers. dietary fibre and low in sugar, mostly from fruits and
This chapter also summarises some available information honey.48 Methods of food preparation include pulverising,
on eight common cancers, irrespective of any recognised links drying, and roasting. These diets are usually high in foods
to food, nutrition, and physical activity; these factors are dealt of animal origin (ranging from large animals to insects, and
with later in the Report. Four are endemic in middle- and low- also fish and other seafood, depending on location), and
income countries: cancers of the oesophagus, stomach, liver, thus in animal protein. It is sometimes thought that
and cervix. Four are endemic in high-income countries, and gathererhunter diets are high in fat, which is not the case
are in general increasing in middle- and low-income countries: because wild animals are lean. Recent analyses suggest that
cancers of the lung, colon and rectum, breast, and prostate. gathered food generally provides rather more dietary bulk
Information on the trends and projections of levels of physi- and energy than hunted food.49
cal activity, and overweight and obesity, is summarised. People in gathererhunter societies are necessarily physi-
Descriptive epidemiology, including studies of changing dis- cally active, and are often tall and usually lean (only chiefs,
ease patterns in migrant populations, is covered. These stud- or old or incapacitated people might be overweight or
ies can generate hypotheses about relationships between food, obese). The diets of food-secure gathererhunter societies
nutrition, physical activity, and the risk of cancer. However, may be diverse and high in micronutrients.50 51 Diets are
they serve mainly as a foundation for studies that provide liable to become monotonous and deficient in various nutri-
stronger evidence. ents, as well as in energy, when food supplies are chronically
The 12 national examples provided throughout this chap- insecure, or at times of acute food shortage. It is sometimes
ter summarise some of the trends in foods and drinks, obesi- claimed that gathererhunter food systems generate diets to
ty, physical activity, and cancer in countries around the world. which the human species is best adapted.48 However, life
These are Egypt and South Africa (Africa); China, India, and expectancy in gathererhunter societies is and has been usu-
Japan (Asia); the UK, Poland, and Spain (Europe); the USA, ally relatively low.
Brazil, and Mexico (the Americas); and Australia (Asia- Evidence of cancer has occasionally been found in human
Pacific).1-47 and other fossil and ancient remains.52 Historically, cancer



In 2004 Egypt had a population of just over total area. Egypt has a lower-middle-income men and 70 for women (figure 1.1).46
74 million. Nearly the whole population economy, with a gross domestic product of Chronic diseases account for 83.6 per
lives within the Nile Valley and the Nile 4274 international dollars per person (figure cent of deaths, while infectious diseases,
Delta, less than 4 per cent of the countrys 1.3). Life expectancy at birth is 66 years for maternal, perinatal, and nutritional condi-

Non-communicable causes of death Egypt Age-standardised rates of common cancers Egypt

Age-standardised rate per 100 000

Per cent of deaths
Cardiovascular disease
Men Women
24 Cancer

2 58 Respiratory disease

7 10








Data from World Health Organization46 Data from International Agency for Research on Cancer20

of any type seems to have been uncommon among gather- cereals are refined). They include varying amounts of foods
erhunter peoples, if only because their average life of animal origin, and of vegetables and fruits, depending on
expectancy was low. In modern gathererhunter societies, relative food security. Surplus food is stored for consump-
the incidence of cancer rises after contact with industrialised tion in winter and during hard times, and methods of food
and urbanised ways of life, which usually involve shifts in preparation also include fermentation, used for foods as well
patterns of diet and physical activity.53 These points gener- as for the production of alcoholic drinks (see chapters 4.8
ally also apply to pastoralist societies. and 4.9).
The dominant indigenous cereal crop varies in different
1.1.2 Peasantagricultural parts of the world: wheat is grown in the Middle East; bar-
In recent millennia, and until very recently in history, almost ley, rye, and oats in colder, northern climates; millet and rice
all human populations have been rural and mostly peas- in Asia; maize (corn) in the Americas; and sorghum and teff
antagricultural, and the majority still are in most regions of in Africa. Indigenous staple crops also include roots and
Asia, many regions of Africa, and some parts of Latin tubers such as cassava (manioc), yams, potatoes, and also
America. Peasantagricultural food systems involving the cul- plantains. Pulses (legumes) are also farmed to ensure agri-
tivation of wheat may have first developed around 9000 cultural and nutritional balance; and other crops such as
years ago in the Fertile Crescent of the Middle East, includ- vegetables and fruits are also cultivated. Birds and animals
ing the region between the Tigris and Euphrates rivers (with- are domesticated and bred for food, and fish and seafood
in modern Iraq). These systems also developed contribute to the diets of communities living beside water.57
independently in Asia, with rice as the staple food, and in As with gathererhunters, the diets of peasantagricul-
the Americas, with corn (maize) as the staple.54 The key fac- tural societies may be diverse and high in micronutrients.
tor in these systems is land settlement, itself determined by Again, when food supplies are chronically insecure, or at
the cultivation and breeding of crops and also animals, birds, times of acute food shortage (including times of war), diets
and fish for human consumption and use.55 In and around are liable to become monotonous and deficient in various
Egypt, people began to make bread from wheat about 6000 nutrients, as well as in energy.
years ago.56 Peasantagricultural societies are necessarily physically
Typically, diets derived from these systems are plant-based: active, although not constantly so: the main times of inten-
they are high or very high in cereals (grains), complement- sive physical work include building field systems, sowing,
ed with animal sources of protein. These diets are therefore harvesting, and storing. The level of energy balance and of
high in starchy foods and usually in dietary fibre (unless the physical activity varies greatly, depending in part on how dif-


tions account for 12.2 per cent; 4.2 per cent 23.5 per cent of all women had a BMI of Consumption of sugars and oils increased
of deaths are due to injuries. The first fig- over 30. By 2000, this figure had risen to 41 substantially and pulses (legumes)
ure gives a breakdown of deaths caused by per cent.46 Fewer data are available for decreased in importance.15 Since the 1970s,
chronic diseases.46 men, but in 1994, the mean BMI for men consumption of all major food groups has
Bladder cancer is the most common type aged 2044 was 26.6, rising increased. However, between 1990 and
of cancer in men, followed by cancers of the to 28.4 for men over 45. In 2002, among 1994 there was a 20 per cent decrease in
lung and liver.20 In women, the dominant all men, 45 per cent had a BMI of between total household food consumption, because
cancers include those of the breast, cervix, 25 and 29.99 and 20 per cent had a BMI of subsidies were removed and food prices
and bladder (for age-standardised rates of over 30.15 See figure 1.4 for projections of rose sharply.15 A national study in 1981
these cancers, see the second figure).20 The the proportions of men and women pre- found that only 24 per cent of urban and
high incidence of bladder cancer is likely to dicted to have a BMI of 30 or more in 15 per cent of rural households ate ready-
be related to bilharzia, a parasitic infection 2015.46 made foods, and meat was eaten more fre-
of the bladder.20 There is also a high inci- The average amount of available food quently in urban households compared
dence of hepatitis C virus, a cause of liver energy rose between 1964 and 2004, from with rural households.15 A repeat survey in
cancer.20 Also see box 7.8.1 It is predicted around 2240 to 3290 kcal/person per day 1998 found that poultry had become the
that there will be a 3.5-fold increase in liver (9400 to 13 800 kJ/person per day).1 Early main source of animal protein and that
cancer by 2030.12 dietary studies in Egypt demonstrated that wheat bread was the most popular type,
For the period 19911994, 46 per cent of corn bread was the staple food and that although homemade wheat-maize bread
men and 48 per cent of women between protein intake was about 100 g/day.15 was common in rural areas. Another study
the ages of 20 and 44 were classified as People from higher-income households highlighted differences in dietary fat
sedentary.46 In 2003, women aged 1549 consumed more dairy products and those intake: in urban women, 27.5 per cent of
had a mean body mass index (BMI) of 28.6. from urban households consumed a wider dietary energy came from fat (mainly as
In teenagers (1319 year olds), average BMI variety of foods. Between 1950 and 1990, vegetable oil) compared with 22.5 per cent
was 23.9; women in their 30s had a mean there was a shift towards a dependence on in rural women.15 Between 1981 and 1998,
BMI of 29.0, while those over 45 had a BMI wheat rather than other cereals (grains), people increasingly ate meals away from
of 31.3. In total, 77.3 per cent of women and a sustained rise in the consumption of home (20.4 per cent of all meals in 1981
aged 1549 had a BMI of over 25. In 1992, meat, fish, and dairy products.15 compared with 45.8 per cent in 1998).15

ficult it is to cultivate the land. The degree of physical activ- are less reliable than those kept in urbanised societies. But
ity and so of body mass in peasantagricultural communi- there is reasonable evidence that relatively common cancers
ties depends mostly on relative food security.58 in peasantagricultural societies include those causally
People in these societies who are prosperous, especially associated with chronic infections, such as cancers of the
those who own land farmed by others, may quite often stomach, liver, and cervix.60
become overweight or obese. But in general, and largely
because of the nature of their dietary staples, peasantagri- 1.1.3 Urbanindustrial
culturalists are usually short and lean. This is still evident in Indigenous or traditional peasantagricultural systems have
rural peasant communities whose food systems remain tra- coexisted with urbanindustrial food systems in most coun-
ditional: for instance, in Africa, Latin America, and in Asia, tries since the creation and growth of cities, and the begin-
notably India and China.58 ning of the industrial revolution. This movement started in
Agriculture enabled the development of towns and then Europe in the 18th century, and then spread to North
cities: throughout the world, walled, urban settlements America and elsewhere. Britain is one exception to this coex-
became surrounded by fields cultivated by peasants. These istence: it was the first country to become mostly urban, with
people subsisted on the food they produced, and the surplus hired workers replacing peasants on increasingly large and
fed the community living within the walls. In times of war, relatively mechanised farms. The Americas are another
the fortified settlement became a refuge for the farmers. This exception: settlers, mostly from Europe, displaced native
crowding of populations into towns and cities caused a sharp populations and developed mechanised agricultural sys-
rise in the rates of infectious diseases, mostly notably among tems.61 In continental Europe, some balance between rural
infants and young children, pregnant and lactating women, and urban ways of life has been preserved. Throughout the
and infirm and old people.59 Mediterranean coastal regions, and in the Middle East, mod-
The average life expectancy of peasantagriculturists ern food systems have deep, historical roots.62 In most of
in general is probably a little longer than that of Africa and Asia, including countries with large cities, the
gathererhunters, with a greater percentage of people basic economies and cultures have remained predominantly
surviving into what would be regarded as late-middle and rural, but this is changing.63
old age. Urbanindustrial food systems have characteristics distinct
The prevalence and incidence of various cancers in tradi- from peasantagricultural and gathererhunter systems.
tional rural societies is often uncertain, even following the Their original purpose was to ensure reliable and adequate
establishment of cancer registers in many countries: records supplies of food of an agreed minimum nutritional quality


South Africa

In 2001 South Africa had a population of women (figure 1.1).46 The most common cancers in men are
nearly 47.5 million.3 The country has a mid- Chronic diseases account for 53.9 per cent those of the prostate, lung, oesophagus,
dle-income economy, with a gross domestic of deaths, while infectious diseases, maternal, colorectum, and bladder.20 Since HIV and
product of 8506 international dollars per perinatal, and nutritional conditions account AIDS became epidemic, Kaposis sarcoma
person (figure 1.3), which masks extreme for 40.2 per cent; 5.9 per cent of deaths are has become more common in both men
socioeconomic inequalities.46 Life expectan- due to injuries. The figure below gives a break- and women. For women, the most common
cy at birth is 47 years for men and 49 for down of deaths caused by chronic diseases.46 cancers are those of the cervix, breast,

Non-communicable causes of death South Africa Age-standardised rates of South Africa

common cancers

Per cent of deaths Age-standardised rate per 100 000

Cardiovascular disease 40
Men Women
13 35

Cancer 30
10 51 Respiratory disease
Diabetes 5










Data from World Health Organization46 Data from International Agency for Research on Cancer20

to entire populations. Technology has been the main driving lic health recommendations, notably when food security has
force behind these systems. For instance, various food-preser- been threatened by wars.69
vation techniques were developed as part of the industrial Urbanindustrial food systems generate relatively energy-
revolution, and there has been further innovation since that dense diets. These are fairly high in meat, and milk and their
time. These include bottling, canning, refrigeration, and products, and in total fats, hardened fats, processed starch-
packaging; the extensive use of sugar and salt; and tech- es and sugars, salt, baked goods, soft drinks, and often also
nologies that suppress, convert, or eliminate perishable qual- alcoholic drinks. These diets are relatively low in both
ities in fresh foods (see chapters 4.6 and 4.9). The clearing dietary fibre and starchy staple foods, other than products
of land to rear cattle and sheep, and the development of rail- made from wheat, which has become the dominant cereal
ways, refrigeration, and other technologies, have made meat, in most countries outside Asia and Africa. Recent advances
milk, and their products cheap and plentiful all year round. in food technology have further altered patterns of food pro-
Sugar derived from cane is the most profitable edible cash duction and consumption, particularly in high-income
crop, and sugars and syrups made from cane, beet, and now countries. Patterns of production and consumption of
also corn are used to sweeten and preserve breakfast foods, vegetables and fruits and fish vary between different
baked foods, desserts, soft drinks, and a vast array of other urbanindustrial food systems, depending on factors such as
manufactured products.64 65 Steel roller mills, invented in the climate and geographical location.70
1870s, separate the components of wheat and enable pro- Efficient urbanindustrial food systems can ensure the
duction of uniform quality white bread, which has become constant supply of food to all sections of the population,
a staple food.66 Hydrogenation, which converts oils to hard even to the lowest-income and marginalised groups. In high-
fats (see chapter 4.5), has made margarine a basic item of er-income countries and regions, this, together with basic
food, and provides ingredients used in the manufacturing of public health initiatives, has helped to greatly reduce rates
many processed foods.67 Perhaps the most remarkable of nutritional deficiencies and other diseases, which people
change following the industrialisation of food systems has are more vulnerable to if they have inadequate food supplies.
been the precipitate drop in breastfeeding.68 At various times, As a result of these food systems, people have become gen-
urbanindustrial food systems have been adjusted in erally taller and heavier.
response to the then current knowledge of nutrition and pub- Since the industrial revolution, as populations have moved


colorectum, lung, and oesophagus (for age- also prevalent. There has been a miscon- BMI, and commonly consumed foods were
standardised rates of these cancers, see the ception of benign obesity: being thin is maize, sugar, tea, milk, and brown bread.
second figure).20 Diseases of poverty and associated with HIV and AIDS, and moder- Urban households ate less maize porridge
chronic diseases coexist, but it is predicted ately overweight women are thought of as but more vegetables and fruits, animal-
that by 2010, deaths from AIDS will account attractive and affluent.24 Overall, in 1998, based products, and fats and oils. It was
for twice as many deaths as those from all 21.1 per cent of men and 25.9 per cent of only in urban areas that fruits and milk
other causes combined.5 41 women had a BMI of at least 25; 10.1 per appeared in the top 10 list of foods and
For the period 20022003, 44 per cent of cent of men and 27.9 per cent of women drinks consumed by more than 85 per cent
men and 49 per cent of women aged 1869 had a BMI of at least 30. See figure 1.4 for of people. In men, alcoholic drinks made a
were classified as sedentary (figure 1.6).46 projections of the proportions of men and significant contribution to dietary energy
Some regional studies suggest that young women who will have a BMI of 30 or more (1014 per cent). People living in rural areas
women who did not finish school have low in 2015. obtained a higher proportion of total
levels of physical activity.24 There is a lack of The average amount of available food dietary energy from carbohydrates, where-
physical education in schools, and poor energy rose between 1964 and 2004, from as the most urbanised populations derived
environment and high crime rates prevent around 2700 to 3000 kcal/person per day one third of their energy from animal foods
leisure activity outside school.24 (11 400 to 12 600 kJ/person per day). In the high in protein.
In 1998 men aged 1524 had an average same period, sugar consumption dropped Urbanisation is generally accompanied
body mass index (BMI) of 21.1; for those from 420 to 370 kcal/person per day (1800 by an improvement in micronutrient
aged 3565, average BMI remained con- to 1500 kJ/person per day).1 The National intakes, but this way of life is also associat-
stant at around 25. Just 7.8 per cent of men Food Consumption survey of 1999 found ed with increases in overweight and obesi-
aged 2534 had a BMI of over 30 compared that stunting was the most common nutri- ty.44 Other studies have suggested shifts
with 17.3 per cent of men aged 4554. In tional disorder, affecting almost one fifth of towards a Western dietary pattern in peo-
the same year, women aged 1524 had an children aged 19, with the lowest levels in ple living in both urban and rural areas, typ-
average BMI of 23.7; for those aged 3564, urban areas. There was a similar pattern for ified by a decrease in starchy foods and
average BMI remained constant at around underweight, where 10 per cent of children dietary fibre consumption, and an increase
29. In women aged 2534, 27 per cent had aged 13 consumed less than half of their in fat. They have also shown that half of
a BMI of over 30 compared with 45 per cent suggested daily dietary needs, and 26 per the population does not eat the locally
of women aged 4564.46 Although under- cent consumed less than two thirds.25 In recommended four portions of fruits
nutrition remains a problem among rural rural areas, adults from lower-income and vegetables each day, while a quarter
children, obesity and associated diseases are households were shorter and had a lower eats none.10

from rural to urban areas, there have been rapid and pro-
Figure 1.1 Life expectancy at birth
found changes in both the nature and quality of their foods
Years and drinks, and the patterns of diseases they suffer.71
Asia-Pacific North America
90 Urbanindustrial food systems have evidently improved peo-
Africa Asia Europe Latin
America ples strength and health in early life. They are also a factor
in the doubling of average life expectancy since 1800, and
the increase in global population from around 1 billion in
1800 to 6.5 billion in 2006.72 The range of current life
60 expectancy in selected countries is illustrated in figure 1.1.
In the second half of the 20th century, attention focused
on the apparent ill-effects of these food systems on people,
40 mostly in later life. By the 1980s, it was generally agreed that
these industrialised diets increase the risk of some chronic
30 diseases, usually of later life, which had become common or
epidemic in higher-income industrialised countries. These
included obesity, type 2 diabetes, and coronary heart disease.
10 At the same time, in examining patterns of both diet and can-
cer across the world, and among migrants, it was increas-
ingly thought that these diets were partly responsible for

South Africa











some cancers, notably those of the colon and rectum, breast,

ovary, endometrium, and prostate.73-75
In the last decades of the 20th century, the demographic,
Men Women nutritional, and epidemiological transitions that had, until
then, largely been apparent only in higher-income countries
Data from World Health Organization46
became global. Since the 1990s, and outside Europe, North
America, and other high-income countries, economic glob-



In 2004 China had a population of over 1.3 1.3). Life expectancy at birth is 70 years for 2004 found that there has been a shift
billion. The one-child policy introduced in men and 74 for women (figure 1.1).46 towards nutrition-related chronic diseases
1979 has reduced annual population Chronic diseases account for 78.9 per cent such as type 2 diabetes, cancer, and cardio-
growth to 1.07 per cent. The United Nations of deaths, while infectious diseases, mater- vascular disease.14
estimates that the population will have nal, perinatal, and nutritional conditions Stomach cancer is the most common type
increased to nearly 1.5 billion by 2025.46 The account for 11.7 per cent; 9.3 per cent of of cancer in men, although it has declined
country has a lower-middle-income econo- deaths are due to injuries.46 The figure slightly since 1980.20 Lung cancer has risen
my, with a gross domestic product of 5581 below gives a breakdown of deaths caused steadily over the same period.20 Liver cancer
international dollars per person (figure by chronic diseases.46 A study published in has risen since 1990, although levels are now

Non-communicable causes of death China Age-standardised rates of common cancers China

Age-standardised rate per 100 000

Per cent of deaths
Cardiovascular disease Men
35 Women
2 10 30
Cancer 25
22 44
Respiratory disease 15










Data from World Health Organization46 Data from International Agency for Research on Cancer20

alisation is thought to be the single main force shifting tial to sustain young populations that walk or cycle to work
populations from the countryside into cities, changing the that is physically demanding, are unsuitable for ageing pop-
dominant food systems from peasantagricultural to ulations who sit for most of the day, even if they engage in
urbanindustrial, and transforming patterns of disease. This some recreational physical activity.
phenomenon includes the unprecedented and accelerating There is some evidence that these very recently introduced
movement of money, goods, and ideas. All this has been urbanindustrial food systems have lowered the rates of
made possible by new international, political, and econom- nutritional deficiencies and infectious diseases of early life
ic policies, by the creation of supranational regulatory bod- in middle- to low-income countries and regions. But the
ies such as the World Trade Organization, and by the apparent impact on the rates of chronic diseases in these
electronic revolution.76-78 areas is of increasing public health concern. In most of these
Peoples levels of physical activity have also changed dra- regions, with the exception of sub-Saharan Africa, childhood
matically as a result of the move from peasantagricultural overweight, obesity, and type 2 diabetes have become com-
to urbanindustrial ways of life. In 1950, the UN reference mon and, in some countries, epidemic.70
man weighing 65 kg (143 lbs) was estimated to be in ener- The Panel emphasises that there is no reason to think that
gy balance at an average of 3200 kilocalories (kcal)/day urbanindustrial food systems are intrinsically harmful. They
(13398 kilojoules [kJ]/day); the reference woman weighing were first developed using relatively crude technologies, and
55 kg (121 lbs) was estimated to be in energy balance at 2300 at a time when something was known of their positive
kcal/day (9630 kJ/day). Today in the USA, average weights impact on growth and strength, but little of their long-term
are much higher, yet the figure for the reference person (men impact on health. Since then, many new technologies have
and women combined) is taken to be 2000 kcal/day (8374 been developed, and there is a clearer understanding that
kJ/day) for the purposes of nutrition food labelling. The rea- some methods of preserving and processing food are bene-
son for this drop in human energy requirements is because ficial, whereas others are a factor in increasing the risk of
three of the four settings for physical activity occupation- disease. Future developments can ensure universal food
al, household, and transport (see Chapter 5) have become security, avoid earlier mistakes, and reduce the risk of chron-
increasingly mechanised. Energy-dense food systems, essen- ic diseases, including cancer.


stabilising.20 The incidence of cancer of the energy rose between 1964 and 2004, from varied, resulting in a diet low in saturated
oesophagus has remained stable since the around 1850 to 2940 kcal/person per day fatty acids and high in n-3 polyunsaturated
1980s and cancers of the colorectum are also (7760 to 12 290 kJ/person per day). This is fatty acids; people there eat green, leafy
relatively common.20 For women, the most largely due to an increase in the availabili- vegetables with almost every meal.26 The
common cancers are those of the lung, ty of fats and oils, meat, and sugar.1 The incidence of nutrition-related diseases and
stomach, and breast, which have risen 19571962 famine was followed by a liber- deaths from these diseases are lower in this
steadily since the 1980s; of the liver, which alisation of food production. Economic region than in other parts of China.
has risen since 1990; and of the oesophagus. growth has reduced poverty and Chinese Snacking contributes minimal energy
For age-standardised rates of these cancers, diets now are influenced by the Western intake to Chinese diets (0.9 per cent).
see the second figure.20 pattern: cereals (grains) and lower-fat mixed However, snacking and eating food away
For the period 20022003, 10 per cent of dishes are being replaced with animal foods from home are increasing among children
men and 12 per cent of women aged 1869 and edible fats.14 Recent national nutrition- from middle- and high-income groups.
were classified as sedentary (figure 1.6).46 al surveys show that energy intake from ani- Foods commonly eaten away from home
These figures are likely to increase, with fur- mal sources increased from 8 per cent in include cereals (grains), vegetables and
ther urbanisation and greater use of vehi- 1982 to 25 per cent in 2002, and that ener- fruits, meat, eggs, and fish. Between 1991
cles for transport. Between 1980 and 2003, gy from fat, particularly among people liv- and 1997, the proportion of children from
the number of cars produced in China ing in urban areas, increased from 25 to 35 low-income households eating foods away
quadrupled to more than 2 million.84 per cent over the same period.84 Intake of from home did not change, but there was
In 1997 men aged 2464 had an average cereals (grains) has also decreased substan- an increase among children from higher-
body mass index (BMI) of around 25; just 2.1 tially since the mid-1980s among urban and income groups, with a 10 per cent increase
per cent of men aged 2074 had a BMI of rural populations, with a larger decrease in in the consumption of foods from animal
over 30. In the same year, women aged the consumption of coarse grains compared sources eaten away from home. Eating food
2529 had an average BMI of 22.2, and with refined varieties. The biggest drop in prepared away from home accounted for 15
those aged 3564 had a BMI of around 25. cereal intake has been among people in the per cent of total energy intake for all
Just 3.7 per cent of women had a BMI of lowest-income groups. Vegetable and fruit Chinese children during this period.27
over 30.46 In 2002, 18.9 per cent of men and intakes have decreased since 1989, although Despite these statistics, only 10 per cent of
women aged 18 and above had a BMI of they are highest in urban populations. Fat Chinese children and young people con-
over 25, and 2.9 per cent of them had a BMI intake is also increasing and many adults sumed any snacks during the study period,
of over 30. See figure 1.4 for projections of obtain 30 per cent or more of their overall and there was little evidence then that they
the proportions of men and women who energy intake from fat.14 Regional varia- consumed significant amounts of soft
will have a BMI of 30 or more in 2015. tions also exist: for example, the dietary pat- drinks, although this is now changing
The average amount of available food tern around the city of Hangzhou is very rapidly.

1.2 Foods and drinks, physical activity, plant sources such as roots and tubers, cereals (grains), and
body composition fruits.
However, this pattern is changing, with proportionally
1.2.1 Foods and drinks more dietary energy available for consumption now coming
Substantial changes have occurred in the patterns of foods from animal sources. Since the 1960s, estimates for animal
and drinks supplied and consumed throughout the world, sources for low-income countries have risen from around
and these changes are becoming increasingly rapid. Also see 160 to 340 kcal/day (670 to 1400 kJ/day). During the same
Chapter 4. period, estimates of the energy available for consumption
Economic development is generally accompanied by from plant sources have also risen, from 1900 to 2340
quantitative and qualitative changes in food supplies and kcal/day (7900 to 9800 kJ/day) (figure 1.2). There have
therefore in diets. This nutrition transition may reduce the been similar changes in the availability of both animal and
risk of some dietary deficiencies and improve overall nutri- plant sources of energy in high-income countries. However,
tion. But it can also be accompanied by adverse shifts in the in these cases, the proportion of energy from animal sources
composition of diets, for instance, with a greater proportion is much greater: around one third or 940 kcal/day (3900
of energy coming from fats and oils, and added sugars. Over kJ/day).81 The proportion of dietary energy available from
recent years, such dietary changes have been rapid in the cereals (grains) has remained constant at around 50 per
middle- and low-income countries of Asia, Africa, the Middle cent, though dietary energy available from cereals (grains),
East, and Latin America.63 79 80 in particular rice and wheat, have decreased slightly in low-
The Food and Agriculture Organization (FAO) of the UN income countries. This trend is likely to continue until the
records global differences in the availability of food crops and 2030s in middle- and low-income countries.81
commodities (box 1.1). These data provide information on Large variations exist across the world in the amounts of
the average amounts of food available for consumption, fat available for consumption. The highest availability is in
rather than actual food consumption. Animal products have Europe and North America; the lowest is in Africa. The quan-
traditionally made up a small proportion of food availabili- tity of available fat in diets has increased globally since the
ty in low-income countries; most dietary energy comes from 1960s, with the exception of sub-Saharan Africa.81 These


Figure 1.2 Changes in available energy from animal Figure 1.3 Gross domestic product per capita, 2004
and plant food sources
International $ Asia-Pacific North America
kcal Low income Middle income High income
40 000

low income transition 35 000

Africa Asia Europe Latin
3000 America
30 000
25 000

20 000

15 000

1000 10 000

500 5000












South Africa









19671969 197779 198789 199799

Data from Food and Agriculture Organization1 Data from World Health Organization46

Food energy from animal and plant food sources in selected low-, Current gross domestic product per capita for selected countries in
middle-, and high-income countries, 19671999 international dollars

changes are accounted for by an increase in the availability kcal/day (5400 kJ/day) from these foods.83 In low-income
and consumption of plant oils in lower-income countries.82 countries between the 1960s and 1990s, consumption of
Palm oil intake is increasing in South-East Asia, and olive oil meat rose by 150 per cent, and of milk and dairy products
is now consumed widely in Europe and not just in by 60 per cent. By 2030, it is predicted that consumption of
Mediterranean countries. animal products could rise by a further 44 per cent, with the
Analysis of food balance sheet data suggests that available biggest contribution coming from poultry. If stocks of fish can
energy for consumption has increased steadily on a world- be maintained, fish consumption is likely to rise by 1920
wide basis. Since the 1960s, this has increased globally by kg/person in the same period. Owing to decreases in the cost
approximately 450 kcal/person per day (1900 kJ/person per of these foods in real terms, low-income countries have high-
day), and by more than 600 kcal/person per day (2500 er levels of meat and fat consumption at much lower levels
kJ/person per day) in low-income countries.81 Regional dif- of gross domestic product (GDP) than was the case in coun-
ferences exist. For example, there has been little change in tries that underwent socioeconomic transition in the 1960s
sub-Saharan Africa, and in Asia the amount of available ener- and 1970s. Figure 1.3 shows the GDP of selected countries.
gy has risen dramatically: in China by almost 1000 kcal/per- According to food consumption surveys, only a minority of
son per day (4200 kJ/person per day). These data need to the worlds adult population consumes the commonly rec-
be interpreted with caution, as they do not relate directly to ommended minimum daily amount of vegetables and fruits
energy consumption (box 1.1). Global average available of 400 g/person. Low-income countries have the lowest
energy is predicted to rise from around 2800 kcal/person per intakes of vegetables and fruits, and vegetables are general-
day (11700 kJ/person per day) (19971999 average) to ly more readily available than fruits.43 In India, for example,
2940 (12 300 kJ) in 2015, and to 3050 (12 800 kJ) in 2030. levels of vegetable and fruit intake have remained static at
Again, see box 1.1. 120140 g/day. Australia, Japan, and North America have
With increasing socioeconomic status, the proportion of high levels of intake, for example 300 g/day in Australia. In
energy in diets from staples such as cereals (grains) and roots Europe, average consumption is between 250 and 350 g/day
and tubers declines, whereas the proportion of energy from often much higher in Mediterranean countries, for
fats and oil, and animal protein (including from meat, milk, instance 550 g/day in Spain and Scandinavian countries
and eggs) increases. For example, in China, energy intake have particularly high fruit intakes.43 Countries in Europe,
from foods of animal origin has increased significantly: the Latin America, North America, and South-East Asia have seen
average Chinese adult now consumes more than 1300 an increase in the availability of vegetables and fruits for con-


Box 1.1 Measurement of food supply and consumption

The data here on energy, foods, and drinks are those of energy. Balance sheets over- surveys or household income/expenditure
are taken from food balance sheets com- estimate food consumption in high-income surveys.
piled by the Food and Agriculture countries, where substantial amounts of Household income/expenditure surveys,
Organization of the United Nations. These food are wasted or fed to pets. They such as the World Banks Living Standards
are statistical data on the production, underestimate consumption in countries Measurement Study, look at multiple
trade, and use of agricultural commodities that are not dominated by urbanindustri- aspects of household welfare and behav-
for all countries. Food balance sheets are al food systems, and where many people iour, and collect data on the quantities of
the most common and widely used data grow their own food, raise animals, or food purchased by a representative sample
sets for food supply estimates. A food bal- gather wild food such as fungi and berries. of households. These surveys provide
ance sheet provides estimates of the food It follows that balance sheet data showing detailed information about foods con-
available for human consumption, and an increases in food energy over time tend to sumed in and away from the home over a
overall picture of a countrys food supply reflect economic development and greater limited time period, and can be used to
during a specified period of time, which use of money, rather than actual increases document differences in regional, geo-
can be compared between countries. in availability. graphical, or household socioeconomic
It follows that these estimates of avail- The accuracy of a food balance sheet characteristics. While these surveys are
ability are not measures of consumption. also depends on the reliability of the generally more useful than food balance
They record information about the supply underlying statistics of supply, use, and sheets for assessing household consump-
of food (production, imports, stock population. Also, the data do not take into tion, they are less readily available. Balance
changes, and exports) and about how it is account regional differences, so the infor- sheets are often available for a large num-
used (as feed and seed, in processing, and mation may not be representative of the ber of countries and for most years.
wastage, as well as food). The amounts of entire country. In countries where there is Food balance sheets, household income/
foods and drinks recorded on these bal- wide variation in income and food access, expenditure surveys, and methods of assess-
ance sheets are expressed per person (in for example, the overall supply picture pro- ing individual dietary intakes (see Chapter
kg/person per year or kcal/person per day). vided by the balance sheet is of limited use. 3) all provide information on food supply
The estimates in food balance sheets In such cases, food balance sheets can be and consumption, and they have different
that need to be treated with most caution complemented with national nutrition purposes, uses, and limitations.

sumption since the 1960s. In contrast, in eastern and cen- are either overweight or obese have risen considerably. In
tral Africa, availability has decreased since the mid-1980s. 2002, there were 184 million overweight and 31 million
Studies in children suggest that their eating patterns vary obese people in China, out of a population of 1.3 billion.14
around the world. For instance, children living in the USA The prevalence of overweight and obesity among 718 year
and the Philippines consume one third of their daily energy olds increased substantially between 1985 and 2000.84
away from home, and snacks provide one fifth of their daily Between 1989 and 1997, one study estimated that the pro-
energy. In contrast, children living in Russia and China eat portion of overweight and obese men in China rose from 6.4
very little food away from the home. Snacks provide about to 14.5 per cent, and in women from 11.5 to 16.2 per cent.85
16 per cent of dietary energy for Russian children, but Another study, in nine Chinese provinces, found that
account for only 1 per cent in Chinese children.2 between 1989 and 2000 there was a 13.7 per cent increase
A US study showed that between 1977 and 2001, con- in the proportion of men, and a 7.9 per cent increase in the
sumption of sweetened drinks increased by 135 per cent. proportion of women, who were overweight or obese.
During the same period, milk consumption decreased by 38 During the same period, there was an average 2 per cent
per cent, resulting in an overall daily increase of 278 kcal decrease in the number of men and women who were clas-
(1164 kJ) from drinks.31 sified as underweight.86
The World Health Organization MONICA Project moni-
1.2.2 Overweight and obesity tored 10 million adults in 21 countries over a 10-year peri-
There have been rapid changes in rates of overweight and od in the 1980s and 1990s. During this time, the mean body
obesity throughout the world since the 1980s, at the same mass index (BMI) increased in most populations, with the
time as the urbanisation and industrialisation of middle- and largest increases in regions of Australia and the USA. Over
low-income countries. Such countries often experience the the course of the project, the overall average BMI increased
dual burden of nutritional deficiencies and chronic diseases. by 1.5.87 However, average BMI decreased in Russia and
Also see Chapters 6 and 8. Central Europe, and in certain regions of Italy and
The most recent estimates suggest that in 2002 there were Switzerland. The UK has one of the highest rates of excess
1 billion overweight or obese people worldwide, with weight in Europe. This has increased threefold since 1980;
Chinese people accounting for approximately one fifth. The in 2003, 65 per cent of men and 56 per cent of women were
example of China is remarkable. Historically, China, which overweight, with 22 per cent of men and 23 per cent of
is classed as a lower-middle-income economy by the World women classified as obese.88
Bank, had a lean population. But the prevalence of under- Historically, food insecurity, undernutrition, and under-
weight adults has decreased and the numbers of people who weight, and their likely contribution to infection, have been



In 2004 India had a population of over 1.1 China to reach the 1-billion mark.46 India dollars per person (figure 1.3). Life
billion, growing at a rate of about 1.2 per has a low-income economy, with a gross expectancy at birth is 61 years for men and
cent a year; it was the next country after domestic product of 1830 international 63 for women (figure 1.1).46

Non-communicable causes of death India Age-standardised rates of common cancers India

Age-standardised rate per 100 000

Per cent of deaths

Cardiovascular disease 30
Men Women
3 Cancer 20

57 Respiratory disease 10

15 Diabetes









Data from World Health Organization46 Data from International Agency for Research on Cancer20

the main nutrition-related public health issues in middle- and

low-income countries. This is no longer the case. Thus, Figure 1.4 Projected increases in obesity
surveys of women between 1992 and 2000 found that
overweight exceeds underweight in most middle- and low- Per cent of adults with BMI 30 Projections based on
income countries, including those in North Africa and the 70
adults aged 30 100

Middle East, Central Asia, China, and Latin America. Indeed,

there has been a disproportionate increase in, and prevalence 60
of, obesity among the lowest-income groups in most coun-
tries. It is more likely that people will be overweight if they 50
live in urban areas compared with rural areas, and countries
with a higher GDP have a greater ratio of overweight to 40
underweight women.89 North Africa and the Middle East are
two areas of the world with middle- and low-income coun-
tries that are experiencing very high rates of overweight and
obesity, often higher in women than in men.82
The rise of overweight and obesity since the mid-1970s has
been much faster in lower-income countries.63 In Europe and
the USA, the prevalence has risen relatively slowly, by
0.30.5 per cent each year; but the figures are two- to four-
fold higher in many low-income countries.90 Projections from 0

South Africa











existing data suggest that by 2015, levels of obesity could

be as high as 50 per cent in the USA, between 30 and 40
per cent in the UK and Australia, and more than 20 per cent
in Brazil.46 See figure 1.4. It is estimated that more than 12
million adults in England will be obese by 2010, while 25 2002 men 2002 women 2015 men 2015 women

per cent of children who live in households with obese par-

ents will become obese themselves.88 Data from World Health Organization46

1.2.3 Physical activity

Changes in degrees of physical activity throughout the world Projected increases in obesity (BMI of above 30 kg/m2) in
selected countries, 20022015
have also been rapid since the 1970s, as paid and household


Chronic diseases account for 58.1 per aged 4054 had an average BMI of 23.6.46 energy rose between 1964 and 2004, from
cent of all deaths, while infectious diseases, For men aged 2070, 25.4 per cent had a around 2050 to 2470 kcal/person per day
maternal, perinatal, and nutritional condi- BMI of over 25.46 Women aged 2024 had (8580 to 10 360 kJ/person per day).1
tions account for 32.9 per cent; 9 per cent an average BMI of 20.9, rising to an aver- Recently, though, there have been large
of deaths are due to injuries.46 The first fig- age of 24 at age 30; and women aged increases in the consumption of animal
ure gives a breakdown of deaths caused by 3069 also had a BMI of around 24.46 In products, fats, and sugars. The proportion
chronic disease.46 total, 35.8 per cent of women aged 2070 of energy from fat has increased each year,
Common cancers in men include those of had a BMI of over 25.46 A review from 2002 although within India there are differences:
the oral cavity and pharynx.20 Although established that the prevalence of for higher-income groups, 32 per cent of
these cancers have declined since the late preschool obesity was about 1 per cent, but energy comes from fat, compared with
1970s, the incidence remains high.20 Cancers stunting remained a problem in over half 17 per cent in lower-income groups.39
of the oesophagus and lung have also of all children.39 Obesity has been uncom- Since 1975 there has been a reduction in
decreased slightly in the same period.20 In mon in India and varies with socioeconomic cereal (grains) consumption, particularly
women, cancer of the cervix is the most com- status, being more common in high-income coarse grains, although this has not affect-
mon type, and has been since the 1970s; households. In the 1970s, 2.1 per cent of ed overall energy consumption. This is
breast cancer has increased steadily during men and 2.9 per cent of women had a BMI probably due to large increases in intakes
this time.20 Cancers of the oral cavity and of 25 or more, while less than 0.5 per cent of fats and animal protein, and also of milk
oesophagus have declined slightly since the of men and women had a BMI of 30 or and milk products. In lower-income house-
late 1970s (for age-standardised rates of more.46 By 1998 these figures had risen: 4.4 holds, fat comes mainly from vegetable
these cancers, see the second figure).20 per cent of men and 4.3 per cent of women foods, with very little consumption of ani-
For the period 20022003, 10 per cent of had a BMI of 25 or more.46 See figure 1.4 mal fats, whereas in the highest-income
men and 16 per cent of women aged 1869 for projections of the proportions of men households, the majority of fat is from ani-
were classified as sedentary (figure 1.6).46 and women who will have a BMI of 30 or mal sources. India is a major producer of
In 2000 men aged 2024 had an average more in 2015.46 vegetables and fruits, much of which are
body mass index (BMI) of 20.7, while those The average amount of available food exported.

work has become increasingly mechanised, and vehicles are not do enough physical activity getting from one place to
used more often for transport. Occupational and household another.91 In Europe, people living in more northerly regions
physical activity has reduced dramatically in high-income such as Scandinavia have higher levels of activity than those
countries. Also see Chapters 5, 6 and 8. living further south, for example, in Mediterranean coun-
There is as yet no globally accepted, quantified definition tries. Women tend to exercise less than men and this differ-
of physical inactivity, or of the extent to which populations ence is greatest in southern European countries.42
or people should be physically active. In 2002, the WHO A study conducted in 1953 demonstrated that more than
recommended a minimum of 30 minutes moderate-intensi- half of US school children failed a minimum standard of
ty physical activity most days; it found that at least 60 per
cent of the worlds population fails to achieve this level of
physical activity.91
The proportion of people employed in agriculture can Figure 1.5 Projected levels of inactivity in selected
regions in 2020
reflect the level of work-related activity undertaken in a
country, and there may be a linear relationship between the
two.91 Thus, it is likely that, compared with high-income Insufficient Inactive
countries, transport-related and occupational and household
Africa 4555 1020
physical activity is higher in low-income countries.
USA/Canada 3550 1730
Transport-related physical activity (cycling, walking) is high-
Latin America 2545 1747
er in those countries with the lowest gross domestic prod-
uct and low car ownership, and this differs little between Middle East 3042 1530

men and women. Europe 3060 1540

Data on physical activity in Africa are limited. Several stud- India/Bangladesh 3042 1425
ies are available for South Africa, but these cannot be used New Zealand/Australia/Japan 4856 1520
to predict or generalise about activity levels across the entire China 40 1522
continent. Some small regional studies have been performed,
in Ethiopia, for example, which provide useful local infor-
mation, but they are not nationally representative. Data from Bull et al93
Data from Europe, where recreational physical activity
accounts for a greater proportion of total activity, suggest
that approximately half of all walking and cycling trips are Percentage of adults projected to have insufficient levels of
physical activity or to be inactive in 2020 in selected regions
less than 3 km. Therefore, almost half of European adults do



In 2004 Japans population was just over 30 039 international dollars per person (fig- cent of all deaths, while infectious diseases,
128 million, with 79 per cent living in urban ure 1.3). Life expectancy at birth is 79 years maternal, perinatal, and nutritional condi-
areas.46 The country has a high-income for men and 86 for women (figure 1.1).46 tions account for 10.5 per cent; 10.8 per
economy, with a gross domestic product of Chronic diseases account for 78.7 per cent of deaths are due to injuries. The first

Non-communicable causes of death Japan Age-standardised rates of common cancers Japan

Age-standardised rate per 100 000

Per cent of deaths

Cardiovascular disease 60
Men Women
2 14
Cancer 40
6 37 30
Respiratory disease 20

44 0



Data from World Health Organization46 Data from International Agency for Research on Cancer20

fitness, compared with less than 10 per cent of European

Figure 1.6 Sedentary behaviour in adults in selected
children.92 Another study in 2001 found that only 0.2 per countries (age 1869)
cent of US adults were physically active in both occupational
and transport settings, compared with 29 per cent of Chinese Per cent of adults classified as sedentary
adults.91 In the USA, while socioeconomic status has a large
impact on whether someone participates in recreational Men Women

physical activity, there are only small differences between Brazil 28 31

men and women, and activity levels decline with age. Similar China 10 12
trends exist for men and women and for socioeconomic sta- India 10 16
tus in Australia.42 Mexico 17 18
A number of factors will affect levels of physical activity in
South Africa 44 49
future. Economic development has the effect of reducing lev-
Spain 27 33
els of occupational, household, and transport physical activ-
ity. It reduces the amount of physical activity in the
workplace, often because of a shift from agriculture to man-
Data from World Health Organization46
ufacturing and service industries. Improved public transport
in middle- and low-income countries reduces transport-
related activity. Similarly, as people gain more disposable Sedentary behaviour is defined as less than 30 minutes of
income, they are more likely to own a car, which means that moderate physical activity (equivalent to brisk walking) on fewer
than 5 days/week, or less than 20 minutes of vigorous physical
they will make fewer journeys by bicycle or foot. Recreational activity (equivalent to running) on fewer than 3 days/week. Also
activity is the only area in which physical activity may defined in terms of metabolic equivalents (METs) as achieving
increase as economies develop and countries become increas- less than 60 METhours/week of any combination of activity on
ingly urbanised and mechanised, although people may not fewer than 5 days/week (also see chapter 5.2)
necessarily use their leisure time for active pursuits.91
Other factors constrain physical activity in cities, such as there are sidewalks/pavements, parks, or other areas where
personal safety: crime rates are often high and it may be they can move around freely and safely. Cultural and reli-
unsafe to walk, jog, or cycle in the streets. Furthermore, city gious customs may also limit activity levels, particularly for
and town planning may not encourage people to be active women.
for example, people can only walk, run, ride, and play if By 2020 it is predicted that more adults will be physically


figure gives a breakdown of deaths caused indicate that in 2015 almost 900 000 people decreased, from 75 per cent of energy in
by chronic diseases.46 will develop cancer and 450 000 will die the 1940s to 41.3 per cent in 2000. Energy
In men, cancer of the stomach is the most from cancer.47 from fat increased from 6.9 per cent in 1949
common type of cancer. This is followed by Regional studies suggest that 6870 per to 25 per cent in 1988, and to 26.5 per cent
colorectal, lung, liver, and prostate cancers, cent of men and 7082 per cent of women, in 2002. Total fat intake has increased sig-
which have increased since the 1960s. aged 2070, are physically inactive. nificantly following the countrys econom-
Cancer of the oesophagus has remained In men there has been a steady increase ic growth, from 15 g/person per day in the
steady since the 1960s, although the inci- in body mass index (BMI) since the mid- 1940s to 59 g in 1983, remaining at around
dence of cancer of the pancreas has 1970s. In 2002, 17.5 per cent of men aged this level in 2002. The percentage of ener-
increased since then.20 In women, breast 2029 and around 30 per cent of those gy in diets from protein has risen from 12.4
cancer is the most common type and its inci- aged 3060 had a BMI of over 25. Only 7 per per cent in the 1940s to 15.9 per cent in
dence has risen since the 1970s.20 Colorectal cent of women aged 2029, 19 per cent of 2000. However, there has been a larger
cancer is the next most common type, and those aged 4049, and 25.6 per cent of increase in the percentage of protein from
this has also increased. Stomach cancer inci- those aged 5059 had a BMI of over 25. See animal sources: from 18.6 per cent in the
dence has decreased since the 1960s, but figure 1.4 for projections of the proportions 1940s to more than 50 per cent in 2000. In
the rate remains high; lung cancer has risen of men and women who will have a BMI of 2002, people were continuing to eat more
steadily since the 1960s.20 The incidence of 30 or more in 2015. green and yellow vegetables, with people
cancer of the cervix increased during the The average amount of available food over 50 tending to eat the most vegetables.
1960s and remained high in the 1970s, but energy rose between 1964 and 2004, from Fruit intake peaked in 1975 and has since
has since declined (for age-standardised around 2570 to 2760 kcal/person per day decreased and stabilised. In 2002, Japanese
rates of these cancers, see the second fig- (10 780 to 11 540 kJ/person per day). Meat diets did not provide the recommended
ure). 20 However, the total numbers of new consumption also increased during this intake of calcium: although consumption of
cancer cases and cancer deaths are set to time.1 Steamed rice was the staple food milk and dairy products had increased, con-
rise because Japan has an ageing popula- until 1950, and accounted for 80 per cent of sumption of fish and shellfish had declined
tion. Cancer has been the leading cause of energy intake before 1935. Dietary intake slightly. Salt intake remained high, at over
death in Japan since 1981 and projections of cereals (grains), almost all rice, has 12 g/person per day. 23

inactive.91 Clearly, levels of physical activity will vary in dif- ageing world population; improved screening, detection, and
ferent areas and countries around the world. In Europe, for treatment, which increases the number of people living with
example, the former Soviet Union states and countries in a diagnosis of cancer (see Chapter 9); the projected increas-
eastern Europe are at the lower end of the estimates, with es in tobacco smoking in many countries; and the increase
western European countries at the higher end. Indeed, these in the number of people with HIV/AIDS in some countries.
figures are expected to rise further in western Europe and it The global age-adjusted incidence of cancer is also likely to
is estimated that 5060 per cent of adults will not be suffi- increase. Also see box 1.2.
ciently physically active by 2020. Also see figure 1.5. The per- Globally, the most commonly diagnosed cancers (exclud-
centage of adults currently classified as sedentary in selected ing all types of skin cancer) are those of the lung, colon and
national examples is shown in figure 1.6. Using different def- rectum, and breast, with lung cancer being the leading cause
initions, the amount of adults (aged over 16 years) classi- of cancer death.94 95 Geographical and socioeconomic dif-
fied as sedentary for the UK are 60 per cent of men and 66 ferences exist for the most common cancers. In low-income
per cent of women. For the USA 52 per cent of men and 65 countries, the most prevalent cancers include those of which
per cent of women (aged over 18 years) are classified as infectious agents are a major cause, while in high-income
sedentary.46 countries, they include hormone-related cancers. In high-
income countries, and among men, prostate cancer is the
1.2.4 Cancer most common type, followed by cancers of the lung, stom-
Patterns of cancer and trends, incidence, and projections vary ach, and colon and rectum. In low-income countries, and
greatly in different parts of the world. Also see Chapter 7. among men, lung cancer is the most common type, followed
In 2002 there were more than 10 million new cases of can- by cancers of the oesophagus, stomach, and liver. Breast can-
cer (excluding non-melanoma skin cancers) recorded world- cer is the most common type among women living in high-
wide, and nearly 7 million cancer deaths. By 2020 these income countries, followed by cancers of the lung, colon and
figures are estimated to rise to over 16 million new cases, rectum, and endometrium. Breast cancer is also the most fre-
with 10 million deaths. There may be more than 20 million quent type among women living in low-income countries,
new cases of cancer in 2030.93 Indications suggest that, at followed by cancers of the lung, stomach, and cervix.94 96
that time, 70 per cent of cancer deaths will be in low-income
countries. Oesophagus
This projected increase is accounted for by a combination Oesophageal cancer is the seventh most common type of
of factors: the projected increase in global population; an cancer worldwide, with more than 460 000 new cases record-



In 2001 the UK population was nearly 60 of deaths are due to injuries. The first fig- ing steadily since the 1960s, is now decreas-
million.30 The UK is a high-income economy, ure below gives a breakdown of deaths ing, and stomach cancer incidence has
with a gross domestic product of 31 300 caused by chronic diseases.46 declined since the 1960s.20 In women, breast
international dollars per person (figure Prostate cancer is the most common can- cancer is the most common type, and
1.3). Life expectancy at birth is 76 years for cer type in men and has increased steadily although rates were fairly constant during
men and 81 for women (figure 1.1). since the 1970s.20 Lung cancer incidence the 1960s and 1970s, they have risen steadi-
Chronic diseases account for 84 per cent peaked in the 1960s, remained high until ly since then.20 The incidence of colorectal
of all deaths, while infectious diseases, the mid-1980s, and is now declining.20 cancer has remained steady since the
maternal, perinatal, and nutritional condi- Colorectal cancer has risen steadily since 1960s.20 Lung cancer rose from the 1960s to
tions account for 11 per cent; 4.9 per cent 1960.20 Bladder cancer, which had been ris- 1980s, and has remained steady since

Non-communicable causes of death UK Age-standardised rates of common cancers UK

Age-standardised rate per 100 000

Per cent of deaths 90
Cardiovascular disease Men Women
16 60
1 Cancer 50
8 42
Respiratory disease
33 Diabetes
0 Prostate










Data from World Health Organization46 Data from International Agency for Research on Cancer20

Increases in BMI UK
ed in 2002. Because it has a poor survival rate, it is the fifth more common in parts of China than in Europe and
most common cause
Per cent adults ofBMI
with cancer
2529.9death, responsible
Per cent for 30
adults with BMI nearly North America.94 97 Other areas of high risk include southern
000 deaths in 2002.
39050 and eastern Africa, south-central Asia, and some countries
Incidence rates vary widely between countries. Studies in South America.
suggest that cancer of the oesophagus is 100 times Geographical variability of exposure to established car-

20 1.2 Measurement of cancer incidence and mortality
10 cancer registers were established in cer. These rates are usually expressed as the registries prepare cancer-incidence statis-
various countries in the second half of the number of new cases (or deaths) each year tics. With types of cancer where survival is
20th 0 century, descriptive studies showed for every 100 000 people. high, cancer mortality statistics will not
1980 2003 1980 2003
reliably for (UK)the first time that rates and
(England) (UK) Cancers are not usually diagnosed until
(England) reflect occurrence rates. But globally, it
trends of different cancers vary, sometimes they produce symptoms, so there is a peri- is easier to obtain statistics for mortality
substantially across different countries. od of time Women
between the first stages of can- than for incidence, so these are often used
This variation suggested that cancer is not cer development and its identification. This for comparisons between population
from Department of Health48
inherited and et al49
that different
and Rosenbaum length of time can vary greatly, and there groups.
cancers have different causes. are also considerable differences in survival It can also be difficult to compare can-
Many countries publish annual incidence times and how types of cancer respond to cer incidence globally: not all countries and
and mortality rates for cancer. The inci- treatment. regions are covered by cancer registries,
dence rate refers to the number of new Many countries and international agen- and these organisations may use different
cancer cases reported; the mortality rate cies track mortality statistics with causes definitions and collect different data, both
refers to the number of deaths from can- of death, and national or regional cancer geographically and over time.

Data from World Health Organization46 Data from Inte

then.20 Cancer of the ovary has increased from around 3280

Increases in BMI UK
slightly since the 1960s, and rates of cancer to 3480 kcal/person
of the endometrium have remained the per day (13 730 to Per cent adults with BMI 2529.9 Per cent adults with BMI 30
same since the 1960s (for age-standardised 14 570 kJ/person 50
rates of these cancers, see the second fig- per day). Consump-
ure).20 The incidence of childhood cancer tion of pasta, rice, 40
has been rising at an average rate of 1.1 cereals (grains),
per cent each year and, between 1978 and yogurt, soft drinks, 30
1997, the age-standardised incidence savoury snacks, and
increased from 120 to 141 cases/million chil- nuts has increased 20

dren.22 Children in the British Isles have the since the mid-1980s.
highest rates of skin cancer in Europe.11 Over the same peri-
In 2003, 64 per cent of men and 76 per od, intakes of fish
cent of women aged 1669 were classified and fish dishes and 1980 2003 1980 2003
as sedentary.46 A study to examine exercise eggs and egg dishes (UK) (England) (UK) (England)
patterns in adults in 1991 and again in 1999 have decreased.19
Men Women
found that only 4 in 10 adults had man- Large studies sug-
aged to meet and maintain the current rec- gest that fat intake
Data from Department of Health48 and Rosenbaum et al49
ommended level of activity, or to increase has decreased be-
their level. During the study period, the cause people now
majority either reduced their activity level consume less whole
or maintained it below the recommended milk, butter, and red meat, and more veg- portions of vegetables and fruits. Instead,
level, and 15 per cent of the sample was etables and fruits.35 Men surveyed between men ate an average of 2.7 portions while
inactive, both in 1991 and 1999.33 2000 and 2001 were more likely to eat women had 2.9.19 The survey also showed
In the UK, body mass index (BMI) has foods containing fats, oils, and sugars, as that vegetable and fruit consumption was
risen steadily since the mid-1970s. For the well as meat and meat products, and soft particularly low in young adults, and that
proportions of men and women in 1980 and alcoholic drinks. people from low-income households were
and in 2003 with a BMI of between 25 and In the same survey, and compared with less likely to eat fruit and yogurt.19 It
29.9, or of over 30, see the figure on this older men and all women, young men were appears that more-educated adults put
page. Also see figure 1.4 for projections of more likely to eat savoury snacks and soft dietary guidelines into practice, reducing
the proportions of men and women who drinks, and less likely to eat eggs, fish, and the amount of fat in their diets and increas-
will have a BMI of 30 or more in 2015. fruit.19 Women ate more fruit, although ing the amount of vegetables and fruits
The average amount of available only 13 per cent of men and 15 per cent of they eat.3
food energy rose between 1964 and 2004, women ate the recommended five daily

cinogens can explain some of these differences. In high- high-income countries (almost 70 per cent in 1980); almost
income countries, alcohol and smoking tobacco are the main half were predicted to occur in middle- and low-income
carcinogenic agents, whereas chewing tobacco is more com- countries in 2005.94 95 The USA and Europe have the high-
mon in India. Pockets of high incidence occurring in parts est numbers of lung cancer cases for both men and women,
of China and in the Caspian littoral of Iran may be due to but the incidence appears to have peaked, and may now
general nutritional deficiencies. Incidence of adenocarcino- be declining in the USA and in parts of northern Europe. It
ma of the lower third of the oesophagus is steadily increas- is, however, still increasing in southern and eastern Europe.
ing in the USA and Europe, which is likely to be linked to Men are more likely to develop lung cancer than women,
an increasing incidence of acid reflux from the stomach due almost certainly because, on average, they smoke more than
to obesity.94 95 Also see chapter 7.3. women. Worldwide, 1 billion men and 250 million women
currently smoke tobacco. It is estimated that throughout the Lung 20th century, 100 million people died from tobacco use.93
Lung (pulmonary) cancer has been the most common type Also see chapter 7.4.
of cancer in the world since 1985, with around 1.35 million
new cases recorded in 2002. It is also the most common Stomach
cause of cancer death. In 2002, 1.2 million people died from Stomach (gastric) cancer is now the fourth most common
lung cancer. type of cancer worldwide, with around 925 000 new cases
Between 1985 and 2002, the estimated number of lung recorded in 2002. It is the second most common cause of
cancer cases worldwide rose by 51 per cent, and the num- death from cancer, with around 700 000 deaths annually.
ber of cases in middle- and low-income countries has Until about the mid-1980s, stomach cancer was the most
increased steadily over recent years. Previous estimates indi- common type in the world. Since then, rates have fallen sub-
cated that the majority of lung cancer cases occurred in stantially in all high-income countries, and overall rates are



In 2004 Poland had a population of around birth is 71 years for men and 79 for women tions account for 3.9 per cent; 7.9 per cent
38.5 million. The country has an upper-mid- (figure 1.1).46 of deaths are due to injuries. The first fig-
dle-income economy, with a gross domestic Chronic diseases account for 88.2 per ure gives a breakdown of deaths caused by
product of 12647 international dollars per cent of all deaths, while infectious diseases, chronic diseases.46
person (figure 1.3).46 Life expectancy at maternal, perinatal, and nutritional condi- Lung cancer is the most common type of

Non-communicable causes of death Poland Age-standardised rates of common cancers Poland

Age-standardised rate per 100 000

Per cent of deaths
Cardiovascular disease Men
70 Women
2 11
Cancer 50
55 Respiratory disease 30
30 20







Data from World Health Organization46 Data from International Agency for Research on Cancer20

now about 15 per cent lower than in 1985.94-96 Exposures to the hepatitis B and C viruses are known to
Stomach cancer is now much more common in Asia than increase the risk of developing liver cancer; 85 per cent of
in the USA or Europe. Indeed, 42 per cent of cases occur in cases in low-income countries are attributed to exposures to
China alone.94 97 High-risk areas are China, Japan, eastern these two viruses. Also see chapter 7.8.
Europe, and Central and South America. Low-risk areas are
South-East Asia, northern and eastern Africa, the USA, and Colon and rectum
Australia and New Zealand. In most countries, incidence has Colorectal cancer (of the colon or rectum) is the third most
dropped by about 15 per cent compared with 1985. common type of cancer worldwide, with just over 1 million
The bacterium Helicobacter pylori is an established cause new cases recorded in 2002. Mortality is approximately half
of stomach cancer. Reduction in stomach cancer rates can be that of the incidence, and nearly 530 000 deaths were
explained partly by reduced exposure to H pylori and part- recorded in 2002, making it the fourth most common cause
ly by increased use of refrigeration to preserve foods.94 97 Also of death from cancer.
see chapter 7.5. There is a large geographical difference in the global dis-
tribution of colorectal cancers. Incidence varies up to 25-fold Liver between countries with the highest rates (the USA, Australia
Liver (hepatic) cancer is the sixth most common type of can- and New Zealand, and in parts of Europe) and those with
cer worldwide, with around 625 000 new cases recorded in the lowest rates (in Africa and Asia). Intermediate levels
2002. The poor prognosis makes it the third most common occur in South America.
cause of cancer death, with around 600 000 deaths in 2002. Incidence of colorectal cancer may be stabilising in parts
In most countries, the incidence of liver cancer is stable of northern and western Europe, and possibly declining
and there is little difference in survival rates between high- gradually in the USA. Elsewhere, however, the incidence is
and low-income countries. More than 80 per cent of cases increasing rapidly, particularly in Japan and in middle- and
occur in middle- and low-income countries. Areas with a low-income countries.94 96
high incidence are China (55 per cent of all new cases), sub- As shown in 1.3, the incidence of colorectal cancer increas-
Saharan Africa, and eastern and south-eastern Asia. es quickly when people migrate from low- to high-risk areas
Incidence is lower in high-income countries and in Latin of the world. Indeed, the incidence rate is higher in Japanese
America, although Japan and areas of southern Europe have people born in the USA than in white people born in the
intermediate incidence levels.94 96 USA. Also see chapter 7.9.


cancer in men and age-adjusted incidence aged 3044 had a BMI of 24.1, while those Tobacco smoking and alcoholic drink con-
has remained stable since the 1970s.20 The aged 4559 had a BMI of 26.7.46 Only 1.5 per sumption are underlying factors in overall
incidence of colorectal cancer has increased cent of women aged 1529 had a BMI of mortality trends in eastern Europe. An
since 1990, and both prostate and bladder over 30, rising to 22.5 per cent of those aged analysis of national household budget and
cancers have increased slightly since the 4559, and 23.7 per cent of women aged individual dietary surveys carried out in the
1970s.20 Stomach cancer incidence peaked in 6074.46 In a study of adults in Warsaw, the 1990s found that, each day, the average
the late 1970s and has declined steadily since. average adult BMI remained stable between person ate around 300 g of dairy products
Breast and colorectal cancers are the most 1983 and 1993 at approximately 27.46 Overall and the same amount of cereals (grains)
common types in women and their rates in 1996, 10.3 per cent of men and 12.4 per and roots and tubers, although consump-
have risen steadily since the 1970s.20 Cancer cent of women had a BMI of 30 or more.46 tion of pulses (legumes) was very low.38 A
of the cervix has remained steady since the See figure 1.4 for projections of the propor- study of students found that women ate
mid-1970s, whereas cancers of the lung, tions of men and women who will have a meat and drank beer less frequently than
ovary, and endometrium have increased in BMI of 30 or more in 2015.46 men, and they ate more fruit and drank
this period (for age-standardised rates of The average amount of available food more milk.8 Another local study, in Warsaw,
these cancers, see the second figure).20 energy rose between 1964 and 2004, from reported decreases in intakes of total ener-
In 1996, 31 per cent of men and 32 per around 3310 to 3520 kcal/person per day gy, dietary cholesterol, and dietary animal
cent of women aged 1575 were classified (13 850 to 14 730 kJ/person per day). The fats, and an increase in vegetable oil intake
as sedentary.46 amount of energy available from sugars between 1984 and 2001.45 Another study
In 1996 men aged 1529 had an average and meat increased during this period, found that between 1990 and 2000, the
body mass index (BMI) of 23.1.46 This rose to while the energy available from animal fats proportion of men eating fruit each day
25.9 for men aged 3044, while those aged fell substantially.1 In 1989 Poland began the increased from 36 to 42 per cent. Levels of
4575 had a BMI of between 26 and 27.46 transition from a centrally planned to a intake were stable in women, with around
Only 2.4 per cent of men aged 1529 had a market economy. This resulted in dramatic 60 per cent eating fruit every day. In con-
BMI of over 30, rising to 10.8 per cent of increases in food prices, and although the trast, over the same decade, only 2223 per
those aged 3044, and 17.5 per cent of 4559 transition gave people a better choice of cent of men limited their fat intake,
year olds.46 In the same year, women aged foods, there was a decline in food demand although more women did during this peri-
1529 had an average BMI of 21.2.46 Women and alterations in dietary patterns.37 od (an increase from 23 to 45 per cent).40 Breast Cervical cancer is the seventh most common cause of death
Breast cancer is the most common type of cancer in women, from cancer overall, and the third most common in women,
and the third most common cancer overall. Incidence rates and was responsible for nearly 275 000 deaths in 2002.
are increasing in most countries, with an estimated 1.15 mil- Over 80 per cent of cases occur in low-income countries.
lion new cases recorded in 2002. Breast cancer is the sixth Areas with the highest incidence rates are sub-Saharan Africa,
most common cause of death from cancer overall. However, the Caribbean, Central and South America, and south-cen-
it is the second most common cause of cancer death in tral and South-East Asia. Incidence rates are lowest in Europe,
women, with just over 410 000 deaths recorded in 2002. the USA, Japan, China, and Australia and New Zealand.
The incidence of breast cancer is highest in high-income The incidence has dropped substantially in high-income
countries (although not in Japan) and more than half of all countries following the introduction of cervical screening
cases occur in these countries. Although breast cancer has programmes. The major established cause of cervical cancer
been less common in women living in low-income countries, is infection with certain subtypes of human papilloma
age-adjusted incidence is increasing, and the rates of viruses (HPV). Other cofactors (parity, contraception, HIV
increase are often greater in these countries. infection, and smoking) can also modify the risk of this
Globally, estimates indicate that breast cancer incidence cancer in women infected with HPV.94 96 97 Also see
has increased by 0.5 per cent annually since 1990. However, chapter 7.13.
certain cancer registries, such as those in China and other
parts of Asia, are recording annual increases in incidence of Prostate
up to 4 per cent. Rates are low in Africa, with the lowest inci- Prostate cancer is the third most common type of cancer in
dence in central Africa.94 96 men, and the sixth most common cancer overall, with near-
Migrant studies show that breast cancer rates change when ly 680 000 new cases recorded in 2002. The majority of cases
women move to a new country. See 1.3. Also see chapter are diagnosed in men over the age of 65, and this cancer
7.10. accounted for just over 220 000 cancer deaths in 2002. This
made it the eighth most common cause of death from can- Cervix cer overall, and the sixth most common in men.
Cancer of the cervix is the second most common type of can- Prostate cancer is more common in high-income countries,
cer among women, and the eighth most common cancer but the incidence remains low in Japan. Incidence rates have
overall, with around 500 000 new cases recorded in 2002. been influenced by screening programmes, which increase



In 2004 Spain had a population of over ure 1.3).46 Life expectancy at birth is 77 years maternal, perinatal, and nutritional condi-
43 million. The country has a high-income for men and 83 for women (figure 1.1).46 tions account for 5.5 per cent; 6.9 per cent
economy, with a gross domestic product of Chronic diseases account for 87.4 per cent of deaths are due to injuries. The first figure
24 325 international dollars per person (fig- of all deaths, while infectious diseases, below gives a breakdown of deaths due

Non-communicable causes of death Spain Age-standardised rates of common cancers Spain

Age-standardised rate per 100 000

Per cent of deaths
Cardiovascular disease
50 Men Women

19 40
35 30
Respiratory disease 20
33 0











Data from World Health Organization46 Data from International Agency for Research on Cancer20

diagnosis rates. This has resulted in a huge increase in the

Figure 1.7 Mortality from stomach and colorectal
number of recorded cases in the USA in recent years, cancer in European migrants to Australia
although the incidence in several high-income countries has
declined since the 1990s. Prostate cancer incidence is
Stomach cancer Colon cancer Rectal cancer
increasing rapidly in low-income countries, particularly in Length of
Latin American countries (such as Costa Rica, Colombia, residence
(years) <16 >16 <16 >16 <16 >16
and Ecuador) and in China. Again, this may partly be due Country of
to increased awareness and screening. origin
Mortality from prostate cancer is lower (5.8 per cent of Yugoslavia 2.22 1.23 0.47 0.66 0.46 1.34
cancer deaths in men),94 and may give a better indication England 1.47 1.24 0.99 1.04 1.23 1.04
of actual disease patterns.97 Even so, mortality is approx- Scotland 1.84 1.46 1.47 1.24 1.05 1.08
imately 10 times more common in the USA and Europe Ireland 1.77 1.21 0.62 1.06 1.17 1.18
than in Asia. Also see chapter 7.14.
Poland 1.69 1.71 1.02 1.14 0.43 1.34

Greece 1.35 1.15 0.36 0.69 0.34 0.7

1.3 Migrant and other ecological studies Italy 1.43 1.49 0.37 0.7 0.48 0.8

Australia 1.0 1.0 1.0 1.0 1.0 1.0

Ecological studies (also called correlation studies) exam-

ine the relationships between environmental factors and Data from McMichael et al102

disease outcomes, often in different countries, at an aggre-

gate level (see chapter 3.1.2). These provided the first sys-
Relative risk of death from cancer of the stomach, colon, and
tematically gathered evidence suggesting that the principal rectum in European migrants to Australia (19621976) compared
causes of cancer are environmental, and that food, nutri- with people born in Australia
tion, and physical activity are among these factors. Early
studies showed strong correlations among countries
between, for instance, dietary fat intake and breast can- specific nutritional factors and cancers at particular sites,
cer rates.98 and for the general proposition that patterns of cancer might
While not providing strong evidence for causation, such be altered as a result of changing patterns of eating and
studies generated hypotheses for possible links between other ways of life. Part 2 of this Report explores the degree


to chronic diseases.46 BMI of 26, and 4575 year olds had a BMI energy rose between 1964 and 2004, from
Lung cancer is the most common cancer of 27.46 In total, 35 per cent of men aged around 2700 to 3480 kcal/person per day
type in men and incidence rates have 2564 had a BMI of over 27, and 12.2 per (11 330 to 14 590 kJ/person per day), due
increased dramatically since the 1970s.20 cent had a BMI of over 30.46 Women aged largely to an increase in the availability of
Rates of colorectal, prostate, bladder, and 2534 had an average BMI of 23, while sugars and meat.1 The Mediterranean-style
oral cancers have risen since records began those aged 3544 had a BMI of 25, and diet is often seen as the healthiest in
in the 1970s.20 Breast cancer is the most 4574 year olds had a BMI of between 27 Europe, but Spanish diets have recently
common type in women and the rate has and 28.46 Overall, 25.7 per cent of women shifted towards being high in fat and dairy
doubled since the 1970s.20 Colorectal cancer aged 2564 had a BMI of over 27, and 12.1 products, with only moderate amounts of
is the next most common type, which has per cent had a BMI of over 30.46 vegetables. Dairy and fruit intakes are the
seen a steady rise during this period.20 Over the period 19771993, the propor- highest in Europe, but so is the proportion
Cancers of the endometrium and cervix tion of people with energy-intensive jobs of energy in diets from fat.29 Between 1964
have remained steady since the 1970s, but halved. In children aged 67, there has and 1990, consumption of plant-based
cancer of the ovary has risen (for age-stan- been a marked increase in obesity and over- foods decreased from 1289 to 995 g/person
dardised rates of these cancers, see the sec- weight, higher even than in US children of per day. In the same period, intakes of cere-
ond figure).20 the same age. Obesity in adolescents is also als (grains), pulses (legumes), and potatoes
A survey in 1997 found that 76 per cent among the highest in the world.29 Between all halved. While consumption of other veg-
of adults aged 16 and over did no regular 1990 and 2000, 45 per cent of men and 32.2 etables remained stable, fruit intake dou-
exercise during their leisure time29; 46 per per cent of women had a BMI of 25 or bled to 327 g/person per day. Consumption
cent of adults were classified as sedentary, more, and 13.4 per cent of men and 15.8 of animal products increased from 407 to
with only 7 per cent of adults recording any per cent of women had a BMI of 30 or 743 g/person per day due to a large
physical activity each week.29 more. See figure 1.4 for projections of the increase in the amounts of meat, poultry,
Between 1994 and 1997, men aged proportions of men and women who will milk, and dairy products in peoples
2534 had an average body mass index have a BMI of 30 or more in 2015.46 diets, although intakes of animal fats
(BMI) of 25, while those aged 3544 had a The average amount of available food decreased.29

Figure 1.8 Cancer among female Iranian migrants Figure 1.9 Incidence of colorectal cancer in Asian
to British Columbia, Canada migrants to USA and their descendants

Age-standardised incidence in women per 100 000 Ethnicity Birth place Incidence rate per 100 000 people

Cancer Ardabil Kerman Iranian British Men Women

province province migrants to Columbia
(Iran) (Iran) British general White USA 89.9 64.3
Chinese USA 66.9 40.9
Breast 7.6 16.9 68.5 81.4
China 87.8 44.7
Colorectal Not done 5.9 11.6 26.6
Japanese USA 142.5 90.1

Japan 69.3 63.5

Data from Yavari et al104
Filipino USA 57.2 14.2

Philippines 44.4 25.7

Age-standardised incidence of breast and colorectal cancer is
increased in Asian migrants to Canada compared with source
Data from Flood et al106

Age-standardised incidence of colorectal cancer is increased in the

descendants of Japanese migrants to the USA

to which such hypotheses are upheld or refuted by the total- people who have migrated from eastern Asia to the
ity of the relevant published literature, including more robust Americas; from the Indian subcontinent to Africa and the
observational and also experimental types of study. UK; from Europe to Australia; and from Africa to the
The most compelling evidence, suggesting that the main Caribbean, and then to the UK. All of these population move-
causes of cancers of most sites are environmental (due to fac- ments are accompanied by marked changes in patterns of
tors that people are exposed to) rather than genetically diet, physical activity, and disease.
inherited comes from studies of migrant populations. Migrations from Japan to the USA, from the Caribbean to
There are many migrant populations. Examples include the UK, and from European countries to Australia have been



In 2004 the USA had a population of almost Prostate cancer is the most common type the second figure below).20
300 million. The country has a high-income of cancer in men and the incidence rate has In 2003, 22 per cent of men and 27 per
economy, with a gross domestic product of more than doubled since the 1970s.20 Lung cent of women aged 2065 were classi-
39 901 international dollars per person (fig- cancer peaked in the early 1980s and has fied as sedentary; physical inactivity was
ure 1.3), which masks socioeconomic since declined slightly.20 Rates of colorectal more prevalent among people with a low
inequalities.46 Life expectancy at birth is 75 and bladder cancer have remained stable, income.91
years for men and 80 for women (figure although melanoma has increased steadi- In 2002 men aged 2024 had an average
1.1).46 ly.20 Breast cancer is the most common type body mass index (BMI) of 26.2, while those
Chronic diseases account for 84.7 per in women, followed by lung cancer, and aged 2529 had a BMI of 27, and men aged
cent of all deaths, while infectious diseases, both have increased since the 1970s.20 Over 3065 had a BMI of between 27 and 29.46
maternal, perinatal, and nutritional condi- the same period, the incidence of cancers of While 19.7 per cent of men aged 2024 had
tions account for 6.7 per cent; 8.6 per cent the colon and rectum and of the ovary have a BMI of over 30, this rose to 23.6 per cent
of deaths are due to injuries. The first fig- remained stable, while cancer of the of those aged 3034, and 30 per cent of
ure below gives a breakdown of deaths endometrium has decreased slightly (for 4044 year olds.46 Women aged 2024 had
caused by chronic diseases.46 age-standardised rates of these cancers, see an average BMI of 26.2, while those aged

Non-communicable causes of death USA Age-standardised rates of common cancers USA

Age-standardised rate per 100 000

Per cent of deaths

Cardiovascular disease 120

Men Women
Cancer 80
4 41 60
Respiratory disease
8 40

29 0










Data from World Health Organization46 Data from International Agency for Research on Cancer20

studied in some
Increases detail, as have movements of populations
in BMI USA stomach cancer, which corresponded to the length of time
from rural to urban areas within countries. Both types of the migrants stayed in Australia. However, their risk of col-
Per cent result in dietary changes, which are followed, orectal cancer increased proportionally to the length of their
45 one or two generations, by changes in disease pat- stay.100 See figure 1.7. A later study demonstrated that
terns. Patterns of cancer among migrant groups often change deaths from breast cancer among Italian migrants to
faster than they do among people who remain in their home Australia were half that of Australian-born women during
35 or among people native to the host country. the first five years after emigrating. However, after 17 years,
In the 1980s, one study demonstrated that breast cancer Italian migrants had similar death rates (due to breast can-
30 increased almost threefold in first-generation cer) to women born in Australia.101
Japanese women who migrated to Hawaii, and up to five- Following migration, the incidence of certain cancers may
fold in the second generation. Colorectal cancer incidence increase, whereas the incidence of other cancers may
20 almost fourfold in the first generation but did not decrease. Thus among Iranian immigrants to Canada, in
increase further with subsequent generations.99 In this same women, breast cancer incidence rate increased fourfold, and
population, the incidence of stomach cancer dropped by colorectal cancer incidence rate doubled; but there was a
half in the first generation, and dropped further in dramatic decrease in cancers of the stomach and oesopha-
the second generation.99 gus in both sexes.102 See figure 1.8.
5 study, published in 1980, of European migrants Another study showed that breast cancer incidence
to Australia demonstrated a reduction in the death rate from increased threefold within one generation in Polish migrants
190662 197174 1976 198894 19992000

24 Men Women
BMI 2529.9 BMI 2529.9

BMI 30+ BMI 30+

Data from World Health Organization46 Data from Inte


2529 had a BMI of 27.4, and 3065 year and fruits. Between
Increases in BMI USA
olds had a BMI of between 28 and 30.46 1970 and 2000,
While 23.1 per cent of women aged 2024 average population
Per cent
had a BMI of over 30, this rose to 30.9 per consumption of
cent of women aged 2529, and to more meat and poultry
than 40 per cent of those aged 5564.46 The per person increas- 40
obesity epidemic began earlier in the USA ed by 11 per cent:
than in other high-income countries. people ate less beef 35
Between 1906 and 1962, 10.4 per cent of but more poultry.16
men and 15 per cent of women had a BMI Analysis of national 30
of 30 or more.46 By 1999/2000, these figures food surveys de-
had increased to 27.7 per cent of men and monstrated that
34 per cent of women (see figure).46 In 2002 between 1977 and
a US health survey found that almost 75 per 2001, people con-
cent of people were trying to prevent sumed more sweet- 15
weight gain.106 See figure 1.4 for projec- ened drinks and less
tions of the proportions of men and milk. The portion 10
women who will have a BMI of 30 or more sizes of sweetened
in 2015.46 drinks increased and 5
The average amount of available food they contributed
energy rose between 1964 and 2004, from more energy to
190662 197174 1976 198894 19992000
around 2930 to 3750 kcal/person per day diets: an increase
(12 250 to 15 690 kJ/person per day).1 from 50 to 140 Men Women
Between 1977 and 1996, the average pro- kcal/person per day
BMI 2529.9 BMI 2529.9
portion of meals eaten in restaurants or (210 to 600 kJ/per-
fast-food outlets rose from 9.6 to 23.5 per son per day). During BMI 30+ BMI 30+

cent, and fast food now accounts for 20 per the same period,
cent of dietary energy.2 fruit drinks increas-
Data from World Health Organization46
Vegetable and fruit intakes have ed from 20 to 45
increased since the 1980s across all income kcal/person per day
levels, and people now eat more fresh and (80 to 190 kJ/person
frozen vegetables and fruits than canned. per day); energy intake from milk dropped and they consumed one third of their total
Potatoes are commonly eaten. Bananas from 140 to 100 kcal/person per day (600 to energy away from home.2 Energy from soft
are the most popular fruit.34 More than 410 kJ/person per day), with the largest drinks, fast foods, and salty snacks doubled
80 per cent of men and 70 per cent of drop in milk consumption among those between 1977 and 1996, and soft drinks
women aged 2064 fail to eat the recom- aged 218.31 Another study revealed that 93 now provide 8.5 per cent of total energy in
mended five daily portions of vegetables per cent of young people ate snack foods young peoples diets.2

to the USA.103 Japanese, Chinese, and Filipino people who environmental and suggest that patterns of food, nutrition,
migrate to the USA have a higher risk of colorectal cancer and physical activity are important among these causes.
than their counterparts who do not migrate. One study of Migrants share a common genetic background, as do their
US-born Japanese men demonstrated incidence rates of col- parents and children: the genetic pool of any population does
orectal cancer twice as high as Japanese men born in coun- not change within a generation or two. But as shown in
tries other than the USA or Japan and 60 per cent higher Chapter 2, different patterns of environmental exposure can
than in white people born in the USA (figure 1.9).104 and do alter patterns of DNA damage and gene expression,
Data from more recent migrant studies show that cancer and so cancer, in a relatively short time.
incidence rates generally become similar to those of the
adopted country in second-generation immigrants.17 105 This
is illustrated in first-generation immigrants to Sweden, 1.4 Conclusions
where the incidence of all cancers was 5 and 8 per cent lower
for men and women, respectively, compared with native Between the early 2000s and 2030, the global absolute num-
Swedes.17 By the second generation, however, the incidence ber of cancer cases is projected by UN agencies to double,
was only marginally below the figures for people native to most of all in the middle- and low-income countries of Africa
Sweden.105 and Asia. Some of this increase can be attributed to the
Correlation studies, and migrant studies in particular,
prove that the main determinants of patterns of cancer are Continued on page 29



In 2004 Mexico had a population of over be overweight or obese.18 One fifth of higher in the south of the country than the
107 million. The country has an upper-mid- school-age children are overweight or north.
dle-income economy, with a gross domes- obese, and the risk of body fatness rises if Folate intakes were lower in urban
tic product of 10 158 international dollars their mother has a school education and areas, in the north of the country, and in
per person (figure 1.3). Life expectancy at higher socioeconomic status.18 Men aged Mexico City, compared with the south and
birth is 72 years for men and 77 for women 2029 had an average body mass index in rural areas.7 This demonstrates regional
(figure 1.1).46 (BMI) of 25.2, while those aged 3039 had differences in diets, particularly in the
Chronic diseases account for 72.4 per a BMI of 26.9, and men aged 4059 had a amounts of green, leafy vegetables people
cent of all deaths, while infectious diseases, BMI of 27.5. While 11.8 per cent of men eat. Preschool children in the north and in
maternal, perinatal, and nutritional condi- aged 2029 had a BMI of over 30,46 this Mexico City had the highest intakes of fat
tions account for 16.5 per cent; 16.49 per rose to 20.9 per cent of 3039 year olds.46 and the lowest intakes of dietary fibre.
cent of deaths are due to injuries. The first Around 25 per cent of men aged between Children in the south, those indigenous to
figure below gives a breakdown of the 40 and 69 had a BMI of over 30. Women the country, and those of lower socioeco-
deaths caused by chronic diseases.46 aged 2029 had an average BMI of 25.6.46 nomic status had higher intakes of dietary
Prostate cancer is the most common type This rose to 27.9 for those aged 3039, and fibre and starchy foods, the lowest fat
of cancer in men, followed by lung, stom- to 29 for women aged 4069. While 16.7 intakes, and the highest risk of inadequate
ach, and colorectal cancers. Cancer of the per cent of women aged 2029 had a BMI micronutrient intakes for vitamin A, vita-
cervix is the most common type in women, of over 30, this rose to 29.6 per cent of min C, folate, calcium, iron, and zinc.7 In
followed by cancers of the breast, stomach, those aged 3039. Around 40 per cent of women, there was a risk of inadequate vit-
and ovary (for age-standardised rates of women aged 4059 had a BMI of over 30.46 amin A, vitamin C, and folate intake.
these cancers, see the second figure See figure 1.4 for projections of the pro- Consumption of starchy staple foods and
below).20 Women living in rural areas have portions of men and women who will have intakes of folate, calcium, and iron were
a higher risk of cancer of the cervix com- a BMI of 30 or more in 2015.46 significantly higher in rural women com-
pared with those living in urban areas.32 The average amount of available food pared with those living in urban areas.
Stomach cancer incidence has risen since energy rose between 1964 and 2004, from Saturated fatty acid consumption was
1980, and this increase is more evident in around 2470 to 3150 kcal/person per day lower in the south, reflecting the greater
men.20 (10 350 to 13 1780 kJ/person per day).1 contribution of beans and cereals (grains)
In 20022003, 17 per cent of men and 18 In 1999, micronutrient deficiencies re- in diets.6 Women in urban areas, and those
per cent of women aged 1869 were clas- mained a problem.36 Undernutrition was of higher socioeconomic status, consumed
sified as sedentary (figure 1.6).46 more prevalent in the indigenous popula- more cholesterol, saturated fatty acids, and
Between 1995 and 1999, the prevalence tion, in people of lower socioeconomic total fat.6 Across the country as a whole,
of overweight and obesity in children rose status, and in rural areas and the south. dietary fibre consumption was found to be
from 13.6 per cent to 19.5 per cent.18 One in five children under the age of 5 inadequate, but intakes were higher in
Geographically, the highest prevalence of years was stunted, and 2 per cent were central and south Mexico, mainly because
overweight and obesity is in Mexico City classified as suffering from wasting. Rates peoples diets contained beans and cereals
and in the northern region. Across the of stunting and wasting were three times (grains), although their intake of vegeta-
country as a whole, girls are more likely to higher in rural than urban areas, and were bles and fruits was low.

Non-communicable causes of death Mexico Age-standardised rates of common cancers Mexico

Age-standardised rate per 100 000

Per cent of deaths
Cardiovascular disease
25 Men Women

28 31 Cancer

Respiratory disease 10

16 Diabetes









Data from World Health Organization46 Data from International Agency for Research on Cancer20



In 2004 the population of Australia was This rose to 43.8 per cent of those aged sumption was decreasing in adolescents,
over 20 million. The country has a high- 2529, and to between 50 and 70 per cent although the intake of milk products was
income economy, with a GDP of 31 454 of men aged 3075.46 Between 17 and 20 declining only among girls. Fewer adoles-
international dollars per person (figure per cent of men aged 3569 had a BMI of cents ate cereals (grains) compared with
1.3).46 Life expectancy at birth is 78 years over 30.46 The average BMI of women aged other age groups, although cereals and
for men and 83 for women (figure 1.1).46 1824 was 22.8.46 This rose to 24.9 for cereal products contributed 3437 per cent
Chronic diseases account for 86.9 per women aged 3034, and those aged 4075 of their total dietary energy. Cereal products
cent of all deaths, while infectious diseases, had a BMI of between 25 and 27.1.46 The contributed the greatest amount of food by
maternal, perinatal, and nutritional condi- proportion of women with a BMI of over 25 weight to adults diets, followed by milk and
tions account for 4.7 per cent; 8.4 per cent increased with age: 19.9 per cent of women milk products, then pulses (legumes). Fruit
of deaths are due to injuries. The first fig- aged 1824, and 3455 per cent of those consumption increased with age, whereas
ure below gives a breakdown of the deaths aged 3075. Overall, between 15 and 20 per intakes of cereals (grains), milk, meat, and
due to chronic diseases.46 cent of women aged 3075 had a BMI of poultry decreased. Adults in the Northern
In men, prostate cancer is the most com- over 30.46 There has been a steady increase Territory consumed more meat, poultry,
mon type and rates have doubled since the in BMI since 1980, when 40.6 per cent of game, and alcoholic drinks, and less veg-
1970s. Colorectal cancer has increased and men and 20.2 per cent of women had a BMI etables and fruits, than people living in
lung cancer has declined slightly. There has of 25 or more, and 9.3 per cent of men and other areas. Men were slightly more likely
been a large increase in the incidence of 8 per cent of women had a BMI of 30 or to eat food away from home (64 per cent)
melanoma and a slight increase in bladder more.46 In 2000, 48.2 per cent of men and compared with women (57 per cent).
cancer. In women, breast cancer is the most 29.9 per cent of women had a BMI of 25 or Almost one third of adults thought that
common type and the rate has increased more, and 19.3 per cent of men and 22.2 they should be eating more fruit, and 25 per
since the 1970s. Colorectal cancer is the per cent of women had a BMI of 30 or cent thought they should eat fewer high-fat
next most common type, although this has more.46 See figure 1.4 for projections of the foods.4 Fruit and vegetable consumption
remained stable. There has been a large proportions of men and women who will was highest in 1839 year-olds, with 4050
increase in melanoma, and lung cancer has have a BMI of 30 or more in 2015.46 per cent consuming a combination of at
doubled since the late 1970s, although can- The average amount of available food least one portion of fruit and three portions
cer of the endometrium has remained sta- energy remained stable between 1964 and of vegetables each day, although this
ble (for age-standardised rates of these 2004: around 3130 and 3120 kcal/person per amount then declined steadily with age.
cancers, see the second figure below).20 day (13100 and 13 070 kJ/person per day), Only 37.6 per cent of children (1217 year
In 2001, 30 per cent of men and 32 per respectively.1 The most recent National olds) ate this quantity of vegetables and
cent of women aged 1575 were classified Nutritional Survey in 1995 found that 90 per fruits and, in total, 37.2 per cent of people
as sedentary.46 cent of the people surveyed ate cereals over the age of 12 failed to eat this amount.
In the same year, men aged 1824 had (grains) or cereal products, and milk or milk A study of people aged 2075 living in
an average body mass index (BMI) of 24.3, products, the day before the interview. Queensland found that 63 per cent of par-
while those aged 2529 had a BMI of 25.3, However, half of the men and one third of ticipants drank too much alcohol, 40 per
and men aged 3075 had a BMI of the children had not eaten fruit, and 20 per cent were not sufficiently physically active,
between 26 and 27.5.46 Of men aged cent of children had not eaten vegetables and less than half ate the recommended lev-
1824, 34.3 per cent had a BMI of over 25.46 the day before the interview. Fruit con- els of fruits and vegetables.13

Non-communicable causes of death Australia Age-standardised rates of common cancers Australia

Age-standardised rate per 100 000

Per cent of deaths
Cardiovascular disease Men Women
17 60
Cancer 50
7 Respiratory disease 30
35 0











Data from World Health Organization46 Data from International Agency for Research on Cancer20



In 2004 Brazil had a population of over 185 improved provision of healthcare rather than 2015.46 In 1975, for every obese woman,
million. The country has a middle-income changes in incidence. approximately another two were under-
economy, with a gross domestic product of In 1997, 28 per cent of men and 31 per weight. Between 1975 and 1989, the
8140 international dollars per person (fig- cent of women aged 1869 were classified prevalence of underweight almost halved,
ure 1.3), which masks extreme socioeco- as sedentary (figure 1.6).46 while the prevalence of obesity doubled;
nomic inequalities.46 Life expectancy at In 1997 men aged 2065 had an average so by 1997 there were two obese women
birth is 67 years for men and 74 for women body mass index (BMI) of between 23 and for every underweight woman.28
(figure 1.1).46 25.46 Just 2.1 per cent of men aged 2024 The average amount of available food
Chronic diseases account for 75.3 per had a BMI of over 30, rising to 9.1 per cent energy rose between 1964 and 2004, from
cent of all deaths, with infectious diseases, of 3034 year olds, 11.4 per cent of 4044 2313 to 3157 kcal/person per day (9684 to
maternal, perinatal, and nutritional con- year olds, and 12.3 per cent of men aged 13 218 kJ/person per day), largely due to an
ditions accounting for 16.2 per cent; 8.5 5054.46 Women aged 2039 had an increase in the availability of meat and
per cent of deaths are due to injuries. The average BMI of between 22.5 and 24.9, oils.1 In one study, people living in urban
first figure below gives a breakdown of while from age 40 onwards, average BMI areas ate more vegetables and fruits than
deaths caused by chronic diseases.46 remained between 25 and 27.46 Just 5.2 those living in rural areas; intake increased
Prostate cancer is the most common type of per cent of women aged 2024 had a BMI with age, schooling, and income. However,
cancer in men. This is followed by lung, stom- of over 25, rising to 17.4 per cent of those only 41 per cent of adults reported eating
ach, colorectal, and bladder cancers. Breast aged 4044, and 25.5 per cent of women fruit every day and 30 per cent reported
cancer is the most common type in women, aged 6064.46 The proportion of men and daily vegetable intake.21 Wasting and
followed by cancers of the cervix, colorectum, women with a BMI of 30 or more has stunting in children due to undernutrition
stomach, and lung (for age-standardised rates increased steadily since the mid-1970s.46 In have decreased rapidly since 1975,
of these cancers, see the second figure 1975, just 2.1 per cent of men and 6 per although it remains a major problem in
below).20 Age-standardised mortality rates for cent of women had a BMI of 30 or more; the north-eastern region of the country.
childhood cancers have declined since 1980 this rose to 6.4 per cent and 12.4 per cent Obesity among children is low, but those
and there has been a decrease in mortality in 1996/7.46 See figure 1.4 for projections from higher income households in the eco-
from oral and pharyngeal cancers since the of the proportions of men and women nomically developed south-eastern region
early 1980s.9 These figures may represent who will have a BMI of 30 or more in are more likely to be overweight.28

Non-communicable causes of death Brazil Age-standardised rates of common cancers Brazil

Age-standardised rate per 100 000

Per cent of deaths
Cardiovascular disease
50 Men Women
6 30
48 Respiratory disease 20
20 0











Data from World Health Organization46 Data from International Agency for Research on Cancer20


Continued from page 25

general projected increase in global population; to ageing

populations; and to improved surveillance, detection and, in
the case of prostate cancer, screening and diagnosis.
Nevertheless, at the global level, a real, age-adjusted, popu-
lation-adjusted increase in cancer rates is projected. These
projections show that any global war on cancer is not being
won. It follows that soundly based policies and effective pro-
grammes to prevent cancer which is to say, to decrease
the risk of cancer need not be accompanied by a decrease
in the overall numbers of people suffering and dying from
This ominous prospect should be put in context. First, both
actual numbers and also age-adjusted rates of some cancers
are decreasing in high-income countries, and rates of stom-
ach cancer are generally decreasing worldwide. Second, the
remarkable differences in the numbers of different cancers,
and in their incidence over time, show that most cancer is,
at least in principle, preventable. Third, the theme of this
Report, correlations between changes in patterns of diet,
physical activity, body composition, and changes in patterns
of cancer provide evidence that these factors are important
modifiers of cancer risk.
Furthermore, as shown in Part 2 of this Report, overall the
evidence is a reliable basis for recommendations designed to
prevent cancer. The evidence shows that, together with expo-
sure to tobacco smoke, key aspects of food and nutrition,
physical activity, and body composition are or may be caus-
es of important cancers of some sites. Unlike tobacco, the evi-
dence also shows that other aspects protect against a number
of common cancers. This indicates that many cancers are
preventable not only in principle, but potentially also in


C H A P T E R 2

The cancer process

Food and nutrition modify the risk of cancers at a the ability of humans to evolve and adapt.
large number of sites. This means that some foods Ultimately it is both the genetic information
and drinks, dietary constituents (or their balance in (genotype) and its expression that control the
diets), and methods of food production, processing, characteristics (or phenotype) of an individual. Any
preservation, and preparation influence the exposure during the life course that affects the
development of some cancers. More recently, genotype or its expression may also have an effect
evidence has accumulated about the effects of on the phenotype. At any point in time, the
physical activity and body composition on the risk of phenotype is related not only to the genotype but
a number of cancers, suggesting that bioenergetics also to a host of environmental factors, including
is another factor determining cancer risk and tumour nutritional exposures. This accumulated metabolic
behaviour. experience may begin during maternal and early life,
Since the mid-1990s, great progress has been made and proceed throughout a persons lifetime.
in understanding the cancer process, and which The purpose of this second introductory chapter is
internal and external factors modify cancer risk. to summarise current knowledge and thinking on the
Mapping of the human genome has enabled the biology of the cancer process, with special reference
establishment and development of new disciplines to food and nutrition, physical activity, and body
devoted to understanding biological processes at composition. In Part 2 of this Report, epidemiological
the most basic level, including those that prevent and mechanistic evidence is summarised, and the
cancer, those that cause cancer, and those that Panels assessments and judgements are made,
modify its behaviour. based on a balance of all relevant evidence.
Evolution in living organisms depends on the
accumulation of adaptations as a result of changes
in the expression of the genetic information carried
in DNA. Even with no changes in the DNA,
alterations in how the message in the genetic code is
translated can lead to functional changes. More
importantly, the DNA itself is susceptible to
mutation changes in the genetic code itself as a
result of damage from external causes such as
radiation or simply due to the process of
metabolism. Such mutations are the essential basis
for human evolution, by producing adaptations that
are beneficial in particular environmental
circumstances. At the same time, some mutations
can contribute to the harmful changes in cells that
eventually lead to cancer.
The integrity of the genetic information is
protected by many systems that prevent DNA
damage, or remove or repair damaged DNA if it
occurs. Imperfections in these systems limit the
ability to block all damage and allow both helpful
and harmful mutations to occur. Cancers result when
sufficient mutations have accumulated, most
presenting at an age that was rarely reached in the
evolutionary past of human beings. The
development of cancer may be seen as a corollary of

C H A P T E R 2 T H E C A N C E R P R O C E S S

This chapter summarises the wealth of biological evidence Human adults are made up of around 1013 (or
that documents the ability of food and nutrition, physical 10 000 000 000 000) cells, which are renewed and replaced
activity, and body composition to influence several stages of constantly. About 510 per cent of cancers result directly
the process of the development of cancer. from inheriting genes associated with cancer, but the major-
Nutrients and food constituents have effects that can either ity involve alterations or damage accumulated over time to
inhibit several events that lead to cancer, or contribute to the genetic material within cells. The causes of damage are
cancer development, by altering DNA itself, or by altering both endogenous (internal) and exogenous (environmental).
how the genetic message in DNA is translated. Physical Food, nutrition, and physical activity are important envi-
activity and variations in body composition also appear to ronmental factors in the development of cancer.
influence cancer risk. Indeed, overall dietary patterns can Cancer can also be seen as a group of diseases that affects
indirectly influence cell growth by way of changes in gen- many different tissues and types of cell, and can be defined
eral metabolic, regulatory, and endocrine effects. by their tissue of origin. Approximately 85 per cent of adult
The normal functioning of all biological processes, includ- cancers develop from the epithelial cells of the inner and
ing those of the human body, depends on the availability of outer linings of the body and are called carcinomas. Cancers
substrate and nutrients. Good nutrition defined as appro- of glandular tissue such as the breast are called adenocarci-
priate provision of food and nutrients from the level of the nomas; cancers from bone and muscle derived from meso-
whole organism to the cellular and intracellular level is derm cells (embryonic cells that grow to form muscle, blood,
needed for normal structure and function. When a person is bone, and connective tissue), are called sarcomas.
not suitably nourished, either through under- or overnutri- Each type of cancer has different characteristics, but one
tion, this impacts on the tissue microenvironment, compro- feature of all these diseases is unregulated cell growth and/or
mising both structure and function. cell death. Apart from haematological cancers such as
Understanding the mechanisms underlying cancer devel- leukaemias, this results in a tumour or mass, and cancerous
opment is central to improving its prevention and treatment. cells often invade the surrounding tissue. Spread of cancer
The main body of this Report comprises the Panels judge- cells from the primary site to other parts of the body is called
ments on a series of systematic literature reviews (SLRs) on metastasis. Benign tumours do not invade or metastasise.
evidence linking food, nutrition, and physical activity to can- Malignant tumours do not remain localised but can invade
cer. All evidence on the mechanisms of the cancer process is and/or metastasise.
also based on rigorous review criteria. However, the evidence
presented in this chapter is a summary of this literature, and 2.1.2 Genetic material
the references cited are illustrative examples only. Full details The genetic material of mammalian cells is composed of dou-
of the methods used for the reviews are contained in the SLR ble-stranded DNA made from four organic bases cytosine,
specification manual. The full SLRs and this manual are con- guanine, adenine, and thymine within a helical spine
tained on the CD included with this Report. comprising deoxyribose (a sugar) and phosphate. The com-
bination of a base with phosphate and deoxyribose is called
a nucleotide. Humans have 3 billion base pairs in the DNA
2.1 Basic concepts and principles code that encode approximately 30 000 different genes.
The nucleus of a cell contains DNA, and the information
2.1.1 Cancer in the code is read to generate proteins in the cytoplasm of
Cancer is a group of more than 100 diseases characterised by the cell. This is achieved by transcribing the DNA into RNA,
uncontrolled cellular growth as a result of changes in the and then translating the information in RNA to synthesise
genetic information of cells. Cells and tissues are complex sys- protein. For transcription, the two DNA strands separate and
tems with critical stages and checkpoints to ensure normal an intermediary, complementary copy of the DNA is made
growth, development, and function. Normally the division, dif- from mRNA (which differs slightly in structure from DNA
ferentiation, and death of cells are carefully regulated. All can- and is single stranded). For translation, the RNA leaves the
cers start as a single cell that has lost control of its normal nucleus and binds to an organelle in the cytoplasm called the
growth and replication processes. ribosome. The RNA nucleotides encode for 21 different


components in terms of the

dose (quantity), timing, and
duration of exposure required
to bring about effects. To unrav-
el the contribution of nutrition
to cancer, the biological
processes underpinning cancer
development need to be under-
stood. Extensive evidence exists
for nutritional factors in sever-
al processes related to cancer
development (figure 2.2).
However, because of the com-
plexity of the process, it is not
possible to conclude that mod-
ifying any one, or more, of
these processes influences
cancer risk.
The recent expansion of
knowledge in molecular biolo-
gy has allowed new techniques
to be developed to explain
The genetic message in the DNA code is translated into RNA, and then into protein synthesis,
these mechanisms. Nutri-
and so determines metabolic processes. Research methods called -omics address these different
genomics is a new field with
profound implications in cancer
prevention and therapy, since it
amino acids, with the ribosome moving along the RNA mol- seeks to clarify the impact of nutrition in the maintenance
ecule and translating the genetic code into a sequence of of genome stability, and to dissect out the influence of geno-
amino acids that build into a protein. type in determining our response to diet. Nutrigenomics is
The normal metabolic processes in cells are controlled by the study of nutritional influences on the phenotypic vari-
proteins, each of which is a product of a single gene from the ability of individuals based on genomic diversity (figure 2.1).
DNA in the nucleus. Although each cell in the body contains This determines the sequence and functions of genes, and
exactly the same genes, cells from different organs have dif- studies single nucleotide polymorphisms (SNPs), and ampli-
ferent structures and functions because there is a process of fications and deletions within the DNA sequence as modi-
regulation that determines which genes are expressed; that fiers of the response to foods and beverages and their
is, which genes are turned on and which are not (see 2.2.2). constituents. Nutritional epigenomics is another key deter-
This differential gene expression varies not only from tissue minant of gene expression patterns. It includes non-coding
to tissue but also from time to time over the course of a per- modification of genes (such as methylation, changes in his-
sons life, from embryonic and fetal stages onwards (box 2.1). tone homeostasis, miRNA, and DNA stability) in response
Gene expression is regulated by promoter regions of genes in to nutrition. Nutritional transcriptomics is the study of gene
the DNA, as well as by epigenetic factors those that alter expression patterns at the RNA level, and it can identify
gene expression without changing the nucleotide sequence common nutritional response elements in gene promoters
(see 2.2.3). The availability of nutrients within the immedi- that can be modulated by diet. Proteomics studies the pro-
ate environment influences these processes (figure 2.1). teins that can be expressed by a cell, many of which can be
influenced by nutrition. Metabolomics studies the range of
2.1.3 Nutrigenomics and cancer metabolic processes in a cell and metabolic regulation in
Unravelling links between diet and cancer is complex, as cells or tissues, which again are heavily influenced by food,
thousands of dietary components are consumed each day; a nutrition, and physical activity.
typical diet may provide more than 25000 bioactive food
constituents.1 Assessing intakes of some constituents is dif-
ficult due to wide variations in the amounts of bioactive com- 2.2 Cellular processes
ponents within a particular food.2 3 Dietary constituents
modify a multitude of processes in both normal and cancer The role of nutrition in cancer depends on how it impacts
cells.4 5 on fundamental cellular processes including the cell cycle
The response is further complicated since a single, bio- (figure 2.3; also see 2.5.1). To understand cancer biology,
active food constituent can modify multiple steps in the it is important first to understand normal cellular processes.
cancer process. Likewise, many of these processes can be The integrity of tissues and organs depends on a regulated
influenced by several food components. Normal and cancer balance between cell proliferation and death, and appro-
cells also differ in their responses to bioactive food priate cell differentiation. This regulation is controlled by

C H A P T E R 2 T H E C A N C E R P R O C E S S

several types of genes including oncogenes and tumour sup- process called phosphorylation. A regulated process of phos-
pressor genes (box 2.2), and factors in the cellular environ- phorylation and dephosphorylation is necessary for the
ment that influence their expression. Maintenance of the appropriate initiation, transmission, and cessation of signals.
DNA sequence and structure as cells divide is essential: sev-
eral cellular mechanisms exist to ensure this is achieved. 2.2.2 Gene expression
Gene expression is the process by which the information
2.2.1 Cell signalling within a gene is turned on or turned off. The information
Cells detect and respond to external stimuli and send mess- is used to create the associated proteins and modify the
ages to other cells through a molecular mechanism known amounts produced. Also see figure 2.1.
as cell signalling. Transcription factors are proteins involved in the regula-
Cells within a tissue normally communicate with each tion of gene expression and carry the signal from the
other through a network of locally produced chemicals called cytoplasm to the nucleus. They bind to the promoter regions
cytokines (including some growth factors). Cell proliferation of genes and have the effect of either switching gene
is a tightly controlled and coordinated process, and is stim- expression on or off. There are also nuclear receptors, such
ulated by growth factors. These soluble proteins can be pro- as retinoic acid receptors, that function as transcription
duced locally, either from the same cell (autocrine), or from factors by binding directly to specific DNA sequences.
different cells (paracrine), or as hormones (endocrine) pro- Some so-called housekeeping genes are expressed by
duced by a distant tissue and transported in the blood. almost all cell types. These genes generally encode proteins
Growth factors bind to specific receptors on the cell surface that participate in basic cell functions such as metabolic
and transmit a signal into the cell, which is relayed to the pathways and synthesis, and processing of DNA, RNA, or pro-
nucleus. In the nucleus, genes are switched on to produce teins. Other genes have more restricted expression, and
the proteins necessary for cell division. are expressed only in specific cell types, and/or stages of
Getting the growth signal from the outside of the cell to development.
the nucleus requires a series of steps. The shape of the recep- Gene expression can also be influenced by changes outside
tor changes when the growth factor binds to it, which caus- the DNA of genes. DNA is closely organised and tightly pack-
es part of the receptor to become activated, usually by a aged in the nucleus of cells. To achieve this, DNA is spooled
around proteins called his-
tones. Histone structure can
be modified either, like DNA
itself, by methylation, or
more commonly by acetyla-
tion (addition of an acetyl
group). Acetylation and
deacetylation (removal) are
mediated by the enzymes
histone acetyl transferase
(HAT) and histone deacety-
lase (HDAC), respectively.
HATs relax the packaged
DNA structure, which is asso-
ciated with enhanced tran-
scription, whereas HDACs
stabilise the structure with
higher levels of packaging,
and so suppress transcrip-
tion. Butyrate, produced in
the colon by bacterial
fermentation of non-starch
polysaccharide (dietary
fibre), diallyl disulphide
from garlic and other allium
vegetables, and sulphora-
phane, a glucosinolate from
cruciferous vegetables, can
behave as histone deacety-
lase inhibitors,16 and act to
maintain DNA stability or
Food, nutrition, obesity, and physical activity can influence fundamental processes shown here, enhance transcription.
which may promote or inhibit cancer development and progression. Micro RNAs (miRNAs) are


Box 2.1 Nutrition over the life course

The best understanding of the impact of of the fetus to maternal glucocorticoids. metabolic disturbances associated with
endogenous and exogenous factors on the This led to a permanent alteration in this syndrome promote genetic instability.
cancer process will come from studies of hormonal status and metabolic responses in Growth and development are dependent
the whole life course, and particularly of the offspring.8 9 These effects were attrib- on the supply of adequate energy and
critical periods within it. uted to differential methylation of certain nutrients to match a persons needs. Famine
Early nutritional exposure is an impor- glucocorticoid genes and could be mitigat- exposure in women affects breast dysplasia
tant determinant of phenotypic expression ed by maternal folic acid supplementation.4 later in life: girls exposed to famine before
during later life and is likely to affect vul- Lower birth weight, followed by expo- the age of 10 years have less dysplasia in
nerability to chronic diseases, including sure to periods of rapid rates of growth, later life, whereas those exposed after 18
cancer.6 During pregnancy, the nutrient possibly due to energy-dense diets, is asso- years of age have more dysplasia than non-
demands of the fetus have to be satisfied. ciated with development of metabolic exposed women. This illustrates the impor-
Although the mothers dietary intake is syndrome during adult life.6 This is the clus- tance of timing of nutrition during key
important in maintaining her own nutrient tering of several cardiovascular risk factors stages of development.12
reserves, it is her nutrition status on enter- including hypertension, abdominal obesi- Greater body fatness later in life is
ing pregnancy that determines her capac- ty, insulin resistance, and type 2 diabetes. linked to development of metabolic syn-
ity to deliver appropriate nutrients to the Type 2 diabetes has been associated with drome and related health problems,
fetus. Any stress that modifies her nutri- increased prevalence of some cancers, and including insulin resistance. The inflam-
tional state, either by changing appetite or risk of cancer shows a graded relationship matory state associated with obesity also
altering nutrient demand, can impact on with glycated haemoglobin (a measure of promotes cancer.13 In contrast, energy
the availability of nutrients to the fetus.7 In blood glucose control) throughout the nor- restriction (box 2.5) delays the onset
experimental models, pregnant rats fed a mal range.10 Metabolic syndrome may also of many age-related diseases, including
low-protein diet resulted in overexposure increase cancer risk,11 suggesting that the cancer.

RNA molecules that do not encode proteins; instead they division (see 2.5.1), so maintaining the particular structur-
function as negative regulators of gene expression. Some al and functional characteristics of the cells of specific
mutations in miRNAs have been associated with human can- tissues.
cers, and they can function as oncogenes or tumour sup- Dietary constituents contribute to epigenetic modulation
pressor genes (box 2.2). miRNAs are short, single-stranded of promoter regions in both normal and malignant cells. For
RNA molecules of approximately 22 nucleotides. For example, dietary folate and other methyl-donors such as
instance, to silence (turn off) genes, they may bind to com- choline, methionine, and betaine are essential for DNA
plementary mRNA sequences and degrade them before they synthesis and epigenetic regulation of DNA. Appropriate
have been translated. Profiling miRNA signatures within can- gene expression is maintained by appropriate patterns of
cer cells may aid the diagnosis, classification, and treatment methylation; folate is an important determinant of normal
of cancer. For example, a certain miRNA that is downregu- methylation. In addition, dietary constituents such as genis-
lated in lung cancer is associated with decreased survival.17 tein, which do not provide methyl groups, have also been
Research on the interactions between nutrition and non-cod- reported to modify DNA methylation.
ing RNA molecules is at an early stage but is potentially rel- Imbalances or lack of specific dietary constituents may
evant to cancer. potentially increase the risk of cancer by inducing an imbal-
ance in DNA precursors, leading to altered DNA synthesis
2.2.3 Epigenetic regulation and repair, and may impair appropriate patterns of DNA
Gene expression can also be altered without changing the methylation, with consequences for gene expression. For
DNA sequence. This is called epigenetic modulation. example, inadequate folate availability means cells tend to
Methylation of DNA the addition of a methyl group incorporate uracil into DNA in place of thymine. Global
(CH3) plays a role in gene silencing. Methylation occurs hypomethylation of DNA is an early epigenetic event in cer-
only to cytosine residues located next to guanine bases in the vical carcinogenesis, and the degree of hypomethylation
DNA sequence. These CpG dinucleotide sequences are found increases with the grade of cervical cancer.18 19 Global
throughout the genome; in about half of all genes, clusters hypomethylation of DNA and site-specific hypermethylation
of CpG sequences, so-called CpG islands, are located in the may also be relevant in colorectal cancer.20 A host of bio-
promoter region that normally functions to promote expres- active food constituents from alcohol to zinc has been
sion of the gene. Transcription factors cannot bind to these reported to influence DNA methylation patterns.
sites when methylated and so the gene is silenced. For active Nevertheless, it remains to be determined how important
genes, the CpG islands in the promoter regions are general- these changes are in the overall cancer process. The poten-
ly not methylated. tial cancer-protective effects of green, leafy vegetables (see
Controlled DNA methylation also provides a mechanism chapter 4.2) are often attributed to their folate content.
for cell differentiation. In normal cells, genes may be per- Folates function as a coenzyme in the metabolism of single-
manently silenced so that they can no longer be expressed, carbon compounds for nucleic acid synthesis and amino acid
in a way that is transmitted into daughter cells during cell metabolism.

C H A P T E R 2 T H E C A N C E R P R O C E S S

2.2.4 Differentiation to different growth factors. In the cancer process, one char-
Cells become specialised to perform their particular function acteristic of cells that are accumulating DNA damage is that
through a process known as differentiation. they become de-differentiated and undergo epithelial-
Although each cell within the body contains the entire mesenchymal transition (EMT), characterised by loss of
genome with the same genes, only some genes are active in cell adhesion and increased cell mobility. In addition, dur-
any one cell at any one time. Gene expression controls the ing differentiation, other genes can be silenced by chromatin
subset of genes expressed and the timing of expression, and modification, including DNA methylation and histone mod-
this distinguishes one cell type from another by determin- ifications such as methylation and acetylation (see 2.2.2).
ing their particular structure and function. Stem cells are found among differentiated cells in almost
Hundreds of different cell types arise from one fertilised all adult tissues, where they maintain and regenerate tissues.
egg during development; this is achieved by proliferation Examples include haemopoietic stem cells for continuous
and differentiation of stem cells. Stem cells are unspecialised generation of red and white blood cells, and stem cells in
but can give rise to different specialised cell types. A small the basal layer of the epidermis and at the base of hair fol-
population of cells within the tumour mass of several forms licles in the skin, which can give rise to keratinocytes and
of human cancer, which have both the properties of stem hair follicles. Cellular differentiation also continues in solid
cells and also the characteristics of transformed cells, may tissues such as mammary tissue, which during pregnancy dif-
be important for the development of these tumours. As yet, ferentiates for later milk production. A systematic review has
our understanding of cancer stem cells is basic, although the shown that long-chain n-3 polyunsaturated fatty acids
concept was proposed as early as 1875.22 Several groups have (PUFAs) in fish oils promote differentiation of colonic epithe-
now isolated and identified cancer stem cells in both lial cells.30 31 Vitamin D and retinoic acid may also promote
haematopoietic and epithelial cancers, including cancers of the differentiation of cells.
breast,23 brain,24 25 ovary,26 prostate,27 colon,28 and stomach.29
It is clear that subtle changes in exposure to bioactive food 2.2.5 DNA damage and repair
constituents can have a profound effect on the differentiation Each time a cell divides into two new daughter cells, there
of these cells. is potential for an error in replication of the DNA (see 2.5.1).
In early embryos, proliferation of embryonic stem cells These mutations result in non-functioning genes or in pro-
increases the total number of cells. As the organism devel- teins with altered amino acid sequences that can change cell
ops, cells differentiate and become specialised to their par- function.
ticular function. Transcription factors specific to that cell type DNA is continuously exposed to damage from products of
turn on genes for that particular lineage of cells, so deter- normal intracellular metabolism, including reactive oxygen
mining its structure and function. species, hydroxyl radicals, and hydrogen peroxide; and also
At various stages of differentiation, cells become sensitive to damage from external factors such as ultra-violet (UV)

Box 2.2 Oncogenes and tumour suppressor genes

Oncogenes and tumour suppressor genes tions occur in approximately 30 per cent of approximately 300 different genes, thereby
are present in all cells, and in their normal, all human cancers (and in 7590 per cent of having an important regulatory role in var-
non-mutated form contribute to the regu- pancreatic cancers). ious molecular pathways. The p53 pathway
lation of cell division and death. In cancer, Tumour suppressor genes prevent exces- is altered in most cancers, and the mutated
both types of gene are often mutated, and sive growth of a cell, either by controlling protein product cannot protect the
these alterations contribute to the cancer cell proliferation or by controlling DNA genome, allowing mutations to accumulate.
process. The combined effect of activation repair and genomic stability. Mutation of a In the absence of a functional Rb gene,
of oncogenes and inactivation of tumour tumour suppressor gene results in the loss the retinoblastoma protein pRb is not made
suppressor genes is an important driver of of function of the protein product. and its inhibitory role of inactivating cell
cancer progression. Common tumour suppressor genes inc- cycle transcription factors is absent. This
Oncogenes increase the rate of transfor- lude p53 and the retinoblastoma (Rb) gene. leads to increased, uncontrolled prolifera-
mation from a normal to a cancerous cell. The p53 gene was the first tumour sup- tion, and thereby increased risk of further
Oncogene function is changed by mutation pressor gene to be identified, and is con- mutations. More than 2000 inactivating
so that the protein product is produced sidered to be the guardian of the genome.14 mutations of p53 have been identified in
either in greater quantities or has increased Under normal circumstances, p53 is involved human tumours. Both p53 and pRb can also
activity. The normal, non-mutated form of in several processes such as cell prolifera- be inactivated by the human papilloma
an oncogene is called a proto-oncogene. tion, DNA repair, apoptosis, and angiogen- virus. Development of lung cancer is associ-
More than 100 oncogenes have been esis. p53 is activated by cellular stresses that ated with loss of function of both genes. Li-
identified, including ERBB2 in breast and could facilitate tumour development such as Fraumeni syndrome, a result of a germ line
ovarian cancers, members of the Ras family hypoxia, lack of nucleotides, and in partic- mutation in p53, leads to inherited
in cancers of the lung, colorectum, and pan- ular, DNA damage. susceptibility to a wide range of cancers.
creas, and MYC in lymphomas. Ampli- In response, p53 can either halt the cell Patients with Li-Fraumeni syndrome have
fication of the ERBB2 gene occurs in around cycle to allow DNA repair, or induce apop- a 25-fold increased risk of developing cancer
30 per cent of breast cancers, and RAS muta- tosis. It can bind to the promoter regions of compared with the general population.15


light, as well as other environmental factors including food, DNA adducts. These adducts distort the structure of DNA and
nutrition, and physical activity (see 2.4). There are several disrupt its replication, potentially causing mistranslation.
mechanisms for DNA repair (box 2.3), a vital defence in They can also break the DNA strand, which can result in
maintaining cellular integrity and preventing a cell being mutations or deletions of genetic material.
transformed from normal to cancerous. In addition to P450 enzymes, other systems such as per-
Various studies suggest that nutritional status and/or cer- oxidases (including the cyclooxygenases) and certain trans-
tain food constituents may influence DNA repair. Data from ferases, such as N-acetyltransferase and sulphotransferase,
observational studies suggest that severe malnutrition can can influence carcinogen bioactivation.42-44
impair DNA repair.34 Nucleotide excision repair has been Competing with this metabolic activation of carcinogens
found to be lower in adults with the lowest intakes of is the detoxification process catalysed by a second group of
folate.35 Studies of healthy adults consuming kiwi fruits,36 enzymes, known as phase II enzymes.45 They catalyse con-
cooked carrots,37 or supplements of coenzyme Q10 (an impor- jugation reactions, producing molecules that can be excret-
tant cofactor in metabolism)38 have demonstrated improved ed in bile or urine. Examples include acetyltransferases,
DNA repair in vitro. Selenium induces base-excision repair glutathione-S-transferases (GSTs), UDP-glucuronyltrans-
in vitro via p53 activation in cultured fibroblasts.39 ferases, NAD(P)H:quinone oxidoreductase, and sulpho-
Evidence exists that some dietary components can modify transferases. The induction of phase II enzymes is mediated
DNA damage and gene expression in exfoliated colonocytes. by the antioxidant response element, which is located in the
For example, the amount of single-strand breaks in exfoli- promoter region of specific genes.46
ated colorectal mucosal cells was significantly lower in Each of these enzymes represents a potential target for
healthy individuals consuming a vegetarian diet compared dietary components to influence carcinogen activation.
with a diet high in meat.40 Similarly, DNA damage in exfo- Specificity in response is evident since isothiocyanates from
liated lung epithelial cells can be influenced by dietary com- cruciferous vegetables have been shown to induce expression
ponents. Consumption of a lycopene-rich vegetable juice was of phase II detoxification enzymes without affecting expres-
associated with significantly decreased damage to the DNA sion of phase I enzymes.
of lung epithelial cells in healthy volunteers.41 The carcinogenic properties of polycyclic hydrocarbons,
Defects in any of the DNA repair mechanisms can predis- aromatic amines, N-nitroso compounds, and aflatoxins
pose to cancer. Several inherited conditions link such defects result from the metabolism of these compounds, which
to cancer incidence. For example, patients with hereditary produces carcinogenic by-products. Metabolising enzyme
non-polyposis colorectal cancer have defective mismatch activity can be modulated by dietary factors. The enzyme
repair (see chapter 7.9.2). CYP3A4 participates in the metabolism of over half of
Although the processes of cell development, signalling, all pharmaceutical agents, and is especially sensitive to
and DNA repair are tightly controlled, errors will occur dur- dietary effects. For example, interactions have been report-
ing the trillions of cell divisions that occur over a lifetime. ed between grapefruit juice, red wine, garlic, and various
This may result in inappropriate proliferation or failure of drugs.47 Fooddrug interactions involving CYP1A2, CYP2E1,
damaged cells to die. These changes could provide the glucuronysyltransferases, and GSTs have also been
altered cell with a growth advantage over the normal cells documented.48
in the tissue. If additional alterations occur, this can result The activity of phase I and II enzymes, and thus carcino-
in a cell with the potential to become cancerous. gen metabolism and cancer development, varies between
individuals. SNPs in several phase I and phase II enzymes
have been shown to modulate cancer risk.49 50 There is some
2.3 Carcinogen metabolism difficulty in synthesising this evidence and some studies may
give false positive results. Nevertheless, the literature shows
The chance of carcinogenesis occurring and then progress- tantalising evidence that relates genetic diversity to these
ing is modified by many factors, including food and nutri- enzymes in various processes linked to cancer development.
ion. Dietary constituents that modify carcinogenesis include Some specific examples of SNPs in cancer are listed below.
selenium, allyl sulphur, sulphuraphane, and isoflavonoids. GSTs are involved in the metabolism of environmental car-
Most dietary carcinogens require activation to produce cinogens and reactive oxygen species. People who lack these
reactive intermediates that bind to and damage DNA. Phase enzymes may be at higher risk for cancer because of a
I and phase II metabolising enzymes in the liver and in other reduced capacity to dispose of activated carcinogens.51 As
tissues are involved in this process. These are enzymes that shown in chapter 4.1, aflatoxins produced by moulds that
catalyse the metabolic detoxification of xenobiotics, drugs, contaminate certain types of foods such as cereals (grains)
and carcinogens produced through aerobic metabolism, and and peanuts are carcinogens that are activated by phase
thus protect cells against oxidative stress and reactive elec- I enzymes in the liver into reactive DNA metabolites that
trophilic carcinogen metabolites. cause DNA adducts. However, the number of adducts and
Metabolic activation of carcinogens is generally catalysed therefore the potency of aflatoxin B are influenced by other
by the cytochrome P450 (CYP) family of phase I enzymes enzymes such as GSTs, which eliminate carcinogens before
through oxidation, which usually makes the molecule more they can bind to DNA. Individuals with low expression of
water soluble. Some of the intermediates formed during this GSTs are also at higher risk for colorectal cancer; however,
process may be carcinogenic, and can bind to DNA, forming a diet high in isothiocyanates (derived from glucosinolates

C H A P T E R 2 T H E C A N C E R P R O C E S S

Human intervention studies with cruciferous vegetables

Box 2.3 Mechanisms for DNA repair
have demonstrated induction of GSTs by consumption of
Mammals have five types of DNA repair system32 33: Brussels sprouts and red cabbage varieties, but not with
Direct reversal corrects rather than removes damaged DNA white cabbage and broccoli. The particular isoform of the
bases. enzyme induced may protect against bladder cancer.
Base-excision repair corrects DNA damage caused by reactive Cruciferous vegetables also affect drug metabolism in
oxygen species, deamination, and hydroxylation arising from humans, and red cabbage leads to specific changes in the
cellular metabolism and spontaneous depurination.
patterns of meat-derived urinary mutagens. ITCs may also
Nucleotide-excision repair removes lesions that distort the
structure of the DNA helix, such as pyrimidine dimers and
protect against mutations formed by tobacco carcinogens.61
DNA adducts. In a variety of animal studies, certain dietary components
Homologous recombination and non-homologous end- have shown a reduction in experimentally induced cancers.
joining repair double-strand breaks. Homologous Rats fed ITCs developed significantly fewer oesophageal
recombination is used when a second identical DNA copy is cancers due, in part, to inhibition of cytochrome P450-
available, for example after replication; non-homologous mediated bioactivation of the carcinogen. Dietary indole-3-
end-joining re-links the broken ends of a double-strand carbinol inhibited spontaneous occurrence of endometrial
break. adenocarcinoma and preneoplastic lesions in rats.62
DNA mismatch repair detects and repairs copying errors Flavonoids (polyphenolic compounds found in plant
made during replication.
foods) may also alter carcinogen metabolism. Quercetin has
been shown to inhibit the expression of cytochrome P450
and phase I enzymes, and may reduce tobacco carcinogen
in cruciferous vegetables) can ameliorate this.52 activation.63
Epidemiological evidence supports the idea that individ- Exposure of vulnerable populations to excess amounts of
uals possessing these genotypes are predisposed to a num- dietary constituents, irrespective of whether they are nutri-
ber of cancers, including those of the breast, prostate, liver, ents or not, may actually increase the risk of cancer. In
and colon53; it also shows that nutrition influences cancer one example, consumption of beta-carotene supplements
risk. Thus, an SNP in the methylenetetrahydrofolate reduc- in smokers was associated with increased incidence of lung
tase (MTHFR) gene appears to influence folate metabolism cancer.64 For most dietary components, there will be an
by reducing MTHFR activity, and is associated with a reduc- upper threshold beyond which people may be exposed to
tion in the risk of developing colorectal cancer.54 Certain adverse effects. There are also concerns regarding excessive
SNPs in the N-acetyltransferase gene alter the activity of the supplementation with folic acid, iron, copper, iodide, and
enzyme (involved in the metabolic activation of heterocyclic selenium.65
amines from meat cooked at high temperatures), and may
also increase the risk of colon cancer.
Again, as described in chapter 4.3, people whose diets are 2.4 Causes of cancer
high in red and processed meat and who also carry an insert
variant in CYP2E1, a key activating enzyme of many nitro- A number of different types of exogenous (environmental)
samines, have a greater risk of rectal cancer.55 Consumption factors are known causes of cancer. These include some
of cruciferous vegetables protects against lung cancer in indi- aspects of food and nutrition that are established as car-
viduals lacking the GSTM1 gene.56 In addition, genes that cinogenic by the International Agency for Research on
predispose to obesity may promote obesity-related cancers Cancer, although it is difficult to estimate the proportion of
(box 2.4). cancers directly attributable to these.74 Known causes of can-
Dietary components can either be, or be activated into, cer include tobacco smoking and its use, infectious agents,
potential carcinogens through metabolism, or act to prevent medication, radiation, and industrial chemicals, and also
carcinogen damage. For instance, as summarised in chapter some factors within the scope of this report carcinogenic
4.3, high intake of red meat may result in more absorption agents in food and drink.
of haem iron, greater oxidative stress, and potential for DNA Sometimes the extent to which cancer in general, or spe-
damage.57 58 In addition, iron overload can also activate cific cancers, may be modified by any factor, are calculated,
oxidative responsive transcription factors and inflammation and some of these figures are given here. However, these esti-
in the colon.59 Red meat consumption is also associated with mates should be treated with some caution. First, they are esti-
the formation of N-nitroso compounds. This increases the mates, and cannot be exact, and so are best given as ranges.
level of nitrogenous residues in the colon and is associated Second, individual causes of cancer often interact with one
with the formation of DNA adducts in colon cells.60 another to increase or decrease risk, or are modifiers or pre-
Cruciferous vegetables contain glucosinolates, which are cursors of others; and some act together to produce a multi-
transformed by food preparation into isothiocyanates (ITCs), plicative effect. This point is particularly important with food
which alter the metabolism of carcinogens. Indoles and ITCs, and nutrition, which may have a substantial effect on the risk
two major glucosinolate breakdown products, attenuate the of a cancer with environmental causes other than food, nutri-
effects of polycyclic aromatic hydrocarbons and nitrosamines tion, and physical activity. Third, it is sometimes assumed that
via induction of GSTs and inhibition of cytochrome P450 once all factors that decrease (or increase) risk are added
isoenzymes, respectively. together, with allowance for unknown factors, such estimates


should add up to 100 per cent. This is not so. Reasonable esti- chemicals. These include cytokines, growth factors, reactive
mates for a number of separate or interactive factors may add oxygen and nitrogen species, cyclooxygenase, and lipoxyge-
up to well over 100 per cent because of the interactions. nase products. A chronic inflammatory environment can
increase proliferation and differentiation, inhibit apoptosis
2.4.1 Endogenous causes (programmed cell death), and induce angiogenesis (gener- Inherited germ line mutations ation of new blood vessels). Chronic inflammatory condi-
As mentioned, only a minority (510 per cent) of cancers are tions, such as Barretts oesophagus and ulcerative colitis,
linked to single inherited genes. Such inherited alterations predispose to cancer. In Barretts oesophagus, reflux of acid
are termed germ line mutations, and are passed on from egg can cause the cells lining the gullet to undergo EMT, and
or sperm DNA. Individuals with inherited germ line muta- some areas can develop dysplasia and ultimately cancer.
tions will not definitely get cancer but have an increased risk Around 1 per cent of people with Barretts oesophagus will
of developing cancer compared with the general population. develop oesophageal cancer (this is between 30 and 125
Often mutations in tumour suppressor genes (box 2.2) times higher than the general population). Also see chapter
increase the chance of developing cancer at a young age. 7.3.2. Approximately 5 per cent of patients with ulcerative
These include retinoblastoma (a rare cancer of the eye), Li- colitis, a form of irritable bowel disease, will develop colon
Fraumeni syndrome, multiple endocrine neoplasia type 1, cancer. Also see chapter 7.9.2. Epidemiological and experi-
and kidney cancer in Von Hippel-Lindau disease. Mutations mental evidence has demonstrated that long-term use of
in the BRCA1 and BRCA2 (breast cancer susceptibility) genes non-steroidal anti-inflammatory drugs can inhibit cancer
cause 510 per cent of all breast cancer cases. These genes development in a number of tissues including colon,
normally produce DNA repair proteins. Patients with the oesophagus, and breast.76
syndrome familial adenomatous polyposis coli have a pre- Cancer induced by inflammation may be susceptible to
disposition to colorectal cancer due to mutations in the ade- nutritional influences. Thus, dietary constituents could be
nomatosis polyposis coli tumour suppressor gene.75 Other involved in generation of reactive oxygen species, could
common cancers, including those of the ovary, prostate, pan- influence antioxidant defences, or could suppress the inflam-
creas, and endometrium, may be related to inherited muta- matory process. For example, the glucocorticoid receptor
tions, but only in a small percentage of cases. pathway and the vitamin D receptor are capable of sup-
The other type of genetic mutation somatic gene pressing inflammation.
changes develops during the life course. Such somatic The immune system can be divided into innate and adap-
mutations are not passed on to offspring. This DNA damage tive responses. Innate immunity provides initial defences.
is caused by exposure to external factors such as radiation Adaptive immunity develops later and involves activation of
or carcinogens, or harmful products of normal aerobic lymphocytes and their differentiation into effector and mem-
metabolism. ory cells. The immune surveillance hypothesis proposes that
both the innate and adaptive immune systems constantly Oxidative stress survey for and eliminate newly formed cancer cells, and that
Reactive oxygen species generated through normal oxidative onset and progression of cancer are kept under control by
metabolism have the potential to cause extensive DNA dam- the immune system.77 Immunosurveillance requires that the
age. The body has several mechanisms, which can scavenge immune system recognises something different about cancer
reactive oxygen species to prevent such damage occurring, cells compared with normal cells within the same tissue
or block the effects. often different proteins (termed tumour antigens) that are
Reactive oxygen species cause oxidative damage to DNA. expressed on the surface of a cancer cell.
During repair (see 2.2.5), the damaged, oxidised bases are This recognition of altered self allows the immune sys-
excreted in the urine. Levels of urinary 8-hydroxy-2'- tem to generate a response to these tumour antigens. They
deoxyguanosine, an oxidative DNA damage adduct, can be can be proteins that are only expressed by cancer cells, and
used as an indicator of oxidative DNA damage in humans newly expressed during cancer development; or proteins that
and rodents. Antioxidants can scavenge reactive oxygen have become mutated during the cancer process and so are
species. Vitamins C and E can donate electrons to free rad- different from the non-mutated protein; or proteins
icals and block their damaging activity. Dietary constituents expressed due to differentiation of cancer cells (termed dif-
such as ITCs and polyphenols can also activate the signalling ferentiation antigens); or proteins that are normally
pathways that lead to activation of the antioxidant response expressed by cells but that are expressed at much higher lev-
element (see 2.3), and upregulation of the expression of els by cancer cells. Evasion of immunosurveillance is some-
detoxifying enzymes. times referred to as a further hallmark of cancer78 (see 2.5),
although the evidence remains speculative. Inflammation Specialised mucosal cells form the interface between the
Inflammation is a physiological response to infection, foreign inside and outside of the body.79 These are normally an effi-
bodies, trauma, or chemical or other irritation, and in the cient barrier against pathogens. The gut barrier consists of
acute phase can be helpful. However, chronic inflammation gut-associated lymphoid cells that can sense pathogens, and
can result in DNA damage and cancer promotion. participate in innate and adaptive responses.80 The function
Chronically inflamed tissue is infiltrated with a variety of of these cells is dependent on nutrition.81 82 For example, n-
inflammatory cells that produce a wide variety of bioactive 3 PUFAs can enhance immunity, whereas high concentrations

C H A P T E R 2 T H E C A N C E R P R O C E S S

Box 2.4 Body fatness and attained height

As shown in Chapter 6, the overall evidence synthesis in men and postmenopausal alpha,13 interleukin (IL)-6, and C-reactive
that body fatness is a cause of a number of women,13. The increased insulin and IGF-1 protein, compared with lean people,71
cancers is convincing. Some of the mecha- levels that accompany body fatness result as well as of leptin, which also functions
nisms by which body fatness increases the in increased oestradiol in men and as an inflammatory cytokine.72 Such
risk of cancer are well understood. women,13 and may also result in higher chronic inflammation can promote cancer
Obesity influences the levels of a num- testosterone levels in women (extreme development.
ber of hormones and growth factors.66 obesity can lead to polycystic ovary dis- Also as shown in Chapter 6, factors
Insulin-like growth factor 1 (IGF-1), insulin, ease). Increased levels of sex steroids are that lead to greater adult attained height,
and leptin are all elevated in obese people, strongly associated with risk of endome- or its consequences, are a cause of a num-
and can promote the growth of cancer trial and postmenopausal breast cancers,69 ber of cancers. Adult height is related to
cells. In addition, insulin resistance is and may impact on colon and other can- the rate of growth during fetal life and
increased, in particular by abdominal fat- cers. As shown in Chapter 6, body fatness childhood. The number of cell divisions
ness, and the pancreas compensates by probably protects against premenopausal in fetal life and childhood, health and
increasing insulin production. This hyper- breast cancer; this may be because obese nutrition status in childhood, and age of
insulinaemia increases the risk of cancers of women tend to have anovulatory men- sexual maturity can alter the hormonal
the colon and endometrium, and possibly strual cycles and thus reduced levels of microenvironment, and affect circulating
of the pancreas and kidney.13 Increased oestrogen. levels of growth factors, insulin, and
circulating leptin levels in obese individ- Obesity is characterised by a low-grade oestrogens.
uals are associated with colorectal67 and chronic inflammatory state, with up to 40 Taller people have undergone more cell
prostate cancers.68 per cent of fat tissue comprising macro- divisions stimulated by IGF-1 and pituitary-
Sex steroid hormones, including oestro- phages. The adipocyte (fat cell) produces derived growth hormone,73 and there is
gens, androgens, and progesterone, are pro-inflammatory factors, and obese indi- therefore more potential for error during
likely to play a role in obesity and cancer. viduals have elevated concentrations of DNA replication, which may result in
Adipose tissue is the main site of oestrogen circulating tumour necrosis factor (TNF)- cancer development.

of n-6 unsaturated fatty acids can have a suppressive effect.83 cannot be counteracted by nutritional supplements or
Various factors have been shown to modulate both inflam- antioxidants.83 Physical activity does not increase pro-inflam-
mation and immunity, including vitamins A and E, copper, matory cytokines, but instead increases anti-inflammatory
selenium, zinc, PUFAs, and epigallocatechin-3-gallate mediators.
(EGCG) from green tea.84 Zinc deficiency can lead to abnor- Chronic consumption of alcohol alters both innate and
malities in adaptive immune responses.85 adaptive immunity. Heavy drinkers are more vulnerable to
Immune status and chronic inflammation may explain pat- infections and, as shown in chapter 4.8, to liver cancer.92 One
terns of cancer in different parts of the world. Cancers caused possible mechanism is alteration in hepatic metabolism
by infectious agents, such as those of the liver and cervix (see resulting in functional iron deficiency that impairs immune
chapter 7.8 and 7.13) are more common in low-income function.93-95
countries, where undernutrition may impair peoples
immune responses. Undernutrition, with deficiencies in spe- Hormones
cific micronutrients such as vitamin A, riboflavin, vitamin Lifetime exposure to oestrogen increased by early menar-
B12, folic acid, vitamin C, iron, selenium, and zinc, sup- che, late menopause, not bearing children, and late (over
presses most immune functions and may fail to control 30) first pregnancy raises the risk of, and may be seen as
chronic inflammation.86 87 By contrast, hormone-related a cause of, breast, ovarian, and endometrial cancers in
cancers such as those of the breast and prostate are more women. The reverse also applies: a reduction in lifetime
common in developed countries. exposure to oestrogen due to late menarche, early meno-
The cytokine IL-6 can act as a pro- or anti-inflammatory pause, bearing children, and early pregnancy may reduce the
cytokine. In cancer, IL-6 can either stimulate proliferation or risk of hormone-related cancers. Age at menarche and
exert anti-tumour effects by enhancing both innate and menopause are influenced by nutrition, with high-energy
adaptive immunity.88 Dietary phytoestrogens, such as soy diets leading to early puberty and late menopause, and low-
isoflavones, downregulate IL-6 gene expression and thus energy diets delaying puberty and advancing menopause.
potentially influence the development of hormone-related The oral combined contraceptive pill (containing both
cancers.89 oestrogen and progesterone) has been estimated to halve the
Circulating levels of IL-6 increase (up to 100-fold) risk of ovarian cancer, if taken for 5 years or more.96 This
following exercise; this reduces chronic inflammation protective effect can last for up to 15 years after women stop
by reducing pro-inflammatory mediators and elevating taking these oral contraceptives.97 Using any type of oral
anti- inflammatory mediators.90 Regular moderate and contraceptive may also have a slight protective effect against
occasional vigorous physical activity has been associated bowel cancer.98 In contrast, combined oral contraceptives
with enhanced immunity,91 but prolonged and intense can cause a slight and transient increase in breast cancer risk,
physical activity can cause immune suppression that but only for the duration of use.99 There may also be an


increase in the risk of cervical cancer, although this is part- Approximately 1 in 4 cancers in low-income countries are
ly related to sexual behaviour.100 estimated to be attributable to infection. In 2002, this rep-
Studies of women taking oestrogen-only hormone replace- resented some 1.9 million cancers or close to 1 in 5 of all
ment therapy have suggested that the risk of endometrial cancers worldwide104. Inadequate nutrition or dietary imbal-
cancer is doubled after 5 years of use,101 and the risk of ovari- ances can lead to immunodeficiencies and increased sus-
an cancer rises by 25 per cent after the same length of use.97 102 ceptibility to infections.
However, use of hormone replacement therapy that Dietary factors may influence host susceptibility to the
combines oestrogen and progesterone does not increase ovar- viral infection or persistence of the infection. For example,
ian cancer risk and may even protect against endometrial high folate intake is thought to reduce the susceptibility to
cancer.102 97 101 and the persistence of human papilloma virus.107

2.4.2 Exogenous causes Radiation Tobacco use Both ionising radiation and UV radiation damage DNA and
Tobacco causes an estimated 20 per cent of all cancer deaths, act as carcinogens. This includes radiation used in X-ray
and an estimated total of 1.2 million in 2002.75 Smokers radiographs and in the treatment of cancer. In 1982, the
have increased risk of a number of different cancers (see various forms of radiation were calculated to account for
Chapter 7). Worldwide, around 80 per cent of lung cancer 3 per cent of all cancer deaths.74
cases in men and 50 per cent in women are caused Ionising radiation can cause DNA damage, both directly
by tobacco smoking.103 104 In 2002, out of all new cases of by causing breaks in the DNA strands, and indirectly by inter-
cancer in low-income countries, over 1 in 5 in men and acting with water molecules and generating reactive oxygen
almost 4 per cent in women were attributable to tobacco. species that damage DNA.
In high-income countries, one third of all new cancer cases Although sunlight causes DNA damage, it also induces
in men and just over 1 in 8 in women were attributed to production of vitamin D. One of its metabolites, 1-25-hydroxy-
tobacco smoking. vitamin D, has antiproliferative and pro-differentiation
Cigarette smoke contains at least 80 known mutagenic effects in some cells mediated through the vitamin D
carcinogens, including arsenic, cadmium, ammonia, receptor.
formaldehyde, and benzopyrene. Each will have a separate Exposure to ionising radiation comes from cosmic radia-
mechanism for causing cancer. For example, following meta- tion (air travel increases exposure), natural radioactivity
bolic activation, the activated derivative of benzopyrene, present in rocks and soil, medical exposure through X-rays,
benzo(a)pyrenediol epoxide, can form DNA adducts in lung or atomic radiation from weapons and nuclear accidents.
epithelial cells.105 Ionising radiation increases the risk of various cancers, in
Cigarette smoke is a powerful carcinogen and also a source particular leukaemias, and breast and thyroid cancers.
of oxidative stress. Compared with non-smokers, active UV light from sunlight or sunlamps is divided into three
smokers have lower circulating concentrations of several bands of wavelengths: UVA, UVB, and UVC. UVB is the most
antioxidant micronutrients including alpha-carotene, beta- effective carcinogen and is absorbed by bases in the DNA,
carotene, cryptoxanthin, and ascorbic acid.106 causing characteristic patterns of DNA damage. UVA dam-
ages DNA through generation of reactive oxygen species. Infectious agents UVC in sunlight is absorbed by ozone in the atmosphere
Infectious agents, including viruses, bacteria, and parasites, and does not reach the surface of the earth. UV radiation
can induce DNA damage and promote cancer development. causes both malignant melanoma and non-melanoma skin
Some infectious agents, including hepatitis viruses, the bac- cancer.
terium Helicobacter pylori (H pylori), and parasites, also pro- It has been difficult to separate the effects of nutrients on
mote cancer by causing chronic inflammation (see DNA damage from those on repair, at least in animal or
Both DNA and RNA viruses can cause cancer, although the human studies. There is some evidence that vitamin C and
mechanisms differ.75 DNA viruses encode viral proteins that carotenoids can protect DNA against oxidative damage in
block tumour suppressor genes, whereas RNA viruses or some experimental settings.108
retroviruses encode oncogenes (box 2.2). Human papilloma
virus is an established cause of cervical cancer, Epstein-Barr Industrial chemicals
virus of nasopharyngeal cancer and lymphoma, and hepati- Certain industrial chemicals and pesticides persist in
tis B and C infection of liver cancer. the environment and become concentrated in the food
As summarised in Chapter 7, H pylori is associated with chain. Some of these are within the scope of this Report and
stomach cancer, liver flukes (from eating raw or undercooked are summarised in Part 2. In 1982, industrial chemicals were
freshwater fish) with liver cancer, and Schistosoma haema- calculated to account for less than 1 per cent of cancer
tobium infection with bladder cancer. deaths.74
In most cases, infection with these agents by itself does not Polychlorinated biphenyls (PCBs), organic compounds pre-
lead to cancer but is a contributory or necessary factor in viously used in plasticisers, adhesives, paints, and various
the cancer process. Multiple factors are thought to be oils, do not readily degrade. They are soluble in fat rather
important in determining why cancer is the result in only than water and thus accumulate in carnivorous fish such as
some cases. salmon, and can be absorbed by people who eat these types

C H A P T E R 2 T H E C A N C E R P R O C E S S

of fish.109 They also accumulate in human milk, and can be 2.5 Nutrition and cancer
passed to the infant during breastfeeding. There is limited
experimental evidence suggesting that PCBs have sex The majority of cancers are not inherited. Cancer is, how-
steroid activity and alter oestrogen levels, which may con- ever, a disease of altered gene expression that originates in
tribute to breast cancer risk.109 changes to DNA, the carrier of genetic information. For a cell
Arsenic is genotoxic, causes gene mutations, and is car- to be transformed from normal to cancerous, it has to acquire
cinogenic to humans; arsenic in drinking water is absorbed different phenotypic characteristics that result from alter-
from the gastrointestinal tract.75 The Panels judgements on ations to the genotype. Most cancers develop to the stage of
arsenic are summarised in chapter 4.7. Arsenic can modify being clinically identifiable only years or decades after the
the urinary excretion of porphyrins in animals and humans. initial DNA damage.
It also interferes with the activities of several enzymes of the Cancer development, or carcinogenesis, requires a series of
haem biosynthetic pathway. There is clear evidence that sele- cellular changes. No single gene causes cancer. It is a multi-
nium binds heavy metals such as arsenic and thus modifies step process caused by accumulated errors in the genes that
their absorption.75 control cellular processes. One genetic mutation may allow
a single trait (such as increased survival) to be acquired by Medication a lineage of cells, and descendants of these cells may then
A number of medical treatments modify the risk of some can- acquire additional genetic mutations. However, cancer only
cers. As indicated in, X-rays are carcinogenic, as is develops when several genes are altered that confer growth
radiation used as cancer treatment. and survival advantages over neighbouring normal cells.
The most notorious example has been diethylstilboestrol, The capacity of a cell to achieve effective cancer preven-
once prescribed in pregnancy and now withdrawn, which tion or repair is dependent on the extracellular microenvi-
caused cancers of the vagina and cervix of female children ronment, including the availability of energy and the
born to mothers who received this drug. Treatments that presence of appropriate macro- and micronutrients. Tumours
affect hormonal status have been studied extensively and are are not simply masses of cancer cells. Rather, they are het-
described in Chemotherapy as cancer treatment dur- erogeneous collections of cancer cells with many other cell
ing childhood is followed by an increased risk of lymphoma types so-called stromal cells; cancer cells communicate
in adulthood.110 with stromal cells within the tumour. The tumour microen-
vironment comprises many cell types including infiltrating Carcinogenic agents in food immune cells such as lymphocytes and macrophages,
Food may be contaminated with natural or man-made endothelial cells, nerve cells, and fibroblasts. All these cell
carcinogenic toxicants. These are within the scope of this types can produce growth factors, inflammatory mediators,
Report, and are assessed and judged in Part 2. and cytokines, which can support malignant transformation
Moulds and the toxins produced by some moulds cause and tumour growth, and attenuate host responses. In addi-
DNA adducts and are carcinogenic. Aflatoxin B, a product of tion, factors produced by the cancer cells themselves mod-
the Aspergillus fungus and a common contaminant of cere- ulate the activity and behaviour of the tumour stroma.
als (grains) and peanuts, is an established cause of liver Initiation is the exposure of a cell or tissue to an agent that
cancer (see chapter 4.1). Fumonisin B, a toxin produced by results in the first genetic mutation. This can be an inherit-
the fungus Fusarium verticillioides, may be found on maize ed mutation or an exogenous or endogenous (produced
and may be carcinogenic, although epidemiological studies through oxidative metabolism) factor. Even without exter-
are lacking. nal oxidative stress, hundreds of sites within DNA are dam-
Some carcinogenic compounds are formed during food aged each day but are normally repaired or eliminated.
preparation (see chapter 4.3 and 4.9). Heterocyclic amines Exposure to the carcinogen initiates DNA damage, usually
are formed by cooking meat at high temperatures, and poly- via the formation of DNA adducts. If left uncorrected, these
cyclic aromatic hydrocarbons can be produced in meat and adducts can be transferred to daughter cells during division
fish that has been grilled (broiled) or barbecued (charbroiled) and confer the potential for neoplastic (new and abnormal)
over a direct flame. Also see box 4.3.4. High environmental growth.
concentrations of polycyclic aromatic hydrocarbons, which Initiation alone is insufficient for cancer to develop. An ini-
also come from pollution caused by traffic and industry, tiated cell must go through a process of clonal expansion
can contaminate other foods such as cereals, vegetables, during promotion to become neoplastic; the larger the num-
and fruits. ber of initiated cells, the greater the risk of progressing to
Some N-nitroso compounds are carcinogens, and are cancer. Promotion involves exposure of the initiated cell to
formed in foods containing added nitrates or nitrites; exam- a promoting agent. This may allow alterations in the rate of
ples include fish and meat preserved with salting or preser- proliferation or additional DNA damage to occur, leading to
vatives, and smoking or drying. These carcinogens can also further mutations within the same cell, which alter gene
be generated from ingested foods containing nitrate or expression and cellular proliferation. Finally, these initiated
nitrite. N-nitroso compounds are also produced endoge- and promoted cells grow and expand to form a tumour mass.
nously in the stomach and colon of people who eat large DNA damage continues at this stage and cancer cells often
amounts of red meat or take nitrite supplements. Also see contain multiple copies of chromosomes. This clear, sequen-
box 4.3.2. tial process is typical of experimentally induced cancers but


may be less clear in sporadic cancers in humans. Growth signal autonomy
At the end of the multistage process of carcinogenesis, the Unlike normal cells, cancer cells are not dependent on exter-
cell will bear some or all of the hallmarks of cancer111 (figure nal growth factors to stimulate their division. Instead, they
2.4). Several genes can contribute to each hallmark and one can generate their own signals or respond to lower concen-
gene (for example p53) can contribute to several of the hall- trations of external signals. This frees cancer cells from the
marks. These hallmarks or traits are shared by most, if not growth constraints of normal cells.
all, cancer cells. The six hallmarks of cancer cells are self-
sufficiency in growth signals; insensitivity to antigrowth Insensitivity to antigrowth signals
signals; limitless replicative potential; evasion of apoptosis; Normal cells also receive growth inhibitory signals. Indeed,
sustained angiogenesis; and tissue evasion and metastasis.111 most cells of the body are quiescent and not actively divid-
Food, nutrition, and physical activity-related factors influ- ing. Cells respond to negative environmental signals such as
ence cellular processes and lead to cells accumulating these contact with other cells. Cancer cells have acquired muta-
traits (figure 2.5). tions that interfere with these pathways and so do not
respond to growth inhibitory signals.
2.5.1 Cell proliferation
Three hallmarks of cancer, namely growth signal autonomy, Limitless replicative potential
evasion of growth inhibitory signals, and unlimited replication, Normal cells can divide a finite number of times. Once they
promote enhanced cell proliferation. Over a normal human have replicated 60 or 70 times they stop a process termed
lifetime, approximately 1016 (10000 000 000 000 000) cell senescence, which is thought to constitute a protective mech-
divisions will take place. The sequence of stages of a cell anism against unlimited proliferation. This preordained num-
dividing into two daughter cells is called the cell cycle (figure ber of cell doublings is controlled by telomeres. Telomeres
2.3). Normal cells require external signals from growth fac- are segments of DNA on the ends of chromosomes, which
tors, which stimulate them to divide. Proliferation of normal are shortened during each round of DNA replication.
cells depends, in part, on the presence in the cellular envi- Eventually when the telomeres are too short, the cell can no
ronment of signals that both promote and inhibit growth, and longer divide and it undergoes apoptosis.
the balance between them. By contrast, cancer cells have acquired the ability to main-
Most cells in adults are not in the process of actively divid- tain the length of their telomeres, which means they can
ing, but are in an inactive or quiescent state termed G. To replicate endlessly. Recent work has suggested that senes-
re-enter the cell cycle, cells must be stimulated with growth cence can be induced prematurely, particularly in premalig-
factors and have sufficient space and nutrients for division. nant cells, by activation of the normal, non-mutated forms
During the G1 phase, the cell increases in size, and syn- of genes such as p53 and Rb.111 This senescence is a normal
thesises RNA and proteins. At the end of the G1 phase, cells active process involving genetic and phenotypic changes that
must pass through the G1 checkpoint, which arrests the cycle may protect against cancer development; for example, it may
if damaged DNA is detected, ensuring that it is not replicat- be one mechanism preventing benign moles from progress-
ed. During the S phase, DNA is replicated. The S phase ends ing to malignant melanoma. However, in malignant
when the DNA content of the nucleus has doubled and the melanoma, cell markers of senescence are lost.112
chromosomes have been replicated. In experimental conditions, many constituents of food
When DNA synthesis is complete, the cell enters the G2 such as retinol, calcium, allyl sulphide, n-3 fatty acids, and
phase, during which the cell continues to increase in size and genistein are known to influence progression of cells
produce new proteins. The G2 checkpoint leads to arrest of through the cell cycle. These studies, when conducted in cells
the cell cycle in response to damaged or unreplicated DNA; in culture, need to be assessed cautiously because they may
otherwise the cell divides into two daughter cells during the not always adequately reflect events in vivo. However, they
M (mitosis) phase, and the M checkpoint ensures each can and do provide evidence additional to that gained from
daughter cell receives the correct DNA. The cell cycle is con- epidemiological studies. Also see chapter 3.1 and 3.2.
trolled by a set of proteins called cyclins and their specific Specific dietary components have effects on cell cycle pro-
cyclin-dependent kinases (CDKs). These join to form gression and proliferation in experimental settings. Some
cyclinCDK complexes, which activate transcription factors. known or hypothesised benefits of some dietary constituents
This in turn activates transcription of the genes required for are summarised here.
the next stage of the cell cycle, including the cyclin genes. Vitamin A (in the form of retinol) can lead to cell cycle
In this section, examples are given to illustrate the known arrest.113 Retinoids and carotenoids inhibit proliferation by
interactions between nutritional factors and these physio- binding retinoid receptors on the cell surface. Reduced
logical processes. expression of retinoid receptors occurs during development
Among the modulators of cell-cycle progression are of lung cancer114; retinoic acid receptor silencing is also com-
specific nutrients, which can either function as energy mon in other malignancies. Retinoic acid, a metabolite of vit-
sources or regulate the production and/or function of pro- amin A, has been used as a chemopreventive and therapeutic
teins needed to advance cells through the replicative cycle. agent in cervical cancer.115 Retinoids can inhibit proliferation
Vitamin A, vitamin B12, folic acid, vitamin D, iron, zinc, and of initiated cells by inducing apoptosis or inducing differen-
glucose all contribute to the control of cell cycle progression tiation of abnormal cells back to normal.116 Retinoids may
(figure 2.3). also cause regression of precancerous lesions in the cervix.117

C H A P T E R 2 T H E C A N C E R P R O C E S S

responsive genes.122
Calcium has a growth
inhibiting action on normal
and tumour gastrointestinal
cells.123 However, certain
dietary compounds can also
stimulate proliferation in
experimental cell lines, for
example, colonic cells can be
induced to hyperproliferate
by dietary haem iron.124
In a variety of animal
studies, certain dietary com-
ponents have shown reduc-
tions in experimentally
induced cancers. Allyl sul-
phides in garlic inhibit
experimentally induced
colon tumour formation.
Although this is not com-
pletely understood, experi-
ments with diallyl disulphide
suggest a block in the G2/M
phase in the progression of
the cell cycle, and induction
of apoptosis.125
Fish oil supplements
decrease the number of
tumours in experimental
models of colorectal can-
Nutrition may influence the regulation of the normal cell cycle, which ensures correct DNA cer.126 Long-chain n-3 PUFAs
replication. G0 represents resting phase, G1 the growth and preparation of the chromosomes for in fish oils can limit tumour
replication, S phase the synthesis of DNA, G2 the preparation of the cell for division, and M cell proliferation30 31 by
represents mitosis. modifying signalling path-
ways,127 128 129 for example,
by decreasing signalling of
Butyrate and diallyl disulphide can act as histone deacety- activated oncogenes.130 Animals that receive a diet supple-
lase inhibitors16 (see 2.2.2), and arrest the cell cycle. Folate mented with n-3 fatty acids have fewer colonic tumours than
is a necessary cofactor for DNA synthesis, and deficiency can those fed a diet supplemented with corn oil,131 due to dietary
reduce cell proliferation due to decreased DNA synthesis. fibre-altering, fatty acid-binding, protein expression in
Phenolic compounds, including genistein and EGCG, can colonocytes during tumour development.
inhibit some cyclins and cyclin-dependent kinases.118 Various growth factors and hormones involved in normal
Specifically, in people with oral leukoplakia, green tea (which cell processes can be used or produced by cancer cells to
contains EGCG) has been associated with significant decreas- maintain or augment uncontrolled cell proliferation. The
es in the size of cancers and of micronuclei formation in exfo- receptor for IGF-1 is overexpressed on many cancer cells.
liated oral cells.119 IGF-1 can enhance the growth of a variety of cancer cell
Phytoestrogens are found in high concentrations in soya lines132 by stimulating progression of the cell cycle from G1
beans, and have been shown in vitro to exhibit a plethora of to S phase.13
different anti-cancer effects, including inhibiting prolifera- Insulin itself can also act as a growth factor for tumour cell
tion.120 121 Glucosinolates from cruciferous vegetables are con- proliferation, both by binding to the insulin receptor on can-
verted in the liver to ITCs, which can arrest cell cycle cer cells and by stimulating increased IGF-1 production by the
progression, as well as induce phase 2 enzymes, which can liver.13 133 Insulin resistance increases with body fatness, in
promote carcinogen excretion (see 2.3). Only about one third particular abdominal fatness, and the pancreas compensates
of people have the types of microbial flora in their gut that by increasing insulin production. This hyperinsulinaemia is
are capable of metabolising the dietary isoflavone daidzein associated with a risk of cancers of the colon and endometri-
to equol. Compared with Western populations, Asian popu- um, and possibly of the pancreas and kidney.13 Leptin, a hor-
lations are more likely to produce equol, and this affects the mone produced by fat cells, can also stimulate proliferation
expression of genes involved in cell signalling and differen- of many premalignant and malignant cell types,134 as can a
tiation, and cell division. Equol can also modulate oestrogen- number of sex steroid hormones (see Chapter 6, and box 2.4).


age, disruption of the cell

cycle, hypoxia, reactive oxy-
gen species, and physical or
chemical insult. Two non-
exclusive pathways, the
intrinsic (mitochondrial)
pathway or the extrinsic
(death-receptor) pathway,
can be activated. Both
involve activation of caspas-
es, a family of protease
enzymes that cleave intra-
cellular proteins.143 In apop-
tosis, p53 functions as a
transcriptional activator of
genes encoding apoptosis
effectors. p53 can also exert
a direct apoptotic effect by
damaging mitochondria.144
Cancer cells have acquired
mutations in genes regulat-
ing apoptosis and therefore
can evade apoptotic signals.
Cancer cells have different characteristics from normal cells. The six hallmarks shown here are the Defects in apoptosis are
phenotypic changes that need to be accumulated over time as a result of genetic changes (mutations often observed in established
and epigenetic factors) in order for a cell to become cancerous. cancers. In cancer cells,
many signals that normally
induce apoptosis, such as
Physical activity improves insulin sensitivity and decreas- damaged DNA or expression of activated oncogenes, are pre-
es levels of insulin.135 However, exercise has little or no long- sent but apoptosis is not induced. This avoidance of apop-
term effects on circulating IGF-1 levels.136-138 IGF binding tosis allows further opportunity for additional mutations to
activity may increase with physical activity, and thus over- develop. In cancer cells with mutations in p53 or other
all IGF-1 bioavailability and activity may decrease. Physical members of this family, apoptosis may not occur.
activity decreases serum oestrogen and androgens in post- Additionally, mutations in genes that would normally
menopausal women. In premenopausal women, it decreas- activate p53 or regulate its activity, or in genes that should
es circulating oestrogens, increases cycle length, and be switched on as a result of p53 activation, can have the
decreases ovulation, all of which provide a protective effect same effect. Cancer cells with upregulated expression of
for breast and endometrial cancers. Also see Chapter 5. IGF-1R and increased responses to IGF-1 have decreased
In experimental animals, energy restriction leads to a apoptosis.133
reduction in cell proliferation66 (box 2.5). At the molecular In experimental settings, energy restriction creates a pro-
level, dietary energy restriction affects levels of cell cycle con- apoptotic environment, adjacent to premalignant and
trol proteins (decreased cyclins, increased levels of cyclin- malignant breast pathologies.145 Long-chain n-3 PUFAs in
dependent kinase inhibitors, and decreased cyclin-dependent fish oils limit tumour cell proliferation, increasing apoptot-
kinases), leading to reduced Rb phosphorylation and inhib- ic potential along the crypt axis, promoting differentiation
ited cell cycle progression.139 This, in turn, may directly and limiting angiogenesis.30 31 126
inhibit tumour growth and/or indirectly reduce cancer devel- Reactive oxygen species can induce apoptosis, but it is
opment by reducing the number of cell divisions, thus reduc- also possible that scavenging of reactive oxygen species by
ing the chances for incorrect DNA replication or preventing dietary antioxidants can delay or inhibit apoptosis, and thus
damaged DNA from being replicated. favour survival of premalignant cells. Indeed, this could
explain why dietary antioxidant intervention trials have pro-
2.5.2 Evasion of apoptosis duced mixed results.146 147
Apoptosis is the tightly regulated process of cell death that Many dietary components have been shown to induce
controls cell numbers, removes damaged cells, and prevents apoptosis in cultured cancer cells and in experimental mod-
damaged cells being replicated, thereby maintaining
tissue integrity and protecting against cancer. Ultimately,
Maintenance of healthy cells depends on regulated processes,
cells break into small membrane-surrounded fragments which can be influenced by factors related to food, nutrition, and
(apoptotic bodies) that are phagocytosed without inducing physical activity, either to protect the cell from or to promote
inflammation. cancer. The evidence for what is shown here comes from a variety
Triggers for apoptosis in normal cells include DNA dam- of experimental studies.

C H A P T E R 2 T H E C A N C E R P R O C E S S


Box 2.5 Energy restriction

Restriction of energy intake from food is species, and therefore less exposure of intake are associated with increased DNA
the most effective single intervention for DNA to damaging oxygen radicals. synthesis and levels of some cyclins and
preventing cancer in experimental animals. Dietary energy restriction reduces levels CDKs.113 Energy restriction may also create
It increases the lifespan of rodents, and of circulating IGF-166 140 and insulin, which a pro-apoptotic environment and reduce
suppresses tumour development in mice. In are growth factors for many cells, includ- blood vessel density, as shown in pre-
addition, energy restriction can suppress ing breast cancer.141 IGF-1 stimulates pro- malignant and malignant breast path-
the pro-cancer effects of many carcinogens gression through the cell cycle from G1 to ologies.139 It may also activate other
in experimental animal models.66 S phase, and high levels of insulin increase protective pathways, such as the activation
Energy restriction leads to a reduction in production of IGF-1.13 Energy restriction of protein deacetylases.142
cell proliferation.66 This may directly inhib- also decreases expression of cyclins and The data on energy restriction must be
it tumour growth, and also indirectly cyclin-dependent kinases (CDKs), and interpreted with caution, as all studies
reduce cancer development by reducing increases levels of CDK inhibitors, leading have been performed in experimental ani-
overall proliferation, thus reducing the to reduced Rb phosphorylation and inhib- mals and there is an absence of epidemio-
chances for incorrect DNA replication, or ited cell cycle progression.139 Energy restric- logical and mechanistic data in humans.
by preventing damaged DNA from being tion also decreases other inflammatory Therefore the relevance of these findings
replicated. Reduced metabolism results in markers.66 Conversely, increased glucose in experimental animals to the human
reduced generation of reactive oxygen levels associated with increased energy condition is not yet clear.

els of cancer.148 These include EGCG, curcumin, genistein, inhibitor in healthy people, which could decrease plasma
indole-3-carbinol, resveratrol, ITCs, lycopene, capsaicin, levels of VEGF.157
and organosulphur compounds.149 In premalignant cells,
retinoids, polyphenols, and vanilloids stimulate apoptosis.150 2.5.4 Tissue invasion and metastasis
Alpha-tocopherol (a form of vitamin E) has been shown both Normal cells in solid tissues maintain their position in the
to induce151 and to protect against apoptosis.152 body and generally do not migrate. As a cancer increases, it
eventually reaches the membrane encapsulating the organ.
2.5.3 Sustained angiogenesis Tumour cells secrete enzymes such as matrix metallopro-
Angiogenesis, the formation of new blood vessels, is essen- teases (MMPs), which digest the membrane and allow the
tial for the supply of nutrients and oxygen to any growing cancer to invade adjacent tissue. Once through the mem-
tissue, including tumours. Most cells within tissues reside brane, cancer cells can access other sites via the blood and
within 100 mm of a capillary blood vessel. The generation lymphatic systems. This migration of cancer cells, or metas-
of blood vessels in adults is fairly constant and tightly tasis, is a common characteristic of most cancer deaths.
controlled by a balance of angiogenesis inducers and There is limited evidence for dietary components to influ-
inhibitors. For a cancer to progress to a larger size, it must ence these late stages of cancer, although in vitro, EGCG,
acquire the ability to induce angiogenesis. Currently about resveratrol, quercetin, curcumin, and genistein can inhibit
35 proteins have been identified as angiogenesis activators one or more MMPs. Vitamin C can inhibit MMP production
or inhibitors.153 by a number of human cancer cell lines and prevent inva-
In experimental settings, one of the first dietary compo- sion of these lines in vitro.158 Vitamin E can inhibit metas-
nents for which a beneficial anti-angiogenic effect was tasis of pre-established tumours in mouse models of breast
clearly demonstrated was EGCG from green tea.154 Now cancer.159
some 20 different compounds consisting mainly of flavonoids
and isoflavones (including genistein) are documented as
being able to modulate the angiogenic process. Diets high 2.6 Conclusions
in n-6 fatty acids are associated with poor prognosis in breast
cancer patients, whereas those high in n-3 fatty acids appear Great progress has been made since the mid-1990s in the
to suppress angiogenesis.155 Long-chain n-3 PUFAs in fish understanding of the cancer process. Evidence is accum-
oils limit angiogenesis in other experimental cancers.30 31 ulating that shows or suggests that food and nutrition, and
Curcumin, quercetin, and resveratrol have all been shown to physical activity and associated factors, are important in
inhibit the angiogenic factor, vascular endothelial growth modification of the cancer process. Moreover, there is
factor (VEGF), in cultured cancer cells. Garlic extract may increased evidence that specific dietary patterns, foods and
inhibit experimentally induced angiogenesis, as it can sup- drinks, and dietary constituents can and do protect against
press endothelial cell motility, proliferation, and tube for- cancer, not only before the process starts, but also after-
mation.156 Phytoestrogens found in high concentrations in wards.
soya beans have also been shown to inhibit angiogenesis.120 Understanding the mechanisms that control cell structure
Energy restriction reduces blood vessel density in pre- and function, and so influence the cancer process, will aid
malignant and malignant breast pathologies.145 Exercise not only understanding of cancer as a whole, but also the
increases the levels of a circulating endogenous VEGF- development of preventive strategies.


C H A P T E R 3

Judging the evidence

The task of expert committees responsible for Judgements of the Panel are shown in the form of
reports such as this is to collect, discuss, and judge matrices at the beginning of the chapters of Part 2 of
scientific evidence, as a basis for recommendations this Report. Two key judgements in these matrices,
made in the public interest. The purpose of this third and in the text, are those of convincing and
introductory chapter is to summarise the process the probable. These denote the Panels judgements that
Panel has used in the five years of its work, in order the evidence of causality that a factor either
to ensure that its assessments, judgements, and decreases or increases the risk of cancer is strong
recommendations made in the chapters that follow enough to justify population goals and personal
are reliable. recommendations, which are made in Part 3. The
As shown in the previous chapters, while cancer is criteria agreed by the Panel for grading the evidence
a disease of cells and tissues, the main determinants convincing, probable, or limited, or else showing
of many cancers are environmental, which means that any substantial effect on the risk of cancer is
that most cancers are, at least in theory, unlikely, are also specified in this chapter.
preventable. Environmental factors that modify the
risk of cancer include food and nutrition, physical
activity, and body composition. One purpose of
scientific research in this field is to determine which
aspects of these factors protect against, and which
are causes of, cancers of various sites. Such research
is also concerned with which aspects are most
important that is, which have the most powerful
or general effects.
Some of the methods used by the Panel
responsible for this Report are new. Others are
developed from those used elsewhere, including in
the previous report. The best evidence that aspects
of food, nutrition, physical activity, and body fatness
can modify the risk of cancer does not come from
any one type of scientific investigation. It comes
from a combination of different types of
epidemiological and other studies, supported by
evidence of plausible biological mechanisms. Such
comprehensive evidence has been collected in the
form of 20 systematic literature reviews specially
commissioned as the basis for this Report, compiled
by nine independent centres of scientific excellence,
covering 20 cancer sites, the determinants of
obesity, and recommendations made by other
authoritative reports. These reviews amount to a
comprehensive examination of the relevant types of
epidemiological and experimental evidence,
organised using a common methodology. Their
findings, as summarised for and then assessed and
judged by the Panel, are shown in Part 2 of this
Report. The full systematic literature reviews are
contained on the CD included with this Report.

C H A P T E R 3 J U D G I N G T H E E V I D E N C E

Since the mid-1990s, the discipline of epidemiology has the introductory passages of the sections of Chapter 7, and
placed increasing emphasis on the use of meta-analyses and box 7.1.1.
systematic reviews, both of which have informed the Panel The International Agency for Research on Cancer (IARC),
responsible for this Report. Taken together with lines of evi- a branch of the World Health Organization, compiles inter-
dence from other types of study, these provide a reliable basis national cancer statistics using data from national and
for judgements and recommendations designed to improve regional cancer registries around the world.1 Cancer inci-
public health. dence rates are usually specified by gender and age. Cancer
The task of the Panel, with the supporting Secretariat, has mortality rates, generally derived from data collected rou-
been to commission, summarise, and display a comprehensive tinely on causes of death, are more widely available than
range of evidence; to assess and judge this evidence; and to cancer incidence rates.
draw conclusions and make recommendations based on this Descriptive epidemiology informs cancer surveillance
systematic and transparent process. Also see Appendix A. programmes, and is a basic tool for determining patterns
This chapter details the nature of the science relevant to of cancer, relative rates of cancer and other diseases, and
the work of the Panel. It also summarises the processes devel- changes in patterns and trends over time. Remarkable
oped from initial work done by a Methodology Task Force. changes in the incidence of cancers (for example, the
Its findings have been a foundation for the work of the Panel. general drop in rates of stomach cancer or the increase
The two most important parts of this process are the sys- in rates of lung cancer in many middle- and low-income
tematic literature reviews (SLRs), explained in box 3.4, and countries) provide first lines of evidence pointing to causa-
the criteria the Panel has agreed for grading the evidence, tion due to corresponding changes in environmental
described in box 3.8. These and other methods have deter- circumstances.
mined the Panels approach to gathering, summarising, Like all types of study, descriptive epidemiology has limi-
assessing, and judging evidence, and agreeing on population tations. Apparent trends in cancer incidence and mortality
goals and personal recommendations as a basis for the pre- may be due in part to changes and developments in screen-
vention of cancer. ing, diagnosis, or treatment. For example, the rapid rise in
the recorded incidence of prostate cancer in the USA, the UK,
and other higher-income countries is largely due to wide-
3.1 Epidemiological evidence spread use of diagnostic techniques that identify early
evidence of this cancer.
Epidemiological research describes and seeks to explain the
distribution of health and disease within human populations. 3.1.2 Ecological studies
The methods used are based mainly on comparative obser- Ecological studies are designed to explore relationships
vations made at the level of whole populations, special between environmental factors and disease amongst popu-
groups (such as migrants), or individuals within populations. lations rather than individuals.
This type of investigation is known as observational. By relat- Within the scope of this Report, ecological studies compare
ing differences in circumstances and behaviour to differences relationships between estimated levels of consumption of
in the incidence of disease, associations are identified that foods and drinks, levels of physical activity, and degrees of
may be causal. In epidemiological studies, an exposure is a body fatness with rates of cancer for populations. For exam-
factor or condition that may increase or decrease the risk of ple, as already mentioned, early observations found impres-
disease. In this Report, food, nutrition, physical activity, and sive correlations between national per capita intake of total
body composition are the exposures investigated. The meth- dietary fat and rates of breast cancer mortality, mapped
ods summarised here and applied to cancer are also used to across many countries,2 leading to the hypothesis that rela-
study and understand other diseases. tively high consumption of total fat was an important cause
of breast cancer.
3.1.1 Descriptive studies The findings of ecological studies, together with those
The most basic information about cancer comes from sta- from migrant and laboratory studies (see 3.1.3 and 3.2),
tistics on cancer incidence and mortality. See chapter 1.2.4, were important factors leading to judgements and recom-


Box 3.1 Issues concerning interpretation of the evidence

Interpretation of epidemiological evidence middle- and low-income countries, but also Confounding. A confounder is a factor
on any and all aspects of foods and drinks, globally. associated with both the outcome (in this
physical activity, body composition, and case, cancer) and the exposure being stud-
associated factors, with the risk of cancer, Measurement. Many study exposures are ied, but is not an effect of the exposure. It
is never simple. General considerations difficult to determine and are thus mea- is never possible from observational studies
include the following, which need to be sured imprecisely. It is easier to measure to eliminate completely the possibility that
taken into account when evidence is assem- food intakes than intakes of dietary con- an evident effect of a constituent, or aspect
bled and assessed. This emphasises that stituents of foods. This can lead to an of a food or drink, is at least in part caused
expert judgement is essential. undue degree of importance being given to by another factor.
studies of aspects of food and nutrition
Patterns and ranges of intakes. Most stud- that happen to be more easily measured. Reporting bias. Studies reliant on self-
ies are carried out in high-income countries. reporting of dietary intake are prone to sys-
Their findings may have limited application Terminology. For some foods and drinks, tematic bias. People tend to over-report
in countries where dietary and physical and dietary constituents, there are no gen- consumption of foods and drinks they
activity patterns are different. They may erally agreed definitions. Examples include believe to be healthy, and under-report
also be unrevealing in their own countries dietary fibre and processed meat. Also, foods and drinks they believe to be
if the ranges examined are relatively nar- some common definitions may disguise real unhealthy. Under-reporting of energy
row. Some foods that are important dietary differences: different types of dietary fibre intake has been shown to be associated
constituents outside high-income countries have different biological effects. with factors such as age, overweight and
are often not examined. obesity, perceived body size, and other per-
Study design. The relative merits of differ- sonal characteristics.7-14 Allowance for this
Classification. Following from the above, ent types of epidemiological study design, is an inexact science. Also see 3.3.
studies usually classify foods and drinks, and the relative value of epidemiological
and physical activity in ways that corre- evidence compared with experimental evi- Production, preservation, processing,
spond to the patterns of high-income coun- dence, are likely to remain to some extent preparation. Studies of foods and drinks,
tries. Their findings may over-emphasise a matter of opinion. The special power of and of food groups, may neglect the
the significance (or insignificance) of foods randomised controlled trials (see 3.1.6), effects of methods of production, preser-
and drinks commonly consumed in high- most often used to test the effects of vation, processing, and preparation (includ-
income countries, and they may overlook dietary constituents as opposed to whole ing cooking). They are also inclined to
other foods and drinks consumed in diets, could lead to over-emphasis of the underestimate the significance of foods
other parts of the world. The same points importance of isolated constituents which, and drinks combined in dishes or meals,
apply to types of physical activity. This within the context of food and diets, may and as components of whole dietary
may impede understanding, not only in have other effects. patterns.

mendations made in reports on diet and cancer published in 3.1.3 Migrant studies
the 1980s.3-5 Migrant studies compare cancer rates for migrants, and for
While ecological studies, like other observational studies, their offspring, in their current country of residence, with
may suggest a relationship between a specific environmen- rates in their country of origin.16 These studies show that
tal factor (such as an aspect of food and nutrition) and populations migrating between areas with different cancer
disease, the actual causal relationship may be with a differ- incidence rates acquire the rates characteristic of their new
ent confounding factor, which may or may not be associat- location for some cancers, often after only one or two gen-
ed with the environmental factor being investigated.6 The erations. This shows that environmental, rather than inher-
example of total fat consumption and breast cancer is a case ited, factors are primarily responsible for the large differences
in point: total fat consumption, disposable income, and con- in cancer rates in different regions and countries (see chap-
sumption of alcoholic drinks might all correlate with one ter 1.3).17 Those diseases for which incidence shifts with
another, and also with breast cancer. See box 3.1. migration, such as cancer, are diseases with evidently impor-
Ecological studies are often used to identify associations tant environmental causes.
or trends that warrant further investigation. They have spe-
cial strengths, particularly when conducted between popu- 3.1.4 Case-control studies
lations, either internationally, or cross-culturally among In case-control studies, individuals diagnosed with a specif-
different populations within a country. Thus, the contrast in ic type of cancer (cases) are compared with otherwise sim-
dietary intake between countries is often much larger than ilar individuals who have not been diagnosed with cancer
the contrast within countries. In addition, average national (controls). The control group is a sample of the population
diets are likely to be more stable over time than the diets of from which the cases arose, and provides an estimate of how
communities, families, or individual people. For most coun- the exposures being studied are distributed in that popula-
tries, the changes in overall national dietary intakes over a tion. Identifying and enrolling appropriate controls is a major
decade or two are relatively small. challenge in case-control studies.18-20

C H A P T E R 3 J U D G I N G T H E E V I D E N C E

Case-control studies are subject to recall bias, which can called intervention and control groups, to receive or not
occur when participants reporting of various exposures receive an experimental intervention. The main use of RCTs
(dietary intake, medication, physical activity, and so on) is has generally been to test the efficacy of drugs and other
differentially affected by whether they are cases or controls medical treatments.
in the study. Selection bias is an increasing problem in high- In a double blind RCT, neither the participants nor the
income countries, where participation rates among both case investigators know to which group (intervention or control)
and control groups may be substantially less than 100 per the participant has been assigned. Blinding is used because
cent, and where participation may be related (in different the knowledge of group assignment might influence study
ways) to various exposures. However, case-control studies results, but it is usually impossible to achieve with trials
are usually less expensive than cohort studies, and can be involving physical activity, or those investigating foods and
completed over shorter periods of time. drinks in their usual form.
A nested case-control study is carried out within an exist- An effective use of RCTs is to test the effects of supple-
ing cohort study (see 3.1.5). In this type of study, all of the mentation with specified doses of dietary micronutrients
cases in the cohort are compared with a sample of the non- (as pills or by other means). However, pharmacological doses
cases. A nested case-control study has the strengths of a of supplements are often studied doses much higher
cohort study notably that diet is assessed among study than can be derived from diets and results may not be
participants prior to the diagnosis of cancer, so avoiding directly relevant to dietary intakes of that micronutrient.
recall bias but is less expensive to conduct, since only a Such trials may yield powerful evidence of the effect of a
sample of the non-cases are included in the analysis. specific dietary constituent. However, they are often con-
ducted as a result of promising epidemiological studies
3.1.5 Cohort studies that have shown protective effects of a particular group of
In prospective cohort studies (usually simply called cohort foods, and there is always a possibility that the actual active
studies), the diets, body compositions, and/or physical activ- agent or combination of agents in the foods has not been
ity levels of a large group (cohort) of people who are used in the trial. Dietary constituents that are or may be
assumed to be healthy are assessed, and the group is fol- protective when contained within foods may have unex-
lowed over a period of time. During the follow-up period, pected effects in isolation, especially at doses higher than
some members of the cohort will develop and be diagnosed those found in normal diets. For example, in the Alpha-
with cancer, while others will not, and comparisons are Tocopherol, Beta-Carotene Cancer Prevention Trial (ATBC
then made between these two groups. Because measure- Trial) of male smokers in Finland, high dose beta-carotene
ments are made before any cancer diagnosis, cohort studies supplementation was associated with increased incidence of
are not subject to recall bias. A single cohort study allows lung cancer.26
examination of the effects of diet and physical activity on RCTs are also used to test interventions designed to change
multiple types of cancer and other diseases. Also, in cohort behaviour, including dietary intakes and physical activity.
studies, blood and tissue samples are often collected and Such trials require a high level of commitment by partici-
stored for future analysis. Finally, cohort studies provide the pants, and learning how to conduct them well is a topic of
opportunity to obtain repeated assessments of participants active investigation.
diets at regular intervals, which may improve the dietary A unique and important strength of sufficiently large RCTs
assessment. is that confounding variables, both known and unknown,
Cohort studies may need to be very large (up to tens or will on average be distributed equally between the treatment
even hundreds of thousands of participants) to have and control groups, and will therefore not bias the study
sufficient statistical power to identify factors that may results.
increase cancer risk by as little as 20 or 30 per cent. Also,
meaningful comparisons between cases and non-cases can 3.1.7 Meta-analysis
be made only for factors that vary sufficiently within the Meta-analysis is a method used to combine the results of sev-
cohort. eral studies addressing similar questions. Unless an epi-
Cohort studies are expensive, so they have been conduct- demiological study is sufficiently large, modest but
ed mostly in high-income countries. The European potentially important associations can be missed, simply
Prospective Investigation into Cancer and Nutrition (EPIC), because of the inadequate statistical power of the individual
started in 1992, is a cohort of more than 520 000 men and study. Meta-analysis is used to provide summaries of select-
women in 10 European countries.21 22 In the US, large cohorts ed collections of studies.
include the Nurses Health Study, established in 1976, and the Study-level meta-analysis provides single estimates of
Nurses Health Study II, established in 1989, each with a effect using information from multiple studies of the same
cohort of more than 100 000 women.23-25 Increasing numbers design. These summary estimates can provide evidence
of cohort studies are now being conducted in middle- and regarding the presence or absence of an association, as well
low-income countries. as examining possible dose-response relationships (box 3.2).
Meta-analysis, often displayed graphically on a forest plot
3.1.6 Randomised controlled trials (box 3.3), can also identify heterogeneity between studies.
A randomised controlled trial (RCT) is an experiment in This heterogeneity can be quantified using a measure called
which participants are randomly assigned to groups, often I2, which ranges from 0 to 100 per cent, and indicates the


Box 3.2 Dose response

Dose response is a term derived from low and high levels of exposure, but the categories. Although this is done for sta-
pharmacology, where it denotes a change other way at intermediate levels of expo- tistical reasons and can make effects easi-
in the effect of a drug according to the sure, shown as J- or U-shaped curves. er to detect, the number and location of
dose used. This concept can be applied to In such cases, the exposure is evidently category boundaries may obscure the true
any exposure, including food, nutrition, beneficial or harmful only within certain relationship between exposure and the
and physical activity. For example, different ranges. outcome, and non-linear effects of expo-
amounts of food and drink consumed, or Throughout Chapters 46, this Report sure may be missed if insufficient cate-
of physical activity taken, may lead to a dif- uses two forms of dose-response graph as gories are used.
ferent likelihood of any particular out- a means of displaying graded responses. Evidence of dose response is important
come, such as cancer. Such a graded These show the direction and shape of the when framing recommendations. For
response, or biological gradient, may show association, and allow estimates to be example, if the evidence for cancer showed
that higher exposure leads to increased made of levels of exposure that may influ- no threshold of effect for alcoholic drinks,
risk, or to reduced risk, and vice versa. ence risk. In order to combine and quanti- such that the risk of cancer increased from
Dose responses take different forms. The fy study results, the dose-response curves having any amount of alcoholic drink,
effect may be linear, shown in graphic form are also presented with the exposure vari- however modest, then a recommendation
as a straight line. There may be a thresh- able displayed per standard unit of based on the evidence for cancer would be
old below which there is no effect, but increase. The demonstration of a biologi- to avoid alcoholic drinks. However, if there
above which there is an effect. This is cal gradient adds weight to evidence that is clear evidence of no effect below a cer-
shown as a horizontal line that inclines an exposure may be causal. Diet and phys- tain level of consumption, then the rec-
once the threshold is reached. Or the effect ical activity exposures are continuous vari- ommendation would differ accordingly.
may be to influence risk one way at both ables, but are often reported in discrete Also see chapter 4.8.

percentage of total variation across studies that is not due 3.2 Experimental evidence
to chance. In general, an I2 of 25 per cent or less indicates
low heterogeneity; around 50 per cent indicates moderate Epidemiological studies all have strengths and limitations. So
heterogeneity; and 75 per cent or more indicates high do laboratory and mechanistic studies; their main strength is
heterogeneity.27 control. The environment of these research studies is defined
For this Report, RCTs and ecological, cohort, and case-con- by chosen experimental conditions: precise manipulations can
trol studies have been the subjects of systematic review, and, be made and relatively exact measures taken. Occasionally
when possible (and separately) of study-level meta-analysis. the test participant is a human volunteer, but usually these
Studies were included in a meta-analysis only when suffi- studies are conducted in animals (in vivo) or using human
cient data were included in the publication. In addition to or animal cells grown in the laboratory (in vitro).
effect measures and their standard errors (or confidence Rodents (usually rats or mice) are the most commonly used
intervals, see box 3.3), key elements of adequate reporting animals in laboratory experiments. Their relatively short life-
included the number of people with and without disease span provides comparatively fast results in cancer studies, and
for each exposure category, and boundaries of exposure they offer a whole body system suited to a wide variety of
categories. tests. Rodent studies can show how nutrients and other com-
The SLRs on which the conclusions of this Report pounds might affect the cancer process. But it is known that
are based include original study-level meta-analyses some interventions that affect rodents do not affect humans,
undertaken by the independent centres of scientific excel- or do not affect them in the same ways or to the same degrees,
lence. The Panel considered all studies identified in the SLR, and vice versa. Also, experiments on animals may be highly
not just the results of the meta-analyses. Full details of artificial, using special breeds of rodents initially given mas-
the methods used for the meta-analyses are contained in sive doses of carcinogenic agents, and then fed nutrients or
the SLR specification manual. The full SLRs and the manu- other substances at levels far higher than humans would nor-
al are contained on the CD included with this Report. Also mally consume, or could ethically be given.
see box 3.4. Human or animal cells, sometimes derived from particu-
Pooled analysis is a type of meta-analysis where original lar cancers, can be grown in vitro in the laboratory and used
individual-level data from various published epidemiologi- in experiments to help researchers understand mechanisms
cal studies of the same type usually prospective cohort that may lead to the development of cancer. The Panels deci-
studies are combined and reanalysed. The combination sion on what types of experimental studies were admissible
of data from multiple studies creates a larger data set and as evidence for this Report is summarised in box 3.5.
increased statistical power. Published studies from pooling
projects, in addition to the SLR study-level results generat- 3.2.1 Human feeding studies
ed specifically for this Report, were taken into account by the Human volunteers can be studied in a controlled environ-
Panel in making its assessments and judgements. ment, such as within a metabolic unit, where their diets and

C H A P T E R 3 J U D G I N G T H E E V I D E N C E

Box 3.3 Forest plots

The graphic device known as a forest

plot is the usual method of presenting
Figure 3.1 Alcoholic drinks and oesophageal cancer;
cohort and case-control studies
There is only one cohort study shown
on the forest plot, and it has a wide CI,
the results of meta-analysis of a num- but the estimate is statistically signifi-
Relative risk (95% CI)
ber of studies. In the forest plot below, cant and consistent with results from
studies are presented that examine the Cohort the case-control studies.
Kjaerheim 1998 1.26 (1.101.44)
relationship between alcoholic drinks The size of each square on the plot
and oesophageal cancer. This plot is Case control represents each studys calculated
also shown as figure 4.8.5. Tuyns 1983 Men 1.12 (1.071.17)
weight (influence) on the combined
Tuyns 1983 Women 1.04 (1.001.09)
This plot shows 1 cohort study and Decarli 1987 1.04 (1.031.05) (summary) estimate (the diamond). The
23 estimates from 20 case-control stud- La Vecchia 1989 1.01 (0.991.03) size of the square is calculated taking
Franceschi 1990 1.02 (1.011.04)
ies. The horizontal axis of the plot De Stefani 1990 Men 1.04 (1.031.06)
a number of factors into account, such
shows the relative risk (RR) and is De Stefani 1990 Women 1.13 (0.991.29) as the number of people in the study,
bisected by the vertical axis, which rep- Sankaranarayanan 1991 Men 1.02 (1.011.03)
and the event rate (here, the rate of
Cheng 1992 1.08 (1.071.09)
resents no difference in effect on risk Tavani 1993 1.03 (1.001.06)
oesophageal cancer occurrence). The
between the exposure categories that Tavani 1994 1.02 (1.001.04) diamond summarises the meta-analysis.
are compared (the RR is 1.00). Also Hanaoka 1994 1.24 (1.141.33)
The width of the diamond represents
Brown 1994 Black men 1.04 (1.031.04)
see 3.4.3. Brown 1994 White men 1.03 (1.051.07) the 95 per cent CI of the overall
The squares represent the results of Castelletto 1994 1.05 (1.031.07)
estimate. Unless indicated otherwise,
Gammon 1997 1.07 (1.051.08)
each individual study. Each square is Bosetti 2000 1.03 (1.031.03)
random effects models, which do not
centred on the point estimate of the Takezaki 2001 1.00 (0.961.04) assume that the links between exposure
RR for that study. The point estimate is Sharp 2001 1.00 (0.951.05)
and outcome are the same in different
Boonyaphiphat 2002 1.05 (1.041.07)
the extent to which any exposure (in Dal Maso 2002 1.03 (1.021.04) studies, were used to generate the
this case, alcoholic drinks) is associated Lee 2005 0.99 (0.971.00) forest plots presented in this Report.
Yang 2005 1.05 (1.031.06)
with the risk of cancer (in this case, of Summary estimate 1.04 (1.031.05)
The Panels judgement for this particu-
the oesophagus). The line running lar example is given in chapter 4.8.5.
through the squares represents the 95 The forest plots presented in this Report
0.8 1 1.25 1.5
per cent confidence interval (CI) of the Relative risk, per drink/week do not contain all of the studies identi-
estimate. Where no line is apparent, fied in the SLRs. Sometimes, more stud-
the CI falls within the square. The CI is ies could be included in a comparison
an indication of how much random between those at the highest levels
error underlies the point estimate; it compared with the lowest levels of
does not take into account confound- of meta-analysis is demonstrated: of the exposure in different studies. This can
ing and other forms of systematic bias.15 A 20 case-control studies, 6 are non-signifi- give an indication as to whether or not
confidence level of 95 per cent indicates a cant or only marginally so, of which 1 sug- there is an association between exposure
95 per cent probability that the true pop- gests a protective effect (that is, it has an and outcome. However, because the actu-
ulation value falls within the CI.28 RR of less than 1.00). But taken together, al levels of exposure vary between studies,
When the CI does not cross the vertical as shown by the summary diamond, an this cannot give a quantified summary
axis representing no difference, the esti- overall significant effect, consistent with a estimate of effect. The Panel discussed all
mate is considered statistically significant. judgement that alcoholic drinks are a studies identified, not just those included
Looking at the example above, the value cause of oesophageal cancer, is shown. in a meta-analysis.

activity levels can be highly regulated and measured. In some questions. Often the animals have tumours produced by irra-
studies, subjects live at the metabolic unit for short periods diation, viruses, chemicals, or other carcinogens, or they may
of time, eating only foods and meals provided as part of the be genetically prone to develop cancer. The effect of dietary
study. Since the nutrient composition of such diets can be or other interventions on the prevention or progression of
controlled and manipulated, investigators can study the such tumours is then investigated.
effects of various changes in nutrient intakes on factors such As indicated in 3.2, the strength of these studies is the tight
as hormone levels or cell proliferation assays, which may be control of experimental conditions. Their limitations are
important predictors of cancer or other diseases. These are their artificiality and the fact that no effect on rodents, how-
intermediate markers, however, and relating the results to ever unequivocal, can be assumed to apply to humans.
cancer occurrence may be problematic. Rather like the results of population ecological studies,
results from animal studies provide first lines of evidence
3.2.2 Live animal models that may prompt more persuasive research; they can also cor-
Laboratory animals can be used to test the effects of food, roborate such findings.
nutrition, and physical activity on the development of can-
cer. Human genes can be added to animals DNA (creating 3.2.3 In vitro studies
transgenic animal models) or key genes can be removed (cre- In vitro studies are conducted using cells or other test sys-
ating knockout animal models) to address specific research tems. Human cells, animal cells, mechanistic test systems,


Box 3.4 Systematic literature reviews

The main basis for the Panels work in com- Current practice, when resources allow, and terms. The papers identified were
ing to judgement on the causal relation- is to separate the process of collecting and assessed for relevance using reproducible
ships between food, nutrition, physical displaying evidence from that of discussing criteria. Study characteristics and results
activity, and cancer is a series of 20 specially and judging evidence. Evidence is collect- were extracted and recorded. Data from
commissioned systematic literature reviews ed systematically, after agreeing criteria for different studies were combined and
(SLRs). These have all been conducted inclusion or exclusion for review. As well as analysed, using meta-analysis when appro-
according to a common, detailed specifi- reducing possible bias, this is a more com- priate. Key features of selected studies are
cation, following recommendations made prehensive and transparent approach. This presented in graphic form in Chapters 46
by the Methodology Task Force, which process was used by the previous report, and 8, to aid comparison and quality assess-
completed its work before the Panels first which at the time of its publication was the ment. Existing SLRs were also identified to
meeting. The SLRs form the main evidence most comprehensive in its and allied fields. ensure, as far as possible, that all relevant
basis for the assessments and judgements This current Report has made a step- papers were included.
made by the Panel in Part 2 of this Report, change in this process, by commissioning An important aspect of an SLR is that all
on which the Panels population goals and independent SLRs, and making full use of stages of searching, selection, assessment,
personal recommendations in Part 3 are electronic resources. and analysis are prespecified, objective,
based. The Panel, in commissioning the SLRs reproducible, openly documented, and
This approach differs from that used to and supplementary work, decided to subject to peer review at critical stages. As
prepare most other expert reports. require evidence from all relevant epi- stated, full details of the approach taken
Previously, expert reports concerned with demiological and experimental studies, can be found in the SLR specification man-
the prevention of disease, and other top- together with biological findings. The ual contained on the CD included with this
ics, have relied on less formal methods to alternative approach would have been to Report (together with the SLRs).
collect and assess relevant literature. Until agree a hierarchy of epidemiological evi- The SLRs included evidence published up
the 1980s, most such reports were assem- dence, perhaps with one study type given to the end of 2005, and the Panels con-
bled by members of panels of scientists pre-eminent importance. Instead, while clusions are based on these SLRs. To ensure
who, assisted by secretariats, wrote drafts allowing for some types of epidemiologi- that the Panels recommendations, which
of chapters themselves using their own cal study being more or less prone to bias are derived from their conclusions and
knowledge, either with or without addi- than others, the Panel has based its con- judgements, take into account developing
tional research. The panel then reviewed clusions and judgements on evidence accu- evidence, a further review of studies pub-
the draft report until consensus was mulated from different types of study. For lished during 2006 was conducted. This
achieved. In some cases, report authors the Panel to be convinced that a relation- review was more limited than the full SLRs:
took total responsibility for assembling and ship between an exposure and cancer is it was confined to exposures that had
judging the evidence. In the 1990s, more causal, or that it is probably causal, consis- been judged convincing, probable, sub-
ambitious reports placed greater emphasis tent evidence from different types of study stantial effect on risk unlikely, and limit-
on secretariats, which tended to take more was required, with the exception of ran- edsuggestive, based on the SLRs. (See box
responsibility for drafting the report, and domised controlled trials. 3.8 for an explanation of these terms.) At
for some original research, as directed by The teams responsible for producing the this second review stage, no further meta-
the panel. More recently, panels have SLRs gathered relevant studies in a com- analyses were performed and a review of
sometimes been informed by narrative mon, systematic fashion, using a protocol study quality was not included. For these
reviews commissioned from specialists and designed to limit the potential for bias in reasons, the results of this 2006 review
prepared independently from the panel deciding which evidence should be includ- have been noted but have not been used
process. Such reviews are usually written ed or excluded from analysis. to alter the Panels judgements based on
from the specialists existing knowledge, The first stage of the SLRs was a com- the full SLRs. A further process has been
and form background substrate for the prehensive search of the scientific literature established for a continuous review of
reports. Narrative reviews may be pub- and other sources catalogued on electron- evidence published since 2006, after pub-
lished separately.29 ic databases, using all relevant keywords lication of this Report.

and bacterial systems can be used. interventions. Therefore extrapolation of results to humans
Cell cultures can be primary, where tissue (such as a is limited.
tumour biopsy) is taken directly from humans or animals and
then cultured; or secondary, where the original cells are cul- 3.2.4 Biological pathways
tured a number of times. Such cell lines are commonly used Epidemiological and experimental evidence indicating a
in laboratory research, and can become immortal cultured causal association between an aspect of diet and cancer is
again and again. The cells or tissues are subjected to poten- strengthened when there is evidence of a plausible biologi-
tial carcinogens, and then markers of damage are measured. cal pathway or mechanism by which the cancer process may
Conducting studies in vitro has two main advantages. be modified. The Panel agreed that simply identifying such
First, specific, well defined interventions can be tested; and plausible mechanisms is not sufficient evidence for a causal
second, intracellular mechanisms can be examined. However, relationship; but it also agreed that the case for a causal rela-
these studies do not allow the study of integrated systems, tionship is strengthened when there is evidence of biologi-
such as how organs or the whole body responds to the cal plausibility.

C H A P T E R 3 J U D G I N G T H E E V I D E N C E

Narrative reviews of experimental studies and of evidence dietary recalls, food frequency questionnaires, and food diary
of plausible biological mechanisms were included in the SLRs or food record methods. Most of the studies included in this
that inform this Report. Summaries of these SLRs are pre- Report used dietary assessment data from individuals,
sented in Part 2, Chapters 410. Also see box 3.4. recorded using food frequency questionnaires.
Diet histories take the form of unstructured enquiries,
more useful in clinical settings than in research studies.
3.3 Methods of assessment Dietary recalls may use structured or unstructured methods,
and are often administered many times over the course of a
Some exposures are easier to measure than others. Thus, it study. In a 24-hour dietary recall, a record is made of every-
is relatively easy to assess the impact of tobacco smoking and thing a person can recall eating or drinking over the previ-
exposure to tobacco on cancer risk. Although tobacco smoke ous 24 hours. Automated systems for collecting and
is a mixture of many chemicals, and its interactions with analysing dietary recalls have been developed, which facil-
the body are complex, tobacco can be considered a single itate the use of this method of assessment in large studies.32
exposure. Food frequency questionnaires collect information on food
By contrast, diets are multidimensional exposures and in consumption patterns, typically over the past year. A record
free-living populations cannot be measured with complete is made of the frequency of consumption of perhaps 100 to
accuracy. Moreover, the foods and drinks people consume 150 items, and often includes information on serving sizes.
every day contain thousands of constituents, some well Food frequency questionnaires may be designed to gain
known, others unknown and unmeasured. The relationships detailed information about specific aspects of diets, such as
between food, nutrition, physical activity, and health and dis- intakes of fats or dietary fibre, leaving other components less
ease are complex and difficult to untangle. The presence or well characterised. A questionnaire for whole diets cannot
absence of effect modification (box 3.6) can create additional adequately capture the full variety and composition of indi-
challenges. vidual diets without becoming unwieldy. Food frequency
questionnaires are inexpensive, however, and are practical
3.3.1 Foods, drinks, and nutrients for use in large-scale epidemiological studies.
Peoples dietary intake varies from day to day and over the Food diary or food record methods rely on the participants
course of their lives. There are interrelationships between in the study recording everything they eat and drink over the
food components, between foods in whole diets, and course of one or more days. Participants may be asked to
between diets and other behavioural characteristics such as estimate portion sizes, or weigh foods and drinks.
physical activity or smoking. There are several methods for All dietary assessment methods that rely on self-reporting
assessing food and drink consumption, all with their own are subject to measurement error. Further errors are intro-
weaknesses and strengths. duced by the conversion of food data to nutrient data, using
tables of the chemical composition of foods, which give aver- Dietary assessment methods age nutrient contents for defined foods. This implicitly
Food intakes can be measured for populations, groups, or assumes that all participants eat foods that have the same
individuals. The most commonly used techniques for assess- standard composition and portion size. But in reality, food
ing food and drink consumption are diet histories, 24-hour composition varies widely, depending on soil quality,

Box 3.5 Experimental findings

The Panel agreed a hierarchy of robust- nificantly influence judgements derived Class 2
ness recommended by the Methodology from consideration of the collective weight In vitro data from studies using human
Task Force, which completed its work of evidence from all other types of study. cells validated with an in vivo model;
before the Panels first meeting. The hier- For these reasons, eight types of exper- for example, a transgenic model.
archy of robustness was designed to deter- imental evidence were identified and split In vitro data from studies using primary
mine which types of human and animal into three classes: human cells.
experimental study are likely to be most In vitro data from studies using human
applicable to human cancer. This was done Class 1 cell lines.
for practical and scientific reasons. The In vivo data from studies in human
body of experimental literature is very volunteers (controlled human feeding Class 3
much larger than the body of epidemio- studies). In vitro data from studies on animal cells.
logical literature, and an exhaustive sys- In vivo data from studies using Data from mechanistic test systems; for
tematic review of this literature would genetically modified animal models example, isolated enzymes or genes.
have been impractical. Also, most experi- related to human cancer (such as gene
mental work, such as that conducted as a knockout or transgenic mouse models). For the systematic literature reviews in this
guide to toxicological regulations, either In vivo data from studies using rodent Report, only class 1 evidence was reviewed.
has no evident relevance to the work of cancer models designed to investigate Illustrative evidence from in vitro studies is
the Panel, or else would be unlikely to sig- modifiers of the cancer process. included only in Chapter 2.


3.3.2 Nutrition status

Box 3.6 Effect modification
Nutrition status is not simply a function of dietary intake. It
Effect modification (or effect-measure modification) occurs includes energy and nutrient intakes, and also body nutri-
when a measure of effect for an exposure changes over levels ent stores and body composition, all of which can be stud-
of another variable (the modifier).30 Effect modifiers can some- ied at various levels of complexity. Nutrition status cannot
times even change the direction of an effect. For example, a be completely measured by any one method, and judging
pooled analysis of seven cohort studies found an association which methods are most useful is an important aspect of the
between body fatness and decreased risk of breast cancer in pre- science in this field.
menopausal women, but with increased risk in postmenopausal Some aspects of nutrition status can be assessed relatively
women.31 In this case, menopausal status modifies the effect of
accurately. These include body fatness and measurements of
body fatness on breast cancer risk.
weight and height at birth, during growth, and in adulthood.
Nutrition status is affected by other biological and
behavioural factors, and also by social and environmental fac-
harvesting conditions, animal feed, storage, and food pro- tors. Social factors include economic and political drivers of
cessing, for example. Furthermore, food tables can be incom- food supplies, availability of food, and tradition and culture.
plete: for instance, they may not include information on
phytochemical or fatty acid levels in foods. In many coun- 3.3.3 Physical activity
tries, there may be no records of the composition of tradi- Study of the effects of physical activity on health requires reli-
tional and indigenous foods. able and valid measurements of physical activity in whatev-
Multiple-day food records or 24-hour dietary recalls have er setting it occurs occupational, household, transport,
been used as reference instruments to check the validity of and recreational and also of frequency, duration, and
food frequency questionnaires.33 34 However, studies using intensity. Effects of physical activity are not just a function
biomarkers (see have shown that food record and of total overall energy expenditure. A person may expend
recall methods are also liable to measurement error, and that the same amount of energy during a short period of intense
these errors are correlated with errors from food frequency exercise or in a longer period of moderate activity, but
data. This means that the use of food record or recall meth- the physiological effects may be different. Also see
ods to validate food frequency data results in an overesti- Chapter 5.
mation of the validity of the food frequency data.35-38 Assessments of physical activity may use objective bio-
Often, estimated intakes of macro- and micronutrients are chemical, physiological, or other methods, but these are
adjusted for energy intake (box 3.7). expensive and not commonly used in large studies.
Epidemiological studies usually rely on self-completed ques- Biomarkers tionnaires. These vary in the duration and type of physical
Recovery biomarkers, such as doubly labelled water or 24- activity, the length and detail of the questionnaire, and how
hour urinary nitrogen excretion, can be used to assess the the physical activity measures are calculated.
accuracy of various dietary assessment methods. The dou- As with food questionnaires, physical activity question-
bly labelled water test accurately measures a persons total naires have limitations. Activities may be over-reported when
energy expenditure, which equals energy intake when a per- participants overestimate their recreational activity, for
son is in energy balance.39 Urinary nitrogen excretion is used example, or under-reported, such as when participants do
as a biomarker of protein consumption.40 Studies using these not take account of everyday activities, such as walking
recovery biomarkers suggest that measurement errors from around their home or office. Many questionnaires ask only
all types of dietary assessment instruments are larger than about occupational activity or recreational activity, and there-
previously appreciated.35 38 41 fore do not provide a comprehensive account of peoples total
Concentration biomarkers, such as blood levels of fatty physical activity.
acids or vitamins, can be used to indirectly estimate dietary Results from questionnaires are commonly reported in
intake of these compounds. However, blood levels are deter- terms of energy expenditure. This is usually done by assign-
mined not only by a persons intake of the compound, but ing an energy cost (derived from published guides) to the
also by factors such as the compounds bioavailability and energy value of various activities, and multiplying this by the
excretion, intakes of other dietary components, personal duration and frequency of the activity. But there are large
characteristics such as smoking, and individual variation in variations in the energy values of different activities depend-
metabolism. These determinants and therefore the rela- ing on age, gender, body mass, skill, and level of fitness;
tions between true intakes and the biomarkers vary these can lead to significant errors in estimates.
among people, and this can bias observed dietcancer
associations.9 3.3.4 Cancer outcomes
For some compounds, such as selenium, biomarkers are a In studies of food, nutrition, physical activity, and cancer,
more accurate indicator of dietary intakes than data from accurate identification of cancer occurrence is as important
food frequency questionnaires.42 43 Many studies examine as making accurate measures of food and drink consumption,
concentration biomarkers as indirect proxies for intake, and and of physical activity. In most epidemiological studies,
there is growing interest in combining these data with results data from cancer registries are used, or else participants
from self-report dietary assessment instruments. report whether they have been diagnosed as having cancer.

C H A P T E R 3 J U D G I N G T H E E V I D E N C E

Studies may also require participants to undergo clinical 3.4 Causation and risk
examination, or provide tissue biopsy samples.
The duration of a study can also affect whether the full One of the Panels tasks has been to devise a transparent and
effects of exposures are identified. Relatively short-term objective method that enables evidence of relationships
prospective studies may miss any late effects on cancers. This between diet, physical activity, body fatness, and associated
is one reason why results from studies with long-term fol- factors, and cancer of one or more sites, to be judged as
low-up periods are particularly valuable. causal, with varying degrees of confidence.
Population-based cancer registries collect cancer incidence
and mortality data for the areas they serve, and produce can- 3.4.1 Inferring causation
cer registers. The Panel endorses the view of the panel responsible for the
In the US, the Surveillance, Epidemiology, and End Results previous report, that causal relationships between food and
(SEER) programme collects population-based data on newly nutrition, and physical activity can be confidently inferred
diagnosed cancers from registries that cover approximately when epidemiological evidence, and experimental and other
23 per cent of the US population.49 The European Network biological findings, are consistent, unbiased, strong, graded,
of Cancer Registries represents population-based cancer reg- coherent, repeated, and plausible. Individually, none of these
istries in Europe.50 Population-based registries are becoming factors is likely to be sufficient to infer a causal relationship
increasingly available in middle- and low-income countries. with confidence. Also, individual relationships may be defi-
In total, 57 countries and 186 cancer registries are repre- cient in various respects, but collectively can still be judged
sented in Volume VIII of the Cancer Incidence in Five as causal because of their cumulative weight.
Continents series published by IARC.1 Also see box 1.2. Many types of evidence can contribute to causal inference.
Cancer incidence data are coded in a standardised way, However strong the evidence from any single study, it will
using International Classification of Disease (ICD) codes that rarely justify a conclusion of causality. Increasing surviv-
have been established for oncology.51 These 10-digit codes ability of an observed relationship, when supported by fur-
specify the site of origin of the tumour (topography), the ther studies, or which produces corroborative evidence in
tumour or cell type (morphology), the behaviour of the other categories, as listed above, strengthens the evidence
tumour (malignant, benign, or in situ), and the tumour for a causal relationship.52 53
grade or degree of differentiation. With regard to food and nutrition, and physical activity,
In addition to providing cancer incidence and mortality single exposures are unlikely to act alone to cause or pre-
statistics, cancer registries are used by researchers to identi- vent cancer. In general, many factors act together as con-
fy people who are eligible to enrol in a case-control study, tributory or component causes, forming a complete causal
or to collect information on cancer diagnosis relating to peo- process. Component causes can interact biologically, even
ple who are enrolled in a cohort study. when exerting their effects at different times.30

Box 3.7 Energy adjustment

One of the basic principles in controlled diets. The total amount of food consumed centage of total energy) or regression
human and animal feeding experiments to is determined primarily by body size and analysis to calculate nutrient residuals. In
evaluate the effect of a specific dietary fac- physical activity. an epidemiological analysis, the nutrient
tor is that the diets should be isocaloric Body size and physical activity are of density does not adequately adjust for
(i.e., total energy intake is the same in both interest in their own right, but their effects total energy intake if energy intake itself
groups). This is because differences in need to be disentangled from the effects is associated with disease risk. In this case,
energy intake between two groups would of a specific nutrient. This can be done by total energy intake must be added as a
cause one group to gain or lose more using energy-adjusted nutrient intakes. separate term to the model. Another
weight than the other. The effects of the Expressed another way, studies designed to method, that of energy partition, has
dietary factor being investigated could not change the intake of a specific nutrient been used in some studies, but this is not
then be distinguished from the effects of usually should do so by changing the com- an isocaloric analysis and thus does not
changes in weight. This is important, position of diets rather than total energy control for total energy intake.
because differences in weight themselves intake. Epidemiological studies therefore An additional advantage of energy-
may have different physiological effects. should adjust for energy and not rely on adjusted nutrient intakes is that they are
In epidemiological studies, there are sim- absolute intakes, which reflect both dietary often measured with less error than
ilar reasons to conduct isocaloric analyses. composition and variation in total energy absolute nutrient intakes. This is because
These use statistical methods to adjust intake due to differences in body size and over- or under-reporting of specific nutri-
intakes of the dietary factor under study physical activity. ents tends to be strongly correlated with
for total energy intake. The rationale for The best method to adjust for total over- or under-reporting of total energy
energy-adjusted intakes is that the energy intake has been a matter of con- intake, especially for macronutrients, being
absolute intake of a nutrient is a function siderable discussion.44-48 The two basic calculated from the same foods. These
of two factors: first, the total amount of approaches are to use the nutrient densi- errors are highly correlated, so tend to can-
food consumed, represented by total ener- ty (for example, expressing intake per unit cel each other when calculating nutrient
gy intake; and second, the composition of of energy or, for macronutrients, as a per- densities or energy-adjusted intakes.


3.4.2 The portfolio approach amount to strong evidence if consistently found in large, well
Many different types of study, all of which have strengths and designed studies.
weaknesses, investigate links between food, nutrition, phys-
ical activity, and cancer. Persuasive evidence comes from dif-
ferent types of epidemiological study, supported by 3.5 Coming to judgement
experimental findings that indicate a relevant biological
mechanism. A crucial part of the process that has informed this Report
The Panels judgements, presented in Part 2 of this Report, is the methods used by the Panel in judging whether the evi-
are based on its assessment of the evidence available in the dence that a relationship between aspects of food, nutrition,
scientific literature, with due consideration given to the physical activity, and body fatness, and cancers of the sites
advantages and disadvantages of each type of study design, specified, is or may be causal. The need for evidence from
and to the quality of individual studies. An inclusive or port- different types of study and the characteristics looked for in
folio approach has been taken, recognising the relative such studies have been outlined already. Here, the precise
strengths and weaknesses of different types of study, but in methods used by the Panel are explained.
which no single type of study design is given pre-eminence. The previous report broke new ground in a number of
In general, the strongest evidence comes from consistent respects. One was to display panel judgements in the form
findings from different types of studies, preferably also in of matrices within which panel judgements on causal rela-
diverse populations. tionships, of different degrees of confidence, were shown,
and repeated in the text of the report. This method has been
3.4.3 Quantification of risk adapted in other reports.29 Another was the specification of
Quantification of the risk of any disease is an essential basis criteria guiding these judgements. The previous report also
for public health policy planning. It also guides people in used explicit statements explaining why, on occasion, the
making their own decisions about how they lead their lives. panel had made judgements that did not obviously derive
It is not enough to know that the risk of cancer is affected from the evidence as presented. One general principle was
by diet. It is also important to know by how much. For exam- that of transparency. Readers and users of the previous report
ple, if consumption of alcohol increases the risk of breast can- have been able to follow its reasoning, to challenge any
cer, and diets high in vegetables decrease the risk of various judgements that might seem questionable, and to modify
cancers, to what extent may the incidence of cancer on a or reinforce judgements in the light of further and better
population basis be affected by these factors? And on a per- evidence.
sonal level, how can people best judge how their current The Panel responsible for this Report decided to adapt
diets and ways of life, and any changes they might want to those innovative approaches, and use them as the basis for
make, are likely to affect their own risk of cancer? its work. Some of the judgements made in this Report are
Quantifying risk helps to answer such questions. different from those based on the evidence available a
The strength of a relationship between any risk factor and decade previously, while others confirm or strengthen pre-
the occurrence of disease is commonly expressed in terms of vious judgements.
relative risk (RR). In cohort studies, this is the ratio of risk
(or incidence) of a disease among people with a particular 3.5.1 The matrix approach
characteristic (say, high consumption of red meat) to that An example of a matrix used in this Report is shown in fig-
among people without that characteristic (in this example, ure 3.1. This particular matrix displays the Panels judge-
low or no consumption of red meat). In case-control stud- ments on the likelihood that physical activity modifies the
ies, the odds ratio is used, which is the ratio of the odds of risk of cancers of specified sites. This matrix is used here as
exposure among cases to the odds of exposure among con- an example, to explain the nature of the matrices displayed
trols. Relative risks below 1.0 imply a protective effect: so a in all chapters in Part 2 of this Report. This matrix and its
relative risk of 0.5 for high compared with low vegetable con- judgements are discussed fully in Chapter 5.
sumption implies a halving of risk. Relative risks above 1.0 The title of the matrix is self-explanatory. In this and other
indicate an increased risk. cases, the introductory words are important: here, the
Absolute risk is also important. Small RR values, when footnote specifies that the physical activity referred to is of
consistent, are important when the number of people affect- all types.
ed is large. A large RR of a rare type of cancer amounts to The matrices display the Panels judgements on whether
only a small absolute risk, which may reasonably be con- particular aspects of food and nutrition, physical activity, and
sidered not significant, either by public health planners or body composition do or may modify (or not modify) the risk
by individuals assessing their own choices. By contrast, a of cancers of specific sites. The matrices are of course short-
small RR may amount to a large number of cases for a com- hand, and the entries cannot convey all nuances. Necessary
mon type of cancer. For example, an increased risk of 10 per clarifications and qualifications are stated in footnotes to the
cent implied by a RR of 1.10 amounts to many extra cases matrices.
of colorectal and breast cancer in Europe and North Matrix entries themselves need to be explained. For exam-
America, where these cancers are common. Assessment ple, an entry fruits or foods containing dietary fibre in a
of small RRs depends on the size and quality of the matrix column headed decreases risk means that the Panel
studies in which such risks are identified. Small RRs may has judged that these foods are or may be protective against

C H A P T E R 3 J U D G I N G T H E E V I D E N C E

risk, is judged to be convincing, or else probably causal. A

PHYSICAL ACTIVITY, AND THE RISK OF CANCER judgement of convincing in turn generally justifies a recom-
mendation designed to inform policies and programmes
In the judgement of the Panel, physical activity1 modifies the risk of the
following cancers. Judgements are graded according to the strength of designed to prevent cancer. A judgement of probable also
the evidence. normally justifies a recommendation. So in the case of the
matrix shown, it follows that the Panel would make a rec-
DECREASES RISK INCREASES RISK ommendation on physical activity designed to reduce the risk
of cancer.
Convincing Colon2 The top two rows of the matrix are separated from the row
Probable Breast below, which shows judgements that the evidence is too lim-
(postmenopause) ited, for a variety of reasons (see 3.5.5), to conclude that a
Endometrium relationship is causal, but that there are enough data to
suggest that such a relationship might exist. Normally, a
Limited Lung
judgement of limited suggestive does not justify any rec-
ommendation. The matrices used in Chapter 7 also include
Breast (premenopause)
a row showing judgements where the evidence is so limited
Substantial (again for a variety of reasons) that no judgement can be
effect on risk None identified made whether any association exists or not. For this reason,
unlikely such judgements of limited no conclusion do not indi-
cate whether the evidence is in the direction of decreasing
1 Physical activity of all types: occupational, household, transport, or increasing risk. The final, bottom row of the matrix, sub-
and recreational.
2 Much of the evidence reviewed grouped colon cancer and rectal cancer stantial effect on risk unlikely, shows judgements for which
together as colorectal cancer. The Panel judges that the evidence is
stronger for colon than for rectum.
the evidence, equivalent to a judgement of convincing or
probable, shows that no causal relationship is likely to exist.
For an explanation of the terms used in the matrix,
please see chapter 3.5.1, the text of this chapter,
Terms used in the text and matrices to refer to foods and
and the glossary. drinks, physical activity, body fatness, and other factors are
necessarily shorthand. Thus, in chapter 4.2, the matrix dis-
Figure 3.1 Example of a matrix plays judgements that non-starchy vegetables probably pro-
tect against a number of cancers. The matrix in chapter 4.8
displays judgements that the evidence that alcoholic drinks
the cancer specified. The judgements are derived from cause a number of cancers is convincing. What is meant by
analysis of studies in which relatively high intakes of (in this non-starchy vegetables and by alcoholic drinks is defined
example) fruits and foods containing dietary fibre are com- in the text of these sections.
pared with relatively low intakes. The same point applies to Further, when non-starchy vegetables is used as a matrix
matrix entries physical activity and body fatness in entry and contained in Panel judgements, it means relatively
columns headed decreases risk and increases risk, respec- high consumption of non-starchy vegetables and/or foods
tively, which are derived from analysis of studies of people containing them. The same point applies to many other
whose physical activity levels or degree of body fatness is matrix entries and also to the accompanying text.
relatively high compared with people whose physical activ- Within all matrix cells, exposures are listed in the order of
ity levels or degree of body fatness is relatively low. the contents of the Report. There are a number of cancer sites
In some cases, analysis may show that any effect begins where a substantial number of related exposures meet the
or ends, or is less apparent, below or above evident thresh- criteria for matrix entry. The Panel has judged that it is often
olds. For example, it has been thought that alcoholic drinks appropriate to aggregate such exposures. For example, if
increase the risk of some cancers only above certain levels both alcoholic drinks and wine are judged as exposures
of consumption. Such amounts would be specified in a foot- that probably increase the risk of a type of cancer, then
note to the relevant matrices, and could be reflected in Panel only alcoholic drinks will appear in the matrix for that
recommendations. When matrices include no such foot- cancer site.
notes, this is because no lower or upper threshold of effect The matrices used in this Report differ from those used in
has been identified. In such cases, matrix entries showing the previous report in a number of respects. The previous
or suggesting a causal association should be taken to mean report used categories of possible and insufficient defined
that the effect is across the whole range of dietary intake, differently from the categories of limited suggestive and
amounts of physical activity, or degrees of body fatness limited no conclusion used here. Also, the previous report
found in the studies analysed. The implications of the nature allowed for different weights of evidence for no causal rela-
of the dose-response relationships for recommendations are tionship, whereas this Report includes just the one judge-
further discussed in Chapter 12, in Part 3 of this Report. ment of substantial effect on risk unlikely. The judgements
of convincing and probable, both agreed to be a sufficient
3.5.2 Levels and types of judgement basis for recommendations, are common to both reports,
The top half of the matrix in figure 3.1 shows that the although the criteria allowing such judgements have been
evidence of causality, either of decreased risk or increased refined for this Report (see 3.5.5).


Box 3.8 Criteria for grading evidence

This box lists the criteria finally agreed by so long as this can be explained mendations designed to reduce the inci-
the Panel that were necessary to support plausibly. dence of cancer. Any exceptions to this
the judgements shown in the matrices and Strong and plausible experimental require special explicit justification.
text of the Part 2 chapters. The grades evidence, either from human studies or All the following were generally
shown here are convincing, probable, relevant animal models, that typical required:
limited suggestive, limited no con- human exposures can lead to relevant Evidence from at least two independent
clusion, and substantial effect on risk cancer outcomes. cohort studies or at least five case-
unlikely. In effect, the criteria define these control studies.
terms. Probable The direction of effect is generally
These criteria are for evidence strong consistent though some unexplained
Convincing enough to support a judgement of a prob- heterogeneity may be present.
These criteria are for evidence strong able causal relationship, which would gen- Evidence for biological plausibility.
enough to support a judgement of a con- erally justify goals and recommendations
vincing causal relationship, which justifies designed to reduce the incidence of cancer. Limited no conclusion
goals and recommendations designed to All the following were generally Evidence is so limited that no firm conclu-
reduce the incidence of cancer. required: sion can be made.
A convincing relationship should be Evidence from at least two independent This category represents an entry level,
robust enough to be highly unlikely to be cohort studies, or at least five case- and is intended to allow any exposure for
modified in the foreseeable future as new control studies. which there are sufficient data to warrant
evidence accumulates. All of the following No substantial unexplained Panel consideration, but where insufficient
were generally required: heterogeneity between or within study evidence exists to permit a more definitive
Evidence from more than one study type. types in the presence or absence of an grading. This does not necessarily mean a
Evidence from at least two independent association, or direction of effect. limited quantity of evidence. A body of evi-
cohort studies. Good quality studies to exclude with dence for a particular exposure might be
No substantial unexplained confidence the possibility that the graded limited no conclusion for a
heterogeneity within or between study observed association results from number of reasons. The evidence might be
types or in different populations random or systematic error, including limited by the amount of evidence in terms
relating to the presence or absence of confounding, measurement error, and of the number of studies available, by
an association, or direction of effect. selection bias. inconsistency of direction of effect, by
Good quality studies to exclude with Evidence for biological plausibility. poor quality of studies (for example, lack
confidence the possibility that the of adjustment for known confounders), or
observed association results from Limited suggestive by any combination of these factors.
random or systematic error, including These criteria are for evidence that is too Exposures that are graded limited no
confounding, measurement error, and limited to permit a probable or convincing conclusion do not appear in the matrices
selection bias. causal judgement, but where there is evi- presented in Chapters 46, but do appear
Presence of a plausible biological dence suggestive of a direction of effect. in Chapters 7 and 8.
gradient (dose response) in the The evidence may have methodological When an exposure is graded limited
association. Such a gradient need not flaws, or be limited in amount, but shows no conclusion, this does not necessarily
be linear or even in the same direction a generally consistent direction of effect. indicate that the Panel has judged that
across the different levels of exposure, This almost always does not justify recom- there is evidence of no relationship. With

3.5.3 The food-based approach rather than simply a marker for the particular foods in which
Terms used in the text of this Report and in the matrices it is found; or for other dietary constituents found in the
reflect the Panels decision that the Report, and its judge- same foods; or other associated health-related factors. In
ments and recommendations, should whenever possible be chapter 4.10, some micronutrients appear in matrices grad-
based on foods and drinks rather than on nutrients. This ed as convincing or probable. These judgements are
food- (and drink-) based approach is also apparent in the derived from the findings of good quality, randomised, con-
overall structure of the Report. Chapter 4, the first chapter trolled trials, sometimes also supported by observational
in Part 2, on foods and drinks, is the longest chapter. This is studies, clearly showing that supplements of these micronu-
in part because dietary constituents associated with foods are trients, rather than the foods containing them, affect the risk
grouped with these foods. Thus, matrix entries in chapter 4.1 of cancer.
identify foods containing dietary fibre (rather than dietary Sometimes the studies that are the basis for the Panels work
fibre), and in 4.8 identify alcoholic drinks (rather than alco- have used markers of exposure. Thus, many epidemiological
hol or ethanol). Chapters 4, 5, and 6 also include material studies use body mass index as a marker of body fatness.
presented graphically, such as the forest plots described in When there is clear evidence of an underlying mechanism for
box 3.3. body fatness, the Panel has agreed that the term body fatness
The food-based approach is also justified because of the best represents the causal factor. Usually, anthropometric
uncertainty that any food constituent is a true causal factor, indices other examples being waist to hip ratio and waist

C H A P T E R 3 J U D G I N G T H E E V I D E N C E

further good quality research, any expo- confidence, the possibility that the on risk unlikely with a lower confidence
sure graded in this way might in the future absence of an observed association than this would not be helpful, and could
be shown to increase or decrease the risk results from random or systematic overlap with judgements of limited sug-
of cancer. Where there is sufficient evi- error, including inadequate power, gestive or limited no conclusion.
dence to give confidence that an exposure imprecision or error in exposure
is unlikely to have an effect on cancer risk, measurement, inadequate range of Special upgrading factors
this exposure will be judged substantial exposure, confounding, and selection These are factors that form part of the
effect on risk unlikely. bias. assessment of the evidence that, when pre-
There are also many exposures for which Absence of a demonstrable biological sent, can upgrade the judgement reached.
there is such limited evidence that no gradient (dose response). So an exposure that might be deemed a
judgement is possible. In these cases, evi- Absence of strong and plausible limited suggestive causal factor in the
dence is recorded in the full SLR reports experimental evidence, either from absence, say, of a biological gradient,
contained on the CD included with this human studies or relevant animal might be upgraded to probable in its
Report. However, such evidence is usually models, that typical human exposures presence. The application of these factors
not included in the summaries and is not lead to relevant cancer outcomes. (listed below) requires judgement, and the
included in the matrices in this printed way in which these judgements affect the
Report. Factors that might misleadingly imply an final conclusion in the matrix are stated.
absence of effect include imprecision of Presence of a plausible biological
Substantial effect on risk unlikely the exposure assessment, an insufficient gradient (dose response) in the
Evidence is strong enough to support a range of exposure in the study population, association. Such a gradient need not
judgement that a particular food, nutri- and inadequate statistical power. Defects be linear or even in the same direction
tion, or physical activity exposure is unlike- in these and other study design attributes across the different levels of exposure,
ly to have a substantial causal relation to a might lead to a false conclusion of no so long as this can be explained
cancer outcome. The evidence should be effect. plausibly.
robust enough to be unlikely to be modi- The presence of a plausible, relevant A particularly large summary effect size
fied in the foreseeable future as new evi- biological mechanism does not necessarily (an odds ratio or relative risk of 2.0 or
dence accumulates. rule out a judgement of substantial effect more, depending on the unit of
All of the following were generally on risk unlikely. But the presence of exposure) after appropriate control for
required: robust evidence from appropriate animal confounders.
Evidence from more than one study models or in humans that a specific mech- Evidence from randomised trials in
type. anism exists, or that typical exposures can humans.
Evidence from at least two independent lead to cancer outcomes, argues against Evidence from appropriately controlled
cohort studies. such a judgement. experiments demonstrating one or
Summary estimate of effect close to Because of the uncertainty inherent in more plausible and specific mechanisms
1.0 for comparison of high versus low concluding that an exposure has no effect actually operating in humans.
exposure categories. on risk, the criteria used to judge an expo- Robust and reproducible evidence from
No substantial unexplained sure substantial effect on risk unlikely are experimental studies in appropriate
heterogeneity within or between study roughly equivalent to the criteria used animal models showing that typical
types or in different populations. with at least a probable level of confi- human exposures can lead to relevant
Good quality studies to exclude, with dence. Conclusions of substantial effect cancer outcomes.

circumference do not appear in the matrices. relationships). Its decisions here have been enlightened by
As exceptions to this approach, the Panel has made judge- the rapid development since the mid-1990s of the technique
ments on adult attained height and greater birth weight, of systematic review, using search techniques enabled by the
as shown in the matrices. Many epidemiological studies have electronic revolution.
reported on height and birth weight. It is thought that asso- Since the mid-1990s, about as many studies in the field
ciations between height, birth weight, and cancer risk reflect of this Report have been published as were published in the
some causal association with a combination of genetic, hor- previous 35 years. This development has not just been one
monal, nutritional, and other factors. Uncertainty as to the of quantity but also of design and quality. In particular, many
mechanisms underlying the observations with adult attained cohort studies have been published in the period analysed
height and birth weight mean that the Panel was not able by the SLRs, and some of these have also been pooled. The
to determine the appropriate causal factors to be shown in Panel agreed that in general, cohort studies provide more
the matrices. Instead, the anthropometric markers have been impressive evidence than case-control and other epidemio-
included, with appropriate footnotes. logical study designs, and this decision affected the criteria
for judgement. For this reason, while the best evidence
3.5.4 The basis for robust judgements comes from a number of different study designs, the Panel
The Panel has been particularly careful in deciding the cri- agreed that reasonably strong and consistent evidence was
teria for judgement on causal relationships (or lack of such needed from studies where biases could reasonably be


excluded. Usually this evidence came from cohort studies public health goals and personal recommendations.
with a prospective design or, where available and appropri-
ate, from randomised controlled trials, to allow a judgement Limited and below
of convincing or probable. This was not a requirement spec- For the two types of limited judgement, the evidence falls
ified for the previous report. See box 3.8. short of a higher judgement for a variety of reasons. There
A consequence is that the same amount of evidence for any are also many exposures for which evidence was so limited
particular association has sometimes led to a different judge- that it did not warrant detailed consideration. In these cases,
ment and level of matrix entry from the previous report. evidence is recorded in the full SLR reports contained on the
Sometimes, in cases where there is a greater quantity of evi- CD included with this Report. However, this evidence is not
dence, this might not lead to a higher classification, and included in the summaries or matrices in this printed Report.
could even possibly lead to a lower one. These refinements
are intended to give as robust an assessment as possible, Absence of a causal relation
given current understanding. In these respects, the criteria The strength of this judgement corresponds to that for con-
used by the Panel are more stringent than those pioneered vincing or probable (and replaces the previous reports no
by the previous panel. relationship category). This judgement does not reflect the
The Panel agreed that the criteria for their judgements absence of evidence, which in itself is not evidence of
should be detailed and precise. But such criteria do not absence of effect. As with judgements of convincing or
lead to automatic judgements. However meticulous, they probable, evidence from both observational studies and ran-
cannot replace expert judgement. If a reviewer or a Panel domised trials contribute to such an inference.
member felt that important considerations had been over-
looked by the overall agreed process, this was discussed,
and the Panels final judgement specified, with reasons 3.6 Conclusions
Reports such as this address issues of public importance.
3.5.5 The grading criteria They are informed by a process of collection, display, dis-
Specification of criteria for the grading of judgements cussion, and judgement of evidence as the basis for recom-
enables a common, transparent approach. But as indicated, mendations made in the public interest.
any such criteria cannot fully capture the sophistication and We, the members of the Panel responsible for this Report,
nuances of all the studies considered, or the nature and qual- have had the responsibility to ensure that the judgements we
ity of different studies. have made in Part 2, and the public health goals and per-
In using the criteria specified here, the Panel has taken into sonal recommendations we have specified in Part 3, are
account additional factors including, but not confined to, the clearly and reliably based on current evidence.
type, number, and size of studies; their design and execu- In the five years of our work, we have built on the work of
tion; the nature of any intervention; the definition of cases the previous report, and have been supported by the findings
and non-cases; the selection of any comparison group; meth- of a preliminary Methodology Task Force; by the evidence
ods of characterising exposure and outcome; length and gathered and presented by independent SLR centres; by
completeness of follow-up; and the methods used to ascer- observers from United Nations and other international organ-
tain cases. isations; and by the Reports Secretariat. As far as we know,
Other factors might lead to one or another grading. Failure the whole process, which has also included eight face-to-face
to achieve a higher grade might result from several small Panel meetings, each lasting up to four days, is the most com-
deficits against a number of standards, or from a major short- prehensive and rigorous of its kind yet undertaken.
fall in one particular aspect of evidence. Panel expertise was As this chapter shows, no method used to ascertain causal
essential in judging whether criteria were met, or upgrad- relationships between food, nutrition, physical activity, and
ing factors (box 3.8) were applicable, as well as deciding cancer is perfect. But we believe that the integrated and
what constituted substantial heterogeneity, high-quality sometimes innovative approaches we have taken, sum-
study design, and so on. The criteria provide a consistent marised here, have enabled us to make sound judgements
basis for judgement, not a set of boxes to be ticked. As well and reliable recommendations. We have also done our best
as these upgrading factors, particular reasons why any spe- to make sure that the methods we have used are explained
cific judgement was reached are presented under the rele- and displayed transparently, so they can be readily accessed
vant exposure and cancer site in Chapters 47. and challenged as science develops, or from different points
The following grading criteria specify the quantity, quali- of view. We believe this best serves science, and also the
ty, and nature of evidence that can lead to associations being cause of cancer prevention.
graded differently.

Convincing, probable
In considering the criteria allowing a judgement that the evi-
dence of a causal relationship was convincing, or that the
evidence showed a probable causal relationship, the Panel
was conscious that both judgements were liable to generate


Part 2

` i Vi > `
` } i i

Chapter 4
Foods and drinks 66

Chapter 5
Physical activity 198

Chapter 6
Growth, development, body composition 210

Chapter 7
Cancers 244

Chapter 8
Determinants of weight gain, overweight, obesity 322

Chapter 9
Cancer survivors 342

Chapter 10
Findings of other reports 348

Chapter 11
Research issues 360
P A R T 2

Introduction to Part 2

The brief of the Panel, and of the systematic literature review teams that
provided the basis for the Panels work, has included the task of presenting a
clear, strong, and reliable foundation for the final recommendations. These in
turn form the basis of sound policies and effective programmes to reduce the
rates of cancer in populations, and the risk of cancer in people, whether as
members of communities, or as families, or as individuals.

In this central part of the Report, seven chapters display the findings of the
independently assembled systematic literature reviews, and the judgements of
the Panel derived from these reviews and other evidence as needed. The Panels
judgements are displayed in the form of matrices that introduce five of these
chapters. Judgements of convincing and probable causal relationships,
shown in the top part of these matrices, are the basis for recommendations
made in Part 3 of the Report.

Chapter 4, the first and longest chapter that follows, is concerned with types of
food and drink. The judgements of the Panel are generally food- and drink-
based, reflecting the evidence. Findings on dietary constituents and
micronutrients are identified as, for example, on foods containing dietary
fibre or foods containing folate. For consistency, findings on methods of food
processing are, where possible, shown as part of the whole evidence on the
associated foods so that, for example, the processing and preparation of meat
is integrated with the evidence on meat. Evidence specifically on dietary
supplements and on patterns of diet is included in the two final sections of this

Chapters 5 and 6 are concerned with physical activity, and with body
composition, growth, and development. Evidence in these areas is more
impressive than was the case up to the mid-1990s; the evidence on growth and
development indicates the importance of a whole life-course approach to the
prevention of cancer. As with the chapter on foods and drinks, these chapters
include detailed summaries of the evidence collected in the systematic
literature reviews together with graphic representations of the most significant

Chapter 7 summarises and judges the evidence as applied to 17 cancer sites,

with briefer summaries based on narrative reviews on cancers of five other
body systems and sites. The judgements as shown in the matrices in this
chapter correspond with the judgements shown in the matrices in the previous

Chapter 8, in which judgements are also based on the evidence from the
systematic literature reviews amplified by knowledge of physiological
processes, concerns the biological and associated determinants of weight gain,
overweight, and obesity. Before work on this chapter began, the Panel agreed
that a comprehensive review of the evidence would be likely to show that

obesity is or may be a cause of a number of cancers. It was therefore important
to identify what aspects of food, nutrition, and physical activity themselves
affect the risk of obesity and associated factors.

Improved screening, diagnosis, and medical services, including therapy and

surgery, are in many countries improving the rates of survival for people with
cancer. The number of cancer survivors people living after diagnosis of cancer
is therefore increasing. The relevance of food, nutrition, physical activity, and
body composition to people living with cancer, and to the prevention of
recurrent cancer, is summarised in Chapter 9.

The Panel agreed that its final recommendations should be principally based on
the evidence concerning cancer, and also should take into account findings on
food, nutrition, physical activity, and the prevention of other chronic diseases,
and of nutritional deficiencies and nutrition-related infectious diseases,
especially of childhood. Chapter 10, which is also based on a systematic
literature review, is a summary of the findings of expert reports in these areas.

The proposals for further research contained in Chapter 11 are, in the view of
the Panel, the most promising avenues to explore in order to refine
understanding of the links between food, nutrition, physical activity, and
cancer, and so improve the prevention of cancer, worldwide.

As expected, a comprehensive assessment of all relevant types of evidence

relating to food, nutrition, physical activity, body composition, and the risk of
cancer has proved to be a massive task. The Panel was impressed not only by
the quantity but also the quality of much of the evidence, and the degree to
which a great deal of the evidence was consistent. As a result,
recommendations designed to prevent cancer in general can be made with
confidence. These are contained in Part 3.

P A R T 2 E V I D E N C E A N D J U D G E M E N T S

C H A P T E R 4

Foods and drinks

This chapter, with the following chapters in Part production, preservation, processing, and
Two, forms the basis for the population goals and preparation (including cooking), that have not been
personal recommendations in Part Three. incorporated in previous sections. The final section
The Panel decided that the evidence on food, summarises evidence on dietary patterns, including
nutrition, and cancer is generally most persuasive being breastfed.
for foods rather than for specific nutrients or other The pattern that emerges, though different in
food constituents; and that the evidence from some important respects, is largely similar to that
epidemiological and experimental studies in this based on the evidence gathered in the mid-1990s,
field, usually undertaken to address questions although the confidence with which various
about cancers of specific or related sites, is most exposures are judged to cause or protect from
usefully synthesised in terms of foods and drinks. cancer has sometimes changed.
The detailed evidence on foods and drinks is
presented in this chapter, and that on physical
activity and on body composition in the following
two chapters. These three chapters include
summaries of the evidence, including meta-analyses
presented in graphic form, as well as the Panels
judgements. Chapter 7 presents the evidence on
cancer sites in more summarised form.
In this chapter, whenever possible and
appropriate, the evidence on dietary constituents,
and on food production, preservation, processing,
and preparation (including cooking), is integrated
with the evidence on foods and drinks. So here, for
example, the evidence on carotenoids is considered
together with the evidence on vegetables and
fruits; the evidence on methods of cooking meats is
considered with the evidence on red meat and on
processed meats; and the evidence on ethanol is
considered with alcoholic drinks.
The result is not perfect. There is no single, ideal
way of categorising the evidence on food and
nutrition. But an approach emphasising foods and
drinks is consistent with the generally accepted
view that food-based dietary guidelines and
recommendations are particularly valuable as a
foundation for policies designed to improve public
The first two sections of this chapter summarise
and judge the evidence on plant foods; the next
two sections that on animal foods; and the
following two sections that on fats and oils, and
sugars and salt. The next two sections concern
drinks, the second of which covers alcoholic drinks.
These are followed by sections concerned with
those aspects of dietary constituents, and with food

C H A P T E R 4 F O O D S A N D D R I N K S

4.1 Cereals (grains), roots, tubers,

and plantains


In the judgement of the Panel, the factors listed below modify the risk of cancer. Judgements are graded according to the strength of the evidence.


Exposure Cancer site Exposure Cancer site

Convincing Aflatoxins1 Liver

Probable Foods containing Colorectum

dietary fibre2

Limited Foods containing Oesophagus

suggestive dietary fibre2

effect on risk None identified

1 Foods that may be contaminated with aflatoxins include cereals (grains), and also pulses (legumes), seeds, nuts, and some vegetables and fruits (see chapter 4.2).
2 Includes both foods naturally containing the constituent and foods which have the constituent added (see chapter 3.5.3). Dietary fibre is contained in plant foods
(see chapter 4.2 and box 4.1.2).

For an explanation of all the terms used in the matrix, please see chapter 3.5.1, the text of this section, and the glossary.

These starchy plant foods have been the staple sources of The Panel judges as follows:
dietary energy and bulk for humans since the Foods containing dietary fibre probably protect against
development of settled communities and agriculture. They colorectal cancer; and there is limited evidence
have to be prepared in some way to make them edible. In suggesting that such foods protect against oesophageal
whole or relatively unprocessed forms, they are also cancer. Dietary fibre is found in plant foods: vegetables,
sources of dietary fibre and various micronutrients. fruits, and pulses (legumes) (see chapter 4.2), as well as
Cereals in whole form contain essential fats. When the in cereals, roots, tubers, and plantains. All these foods are
outer layers of these foods are removed and they are highest in dietary fibre when in whole or minimally
refined, most of what remains is starch and protein. processed form.
In general, with industrialisation and urbanisation, Foods high in dietary fibre may have a protective effect
consumption of these foods decreases, and more is because of being bulky and relatively low in energy
consumed in the form of cereal products, which are density. See chapters 6.1, 7.3, 7.9, and Chapter 8 for
typically more energy-dense and which may contain discussion of the role of energy density in weight gain,
substantial amounts of fat, sugar, or salt. Pure starch from overweight, and obesity, and of weight gain, overweight,
these foods is also used as an ingredient in many and obesity in the risk of some cancers, including those of
processed foods. Wheat, rice, maize (corn), and potatoes the oesophagus and colorectum.
and their products are now the main cereals and The Panel also judges that the evidence that foods
roots/tubers produced and consumed globally. contaminated with aflatoxins are a cause of liver cancer is
Overall, the Panel judges that evidence indicating that convincing. Cereals (grains) and peanuts (see chapter
cereals (grains), roots, tubers, or plantains affect the risk 4.2) are the foods most commonly infested by these
of any cancer, remains insubstantial. fungal toxins. Contamination is most widespread in

P A R T 2 E V I D E N C E A N D J U D G E M E N T S

countries with hot, damp climates and poor storage 4.1.1 Definitions, sources
Within the remit of this Report, the strongest evidence, Cereals (grains)
corresponding to judgements of convincing and Cereals (grains) are the seeds and energy stores of cultivat-
probable, shows that foods containing dietary fibre ed grasses. The main types are wheat, rice, maize (corn),
probably protect against colorectal cancer; and that foods millet, sorghum, barley, oats, and rye. In some countries,
contaminated with aflatoxins are a convincing cause of cereal is also a term for dry foods made from grains and
liver cancer. Also see chapter 4.2 for judgements of other ingredients, often eaten with milk for breakfast.
probable protective effects of foods containing various
micronutrients also found in cereals, roots, and tubers, Roots, tubers, plantains
particularly when relatively unprocessed. Roots and tubers are energy stores of plants. Names and def-
initions can vary around the world potatoes are tubers,
which are the tips of underground stems that swell with
Cereals (grains) are the staple foods in large parts of the world, starch (a polysaccharide) and water. While potatoes are
supplying most of the energy and bulk in diets. In some regions, often classed as vegetables (in the USA, for instance), they
roots, tubers, or plantains are staple foods as well as or instead are grouped separately from non-starchy vegetables in this
of cereals (grains). These generalisations apply to practically Report. Sweet potatoes, sometimes called yams in North
all settled rural and most urban populations. Monotonous America, are a type of storage root rather than a tuber, but
poverty diets containing very high levels of these foods, par- true yams are starchy tubers. Cassava (manioc) and yucca
ticularly if refined, are low and sometimes inadequate in pro- are elongated roots, and sago is a starchy food made from
tein and other nutrients. Gathererhunter and pastoral the pith of some types of palm tree. Taro is cultivated for its
communities usually consume less of these starchy foods. Their edible leaves, as well as its starchy corm, which is similar to
nutrient content is variable, largely depending on the degree a tuber. Plantains are one of several fruits used as vegeta-
to which they are refined. bles: they grow on trees and look like bananas, but only a
Consumption of cereals, roots, and tubers in general gradu- small proportion of the starch is converted to sugar during
ally drops with industrialisation and urbanisation, and an the ripening process, which makes them similar to potatoes
increasing amount of wheat in particular is grown for animal to cook with.
feed. These foods are increasingly used as a basis for or ingre-
dients in processed products that are often energy-dense,
high in fats or sugars, and sometimes salt. In lower-income Box 4.1.1 Wholegrain and refined
countries, total population consumption of these foods cereals and their products
may amount to 6080 per cent of total energy, and in high- Many of the cereals (grains) that we consume are refined. Grains
income countries, usually to less than 30 per cent. Also see are first broken into pieces and then refined, sifting away the
Chapter 1. bran, germ and, usually, the aleurone layer. This removes most
Early reports concerned with nutritional deficiencies gen- of the fibre, oil, and B vitamins, as well as approximately 25 per
erally did not pay much attention to these foods and instead cent of the protein. Polishing, as often performed on rice,
removes additional nutrients. Many high-income countries
gave priority to energy- and nutrient-dense foods of animal
therefore fortify refined cereals, including flour, with B vitamins
origin, such as milk, eggs, and meat. Beginning in the 1970s, and iron. Wholegrain products generally contain the con-
interest in dietary fibre increased, following informal epi- stituents of the grain but, given the absence of an internation-
demiological findings that diets high in dietary fibre were ally accepted definition, intact grains are present to a variable
associated with a lower risk of a number of chronic diseases.1 extent. The extent to which the grain remains intact influences
By the 1990s, it was generally agreed that diets relatively physiological processes in the bowel and hence health.
high in cereals (grains) and other starchy staple foods, prefer- Cereal foods may be eaten in wholegrain form, although con-
ably relatively unrefined, protect against obesity, type 2 dia- sumption in refined forms, such as white rice, bread, or pasta,
betes, coronary heart disease, and perhaps also digestive is generally much more common, particularly in high-income
disorders.3 4 Evidence that such diets protect against cancer countries. Refined grains are considered easier than wholegrains
to cook and to chew; are light in colour which is attractive
of any site has been less impressive, but epidemiological
to many consumers; and also have a longer shelf-life than
studies tend not to distinguish between degrees of refine- wholegrain products, as the oil in bran goes rancid relatively
ment of cereals, roots, and tubers. quickly.
This section (4.1) includes cereal products and dietary Breakfast cereals, particularly in the United States and parts
fibre. It also includes contamination by aflatoxins, though of Europe, also account for a significant proportion of grain
this may also affect other plant foods (also see chapter 4.2). eaten. Many breakfast cereals, although based on grains
Non-starchy root vegetables such as carrots are included in (whole or refined), may contain substantial amounts of added
chapter 4.2. Micronutrients found in plant foods are includ- sugars. Grains are further processed to provide ingredients such
ed in chapter 4.2, though most of these are also found in as corn syrup, starch, or alcohol. They also form the basis of
cereals (grains), roots, tubers, and plantains. many animal feeds.
Processed grains have a higher glycaemic index than
unprocessed grains and, generally, the greater the degree of
processing, the greater the glycaemic index (box 4.1.3).

C H A P T E R 4 F O O D S A N D D R I N K S

Box 4.1.2 Foods containing dietary fibre Box 4.1.3 Glycaemic index and load

The concept of dietary fibre arose from observations of the low The degree to which different foods and meals raise blood glu-
prevalence of colon cancer, diabetes, and coronary heart disease cose depends not only on the nature of the carbohydrate, but
in parts of Africa amongst people whose diets were high in also on the characteristics of the foods consumed. Glycaemic
unrefined carbohydrates and whose stools were typically bulky, index (GI) is a measure of the degree to which a food raises
and often or sometimes semisolid. Considerable efforts have blood glucose compared with a standard food (usually glucose
been dedicated to characterising the dietary components of or white bread) under standard conditions. The test food must
what has come to be called dietary fibre that might confer contain the same amount of available carbohydrate (usually 50
health benefit. Naturally occurring dietary fibre is only derived grams) as the standard. GI was originally used as an aid to food
from plant foods. Pulses (legumes) and minimally processed choice in diabetes and has more recently been applied to peo-
cereals are particularly concentrated sources, but vegetables and ple without diabetes. The rise in blood glucose after consuming
fruits also contain significant amounts. Dietary fibre isolated a food depends not only on the GI but also on the amount of
from plant cell walls and in synthetic forms are increasingly food eaten. A related measure, glycaemic load (GL), takes into
entering the food supply. account both the GI of a food as well as the actual amount of
High intakes of dietary fibre, variously defined, have been carbohydrate consumed. The GL of a food may be measured
associated with reduced risk of cardiovascular disease as well as directly or calculated by multiplying the GI of a food by the num-
of some cancers. Definitions of dietary fibre vary. Some are ber of carbohydrate grams in a serving of the food.
based on chemical analyses of the components of plant cell Factors that influence the GI of a food include the type of car-
walls, such as non-starch polysaccharide, others on physiologi- bohydrate, how the food is processed or cooked, and the other
cal effects the carbohydrates that enter the large bowel hav- components present in the food (for example, fat, protein,
ing escaped digestion in the small intestine being defined as fibre). There is some relationship (inverse) between GI and fibre
dietary fibre. The latter definition includes oligosaccharides and content, although some foods high in fibre have a high GI and
resistant starch. The World Health Organization and Food and vice versa. Factors can affect GI by influencing speed of absorp-
Agriculture Organization have recently proposed that only poly- tion, for instance higher fat foods tend to have a low GI. The
saccharides which form part of plant cell walls should be regard- calculated GI of a mixed meal or whole diet has been shown in
ed as dietary fibre and that the health benefits of resistant some studies to correlate with the actual GI obtained by feed-
starch and oligosaccharides are more appropriately considered ing a mixed meal. Although the concept of GI has been contro-
separately. versial, the GI and GL of diets have predicted risks of type 2
diabetes and coronary heart disease and related biomarkers,
independent of dietary fibre, in prospective epidemiological
studies, suggesting that GI and GL may be useful markers.
The relevance to cancer might lie in the fact that the rise in
This section refers to starchy roots, tubers, and plantains. blood glucose after a meal is closely linked to that of insulin,
Carrots, beets, parsnips, turnips, and swedes are non-starchy which apart from its crucial role in carbohydrate and lipid
roots, and are classified as non-starchy vegetables in this metabolism, is also one of a family of important growth factors
Report. Also see chapter 4.2. (also see Chapter 2).

4.1.2 Composition

Cereals (grains)
The relative amounts of dietary constituents in cereals and Roots, tubers, and plantains
cereal foods depend largely on the degree of refinement and Roots and tubers are less concentrated stores of starch,
other forms of processing (box 4.1.1). Starch makes up although this accounts for almost all of their raw weight
about 70 per cent of the raw weight of the storage tissues apart from water. Starch content varies from around 1520
(endosperm) of unprocessed cereal grains. The outer parts per cent in sweet potatoes to 2530 per cent in cassava and
of the grain (the bran and the aleurone layer) contain non- yams, which translates into around 8095 per cent of the
starch polysaccharide, a type of carbohydrate that charac- dietary energy of these roots and tubers. Cooking sweet
terises dietary fibre (box 4.1.2). potatoes makes them taste sweet because an enzyme con-
Cereals also contain variable amounts of protein, oils, B verts as much as 75 per cent of the starch into maltose (a
vitamins, vitamin E and tocotrienols, iron, and various trace disaccharide). Roots and tubers eaten with the skin on are
elements, as well as phytochemicals, some of which, such high in dietary fibre. These foods are generally poor sources
as the antioxidants, are bioactive (box 4.1.2). The germ is of protein, so although protein deficiency is uncommon,
the embryonic part of cereal plants and contains oils, pro- populations that subsist on these foods, and do not eat pro-
teins, and fibre. Various cereals contain other specific com- tein-rich pulses (legumes), are at risk of deficiency, especially
ponents. Wheat contains gluten (a mixture of proteins). Rye children weaned on thin gruels made with these low-protein
has high levels of pentosans and oats contain beta-glucans, foods. They contain variable amounts of other nutrients.
both of which are non-starch polysaccharides, a character- Potatoes contain vitamin C, for example, and the orange
ising feature of dietary fibre. varieties of sweet potatoes contain carotenoids. Yams con-
Cereals (grains) and pulses (legumes) may be contami- tain many bioactive compounds and taro corms are high in
nated with aflatoxins. See box 4.1.4. vitamin B6, fibre, and manganese.

P A R T 2 E V I D E N C E A N D J U D G E M E N T S

Box 4.1.4 Aflatoxins

Mycotoxins are toxins produced by certain (grain), including wheat, rice, maize (corn), Levels of aflatoxin contamination tend
moulds or fungi. Although moulds that barley, and oats; and pulses (legumes) to be highest in countries where rates of
contaminate foods are usually destroyed notably peanuts. Nuts and seeds may also liver cancer are high, such as some African
by cooking temperatures, the toxins they be contaminated. Feedstuffs for farm ani- countries and South-East Asia, including
produce may remain. Aflatoxins are one mals may also be contaminated with afla- China. In general, rates are low in Europe,
type of mycotoxin. All naturally occurring toxins, which can then be secreted in milk but relatively high rates of contamination
aflatoxins are classified as human carcino- or accumulated in tissues. have on occasion been found in the USA.
gens (group 1) by the International Agency Aflatoxins, which are produced by Aflatoxin exposure levels are low in
for Research on Cancer; other mycotoxins, Aspergillus flavus and A. parasiticus, are Europe and Australia, higher in the USA,
such as fumonisins, are suspected carcino- most problematic in countries with hot, and high in many low-income countries.
gens.5 It is common to find co-contamina- damp climates and poor storage facilities. This is particularly the case in tropical and
tion by more than one species of myco- Under these conditions, foods may become subtropical regions where grains and nuts
toxin-producing fungus. In Europe, the contaminated with fungi and then accu- are stored for long periods under non-ideal
Joint FAO/WHO Expert Committee on Food mulate such toxins. Such foods are mar- conditions.
Additives and Contaminants recommends keted and consumed in the countries in Rates are reduced by inspection, use of
that aflatoxin concentrations in foods be which they are produced; they are also fungicides, and screening of imported
kept as low as possible.6 exported to neighbouring countries and foods. However, monitoring of levels of
The main foods that may be contami- intercontinentally. Aflatoxin contamina- aflatoxin contamination in low-income
nated by aflatoxins are all types of cereal tion is therefore a international issue. countries is generally lacking.

4.1.3 Consumption patterns it is used to make grits, cornmeal (used for polenta as well
as corn breads), corn flour, tortillas, tamales, and corn chips.
Cereals and grains It is also the basis of corn starch (a thickener), corn syrup
As societies moved to more settled, agricultural ways of life (a sweetner), and corn oils. Sweetcorn is also eaten as a veg-
1015 000 years ago, cereals became the main staple foods; etable, either on or off the cob. Rice is usually processed to
the types of cereal crops grown depended largely on climate remove the bran and aleurone layers, turning brown rice
and terrain. Wheat, barley, oats, and rye are traditionally sta- into white. It is also used to make flour (the basis for gluten-
ple foods for people living in the Middle East and Europe; free breads), rice powder, noodles, rice paper, rice milk,
and with rice in Asia; maize (corn) in the Americas; and Japanese mochi, and lao-chao (Chinese fermented rice).
sorghum and millet in Africa. But the market for cereals and Barley is used primarily in Asia (tsampa and miso soya paste)
their products is now global, although some, such as and in North Africa (soups, porridges, and flat breads).
sorghum, remain largely regional. Whole rye grains are milled and used to make bread in some
The importance of starchy staples in food systems and north and east European countries. Whole oats are made
diets is broadly connected to economic and industrial devel- into porridges and used in muesli and baked goods, such
opment. Both in higher-income countries and across the as biscuits. Fonio, millet, sorghum, teff, and triticale are
world, there has been a long-term decline in their con- traditional crops and staples in parts of Africa and Asia.
sumption. With increasing urbanisation in lower-income Many grains are also fermented to make alcoholic drinks (see
countries, wheat and maize are replacing traditional staple chapter 4.8.1).
foods. An important exception is Asia, where rice remains the
staple grain. Cereal cultivation and consumption tends to be Roots, tubers, and plantains
highest in most of Asia and lowest in Oceania, parts of Roots, tubers, and plantains are staple foods in some parts
Europe, and North America. of the world. For instance, populations in some regions
Globally, cereal foods provide more than 45 per cent of of sub-Saharan Africa, Latin America, and Oceania base
dietary energy; diets based on these foods tend to be bulky their diets on these foods. Globally, starchy roots provide
with a low energy density (see chapter 8.8.4). Cereals pro- around 5 per cent of dietary energy. Consumption is high-
vide more than 50 per cent of dietary energy in low-income est in the Pacific islands and parts of Africa, with cassava and
countries, but only around 30 per cent in high-income coun- yams providing more than 40 per cent of dietary energy in
tries. While grains contribute roughly 20 per cent of dietary Ghana. Potatoes can provide as much as 10 per cent of
energy in Australia, North America, and central Europe, they dietary energy in North America and Europe. Globally, plan-
can provide as much as 70 per cent in parts of Asia (main- tains provide less than 0.5 per cent of dietary energy, but
ly from rice). Although more wheat is grown than rice on a they are locally important in some African, Latin American,
global basis, much of it is used for animal feed. Rice is the and Caribbean countries, where they can provide more than
principal food for half of the worlds population. 15 per cent of dietary energy. Some populations do not rely
Cereals are very versatile once they have been processed on any of these foods for instance, pastoralist societies
from the raw grain. Wheat is mainly milled to make flour such as the Maasai hunters in East Africa, and the Inuit and
for bread, pastries, cakes, and pasta. Maize (corn) is a sta- other Arctic populations, maintain their traditional ways of
ple food in Latin America and parts of Asia and Africa, where life and diets.

C H A P T E R 4 F O O D S A N D D R I N K S

Dietary fibre Production, preservation, processing, preparation. Few stud-

Dietary fibre intake, measured as non-starch polysaccharides, ies distinguish between unrefined and refined cereals and
varies from 1013 grams (g)/day in Japan and the UK to their products. Many processed foods grouped as cereal
1520 g/day or more in Africa and India. Intake among indi- products, such as ready-to-eat breakfast cereals, are high in
viduals in a population may vary between 7 and 25 g/day.7 added sugars and sometimes salt. The ways in which cere-
als are processed, prepared, and consumed varies greatly in
different cultures.
4.1.4 Interpretation of the evidence

Interpretation of the evidence on any and all foods and 4.1.5 Evidence and judgements
drinks, their constituents, their methods of production,
preservation, processing and preparation, and other factors, The full systematic literature review (SLR) is contained on
with the risk of cancer, is never simple, for general and spe- the CD included with this Report.
cific reasons. Cereals (grains) General The evidence was too limited in amount, consistency, or
For general considerations that may affect interpretation of quality to draw any conclusions.
the evidence, see chapters 3.3 and 3.5, and boxes 3.1, 3.2,
3.6, and 3.7. Roots, tubers, and plantains
Relative risk (RR) is used in this Report to denote ratio The evidence was too limited in amount, consistency, or
measures of effect, including risk ratios, rate ratios, haz- quality to draw any conclusions.
ard ratios, and odds ratios. Foods containing dietary fibre Specific Colorectum
Some considerations specific to cereals (grains), roots, Sixteen cohort studies9-37 and 91 case-control studies inves-
tubers, and plantains are as follows. tigated dietary fibre and colorectal cancer. The Harvard pool-
ing project also analysed original data from 13 separate
Classification. Cereals is a broad classification. Different cohort studies.38
cereals have different nutritional composition and biological An association was apparent from many, though not all,
effects, as do different types of dietary fibre. Any effects cohort studies. Ten studies showed decreased risk when com-
of specific cereals or their constituents may not become paring high with low intake groups,14 19 21 25-29 33 34 which was
apparent. statistically significant in one (figure 4.1.1).28 Two reported
non-significant increased risk,36 39 one showed no effect on
Patterns and ranges of intake. Little evidence relates to roots, risk,30 and one reported no association.18 One study report-
or tubers other than potatoes, or plantains, some of which, ed non-significant decreased risk in women and non-
such as cassava (manioc) or yams, are staple foods in some significant increased risk in men23; one study reported
parts of the world. non-significant increased risk in women and non-significant
Terminology. Potatoes are usually (as here) defined as tubers, decreased risk in men.37 Meta-analysis was possible on eight
but are sometimes (in the USA especially) included with veg- studies, giving a summary effect estimate of 0.90 (95% con-
etables. Bananas, a significant item in many diets, may be fidence interval (CI) 0.840.97) per 10 g/day increment,
(as here) defined as a fruit, or else with plantains as a starchy with moderate heterogeneity (figure 4.1.2). A dose-response
food. There is no internationally agreed definition for dietary relationship was apparent from cohort data.
fibre (box 4.1.1). Because of the abundant prospective data from cohort
studies, case-control studies were not summarised.
Measurement. Non-starch polysaccharides are measured pre- The Harvard pooled analysis from 13 prospective cohort
cisely by the Englyst method,8 but there are fewer epidemi- studies (725 628 participants, followed up for 6 to 20 years,
ological data on non-starch polysaccharides specifically than 8081 colorectal cancer cases) gave a significant inverse asso-
for dietary fibre. The various analytical techniques used to ciation in the age-adjusted model (0.84, 95% CI
assess the fibre content of foods give widely different results. 0.770.92).38 However, the association was attenuated
and no longer statistically significant after adjusting for
Confounding. In high-income countries, high intakes of other risk factors (0.94, 95% CI 0.861.03). One compari-
wholegrain cereal products tend to go together with other son group was statistically significant when maximally
health-conscious dietary and other habits. Also there is pos- adjusted, others were not. Compared with dietary fibre
sible confounding between dietary fibre and other dietary intake of 10 to < 15 g/day, the pooled effect estimate was
constituents and in general with healthier dietary patterns 1.18 (95% CI 1.051.31) for less than 10 g/day (low com-
and ways of life. Data on dietary fibre come predominantly pared with moderate intake). All other measures were not
from dietary sources, that is, plant-based foods (also see box associated with risk of colorectal cancer. The pooled analy-
4.1.1 and chapter 4.2); therefore, no effect can be attributed sis therefore found that, after accounting for other dietary
to different types and sources of dietary fibre. risk factors, high dietary fibre intake was not associated

P A R T 2 E V I D E N C E A N D J U D G E M E N T S

Figure 4.1.1 Dietary fibre and colorectal cancer; Figure 4.1.2 Dietary fibre and colorectal cancer;
cohort studies cohort studies

Relative risk (95% CI) Relative risk (95% CI)

Wu 1987 Men 1.19 (0.602.11) Heilbrun 1989 Men 0.77 (0.481.24)

Wu 1987 Women 0.64 (0.371.11) Bostick 1993 Women 0.88 (0.771.01)

Heilbrun 1989 Men 0.71 (0.451.13) Fuchs 1999 Women 0.99 (0.831.18)
Giovannucci 1994 Men 1.08 (0.681.71) Terry 2001 Women 1.08 (0.841.39)
Steinmetz 1994 Women 0.80 (0.401.91) Colbert 2001 Men 1.02 (0.881.18)
Gaard 1996 Men 0.82 (0.461.46)
Konings 2002 Men 0.95 (0.841.07)
Kato 1997 Women 0.95 (0.791.24)
Konings 2002 Women 0.85 (0.731.00)
Pietinen 1999 Men 1.00 (0.681.58)
Higginbotham 2004 Women 0.83 (0.611.13)
Soneham 2000 Women 0.96 (0.701.32)
Norat 2005 Men 0.77 (0.710.83)
Bingham 2003 0.75 (0.500.95)
Norat 2005 Women 0.89 (0.820.97)
COL00535 Women 0.94 (0.701.26)
Summary estimate 0.90 (0.840.97)
IARCIM 1977 Men 0.92 (0.641.32)
IARCIM 1977 Women 0.86 (0.521.42)
Baron 1997 Women 0.79 (0.451.38)
0.2 0.5 1 2 5
Relative risk, per 10 g/day

0.2 0.5 1 2 5
Relative risk, highest vs lowest exposure category

with a reduced risk of colorectal cancer. The nine case-control studies produced 13 independent
Fibre exerts several effects in the gastrointestinal tract but effect estimates. Of these, 11 estimates were of decreased
the precise mechanisms for its probable protective role are risk,50-53 55 56 58 61 which were statistically significant in eight.
not clearly understood. Fibre dilutes faecal contents, decreas- One estimate indicated no effect on risk54 and one other gave
es transit time, and increases stool weight.40 Fermentation non-significant increased risk.62 The data were most
products, especially short-chain fatty acids, are produced by consistent when stratified for adenocarcinomas; of six
the gut flora from a wide range of dietary carbohydrates that studies, five reported significant decreased risk; results were
reach the colon. Short-chain fatty acids, particularly less consistent for squamous cell carcinoma. All studies
butyrate, can induce apoptosis and cell cycle arrest, and pro- were adjusted for alcohol and smoking except one, which
mote differentiation. Fibre intake is also strongly correlated was adjusted for alcohol but not smoking.50
with intake of folate. The ecological studies were inconclusive. Neither was sta-
tistically significant, with one in the direction of increased
A clear dose-response relationship is apparent from and the other of decreased risk.
generally consistent cohort studies, supported by There is no evidence of a plausible biological mechanism
evidence for plausible mechanisms, but residual through which dietary fibre reduces the risk of oesophageal
confounding could not be excluded. Foods containing cancer. It is not possible to conclude whether an as yet
dietary fibre probably protect against colorectal cancer. unknown mechanism is responsible for an apparent reduc-
tion in risk, or whether it is due to other components found
The Panel is aware that since the conclusion of the SLR, six in the vegetables and fruits that contain dietary fibre.
cohort studies41-46 and one case-control study47 have been pub-
lished. This new information does not change the Panel judge- There is limited evidence from sparse and inconsistent
ment (see box 3.8). case-control studies only, suggesting that foods
containing dietary fibre protect against oesophageal
Oesophagus cancer.
One cohort study,48 nine case-control studies,49-58 and two
ecological studies59 60 investigated dietary fibre and cancer Aflatoxins
of the oesophagus. (Also see box 4.1.4; chapter 4.9; and chapter 7.8). There are
There was some evidence of an association between dietary two approaches to measuring aflatoxin intake. The first uses
fibre and reduced oesophageal cancer risk. The single cohort local food tables to estimate exposure to aflatoxins from diet.
study reported decreased risk when comparing high with low The second approach uses biomarkers of exposure. These are
intakes, with an effect estimate of 0.50, though no assess- derived from knowledge of aflatoxin metabolism. In humans,
ment of statistical significance was included.48 metabolised products of aflatoxins can be detected in blood,

C H A P T E R 4 F O O D S A N D D R I N K S

urine, or breastmilk. Biomarkers of exposure are more accu- studies and some of the case-control studies.
rate and precise. The areas in the world where there is considerable afla-
toxin contamination of foods coincide with the areas where
Liver primary liver cancer rates are high. The epoxide product of
Five cohort studies63-70 and seven case-control studies71-79 aflatoxin AFB1 is known to be genotoxic and is formed in the
assessed associations between biomarkers of exposure to liver.80 It damages DNA, causing G:C base pairs to become
aflatoxin and hepatocellular carcinoma. T:A. GST enzymes can repair this damage with varying effi-
The cohort studies used a variety of different biomarkers ciency between genotypes. Recent studies have shown that
for exposure to aflatoxin, some in blood and some in urine. aflatoxins can damage the p53 gene, which is an important
Despite this variety, all five studies reported increased risk regulator of normal growth.67 Damage to p53 DNA can lead
for the highest levels when compared to the lowest, and all to increased proliferation of abnormal cells and formation
of these reported at least one measure that resulted in a sta- of cancers.
tistically significant increased risk (figure 4.1.3). Studies that The synergistic effect of hepatitis virus infection and afla-
adjusted for hepatitis virus infection tended to show the toxin exposure might be explained by hepatitis virus increas-
greater effects.65 66 There is some evidence of an interaction ing the production of the enzyme (CYP1A2) that produces
whereby the risk is increased by a multiplicative effect if afla- the genotoxic metabolite of aflatoxin.81 It is also possible that
toxin exposure is combined with hepatitis infection. One the hepatitis virus increases the number of G:C to T:A trans-
study showed that people with hepatitis virus antibodies and versions, or that it inhibits nucleotide repair, or that it acts
biomarkers of aflatoxin exposure had a higher risk than those as a tumour promoter.
with hepatitis virus antibodies alone, with an effect estimate
of 10.0 (95% CI 1.660.9).65 The evidence is ample and consistent and is supported
There is evidence from some of the cohort studies for inter- by strong evidence for mechanisms operating in
action with glutathione-S-transferase (GST) genotype.63 64 humans. A dose response is apparent from both cohort
GST is an enzyme involved in the metabolic pathway that and case-control studies. The evidence that aflatoxins
detoxifies aflatoxins. Different genotypes show varying effi- and aflatoxin-contaminated foods are a cause of liver
ciencies at this task. Two genes (GSTT1 and GSTM1) were cancer is convincing.
assessed separately. For each, it is possible to have a posi-
tive or negative genotype. In each case, a negative genotype
increases risk of hepatocellular carcinoma when exposed to 4.1.6 Comparison with previous report
aflatoxins. There is clear, consistent evidence that
GSTM1/GSTT1 positive genotypes protect against the The previous report concluded that dietary fibre/non-starch
increased risk of liver cancer from hepatitis infection com- polysaccharides possibly protect against cancers of the pan-
bined with aflatoxin exposure, which supports a causal role creas, colorectum, and breast. The previous report also con-
for aflatoxins in hepatocellular carcinoma. cluded that wholegrain cereals possibly decrease the risk of
Four case-control studies showed statistically significant stomach cancer and that refined cereals possibly increase the
increased risk for the highest levels of biomarkers when com- risk of oesophageal cancer.
pared to the lowest.71 74 78 79 Two studies showed no effect Since the mid-1990s, evidence for a protective effect of
on risk.73 77 One study showed a non-significant decreased dietary fibre against colorectal and oesophageal cancer risk
risk.72 Heterogeneity may be explained by the diversity in has become somewhat stronger. The finding of the previous
methods of exposure assessment. report, suggesting that the degree of refinement (other than
A dose-response relationship is apparent from most cohort relative amounts of dietary fibre) may be a factor in modi-
fication of the risk of some cancers, was not found.
The previous report classified bananas as plantains. Here
they are classified as fruits. The previous report considered
Figure 4.1.3 Aflatoxins and liver cancer;
cohort studies dietary fibre separately from cereals (grains) and other plant
foods. Here, dietary fibre is considered in the context of cere-
Relative risk (95% CI) als (grains) and other plant foods.

Qian 1994 Men 5.00 (2.1111.85)

Wang 1996 3.80 (1.1112.96) 4.1.7 Conclusions

Yu 1997 Men 6.00 (1.2229.49)

Sun 1999 Men 4.52 (1.5713.01) The Panel concludes:

The direct evidence that cereals (grains), roots, tubers, or
1 2 5 6
plantains affect the risk of any cancer remains unimpressive.
Relative risk, highest vs lowest exposure category
However, foods containing dietary fibre probably protect
against colorectal cancer; and there is limited evidence sug-
gesting that such foods protect against oesophageal cancer.
Dietary fibre is mostly found in cereals, roots and tubers, and
also in vegetables, fruits, and pulses (legumes) (see chapter

P A R T 2 E V I D E N C E A N D J U D G E M E N T S

4.2). All of these are highest in dietary fibre when in whole

or minimally processed forms.
Foods high in dietary fibre may also have a protective
effect indirectly because they are relatively low in energy
density. See chapters 6.1, 7.3, 7.9, and 8 for discussion of
the role of energy density in weight gain, overweight, and
obesity, and of weight gain, overweight, and obesity in the
risk of some cancers, including those of the oesophagus and
The evidence that foods contaminated with aflatoxins are
a cause of liver cancer is convincing. Cereals (grains) and
peanuts (see chapter 4.2) are the foods most commonly
infested by these fungal toxins. Contamination is most wide-
spread in countries with hot, damp climates and poor stor-
age facilities.

C H A P T E R 4 F O O D S A N D D R I N K S

4.2 Vegetables, fruits, pulses (legumes),

nuts, seeds, herbs, spices
Vegetables and fruits are generally low in energy density oesophagus, lung, and stomach. There is limited evidence
(with a few exceptions) and, when consumed in variety, suggesting that fruits also protect against cancers of the
are sources of many vitamins, minerals, and other nasopharynx, pancreas, liver, and colorectum. There is
bioactive compounds (phytochemicals). Many non-starchy limited evidence suggesting that chilli is a cause of
vegetables, including salad vegetables and fruits, may be stomach cancer.
eaten raw and may also be cooked. Pulses (legumes) are Fruits and non-starchy vegetables are generally low
high in protein. Traditional diets all over the world energy-dense foods. For a discussion of the effect of such
combine cereals (grains) with pulses (legumes) and, in foods and drinks on weight gain, overweight, and obesity,
this way, ensure sufficient protein of adequate quality, and the role of weight gain, overweight, and obesity in the
usually with small amounts of animal foods. Nuts and risk of some cancers, see Chapters 6, 7, and 8.
seeds are concentrated sources of numerous Evidence that vegetables and fruits protect against some
micronutrients and of essential fatty acids. All these foods cancers is supported by evidence on foods containing
are sources of dietary fibre. Many herbs and spices have various micronutrients, found especially in vegetables,
potent pharmacological as well as culinary properties. fruits, and pulses (legumes), and nuts and seeds, as well
Consumption of vegetables and fruits is very variable: as in cereals, roots, tubers, and other plant foods. Foods
high around the Mediterranean littoral and some tropical containing folate probably protect against pancreatic
countries; low in many low-income countries, including cancer, and there is limited evidence suggesting that these
some in which fruits are abundant. Consumption of pulses foods also protect against oesophageal and colorectal
(legumes) is also very variable: beans and chickpeas and cancers. Foods containing carotenoids probably protect
their products are basic foods in a number of Latin against cancers of the mouth, pharynx, and larynx, and
American, Middle Eastern, and Asian countries, but pulses also lung cancer. Foods containing the carotenoid beta-
are insignificant in typical North American and most carotene probably protect against oesophageal cancer; and
European diets. Consumption of nuts, seeds, herbs, and foods containing lycopene probably protect against
spices also varies. Traditional Middle Eastern and Indian prostate cancer. Foods containing vitamin C probably
cuisines use a great variety of herbs and spices; garlic, protect against oesophageal cancer. There is limited
usually classified as a herb, is consumed in remarkable evidence suggesting that foods containing quercetin
quantities in some countries. protect against lung cancer.
In general, the Panel judges that findings from cohort Evidence also relevant to chapter 4.1 is grouped here.
studies conducted since the mid-1990s have made the Foods containing selenium (also found in animal foods)
overall evidence, that vegetables or fruits protect against probably protect against prostate cancer; and there is
cancers, somewhat less impressive. In no case now is the limited evidence suggesting that they protect against
evidence of protection judged to be convincing. However, stomach and colorectal cancers. There is limited evidence
in a substantial number of cases, a judgement of probable suggesting that foods containing pyridoxine protect
is justified. Evidence on legumes (pulses), nuts, seeds, and against oesophageal and prostate cancers; and that foods
(with two exceptions) herbs and spices remains containing vitamin E protect against oesophageal and
insubstantial. prostate cancers.
The strongest evidence, here corresponding to
The Panel judges as follows: judgements of probable, shows that non-starchy
Non-starchy vegetables probably protect against cancers of vegetables and also fruits probably protect against cancers
the mouth, pharynx, and larynx, and those of the of the mouth, larynx, pharynx, oesophagus, and stomach,
oesophagus and stomach. There is limited evidence and that fruits also probably protect against lung cancer;
suggesting that they also protect against cancers of the and that allium vegetables, and garlic specifically,
nasopharynx, lung, colorectum, ovary, and endometrium. probably protect against stomach cancer. The case that
Allium vegetables probably protect against stomach vegetables, fruits, and pulses (legumes) may be protective
cancer. Garlic (an allium vegetable, commonly classed as a against cancers of some sites is supported by evidence on
herb) probably protects against colorectal cancer. There is foods containing micronutrients found in these and other
limited evidence suggesting that carrots protect against plant foods. Thus, foods containing carotenoids probably
cervical cancer; and that pulses (legumes), including soya protect against cancers of the mouth, pharynx, larynx, and
and soya products, protect against stomach and prostate lung; foods containing beta-carotene and also vitamin C
cancers. Fruits in general probably protect against cancers probably protect against oesophageal cancer; foods
of the mouth, pharynx, and larynx, and those of the containing selenium and also lycopene probably protect

P A R T 2 E V I D E N C E A N D J U D G E M E N T S



In the judgement of the Panel, the factors listed below modify the risk of cancer. Judgements are graded according to the strength of the evidence.


Exposure Cancer site Exposure Cancer site


Probable Non-starchy vegetables1 Mouth, pharynx, larynx

Allium vegetables1 Stomach
Garlic1 Colorectum
Fruits1 Mouth, pharynx, larynx
Foods containing folate2 Pancreas
Foods containing Mouth, pharynx, larynx
carotenoids2 Lung
Foods containing Oesophagus
Foods containing Prostate
lycopene2 3
Foods containing Oesophagus
vitamin C2 4
Foods containing Prostate
selenium2 5

Limited Non-starchy vegetables1 Nasopharynx Chilli1 Stomach

suggestive Lung
Carrots1 Cervix
Fruits1 Nasopharynx
Pulses (legumes)7 Stomach
Foods containing folate2 Oesophagus
Foods containing Oesophagus
pyridoxine2 8
Foods containing Oesophagus
vitamin E2 6 Prostate
Foods containing Lung
selenium2 5 Stomach
Foods containing Lung

effect on risk Foods containing beta-carotene9: prostate; skin (non-melanoma)

1 Judgements on vegetables and fruits do not include those preserved by salting and/or pickling.
2 Includes both foods naturally containing the constituent and foods which have the constituent added (see chapter 3.5.3).
3 Mostly contained in tomatoes and tomato products. Also fruits such as grapefruit, watermelon, guava, and apricot.
4 Also found in some roots and tubers notably potatoes. See chapter 4.1.
5 Also found in cereals (grains) and in some animal foods. See chapters 4.1 and 4.3.
6 Also found in plant seed oils. See chapter 4.5.
7 Including soya and soya products.
8 Vitamin B6. Also found in cereals. See chapter 4.1.
9 The evidence is derived from studies using supplements and foods containing beta-carotene: see chapter 4.10.

For an explanation of all the terms used in the matrix, please see chapter 3.5.1, the text of this section, and the glossary.

C H A P T E R 4 F O O D S A N D D R I N K S

against prostate cancer; and foods containing folate here. Chapter 4.1 includes dietary fibre, only found naturally
probably protect against pancreatic cancer. Also see in plant foods. Chapter 4.1 also includes aflatoxins, which
chapter 4.1 for the evidence that foods containing dietary also contaminate pulses (legumes), notably peanuts, nuts
fibre, found in plant foods (particularly when in whole or and seeds, and other plant foods. The micronutrients includ-
relatively unprocessed forms), probably protect against ed here, as contained in vegetables, fruits, pulses (legumes),
colorectal cancer. nuts and seeds, are also found in other plant foods, and some
also in animal foods.

Vegetables and fruits (including berries), nuts and seeds, and

herbs and spices, where they grow and can be cultivated, 4.2.1 Definitions, sources
have always been part of human diets. Gathererhunters and
pastoral peoples probably consumed more than relatively Vegetables and fruits are defined in this Report by their culi-
impoverished urban dwellers: for them, vegetables were the nary use, and are grouped for discussion below as vegeta-
main sources of many vitamins, and fruits were a main bles and fruits, pulses (legumes), nuts and seeds, and herbs,
source of energy, from sugar (also found in wild honey). spices, and condiments.
They are consumed abundantly as part of many long-estab-
lished traditional cuisines, around the Mediterranean littoral, Vegetables and fruits
the Middle East, in many Asian countries, and the Pacific Vegetables are the edible parts of plants, usually including
islands, where substantial amounts of meat, dairy products, fungi. Typical examples include cultivated or gathered
and other animal foods are traditionally consumed only occa- leaves, roots, stalks, bulbs, and flowers. Some foods are culi-
sionally. In contrast, monotonous poverty diets include few nary vegetables but are classified botanically as fruits; these
of these foods. include cucumbers, peppers, squash, and tomatoes. Non-
Globally, consumption of these foods is lower than now starchy vegetables are included here, while starchy root veg-
generally recommended. Vegetables and fruits are sometimes etables are considered in chapter 4.1. Non-starchy vegetables
seen as relatively expensive. Well stocked supermarkets usu- can be further divided into green, leafy vegetables, such as
ally now display a variety of local and imported fresh veg- spinach and lettuce; cruciferous vegetables (the cabbage
etables and fruits, although supplies in smaller stores are family), for example, bok choy, broccoli, cabbage, and water-
more variable. Consumption of fresh vegetables and fruits in cress; and allium vegetables, such as onions, garlic, and
many tropical countries in Africa and Latin America is low: leeks.
on average people in Brazil, for example, consume roughly A fruit is the seed-containing part of the plant; but only
the same as people in Britain. The explanation may be that those that are eaten as fruits are included in the culinary def-
in Africa, many rural communities are obliged to grow cash inition, for example, apples, bananas, berries, figs, grapes,
crops that displace gardens, and that in Latin America knowl- mangoes, and melons. This also includes citrus fruits such
edge of the value and pleasure of many indigenous as oranges, grapefruits, lemons, and limes; and also dried
vegetables and fruits has been lost. Many programmes in fruits, such as apricots, figs, and raisins.
tropical countries are now dedicated to regaining this knowl-
edge.1 Pulses (legumes)
Even before the discovery of vitamins as essential nutri- Leguminous plants produce their fruits as pods and are con-
ents beginning in the early 20th century, vegetables and sidered here separately. The dried, edible seeds of this fam-
fruits have been recommended as protective foods. Early ily are often called pulses, although this term is used
reports concerned with nutritional deficiencies paid less interchangeably with legumes. They include beans, lentils,
attention to pulses (legumes), nuts, and seeds, even though peas, and peanuts (groundnuts). The dried forms, which
these plant foods contain protein, and nuts and seeds are have matured and dried on the plant, are eaten most wide-
nutrient- and also energy-dense, perhaps because they are ly. But some varieties are eaten as a green vegetable, such
not much consumed in the countries where most such as peas; the pods are sometimes eaten like this too, for exam-
reports were compiled. Instead, as already mentioned, pri- ple, green beans and runner beans. Some legumes can also
ority was given to energy- and nutrient-dense foods of ani- be sprouted (germinated) and eaten, such as beanspouts.
mal origin. By the 1980s, most reports concerned with
prevention of chronic diseases recommended relatively high Nuts and seeds
intakes of vegetables and fruits and sometimes also pulses Nuts are edible seeds surrounded by a tough, dry shell. This
(legumes), either because these foods were seen as nour- definition includes true nuts (such as hazelnuts and chest-
ishing substitutes for energy-dense fatty or sugary foods, or nuts), as well as seeds that most people think of as nuts
else because they were identified as positively protective (including Brazil nuts, macadamia nuts, and cashews). Other
against cardiovascular disease.2 Evidence that vegetables seeds commonly eaten include sunflower, sesame, pumpkin,
and fruits might be protective against some cancers emerged and poppy seeds. Nuts and seeds are processed for their oil,
in the 1990s.3 A common recommendation has been for at ground into pastes, used as ingredients, or eaten raw or
least five portions (or at least 400 g) of vegetables and fruits roasted as snack foods. Cereals (grains) are also the seeds
a day.4 of plants, but these are discussed separately in this Report
Non-starchy root vegetables such as carrots are included (see chapter 4.1). Seeds, like nuts, have a relatively high oil

P A R T 2 E V I D E N C E A N D J U D G E M E N T S

Box 4.2.1 Micronutrients and other bioactive compounds and cancer risk

Vegetables, fruits, pulses (legumes), nuts, As a rule of thumb, the greater the inten- breakfast cereals also contain this vitamin.
and seeds are sources of a wide variety of sity of the colour of the fruit or vegetable, Pyridoxine is one of a group of water-
micronutrients and other bioactive com- the more beta-carotene it contains. soluble compounds collectively known as
pounds. Foods containing several of these The most concentrated source of vitamin B6. This vitamin is involved in neu-
constituents have been identified in the sys- lycopene is tomatoes, but it is also present rotransmitter synthesis, red blood cell for-
tematic literature reviews, on which this in watermelon, red (bell) peppers, pink or mation and function, niacin (vitamin B3)
chapter is based, as being associated with red grapefruit, pink-fleshed guava, and formation, steroid hormone function, and
cancer risk. These are carotenoids (includ- persimmons. nucleic acid synthesis (also see chapter
ing beta-carotene and lycopene), folate, The B-vitamin folate is a family of com- Food sources include bananas,
vitamin C, vitamin D, vitamin E, quercetin, pounds essential for human health. Folic fish, poultry, liver, potatoes eaten with the
pyridoxine, and selenium. Mechanisms by acid, the synthetic form, is used to fortify skin, green, leafy vegetables, beans, pulses
which they might affect cancer risk are dis- manufactured cereal products, spreads, (legumes), nuts, wholegrains, and fortified
cussed in chapter 4.2.5. However, it is not and, in some countries, flour or grains. breakfast cereals.
possible to ascribe the association between Folates are involved in a number of meta- Selenium is a mineral element that
these foods and lower cancer risk to a bolic pathways, especially in the synthesis occurs in different chemical forms. It is
causal effect of specific compounds with of purines and pyrimidines, which are toxic in large amounts, but is essential in
confidence, as each food contains a com- important for DNA synthesis and cell repli- the diet at trace levels. It is present at vary-
plex mixture of different constituents, all of cation (also see chapter Sources of ing concentrations in different soils; and
which might also contribute to any effect. dietary folate include liver, beans, spinach, since plants take up selenium from the soil,
Carotenoids are found in varying con- broccoli, romaine lettuce, chicory, oranges, these levels determine the amount present
centrations in all vegetables, particularly and papaya. in vegetables. Thus selenium deficiency is
those that are red or orange. They are a Vitamin C (ascorbic acid) is a water-solu- more prevalent in regions where the soil
family of more than 600 fat-soluble ble vitamin. Humans, like a small number of selenium content is low. Selenium is a com-
red/orange pigments that comprise xan- other animals, cannot synthesise vitamin C, ponent of the amino acids selenocysteine
thophylls (such as lutein) and carotenes so it is an essential part of diets. Vitamin C and selenomethionine, which are inte-
(such as alpha- and beta-carotene, and is essential for collagen synthesis and also grated into proteins to form selenopro-
lycopene). Some carotenoids, most impor- has antioxidant activity. Severe deficiency teins. Selenoproteins include antioxidant
tantly beta-carotene, can be converted by causes scurvy. It is added to many foods, enzymes such as glutathione peroxidases,
the body to retinol and are sometimes including bread and soft drinks, in small thioredoxin reductase, which is important
called pro-vitamin A carotenoids. These amounts as an antioxidant preservative. for DNA synthesis, and iodothyronine deio-
compounds tend to be the main dietary Natural dietary sources are vegetables, dinase, which is important for the synthe-
source of vitamin A in low-income coun- tubers, and fruits, including red/yellow sis of thyroid hormones.16 Dietary sources
tries. (bell) peppers, kiwi fruits, broccoli, papaya, of selenium include brazil nuts, fish, whole-
Only about half of the 50 or so citrus fruits, strawberries, and potatoes, but grains, wheatgerm, and sunflower seeds.
carotenoids in human diets can be absorb- it is destroyed by heat or contact with the Quercetin is a flavonoid, which is a type
ed. They have antioxidant and other bioac- air (for instance, when vegetables are of polyphenol; it is not an essential dietary
tivities that are discussed in chapter 4.10. chopped), or lost into cooking water. component. Many studies in cultured cells
Sources of carotenoids include spinach, Vitamin E is a fat-soluble vitamin and a and animals suggest that quercetin has
kale, butternut squash, pumpkin, red (bell) potent antioxidant that occurs as eight dif- antioxidant activity, which could give rise
peppers, carrots, tomatoes, cantaloupe ferent forms: alpha- and gamma-toco- to a range of biological activities, including
melon, and sweet potatoes. pherol are the most common. The most reducing inflammation (also see chapter
Beta-carotene is found in yellow, orange, important dietary sources of vitamin E are Quercetin is found in apples,
and green fruits and green, leafy vegeta- vegetable oils such as palm, sunflower, green and black tea, onions, raspberries,
bles including carrots, spinach, lettuce, corn, soya bean, and olive oils. Nuts, sun- red wine, red grapes, citrus fruits, leafy,
tomatoes, sweet potatoes, broccoli, can- flower seeds, and wheatgerm are also green vegetables, cherries, elderberries,
taloupe melon, oranges, and winter squash sources. Wholegrains, fish, peanut butter, broccoli, blueberries, cranberries, and
(pumpkin). green, leafy vegetables, and fortified bilberries.

content, and the oils produced from them are considered in plants are used as herbs or spices, such as the leaves (sage,
chapter 4.4. bay, or basil), stems (ginger, lemongrass), bark (cinnamon),
rhizomes (ginger), roots (horseradish), flower buds (cloves),
Herbs, spices, and condiments stamens (saffron), seeds (mustard, cumin), kernels (nut-
Herbs and spices, which are generally used to flavour or pre- meg), and fruits (peppers).
serve foods, are of plant origin, although a very small num- A condiment is a substance that adds taste to other foods;
ber of animal products are classed as spices (such as the term is often used for sauces added at the table, which
ambergris). Definitions of herbs and spices vary, but herbs are usually of plant origin. Examples include vinegars,
are usually the fresh or dry leaves or whole plant, while ketchups, chutneys, harissa, mustard, and soy sauce. Salt is
spices are produced from other parts of the plant, such as neither a herb nor a spice, although it is used as a condiment
the seeds, and are usually dried.5 Many different parts of (see chapter 4.5).

C H A P T E R 4 F O O D S A N D D R I N K S

Box 4.2.2 Phytochemicals methods. For instance, the outer leaves of lettuces can have
higher levels of some micronutrients than the inner leaves;
Plants contain a wide range of biologically active compounds, and harvested, unripe fruit that ripens in transit may have
some of which are known as phytochemicals. There may be as lower levels of nutrients than fruits ripened on the plant
many as 100 000 different compounds, which determine par- (box 4.2.1).6
ticular properties in plants, and in the fruits and vegetables they Vegetables and fruits contain vitamins, minerals, dietary
produce, such as flavour and colour. Phytochemicals are classi- fibre, and other bioactive compounds, such as phytochemi-
fied according to their chemical structure and functional char- cals (box 4.2.2). This is a collective term for a variety of plant
acteristics, and include salicylates, phytosterols, saponins,
components that often perform important functions in the
glucosinolates, polyphenols, protease inhibitors, monoterpenes,
plant, such as providing colour, flavour, or protection, but
phytoestrogens, sulphides, terpenes, and lectins.
It is widely believed that the health benefits of diets high in are not essential in the human diet. They include salicylates,
fruits and vegetables are likely to be due partly to the presence flavonoids, glucosinolates, terpenes, lignans, and
of phytochemicals. For instance, several act as antioxidants, pre- isoflavones. All of these groups of compounds have been
venting oxidative damage to cells, proteins, and DNA. It is like- shown either in humans or in laboratory experiments to have
ly that other bioactive phytochemicals have yet to be identified, potentially beneficial health effects when they are included
and those that are known may have additional properties in the in diets. However, the bioavailability of these compounds
body that are not yet understood. But it is thought that nutri- is variable (box 4.2.3) and their ultimate heath effects
ents, phytochemicals, and other, as yet unknown, bioactive com- uncertain.
ponents act together to influence physiological responses.
Plant cell walls are the main source of dietary fibre, and
Although many phytochemicals are bioactive, they are not
all whole fruits and vegetables (but not their juices) contain
essential in the diet and there is no daily requirement, so they
are not classed as nutrients. Humans have developed tastes for varying amounts of fibre (box 4.2.4). Most vegetables and
some phytochemicals, such as the hot flavours of mustard oil, fruits are low energy-dense foods, although there are excep-
bitter alkaloids, and irritating capsaicins. There is genetically tions, for example, avocados, nuts, and seeds.
inherited variation in sensitivity to some tastes, for example, the Some families of fruits and vegetables have characteristic
bitter taste of isothiocyanates in cruciferous vegetables such as components that may confer a particular health benefit (or
cabbage. risk) to the whole family. For instance, cruciferous vegeta-
bles are sources of glucosinolates and their products isoth-
iocyanates and indoles. Allium vegetables and others, such
as chicory and Jerusalem artichokes, store energy as inulin
4.2.2 Composition (chains of fructose sugars) rather than starch (chains of glu-
cose sugars). The body cannot digest inulin, which is called
Vegetables and fruits a prebiotic a substance that is claimed to have health ben-
The composition of fruits and vegetables depends both on efits by promoting the growth of certain types of gut bacte-
species and on subtype, as well as on the environmental, ria. Allyl sulphides and allicin in garlic are distinctive flavour
farming, production, and storage conditions. These include molecules that give vegetables of the onion family their
factors such as sun exposure, soil quality, agricultural prac- sting (box 4.2.3). Green, leafy vegetables are sources of
tices, harvesting time, ripeness, length of time between har- folate, and tomatoes have high levels of lycopene. All of
vest and consumption, and preservation and preparation these components, as well as other phytochemicals (box

Box 4.2.3 Preparation of vegetables and nutrient bioavailability

While some vegetables, often termed carotenes is increased by cooking and processed. Peeling and chopping garlic
salad vegetables, are commonly eaten pureeing vegetables, particularly by releases an enzyme, alliinase, which is
raw, many are cooked before they are adding oil, because these compounds are known to promote the formation of some
eaten. In most cases, whether a vegetable fat soluble.13 Similarly, processing tomatoes sulphur compounds that are not only odor-
is eaten raw depends on personal choice. increases the bioavailability of lycopene, iferous but may provide some health ben-
Most forms of cooking reduce the total another carotenoid: it is four times more efits. Heating garlic without peeling
nutrient content of vegetables, although bioavailable from tomato paste than from inactivates this enzyme and has been
the degree to which this happens varies fresh tomatoes. Thus processed tomato found to substantially reduce or eliminate
between nutrients and with cooking meth- products such as pasteurised tomato juice, the active properties. If garlic is peeled or
ods. However, cooking also increases the soup, sauce, and ketchup provide the most chopped and allowed to stand for 1520
bioavailability of some nutrients.12 bioavailable lycopene. Cooking and crush- minutes, the active agents that are formed
Therefore, although raw vegetables have ing tomatoes (as in the canning process) are not destroyed by normal cooking pro-
higher amounts of nutrients overall, the and including them in oil-rich dishes (such cedures.14
body may absorb more of a nutrient from as pasta sauce or pizza) greatly increases The ways in which vegetables and fruits
the cooked vegetable. lycopene absorption from the digestive are produced and stored may affect nutri-
For instance, carotenoid absorption in tract. ent levels as much as cooking, or more. For
the small intestine is relatively inefficient The biological response to garlic can also example, nutrient levels tend to fall rapid-
(550 per cent); the bioavailability of be influenced by the way that it is ly after harvest.

P A R T 2 E V I D E N C E A N D J U D G E M E N T S

Nuts and seeds

Box 4.2.4 Foods containing dietary fibre Other seeds and nuts are also relatively high in protein and
The concept of dietary fibre arose from observations of the low fat; some contain as much as 60 g fat per 100 g. They are
prevalence of colon cancer, diabetes, and coronary heart disease therefore energy-dense foods (see Chapter 8), as well as
in parts of Africa amongst people whose diets were high in being nutrient-dense. Weight-for-weight, nuts provide more
unrefined carbohydrates and whose stools were typically bulky, calories than either meat or cereals (grains), although chest-
and often or sometimes semisolid. Considerable efforts have nuts are the exception as they are relatively low in fat. Most
been dedicated to characterising the dietary components of nuts contain mainly monounsaturated fatty acids, although
what has come to be called dietary fibre that might confer the exceptions are coconuts, which contain a high propor-
health benefit. Naturally occurring dietary fibre is only derived
tion of saturated fatty acids, and walnuts and pecans, which
from plant foods. Pulses (legumes) and minimally processed
cereals are particularly concentrated sources, but vegetables and
contain mostly polyunsaturated fatty acids (see chapter
fruits also contain significant amounts. Dietary fibre isolated 4.5.2). Nuts and seeds are high in dietary fibre (box 4.2.4),
from plant cell walls and synthetic forms are increasingly enter- especially when they are eaten with their skins or hulls; the
ing the food supply. fibre content is typically 511 g per 100 g. Nuts and seeds
High intakes of dietary fibre, variously defined, have been are also high in vitamins and minerals, particularly the B vit-
associated with reduced risk of cardiovascular disease as well as amins, vitamin E, and folate; and the seed coats contain phe-
of some cancers. Definitions of dietary fibre vary. Some are nolic compounds.
based on chemical analyses of the components of plant cell
walls, such as non-starch polysaccharide, others on physiologi- Herbs and spices
cal effects the carbohydrates that enter the large bowel hav-
Nearly all herbs and spices contain aromatic compounds,
ing escaped digestion in the small intestine being defined as
dietary fibre. The latter definition includes oligosaccharides and
which are volatile molecules that are usually fat- rather than
resistant starch. The World Health Organization and Food and water-soluble. The flavour compounds may make up as much
Agriculture Organization have recently proposed that only poly- as 15 g per 100 g of a spice by weight, although herbs con-
saccharides which form part of plant cell walls should be regard- tain much lower levels typically around 1 g per 100 g.
ed as dietary fibre and that the health benefits of resistant Many plants have evolved to contain these compounds
starch and oligosaccharides are more appropriately considered because they act as deterrents to herbivores. Some of these
separately. aromatic compounds may be bioactive, although possibly not
at the levels found in most diets. Isothiocyanates are respon-
This box also appears as box 4.1.2 in the previous section sible for the spicy/hot flavour of mustard and horseradish,
produced from glucosinolates in cruciferous plants. Chives
and garlic (allium vegetables) contain the distinctive sul-
phides discussed above. Terpenoids are common components
4.2.2), have been shown to have potentially beneficial effects in herbs and spices, providing distinctive flavours. Examples
in laboratory experiments, as detailed in the evidence in include monoterpenes, such as geranial in lemon grass, and
chapter 4.2.5 (also see Chapter 2).7-9 linalool in bergamot; sesquiterpenes, such as bisabolene in
ginger; triterpenoids, such as the saponin glycrrhizic acid,
Pulses (legumes) found in liquorice root; and tetraterpenoids, such as the
Dry pulses are seeds and are higher in protein than most carotenoid, lycopene.
other plant foods. Soya beans and peanuts contain 37 g per
100 g and 26 g per 100 g protein dry weight respectively,
although, on average, pulses contain around 20 g per 100 g 4.2.3 Consumption patterns
protein dry weight.10 These foods are typically high in car-
bohydrates and non-starch polysaccharides (dietary fibre), Fruits and vegetables
and are generally low in fat. Soya beans and peanuts are The global average for vegetable consumption (based on
exceptions, being relatively high in fat with 8 g per 100 g availability and not including vegetable oils) is 2.6 per cent
and 47 g per 100 g fat, respectively (mostly mono- and of total daily energy intake.17 It is generally highest in North
polyunsaturated fatty acids). They also contain oligosac- Africa, the Middle East, parts of Asia, the USA and Cuba, and
charides that are not digested in the gut but are fermented in southern Europe. Although consumption levels are similar
by bacteria in the colon. Soya beans are distinct from other in countries of high and low economic status, vegetables rep-
legumes in that they have a high content of bioactive resent a greater proportion of daily energy intake in the low-
isoflavones, or phytoestrogens, which have hormone-like income countries. Intakes range from 5.3 per cent in parts of
effects in the body. They are also good sources of saponins Asia to as little as 0.2 per cent in sub-Saharan Africa. On aver-
and phytosterols, which decrease cholesterol absorption. age, the availability of vegetables is increasing globally.
Many legumes contain deguelin, which has been shown to The global average for fruit consumption (based on avail-
have anti-tumour effects in laboratory experiments.11 Most ability) is 2.7 per cent of total daily energy intake. Fruit con-
pulses are virtually indigestible and inedible before cooking; sumption is generally higher than vegetable consumption,
immature legumes that are eaten green have higher levels but it shows a greater degree of variability. Fruit consump-
of sugar and lower levels of non-digestible polysaccharides tion is higher in high-income countries, although it repre-
than dried pulses. sents a similar percentage of total available dietary energy

C H A P T E R 4 F O O D S A N D D R I N K S

to that seen in low-income countries. Intakes are highest in energy availability. The highest availability is in the Middle
some parts of Africa, the Middle East, southern Europe, and East and parts of Europe, and the lowest is in South America
Oceania, and lowest in other parts of Africa and Asia. Fruit and parts of Africa; intakes range from 3 per cent of total
consumption also tends to be low in north-eastern Europe. energy in parts of the Middle East to virtually zero in many
Intakes range from as much as 20 per cent of daily energy low-income countries.
in parts of Africa to as little as 0.5 per cent in parts of Asia. Coconuts represent 0.5 per cent of daily energy availabil-
The availability of fruit has increased globally in recent ity globally, although coconuts can be locally important in
decades, although there was a slight decrease in the 1990s. tropical islands, for instance in parts of Oceania, Asia (Sri
Most countries have national recommendations for the Lanka and Indonesia), the Caribbean, and in the African
daily amount of vegetables and fruits that need to be eaten islands. In parts of Oceania, for example, coconuts provide
to maintain optimal health (Chapter 10). These vary, but as much as 20 per cent of energy in the diet.
they tend to recommend three or more servings per day of Sunflower, rape, mustard, and sesame seeds together sup-
vegetables and two or more servings per day of fruits; a serv- ply 0.2 per cent of daily energy intake globally. There are
ing is about 80 g (or half a US cup). In most high-income fewer data available for seeds than for many other foods,
countries for which data were available, daily consumption although sesame seed intake is relatively high in parts of
of vegetables fell short of this target, although this is not due Africa and Asia, providing a maximum of 3.9 per cent of
to lack of availability; indeed, availability is high due to the energy in parts of central Africa. Oils from seed crops are
wide use of refrigeration. Fruit consumption tended to be widely used (see chapter 4.5.3).
closer to national targets. Seasonal availability influences
overall availability, although less so in high-income countries Herbs, spices, and condiments
where vegetables and fruits are more likely to be imported. Although spices are consumed in small amounts to flavour
food, they are such a regular feature of some diets that they
Pulses (legumes) account for a measurable quantity of daily energy intake.
Globally, pulses supply 2 per cent of total energy intake Worldwide, spices provide 0.3 per cent of available dietary
(based on availability) and 3.5 per cent of daily protein energy and in parts of Asia they constitute as much as 1.8
intake.17 The highest availability is in parts of Africa, South per cent. Herbs and spices tend to be part of the tradition-
America, Asia, and the Middle East. In these areas, pulses al diet in the areas from which they originate, and many tra-
are a dietary staple, and can account for as much as 20 per ditional cuisines are characterised by the use of herbs, spices,
cent of daily energy intake and 50 per cent of protein intake. and condiments. Most are now available worldwide,
In societies with high intakes of meat and other foods of ani- although their use still varies greatly in different parts of the
mal origin, pulses are less important in diets, and are usu- world. Many herbs and spices are believed to have medici-
ally consumed infrequently or in small amounts. Peanuts and nal or tonic value and have been used in this way at least
soya beans account for most of the legume products eaten since the times of the earliest medical records. Many mod-
around the world. ern pharmaceuticals are derived from herbs and other plants.
Soya bean availability per person represents 0.5 per cent Many herbs and some spices are biologically very potent:
of daily energy intake globally, but it is notably high in parts the modern pharmacopoeia lists drugs, many of which have
of Asia, and higher than average in parts of Africa and been isolated from herbs, sometimes known as plants with
Central America. In parts of Asia, soya beans account for up healing powers. There are some in vivo experimental data
to 4.9 per cent of daily energy availability and 15 per cent for potentially beneficial effects in the cases of turmeric, saf-
of protein. fron, ginger, pepper, garam masala (a herb and spice mix),
Pulses are eaten in a variety of ways around the world; for and also eugenol and myristin, constituents of a number of
instance, Japanese and Chinese bean curd (tofu), Chinese herbs and spices.
mung bean sprouts, Mexican chilli and refried beans, Indian Conversely, it is at least theoretically possible that some
dahl, Middle Eastern falafel and hummus, Indonesian cul- condiments have adverse effects. Two examples are hot chilli
tured soya bean cakes (tempeh), Cuban black beans and rice, juices and harissa, a fiery condiment; both are consumed in
Boston baked beans, French cassoulet, Brazilian feijoada, substantial quantities in Mexico and the Mahgreb countries
Swedish pea soup, and US peanut butter. Soya beans are par- of North Africa, respectively, and both irritate the mouth and
ticularly versatile and their products are a common feature throat.
in manufactured foods, although they are not commonly
eaten whole. Soya foods include soya drinks and flour, tofu,
tempeh, textured vegetable protein, and the many products 4.2.4 Interpretation of the evidence
that can be prepared from these foods. Fermented soya beans
produce soy sauce and miso. Soya bean oil is also used wide- General
ly (see chapter 4.5.3). For general considerations that may affect interpretation of
the evidence, see chapter 3.3 and 3.5, and boxes 3.1, 3.2,
Nuts and seeds 3.6 and 3.7.
Nuts and seeds were an important part of human diets before Relative risk (RR) is used in this Report to denote ratio
the advent of agriculture and they remain locally important measures of effect, including risk ratios, rate ratios, haz-
in a few areas. Globally, tree nuts supply 0.4 per cent of daily ard ratios, and odds ratios.

P A R T 2 E V I D E N C E A N D J U D G E M E N T S Specific 4.2.5 Evidence and judgements

Considerations specific to vegetables, fruits, pulses
(legumes), nuts, seeds, herbs, and spices include: The full systematic literature review (SLR) is contained on
the CD included with this Report.
Patterns and ranges of intake. Most studies of consumption
of vegetables, fruits, and pulses (legumes) have been con- Non-starchy vegetables
ducted in populations that have relatively homogeneous Mouth, pharynx, and larynx
diets. The limited data on nuts, seeds, herbs, spices, and Thirty-one case-control studies21-50 and 3 ecological studies51-
condiments come mainly from a few human case-control 53
investigated non-starchy vegetables and mouth, pharynx,
studies and some experimental animal studies. or larynx cancers; 1 cohort study54 and 6 case-control stud-
ies33 39 45 55-57 investigated non-starchy vegetables and fruits;
Classification. There is no general agreement on classifica- 23 case-control studies investigated raw vegetables24 27 28 33
tion. Some studies have included cereals such as corn, and 36-43 45 47 50 58-65
; 1 cohort study,66 14 case-control studies,24
tubers such as potatoes, as vegetables, and plantains as fruit. 26-29 39 41 43 45-47 50 63 67
and 1 ecological study68 investigated
Broccoli and green peppers are included as green vegetables cruciferous vegetables; 1 cohort study66 and 10 case-control
in some studies, while only leafy greens are included in this studies24 26-29 39 47 61 67 69 investigated green, leafy vegetables;
category in others; tomatoes are considered yellow-orange 3 cohort studies66 70 71 and 18 case-control studies23 24 26-29 39
vegetables in some but not in others. Some studies report 41-43 46 49 50 63 65 72-75
investigated carrots; and 1 cohort study66
results only for broad categories (for example, all vegetables and 12 case-control studies26-29 39-43 46 50 58 62 65 investigated
or all fruits), whereas others have reported results for more tomatoes.
narrowly defined categories (for example, raw vegetables,
green vegetables, citrus fruits) or for individual food items Non-starchy vegetables
(for example, spinach, carrots, tomatoes). In some stud- Most studies showed decreased risk with increased intake of
ies, vegetables and fruits have been categorised according to non-starchy vegetables. Twenty-two studies reported com-
botanical classification; in others, categorisation has been parisons of high against low intake (figure 4.2.1).22 23 25 26 29-
according to culinary usage. In this report, the terms veg- 31 33 35-46 49 50
Of these, 19 showed decreased risk for the
etables and fruits are used according to their culinary def- highest intake group,22 25 26 30 31 33 35-44 46 49 50 which was sta-
inition. Some studies have included pulses as vegetables tistically significant in 13.22 25 30 31 35 37 38 40 42 43 46 49 50 The
whereas others have classified these as a separate entity or other 3 studies showed non-significant increased risk.23 29 45
not at all. Many older studies have not differentiated
between retinol and carotenoids. Vitamin E intakes are dif-
ficult to quantify since much comes from the vegetable oils Figure 4.2.1 Non-starchy vegetables and mouth, pharynx,
used in food preparation, and intakes within populations are and larynx cancer; case-control studies
usually homogenous because of the widespread occurrence
of vitamin E in commonly consumed foods. Relative risk (95% CI)

Notari 1987 0.42 (0.250.71)

Measurement. Assessment of selenium intake is problemat- Franco 1989 1.70 (0.923.16)
ic because the content of selenium in foods depends to a Oreggia 1991 0.19 (0.050.68)

large extent on the soil selenium content of the area in which Franceschi 1991 0.80 (0.550.68)
Franceschi 1992 0.40 (0.200.80)
the foods were grown. Blood and toenail levels of selenium Zheng 1993 1.73 (0.793.78)
are thought to be fairly accurate indicators of intake and Kune 1993 0.30 (0.110.85)
De Stefani 1994 0.40 (0.190.85)
have been used in several studies. Takezaki 1996 0.80 (0.591.08)
Esteve 1996 0.61 (0.450.81)
Confounding. Smokers consume fewer vegetables and fruits Levi 1998 0.14 (0.080.23)
De Stefani 1999 0.50 (0.290.87)
than non-smokers.18 19 Fat intake inversely correlates with Franceschi 1999 0.50 (0.330.76)
vegetable and, particularly, fruit intake in the USA.20 Recent De Stefani 1999 0.57 (0.301.08)

studies of the effects of fruits and vegetables in cancers Garrote 2001 0.78 (0.401.52)
Bosetti 2002 0.17 (0.110.27)
thought to be caused by smoking have controlled for the Marchioni 2003 0.86 (0.541.38)
effect of smoking. Folate intake is correlated with intake of Lissowska 2003 0.17 (0.070.27)

non-starchy polysaccharide (dietary fibre). Rajkumar 2003 0.44 (0.280.69)

Sanchez 2003 0.54 (0.340.86)
Gaud 2004 1.40 (0.712.76)
Reporting bias. Studies using self-reporting tend to over- De Stefani 2005 0.60 (0.331.10)

report vegetable and fruit consumption. Where an effect

exists, results from such studies are liable to underestimate 0.2 0.5 1 2 5
the extent to which vegetables and fruits modify the risk of
Relative risk, highest vs lowest exposure category

C H A P T E R 4 F O O D S A N D D R I N K S

(international) found no significant association with cancer

Figure 4.2.2 Non-starchy vegetables and mouth, pharynx,
mortality.52 53
and larynx cancer; case-control studies

Relative risk (95% CI) Non-starchy vegetables and fruits

A cohort study that reported results for non-starchy vegeta-
De Stefani 1994 0.53 (0.261.09) bles and fruits in combination reported a statistically signif-
Levi 1998 0.62 (0.530.74)
De Stefani 2000 0.82 (0.641.05)
icant protective effect in the highest consumers (0.55, 95%
Bosetti 2002 0.75 (0.700.81) CI 0.320.95).54 All six case-control studies looking at the
Summary estimate 0.72 (0.630.82) same exposure group reported reduced risk estimates in sim-
ilar comparisons,33 39 45 55-57 which were statistically signifi-
0.5 0.75 1 1.5 cant in four.33 39 55 57 All of these studies adjusted for smoking
Relative risk, per 50 g/day
and alcohol consumption.

Raw vegetables
Twenty-three case-control studies reported separate risk esti-
mates for raw vegetable consumption.24 27 28 33 36-43 45 47 50 58-
All of these reported comparisons of risk between high and
low intake groups, which produced reduced risk estimates
Figure 4.2.3 Non-starchy vegetables and mouth,
pharynx, and larynx cancer; case-control in 2224 27 28 33 36-43 45 47 50 58 60-65; 16 of these were statistical-
studies: dose response ly significant.24 33 36-40 42 43 47 50 60-63 65 No studies reported sta-
tistically significant increased risk estimates. Meta-analysis
of 7 case-control studies gave an effect estimate of 0.71 (95%
De Stefani 1994
CI 0.590.86) per 50 g/day, with moderate heterogeneity.37
39 42 45 59 62 65
These studies also provided evidence of a dose-
Levi 1998
response relationship. The heterogeneity could be partially
explained by variable exposure definitions. These results are
De Stefani 2000 consistent with data for non-starchy vegetables.

Bosetti 2002 Cruciferous vegetables

One cohort study,66 14 case-control studies,24 26-29 39 41 43 45-
47 50 63 67
and 1 ecological study68 reported separate risk esti-
mates for cruciferous vegetable consumption.
The single cohort study showed a non-significant increased
risk for increased intake of cauliflower and a non-significant
0 100 200 300 400
decreased risk for cabbage.66 Four case-control studies
showed statistically significant decreased risk with increased
Non-starchy vegetables (g/day)
intake, either overall or in specific subgroups.24 29 43 47 One
study showed statistically significant increased risk associ-
ated with eating kimchi or pickled cabbage.67 The other nine
studies showed inconsistent and non-significant associa-
tions.26-28 39 41 45 46 50 63 The ecological study showed a sta-
tistically significant decreased risk.68
The remaining studies showed no consistent association,
probably due to varying exposure definitions and study Green, leafy vegetables
design.21 24 27 28 32 34 47 48 Meta-analysis was possible on 4 case- One cohort study66 and 10 case-control studies24 26-29 39 47 61
control studies, giving a summary effect estimate of 0.72 67 69
reported separate risk estimates for green, leafy veg-
(95% confidence interval (CI) 0.630.82) per 50 g/day, with etable consumption.
moderate heterogeneity (figure 4.2.2). All studies adjusted The single cohort study showed no effect for the highest
for sex, smoking, and alcohol consumption. intake group of lettuce when compared to the lowest.66 Nine
A dose-response relationship is apparent from the four case-control studies showed decreased risk with increased
case-control studies that could be meta-analysed (figure intake,24 26 27 29 39 47 61 67 69 which was statistically significant
4.2.3). There is some suggestion that the greatest effect in four.24 39 61 69 One study showed non-significant increased
appears to be with the first increment. That is, any increase risk.28
above the lowest levels of vegetable consumption confers a
protective effect. However, it is not clear that the effect con- Carrots
tinues in a linear fashion with increased dose. Three cohort studies66 70 71 and 18 case-control studies23 24
Of the three ecological studies, one (Hong Kong) study 26-29 39 41-43 46 50 63 65 72-75
investigated non-starchy root veg-
found a significant negative association between vegetable etables and mouth, larynx, or pharynx cancers. There was
consumption and cancer incidence51; the other two variation in the exposure classification in studies. Most

P A R T 2 E V I D E N C E A N D J U D G E M E N T S

assessed carrots, but some looked at tubers and carrots or vegetables and oesophageal cancer. Eight case-control stud-
non-starchy root vegetables or yellow/orange vegetables. ies investigated vegetable and fruit consumption (com-
One cohort study, looking at tubers and carrots, report- bined)95 104 107 114 120-123; 16 case-control studies investigated
ed a non-significant increased risk when comparing high raw vegetables40 60 85 95-97 103 109 113 114 124-129; 1 cohort study66
against low intakes, with a wide confidence interval (1.9, and 5 case-control studies86 93 107 124 125 investigated crucif-
95% CI 0.66.0).66 Another that reported on carotene-rich erous vegetables; 1 cohort study82 and 8 case-control stud-
fruits and vegetables found a non-significant reduced risk ies86 101 103 107 109 111 129 130 investigated allium vegetables; 1
when comparing the highest intake group against the low- cohort study66 and 11 case-control studies86 96 98 111 124 127 131-
est (0.50, p value for linear trend 0.10).70 The third, which 135
investigated green, leafy vegetables; 1 cohort study66 and
evaluated yellow/orange vegetables in postmenopausal US 9 case-control studies58 62 109 111 113 129-132 136 investigated
women, reported a significant reduced risk for the same com- tomatoes.
parison (0.58, 95% CI 0.390.87).71
All of the 18 case-control studies reported comparisons of Non-starchy vegetables
risk between high- and low-intake groups.23 24 26-29 39 41-43 46 Data suggest an association with reduced risk. Of the five
49 50 63 65 72-75
Sixteen reported reduced risk estimates,23 26-29 cohort studies, three reported decreased risk when compar-
39 41-43 50 63 65 72-75
6 of which were statistically significant.49 ing the highest intake group against the lowest, one of which
72 75
The other 2 were non-significant in the direction of was statistically significant (0.66, 95% CI 0.440.99, non-
increased risk.24 27 28 39 43 46 The majority of studies were starchy vegetables82; 0.5, p value for linear trend 0.1, yel-
hospital-based and analysed carrots as a separate exposure.

One cohort study66 and 12 case-control studies26-29 39-43 46 50 Figure 4.2.4 Non-starchy vegetables and oesophageal
58 62 65
investigated tomatoes and mouth, larynx, or pharynx cancer; cohort and case-control studies
Relative risk (95% CI)
The cohort study reported a non-significant increased risk
when comparing the highest intake group against the low- Cohort
est, with a wide confidence interval (1.7, 95% CI 0.83.7).66 Hiryama 1990 1.06 (0.911.24)

Of the 12 case-control studies,26-29 39-43 46 50 58 62 65 10 report- Yu 1993 0.66 (0.440.99)

Guo 1999 0.80 (0.621.53)
ed reduced risk estimates,26 29 39-43 46 50 58 62 65 5 of which were Tran 2005 1.02 (0.851.18)
statistically significant.26 29 39 40 62 Only 2 reported an
increased risk, which was non-significant.27 28 These studies Case control
Cook-Mozalfari 1979 Men 0.85 (0.581.25)
were also the only studies not to adjust for both smoking and Cook-Mozalfari 1979 Women 0.81 (0.491.33)
alcohol intake. Notani 1967 1.06 (5.701.64)
Brown 1968 0.70 (0.391.20)
The general mechanisms through which vegetables could Jun-lao 1989 1.50 (0.191.89)
plausibly protect against cancers of the mouth, larynx, and De Stefani 1990 0.56 (0.311.02)
pharynx are outlined below. Ren 1991 2.57 (1.255.27)
Negni 1991 0.20 (0.090.45)
Although all of the studies mentioned here adjust for Sammon 1992 1.44 (0.622.10)
smoking behaviour and nearly all adjust for alcohol, the rel- Ho 1994 0.60 (0.301.35)

ative risk of smoking is large (particularly when combined Sammon 1998 2.30 (1.064.89)
Lauroy 1998 0.24 (0.113.54)
with alcoholic drinks). It is therefore difficult to eliminate De Stefani 1999 0.50 (0.290.87)
confidently the possibility of residual confounding with ways Takazaki 2000 0.60 (0.450.78)

of life associated with smoking: for instance, smokers con- De Stefani 2000 0.64 (0.341.20)
Levi 2000 0.19 (0.110.33)
sume fewer vegetables than non-smokers. Nayor 2000 0.53 (0.360.87)
Cheng 2000 0.56 (0.221.34)

A substantial amount of consistent evidence on non- Bosets 2000 0.79 (0.471.32)

Takezaki 2001 0.81 (0.401.43)
starchy vegetables, including specific subtypes, mostly Terry 2001 0.60 (0.385.35)
from case-control studies, shows a dose-response Zhany 2001 1.94 (1.222.85)
Chen 2002 0.62 (0.261.35)
relationship. There is evidence for plausible
Cnuk 2002 10.10 (4.4223.09)
mechanisms. Non-starchy vegetables probably protect Xibin 2003 0.44 (0.210.34)
against mouth, pharynx, and larynx cancers. Li 2003 0.76 (0.551.19)
Hung 2004 0.50 (0.310.82)
De Stefani 2005 0.53 (0.271.00)
The Panel is aware that since the conclusion of the SLR, two Yang 2005 0.62 (0.321.18)

cohort76 77 and two case-control studies78 79 have been pub-

lished. This new information does not change the Panel judge-
0.2 0.5 1 2 5
ment (see box 3.8).
Relative risk, highest vs lowest exposure category

Five cohort studies,70 80-83 37 case-control studies22 40 60 84-115
and 6 ecological studies51 52 116-119 investigated non-starchy

C H A P T E R 4 F O O D S A N D D R I N K S

Raw vegetables
Figure 4.2.5 Non-starchy vegetables and oesophageal
cancer; case-control studies Sixteen case-control studies investigated raw vegetables and
oesophageal cancer.40 60 85 95-97 103 109 113 114 124-129
Relative risk (95% CI) All of these studies reported associations with decreased
risk, which were statistically significant in 10.40 60 85 95 97 109
Jun-Lao Li 1989 1.08 (1.031.14) 113 126 127 129
Dose-response meta-analysis was possible on five
De Stefani 2000 0.81 (0.660.99)
Levi 2000 0.66 (0.580.74) studies, giving a summary effect estimate of 0.69 (95% CI
Cheng 2000 0.90 (0.741.11) 0.580.83) per 50 g/day increment (figures 4.2.6 and 4.2.7).
De Stefani 2005
Summary estimate
0.94 (0.861.02)
0.87 (0.721.05)
This exposure category could be less disparate than other
vegetable groupings, as it is clear that preserved vegetables
are not included and variation in cooking methods is
0.5 0.75 1 1.5 2
removed. This may account for the lack of heterogeneity in
Relative risk, per 50 g/day
direction of effect in this subcategory of vegetables.

Figure 4.2.6 Raw vegetables and oesophageal cancer;

case-control studies

low/orange vegetables70; and 0.8, 95% CI 0.601.0 and p Relative risk (95% CI)
value for trend 0.08, stated as not statistically significant,
non-starchy vegetables80). The other two reported a non- De Stefani 2000
Levi 2000
0.38 (0.210.69)
0.64 (0.570.72)
significant increased risk (1.06, 95% CI 0.911.24, non- Cheng 2000 0.83 (0.681.02)
starchy vegetables81; and 1.02, 95% CI 0.881.19, fresh Sharp 2001 0.84 (0.750.95)
De Stefani 2003 0.60 (0.480.76)
non-starchy vegetables83) (figure 4.2.4). Summary estimate 0.69 (0.580.83)
Most (29) of the case-control studies published decreased
risk estimates when comparing the highest intake group
0.25 0.5 0.75 1 1.5
against the lowest,40 60 85-90 94-99 101-105 107-109 111-115 which were
Relative risk, per 50 g/day
statistically significant in 14 (figure 4.2.4).40 88 89 94 97-99 101
102 104 105 109
Five studies reported statistically significant
increased risk.84 91 93 100 106 110 Meta-analysis was possible on
5 of the case-control studies, giving a summary effect
estimate of 0.87 (95% CI 0. 721.05) per 50 g/day incre-
ment, with high heterogeneity (figure 4.2.5). A potential
cause of heterogeneity is the disparate nature of the expo- Figure 4.2.7 Raw vegetables and oesophageal cancer;
case-control studies: dose response
sure definition in different studies, some of which included
pickled and cured vegetables, cooked or uncooked
Two of the ecological studies reported a statistically sig- Castelletto 1994

nificant, positive association between vegetable consumption

Cheng 2000
and cancer incidence116 117; one reported a statistically sig-
nificant, negative association between vegetable consump-
De Stefani 2000
tion and cancer incidence51; and the other three reported no
significant association between vegetable consumption and
Levi 2000
cancer mortality.52 118 119
The Panel is aware that data from the European
Sharp 2001
Prospective Investigation into Cancer and Nutrition (EPIC;
521 457 participants from 10 European countries; 65 cases De Stefani 2003
of adenocarcinomas of the oesophagus), published after the
conclusion of the SLR,140 showed a non-significant reduced
risk (0.72, 95% CI 0.321.64) per 100 g/day increase in veg-
0 100 200 300
etable consumption (adjusted for several variables including
Non-starchy vegetables (g/day)
smoking and alcohol, red meat, and processed meat).

Non-starchy vegetables and fruits

Eight case-control studies investigated vegetable and fruit
consumption (combined) and oesophageal cancer. All
reported a decreased risk with increased consumption.95 104
107 114 120-123
Six of these were statistically significant.95 104 107
114 120 121

P A R T 2 E V I D E N C E A N D J U D G E M E N T S

Non-starchy root vegetables and tubers The general mechanisms through which vegetables could
One cohort study66 and six case-control studies114 122 128 132 plausibly protect against oesophageal cancer are outlined
136 141 142
reported separate risk estimates for consumption of below.
non-starchy root vegetables and tubers.
The single cohort study showed a non-significant increased There is more evidence, including on vegetable
risk for the highest intake group when compared to the low- subtypes, from case-control studies than from cohort
est, after adjustment.66 All six case-control studies showed studies, but both are moderately consistent and there
non-significant decreased risk with increased intake.114 122 128 is some evidence for a dose-response relationship.
132 136 141 142
There is evidence for plausible mechanisms. Non-
starchy vegetables probably protect against
Cruciferous vegetables oesophageal cancer.
One cohort study66 and five case-control studies86 93 107 124 125
reported separate risk estimates for consumption of crucif- The Panel is aware that since the conclusion of the SLR, one
erous vegetables. cohort140 and two case-control studies78 143 have been pub-
The single cohort study showed a non-significant lished. This new information does not change the Panel judge-
decreased risk for increased intake of cauliflower or swede ment (see box 3.8).
and a non-significant increased risk for cabbage, after adjust-
ment.66 Three case-control studies showed decreased risk Stomach
with increased intake,93 107 124 which was statistically signif- Ten cohort studies,71 80 140 144-150 45 case-control studies,109
icant in two.93 124 One study showed a non-significant 151-195
and 19 ecological studies51 52 116-119 196-209 investigated
increased risk125; and one study showed a non-significant total vegetables. Eleven cohort studies,71 144 150 210-218 21 case-
increased risk in women and a non-significant decreased risk control studies,89 165 169 178 179 188 191 219-232 and 8 ecological
in men.86 studies233-240 investigated green-yellow vegetables; 6 cohort
studies70 140 144 146 150 241, 13 case-control studies,162 174 175 179
Allium vegetables 180 187 223 227 229 230 232 242 243
and 2 ecological studies202 240
One cohort study82 and eight case-control studies86 101 103 107 investigated green, leafy vegetables; 3 cohort studies70 146 241
109 111 129 130
reported separate risk estimates for allium and 19 case-control studies58 109 129 152 156 164 171 172 174 232 243-
vegetable consumption. 251
investigated tomatoes; 2 cohort studies150 214 and 6 case-
The single cohort study showed that garlic intake had no control studies157 165 169 226 228 243 investigated white or pale
effect on risk.82 Four case-control studies showed non-sig- vegetables; 6 cohort studies,146 148 214 252-254 25 case-control
nificant decreased risk with increased intake.101 103 107 130 Two studies,109 129 161 162 167 172 174 183 184 191 219 225 226 243 247 248 250 255-
studies showed non-significant increased risk.86 111 One study 264
and 3 ecological studies202 208 238 investigated raw vegeta-
showed a statistically significant decreased risk for garlic and bles; 5 cohort studies144 146 148 253 265 and 6 case-control
that onions/leeks had no effect on risk109; and one study studies158 161 162 164 257 266 267 investigated non-starchy veg-
showed a statistically significant reduced risk for onions and etables and fruits.
a non-significant increased risk for garlic.129
Non-starchy vegetables
Green, leafy vegetables Of 12 independent estimates from the 10 cohort studies that
One cohort study66 and 11 case-control studies86 96 98 111 124 investigated non-starchy vegetable consumption, none was
127 131-135
reported separate risk estimates for green, leafy veg- statistically significant.71 80 140 144-150 Seven studies showed
etable consumption. non-significant reduced risk71 140 144-147 150 and 2 reported
The single cohort study showed no effect for the highest non-significant increased risk.80 149 One study showed non-
intake group of lettuce when compared to the lowest.66 Ten significant increased risk in women and non-significant
case-control studies showed decreased risk with increased decreased risk in men.148 Most effect estimates were close to
intake,96 98 111 124 127 131-135 which was statistically significant 1. Meta-analysis was possible on 9 independent estimates
in five.96 127 132-134 One study showed a non-significant from 7 cohort studies, giving a summary effect estimate of
increased risk in women and a non-significant decreased risk 0.98 (95% CI 0.911.06) per 100 g/day, with moderate het-
in men.86 erogeneity (figure 4.2.8).
Of 45 case-control studies that reported on non-starchy
Tomatoes vegetable consumption, 28 reported statistically significant
One cohort study66 and nine case-control studies58 62 109 111 decreased risk.109 151-153 156-160 163 164 168 169 171 173 176-179 181 182
113 129-132 136
reported separate risk estimates for consumption 184 185 187 190 192
The majority of the 17 remaining studies that
of tomatoes. reported no significant effect on risk were in the direction of
The single cohort study showed a non-significant increased decreased risk.155 162 165-167 170 172 174 183 191 194 195 Four studies
risk for the highest intake group of lettuce, when compared showed non-significant increased risk,180 188 189 193 1 study
to the lowest, after adjustment.66 Eight case-control studies showed no effect on risk,154 and 1 study stated that there was
showed decreased risk with increased intake,58 62 109 111 113 129 no significant association.175 One study showed non-signif-
131 132 136
which was statistically significant in two.62 129 One icant decreased risk in women and non-significant increased
study showed no effect on risk.130 risk in men186; and 1 study showed statistically significant

C H A P T E R 4 F O O D S A N D D R I N K S

Figure 4.2.8 Non-starchy vegetables and stomach cancer; Figure 4.2.9 Green-yellow vegetables and stomach
cohort and case-control studies cancer; cohort and case-control studies

Relative risk (95% CI) Relative risk (95% CI)

Cohort Cohort
Chyou 1990 Men 0.76 (0.531.10) Chyou 1990 Men 0.74 (0.521.05)
Guo 1994 1.05 (0.931.18) Kasum 2002 Women 0.57 (0.191.75)
Botterweck 1998 0.92 (0.781.09) Kobayashi 2002 0.56 (0.330.94)
McCullough 2001 Men 0.98 (0.841.02) Ngoan 2002 0.41 (0.180.94)
McCullough 2001 Women 1.16 (1.001.34) Khan 2004 Men 0.66 (0.113.89)
Fujino 2002 Men 1.12 (0.931.34) Khan 2004 Women 0.18 (0.022.13)
Fujino 2002 Women 1.03 (0.771.36) Summary estimate 0.63 (0.480.82)
Kobayashi 2002 0.86 (0.721.02)
Gonzalez 2006 0.91 (0.661.28) Case control
Summary estimate 0.98 (0.911.06) Lee 1990 0.94 (0.681.29)
Cai 1991 0.44 (0.380.50)
Case control
Negri 1991 0.24 (0.130.44)
Risch 1985 0.84 (0.730.97)
Hoshiyama and Sasaba 1992 0.49 (0.320.74)
You 1988 0.81 (0.750.86)
Ji 1998 Men 0.77 (0.690.86)
De Stefani 1990 0.40 (0.270.60)
Ji 1998 Women 0.89 (0.771.04)
Menik 1992 0.36 (0.160.80)
Ward 1999 0.33 (0.240.46)
Hansson 1993 0.60 (0.460.79)
Hamada 2002 1.09 (0.412.93)
Cornee 1995 0.85 (0.571.28)
Nishimoto 2002 0.62 (0.333.43)
De Stefani 1998 0.39 (0.320.48)
Hara 2003 1.59 (0.733.43)
Ji 1998 Men 0.86 (0.820.90)
Ito 2003 Women 0.55 (0.380.81)
Ji 1998 Women 0.94 (0.861.01)
Lissowska 2004 0.28 (0.120.66)
Ward 1999 0.22 (0.110.43)
Summary estimate 0.59 (0.460.75)
Mathew 2000 0.52 (0.181.46)
De Stefani 2001 0.89 (0.601.32)
Takezaki 2001 0.47 (0.260.81)
Sriamporn 2002 0.86 (0.561.33) 0.5 1 2
Hara 2003 1.06 (0.841.33)
Sipetic 2003 0.07 (0.030.15)
Relative risk, per 100 g/day
Suh 2003 1.04 (0.891.22)
Lagiou 2004 0.51 (0.390.68)
Lissowska 2004 0.90 (0.741.09)
Boccia 2005 1.09 (0.562.12)
Nan 2005 0.90 (0.661.25)
Summary estimate 0.70 (0.620.79)

0.5 0.75 1 1.5 2

Relative risk, per 100 g/day

decreased risk in men and non-significant increased risk in Meta-analysis was possible on 6 independent estimates from
women.161 No studies reported statistically significant 5 studies, giving a summary effect estimate of 0.63 (95% CI
increased risk. Meta-analysis was possible on 20 studies, giv- 0.480.82) per 100 g/day, with no heterogeneity (figure
ing a summary effect estimate of 0.70 (95% CI 0.620.79) 4.2.9).
per 100 g/day, with high heterogeneity (figure 4.2.8). This Of the 21 case-control studies that reported on green-
heterogeneity tended to reflect differences in size, rather yellow vegetable consumption, 16 showed decreased risk,89
than direction, of effect. 165 169 178 179 191 219 220 222-228 230-232
statistically significant in
A dose-response relationship was apparent from case- 12.89 165 169 178 179 191 220 222 223 226 231 232 The remaining 5 stud-
control but not cohort data. ies reported increased risk,111 188 221 229 1 of which was sta-
Results from ecological studies reporting on non-starchy tistically significant.221 Meta-analysis was possible on 12
vegetable consumption were mixed, with almost as many independent estimates from 11 studies, giving a summary
studies reporting increased risk as reported decreased effect estimate of 0.59 (95% CI 0.460.75) per 100 g/day,
risk.51 52 116-119 196-209 with high heterogeneity (figure 4.2.9).
All of the studies adjusted for age and sex; none was
Green-yellow vegetables adjusted for infection with Helicobacter pylori. Nine studies
Eight of the 11 cohort studies that reported on green-yellow were maximally adjusted, seven of which reported a
vegetable consumption showed decreased risk,71 144 150 210 211 significant negative association with higher consumption of
statistically significant in 4.150 210 215 216 Two other stud- green-yellow vegetables, and the other two reported no
ies showed non-significant increased risk212 213 and 1 other significant association.
study reported no statistically significant association.218 A dose-response relationship was apparent from both

P A R T 2 E V I D E N C E A N D J U D G E M E N T S

Figure 4.2.10 Green-yellow vegetables and stomach Figure 4.2.11 White or pale vegetables and stomach
cancer; cohort and case-control studies: cancer: cohort and case-control studies
dose response
Relative risk (95% CI)

Cohort Kobayashi 2002 0.52 (0.231.18)
Khan 2004 Men 0.63 (0.094.12)
Chyou 1990 Men Khan 2004 Women 0.15 (0.012.34)
Summary estimate 0.49 (0.241.01)
Kasum 2002 Women

Kobayashi 2002 Case control

Risch 1985 0.29 (0.071.15)
Ngoan 2002 Cai 1991 0.45 (0.340.60)
Hoshiyama and Sasaba 1992 0.86 (0.321.09)
Khan 2004 Men
Summary estimate 0.57 (0.321.02)

Khan 2004 Women

0.5 1 2
Case control Relative risk, per 100 g/day

Lee 1990

Cai 1991

Negri 1991

Hoshiyama and Sasaba 1992

Ji 1998 Men

Ji 1998 Women
Nine case-control studies showed decreased risk with
increased intake,179 187 223 227 229 230 232 242 243 which was sta-
Ward 1999
tistically significant in three,223 232 243 and in men, but not
Hamada 2002 women, in a fourth study.227 Two further studies showed
Nishimoto 2002 non-significant increased risk174 180; one study showed no
Hara 2003
effect on risk162; and one study stated that there was no sig-
nificant association.175 Meta-analysis was possible on six
Ito 2003 Women
case-control studies, giving a summary effect estimate of
Lissowska 2004 0.90 (95% CI 0.701.16) per 100 g/day, with no hetero-
geneity.162 179 180 187 229 230
One ecological study showed statistically significant
0 100 200 300 400
decreased risk240 with high intake, the other showed non-
Green-yellow vegetables (g/day)
significant increased risk.202
One cohort study146 and 15 case-control studies152 156 164
167 172 231 243-246 261 268-271
also reported separately on lettuce
and salad leaves. The single cohort study showed a non-
significant decreased risk with increased intake. The effect
estimate was 0.88 (95% CI 0.382.60) per 50 g/day.146
Twelve case-control studies showed decreased risk with
cohort and case-control data on green-yellow vegetable increased intake of lettuce or salad leaves,152 156 164 167 231 243
consumption (figure 4.2.10). 246 261 268-271
which was statistically significant in 7.156 243 246
Five out of the eight ecological studies that reported on 261 268-270
Two studies showed non-significant increased
green-yellow vegetable consumption showed decreased risk risk.172 245 One study showed no effect on risk.244 Meta-
with increased consumption,236-240 two showed no associa- analysis was possible on 5 case-control studies that investi-
tion,233 234 and one study showed increased risk.235 gated lettuce or salad leaves, giving a summary effect
This exposure included green-yellow vegetables, green estimate of 0.43 (95% CI 0.240.77) per 50 g/day, with high
vegetables, yellow vegetables, yellow-orange vegetables, car- heterogeneity.152 231 268-270 Heterogeneity was related
rots and pumpkins, and high-carotenoid vegetables. primarily to the size, and not the direction, of the effect.

Green, leafy vegetables Tomatoes

Four cohort studies showed non-significant decreased Two cohort studies showed a non-significant increased
risk with increased intake70 144 146 150; two studies showed risk with increased intake.146 241 One study stated that there
non-significant increased risk.140 241 Meta-analysis was pos- was a non-significant decreased risk (unquantified).70
sible on four cohort studies, giving a summary effect estimate The effect estimates were 1.81 (95% CI 0.853.85) per 100
of 0.85 (95% CI 0.581.25) per 100 g/day, with no hetero- g/day,146 and 1.1 (95% CI 0.761.60) for women and
geneity.140 144 146 150 1.19 (95% CI 0.881.61) for men (both for the highest

C H A P T E R 4 F O O D S A N D D R I N K S

Figure 4.2.12 Raw vegetables and stomach cancer; Figure 4.2.13 Raw vegetables and stomach cancer;
cohort and case-control studies case-control studies: dose response

Relative risk (95% CI)

Jedrychowski 1981
Botterweck 1998 0.51 (0.054.73) Jedrychowski 1986
Galanis 1998 0.81 (0.531.23)
Khan 2004 Men 0.63 (0.162.51) Buiatti 1989
Khan 2004 Women 2.03 (0.1429.46)
Coggon 1989
Summary estimate 0.80 (0.541.18)

Kato 1990 Men

Case control
Jedrychowski 1981 0.10 (0.020.41) Kato 1990 Women
Jedrychowski 1986 1.56 (0.1220.74)
Buiatti 1989 0.56 (0.450.71) Hoshiyama and Sasaba 1992

Caggon 1989 0.13 (0.012.88)

Ramon 1993
Kato 1990 Men 0.59 (0.380.91)
Kato 1990 Women 0.86 (0.481.48) Cornee 1995
Hoshiyama and Sasaba 1992 0.44 (0.290.68)
Ramon 1993 0.35 (0.160.78) Huang 1999
Cornee 1995 0.27 (0.100.72)
De Stefani 2001
Huang 1999 0.77 (0.620.94)
De Stefani 2001 0.46 (0.161.34)
Sriamporn 2002
Sriamporn 2002 0.40 (0.0115.73)
Lee 2003 0.11 (0.040.27) Lee 2003
Lissowska 2004 0.82 (0.521.28)
Summary estimate 0.50 (0.380.65)

0.5 1 2 0 50 100 150

Relative risk, per 100 g/day Raw vegetables (g/day)

intake group when compared to the lowest).241 Raw vegetables

Most case-control studies showed decreased risk with Of seven independent estimates from six cohort studies that
increased intake, which was statistically significant in 10.58 reported on raw vegetables, four reported non-significant
109 152 156 164 171 232 246-248
No studies showed statistically sig- reduced risk,146 214 253 254 two reported non-significant
nificant increased risk. Meta-analysis was possible on 6 case- increased risk,214 252 and the other reported a significant
control studies, giving a summary effect estimate of 0.40 increased risk.148 Two of the increased risk estimates, includ-
(95% CI 0.190.82) per 100 g/day, with high hetero- ing the one that reached statistical significance, were strat-
geneity.109 152 171 232 244 250 ified for women only. Meta-analysis was possible on four
estimates from three studies (not including the one that was
White or pale vegetables statistically significant), giving a summary effect estimate of
This incorporates a wide range of vegetables. For example, 0.80 (95% CI 0.541.18) per 100 g/day, with no hetero-
in Japan white vegetables such as daikon (radish) are com- geneity (figure 4.2.12).
monly consumed. Descriptions used for this exposure were Of the 25 case-control studies that reported on raw veg-
white vegetables, pale green or light green vegetables, and etables, 21 reported decreased risk,109 129 161 162 167 172 174 183
raw chicory. 184 191 219 225 226 243 247 248 250 255 256 258 260 261 264
which was sta-
Both cohort studies showed non-significant decreased risk tistically significant in 13.129 161 172 174 225 226 243 247 248 256 260 261
with increased intake.150 214 Meta-analysis was possible on 264
None of the remaining 4 studies that reported increased
both studies, giving a summary effect estimate of 0.49 (95% risk reached statistical significance.257 259 262 263 Meta-analy-
CI 0.241.01) per 100 g/day, with no heterogeneity (figure sis was possible on 14 independent estimates from 13 case-
4.2.11). control studies, giving a summary effect estimate of 0.50
All six case-control studies showed decreased risk with (95% CI 0.380.65) per 100 g/day, with moderate hetero-
increased intake,157 165 169 226 228 243 which was statistically sig- geneity (figure 4.2.12).
nificant in three.165 169 243 Meta-analysis was possible on three A dose-response relationship was apparent from case-
studies, giving a summary effect estimate of 0.57 (95% CI control but not cohort data (figure 4.2.13).
0.321.02) per 100 g/day, with high heterogeneity, which Of the three ecological studies, two reported statistically sig-
was caused by varying size, not direction of the effect (figure nificant reduced risk208 238 and one reported a non-significant
4.2.11). increased risk with increased raw vegetable consumption.202

P A R T 2 E V I D E N C E A N D J U D G E M E N T S

Non-starchy vegetables and fruits starchy vegetables from this grouping); in addition, there
All five cohort studies showed decreased risk for the high- was 1 relevant pooling project publication.336 Three cohort
est intake group when compared to the lowest,144 146 148 253 studies337-339 and 1 case-control study321 investigated non-
which was statistically significant in two,253 265 and in starchy vegetables specifically; 5 cohort studies285 292 299 340
men, but not women, in a third study.148 Meta-analysis was 341
and 17 case-control studies65 301 307 312 320-322 326 330 342-350
possible on two cohort studies, giving a summary effect esti- investigated green, leafy vegetables (excluding cruciferous);
mate of 0.81 (95% CI 0.581.14) per 100 g/day.146 253 All 2 cohort studies investigated non-starchy root vegetables and
six case-control studies showed decreased risk with increased tubers289 291; and 6 cohort studies,285 289 293 299 339 341 21 case-
intake,158 161 162 164 257 266 267which was statistically significant control studies,65 261 304 307 313 320-322 325-327 342 344 346-348 351-358
in four.158 162 164 257 Meta-analysis was possible on two case- and 1 ecological study333 investigated carrots specifically.
control studies, giving a summary effect estimate of 0.79
(95% CI 0.630.99) per 100 g/day.162 267 Total vegetables
The stomach is a particularly unusual chemical environ- Out of 19 effect estimates from 17 cohort studies, 14 showed
ment and it is possible that, in addition to the general mech- reduced risk with higher levels of vegetable consumption,282
anisms described below, specific mechanisms apply, for 283 286-289 291-297 299 300
which was statistically significant in 3286
instance, in relation to nitrosamine formation. 297 299 300
and in women only in another285; 1 reported no
effect on risk,298 2 showed increased risk,284 290 none of which
A substantial amount of evidence is available, was statistically significant, and 2 showed non-significant
including on specific subtypes, particularly green- increased risk in men but not women.285 296 Meta-analysis
yellow vegetables, with a dose-response relationship in was possible on 10 studies, all of which adjusted for smok-
case-control, but not cohort data. There is evidence for ing, giving a summary effect estimate of 0.95 (95% CI
plausible mechanisms. Vegetables probably protect 0.920.98) per 80 g serving/day, with no heterogeneity.282
against stomach cancer. 283 285-287 292 296 297 300
Two studies did not adjust for smoking,
1 of which showed a non-significant lower vegetable intake
Nasopharynx in cases than in controls,295 and the other reported no effect
Five case-control studies272-276 and two ecological studies51 on risk.298
investigated non-starchy vegetables and nasopharyngeal Pooled analysis from 8 cohort studies (over 430 000 par-
cancer; a further four case-control studies investigated green ticipants, followed up for 6 to 16 years, more than 3200 lung
vegetables.278-281 Preserved vegetables were excluded from cancer cases) showed a non-significant reduced risk when
all categories. comparing high against low intake groups (0.88, 95% CI
Eight of the case-control studies reported reduced risk 0.781.00), with a p value for trend of 0.12.336
when comparing high against low intake groups,272 273 275 276 Out of 27 case-control studies, 17 showed reduced risk
which was statistically significant in three of the non- with higher levels of vegetable consumption,301-304 306-312 314
starchy vegetable studies272 275 276 and in two of the green 316 317 319 322 325-331
which was statistically significant in 8303
vegetable studies.279 280 One other study stated that there was 304 306 308-310 314 316 319 325-328
; 7 studies showed non-significant
no significant association.274 All studies were based in China. increased risk305 313 315 318 320 323 324 and 1 study showed no
The ecological studies produced mixed results. One effect on risk.321 Meta-analysis was possible on 10 studies,
showed significant correlations between the consumption all of which adjusted for smoking, giving a summary effect
of fresh vegetables and reduced risk of nasopharyngeal car- estimate of 0.67 (95% CI 0.530.86) per serving/day, with
cinoma after adjusting for age (r2 = -0.77, p = 0.009 among high heterogeneity.303 308 309 313 316 317 323 325 326 328 329 Three
men and r2 = -0.75, p = 0.013 among women).51 The second studies did not adjust for smoking, all of which showed
study showed an increasing risk with increases statistically significant decreased risk.306 316 319
in local consumption of non-starchy vegetables (r2 = 2.36).277 A dose-response relationship was apparent from both
This study did not report any adjustments for potential cohort and case-control data.
confounding variables or whether the finding was signifi- Most of the ecological studies are suggestive of an associ-
cant. ation between increased vegetable consumption and
The general mechanisms through which vegetables could decreased risk.
plausibly protect against nasopharyngeal cancer are outlined
below. Non-starchy vegetables
All three cohort studies reported non-significant reduced risk
The evidence for non-starchy vegetables is sparse but when comparing highest and lowest vegetable intakes, with
generally consistent. There is limited evidence effect estimates of 0.9 (lung cancer mortality, 95% CI
suggesting that non-starchy vegetables protect against 0.611.33),337 0.75 (95% CI 0.411.37),338 and 0.54 (p value
nasopharyngeal cancer. for trend 0.2, squamous and small-cell carcinomas only)
when comparing the highest with the lowest intake
Lung groups.339 The single case-control study reported a non-
Seventeen cohort studies,282-300 27 case-control studies,301-331 significant increased risk when comparing high and low
and 6 ecological studies52 116 332-335 investigated total veg- vegetable intakes.321
etables and lung cancer (some studies did not separate non-

C H A P T E R 4 F O O D S A N D D R I N K S

Green, leafy vegetables Colorectum

All five cohort studies reported reduced risk when compar- Seventeen cohort studies81 359-379 and 71 case-control
ing high to low intake groups,285 292 299 340 341 which was sta- studies investigated non-starchy vegetables and colorectal
tistically significant in one.299 Dose-response meta-analysis cancer.
was possible on three cohort studies, giving a summary effect Of 20 effect estimates from 17 cohort studies that report-
estimate of 0.91 (95% CI 0.890.93) per serving/day, with ed comparisons of the highest and lowest intake groups, 11
no heterogeneity.285 340 The two non-included studies report- were in the direction of reduced risk,81 362 364 366 371-374 376-378
ed high-versus-low effect estimates of 0.89 (95% CI 3 of which were statistically significant.81 366 371 377 One study
0.661.19)292 and 0.45 (95% CI 0.260.78).299 All five showed non-significant decreased risk in women and non-
cohort studies adjusted for smoking. significant increased risk in men.360 The other 8 reported
Of the 17 case-control studies, 12 reported decreased risk65 non-significant increased risk.359 361 363 365 367-370 375 One study
301 307 320 321 326 330 342 343 345-348
(reaching statistical significance stated that there was no significant association.379 Meta-
in 2343 345 348 and 5 reported non-significant increased risk.312 analysis was possible on 9 independent estimates from 6
322 344 349 350
Dose-response meta-analysis was possible on 8 studies, giving a summary effect estimate of 1.00 (95% CI
case-control studies, giving a summary effect estimate of 0.901.11) per 2 servings/day increment, with moderate to
0.96 (95% CI 0.911.02) per serving/day, with moderate- high heterogeneity.360 362-364 366 370
to-high heterogeneity.65 322 326 343 346-349 Some of this hetero- Because of the abundant prospective data from cohort
geneity may be due to variation in exposure classification, studies, case-control studies were not summarised.
with some studies listing green vegetables being included The general mechanisms through which vegetables could
in this category. plausibly protect against colorectal cancer are outlined
Total non-starchy root vegetables and tubers
Both cohort studies reported reduced risk with increased con- A substantial amount of evidence is available but it is
sumption,289 291 with effect estimates of 0.56 (95% CI inconsistent. There is limited evidence suggesting that
0.360.88)289 when comparing the highest with the lowest non-starchy vegetables protect against colorectal
intake groups, and 0.70 (95% CI 0.530.93) when compar- cancer.
ing the third highest quartile with the lowest (the highest
intake group had a non-significant decreased risk).291 Both The Panel is aware that since the conclusion of the SLR, three
studies adjusted for smoking. case-control studies78 261 380 have been published. This new
information does not change the Panel judgement (see box
Carrots 3.8).
All six cohort studies reported reduced risk,285 289 293 299 339 341
which was statistically significant in one (0.4, p value for Ovary
trend 0.003).341 The other, non-significant, risk estimates Five cohort studies,381-385 eight case-control studies,89 386-392
ranged from 0.61 to 0.82.285 289 293 299 339 and two ecological studies393 394 investigated non-starchy
Twenty of the 21 case-control studies showed decreased vegetables, and three cohort studies381-383 and two case-
risk when comparing high against low intake groups,65 261 304 control studies395 396 investigated green, leafy vegetables.
307 313 321 322 325-327 342 344 346-348 351-358
which was statistically
significant in 8.261 304 321 325 327 346 347 351 353 356-358 One study Non-starchy vegetables
reported no effect.320 Meta-analysis on studies that adjusted All of the cohort studies reported decreased risk with
for smoking was possible on 11 studies, giving a summary increased vegetable consumption.381-385 Meta-analysis was
effect estimate of 0.81 (95% CI 0.730.89), per serving/day possible on four cohort studies, giving a summary effect esti-
increment, with high heterogeneity.65 307 313 322 325-327 347 348 351 mate of 0.64 (95% CI 0.330.97) for an increase of one serv-
352 354-357
ing/day, with no heterogeneity.381 383-385 The study that could
There was some evidence of publication bias for both not be included reported an effect estimate of 0.76 (95% CI
cohort and case-control studies. 0.421.37) for the highest intake group when compared with
The single ecological study reported lower mean intake of the lowest.382
carrots in an area of high lung cancer risk.333 Pooled analysis from 12 cohort studies (over 560 000 par-
The general mechanisms through which vegetables could ticipants, followed up for 7 to 22 years, more than 2100 lung
plausibly protect against lung cancer are outlined below. cancer cases) showed a non-significant reduced risk when
comparing high against low intake groups (0.90, 95% CI
A substantial amount of evidence is available but some 0.781.04), with a p value for trend of 0.06.397
studies were not adjusted for smoking. A dose- All of the case-control studies reported reduced risk,89
response relationship is apparent from both cohort and 386-392
which was statistically significant in five.89 386 387 391 392
case-control studies. There is limited evidence One ecological study reported a non-significant positive
suggesting that non-starchy vegetables protect against regression/correlation between continents393 and the
lung cancer. other reported a negative regression/correlation between

P A R T 2 E V I D E N C E A N D J U D G E M E N T S

Green, leafy vegetables Evidence comes from case-control studies only.

All three cohort studies reported decreased risk with There is limited evidence suggesting that non-starchy
increased green, leafy vegetable consumption.381-383 Meta- vegetables protect against endometrial cancer.
analysis was possible on two cohort studies, giving a sum-
mary effect estimate of 0.96 (95% CI 0.881.03) per two General mechanisms non-starchy vegetables
servings/day, with no heterogeneity.381 383 The third cohort Also see Chapter 2. Non-starchy vegetables provide a pletho-
study reported a statistically significant reduced risk (0.44, ra of potentially cancer-preventive substances, including sev-
95% CI 0.250.79) when comparing the highest and lowest eral antioxidant nutrients (such as carotenoids and vitamin
intake groups.382 C), dietary fibre, as well as phytochemicals (such as glu-
Both case-control studies reported reduced risk from cosinolates, dithiolthiones, indoles, chlorophyll, flavonoids,
increased consumption of green, leafy vegetables,395 396 one allylsulphides, and phytoestrogens). Phytochemicals might
of which was statistically significant.396 influence cancer risk through their antioxidant activities,
The general mechanisms through which vegetables could modulation of detoxification enzymes, stimulation of the
plausibly protect against ovarian cancer are outlined below. immune system, antiproliferative activities, and/or modula-
tion of steroid hormone concentration and hormone metab-
Evidence from cohort and case-control studies is olism, to name a few possible mechanisms. Non-starchy
sparse. There is limited evidence suggesting that non- vegetables are also a source of folate, which plays an impor-
starchy vegetables protect against ovarian cancer. tant role in synthesis and methylation of DNA. Abnormal
DNA methylation has been linked to aberrant gene expres-
The Panel is aware that since the conclusion of the SLR one sion and also to cancers at several sites, and may be partic-
case-control study78 has been published. This new information ularly important in rapidly dividing tissues. It is difficult to
does not change the Panel judgement (see box 3.8). unravel the relative importance of each constituent and like-
ly that a protective effect may result from a combination of
Endometrium influences on several pathways involved in carcinogenesis.
Ten case-control studies investigated non-starchy vegetable Carrots are a source of carotenoids, particularly alpha-
consumption.398-407 Seven case-control studies investigated carotene and beta-carotene, as well as other vitamins and
cruciferous vegetables and endometrial cancer.398-400 405 407-410 phytochemicals with potentially protective effects. Tomatoes
Of the 10 studies that reported on non-starchy vegetables, are a source of vitamin C and carotenoids, particularly
7 showed decreased risk when comparing the highest with lycopene. Potential mechanisms of inhibition include the
the lowest intake groups,400-405 407 which was statistically sig- antioxidant properties of carotenoids and ligand-dependent
nificant in 5.400 402-404 407 Two reported a non-significant signalling through retinoid receptors (see chapter
increased risk398 406 and the other showed no effect on risk.399 There is a complex mixture of phytochemicals present in
Meta-analysis was possible on 8 studies, giving a summary whole vegetables and these may have additive and syner-
estimate of 0.90 (95% CI 0.860.95) per 100 g of vegetable gistic effects responsible for anti-cancer activities.
intake/day, with low heterogeneity.399 401-407 A dose-response
relationship was apparent from these data. Allium vegetables
Five out of the seven case-control studies that investigat- Stomach
ed cruciferous vegetables reported reduced risk when com- Two cohort studies,413 414 27 case-control studies,109 129 152 162
paring high to low intake groups,399 405 407-410 which was 164 171 178 182 185 187 191 194 195 232 243-245 247 248 251 266 270 415-419
statistically significant in one.405 The other two studies 2 ecological studies202 208 investigated allium vegetables and
reported non-significant increased risk.398 400 Meta-analysis stomach cancer; 1 cohort study,413 16 case-control studies,109
was possible on five studies, giving a summary effect esti- 129 182 184 195 232 246 247 251 262 418 420-422
and 2 ecological studies203
mate of 0.79 (95% CI 0.690.90) per 100 g/day, with no het- 208
investigated garlic and stomach cancer. There was also 1
erogeneity.399 405 407 409 410 The two studies that could not be relevant intervention study that combined allitridium and
included suggested increased risk, though not statistically selenium supplements.423 424
significant.398 400
A dose-response relationship is apparent from case-control Allium vegetables
data. Both cohort studies reported decreased risk,413 414 which was
The general mechanisms through which vegetables could statistically significant in one.413 Meta-analysis was possible
plausibly protect against endometrial cancer are outlined on both, giving a summary effect estimate of 0.55 (95% CI
below. Cruciferous vegetables contain glucosinolates. Certain 0.350.87) per 100 g/day, with no heterogeneity (figure
hydrolysis products of glucosinolates, including indoles and 4.2.14).413 414
isothiocyanates, have shown anti-carcinogenic properties in Twenty of the case-control studies showed reduced risk
laboratory experiments and in diets in live experiments in when comparing high with low intake groups,129 152 162 164 171
animals.411 The human genotype of glutathione S-transferase 178 182 185 187 191 194 195 232 243 247 248 270 416 418 419
which was sta-
has been shown to have a significant role in the metabolism tistically significant in 12.129 152 162 164 182 187 194 243 248 270 416 418
of these phytochemicals and may therefore influence poten- Four studies showed increased risk,109 245 266 415 which was sta-
tial anti-cancer properties.412 tistically significant in 2,245 and the remaining 3 reported no
significant effect on risk.244 251 417 Meta-analysis was possible

C H A P T E R 4 F O O D S A N D D R I N K S

was a combined selenium/allitridium supplement.423 424 The

Figure 4.2.14 Allium vegetables and stomach cancer;
cohort and case-control studies 5-year follow-up suggested that the intervention was
effective in reducing stomach cancer incidence in men (0.36,
Relative risk (95% CI) 95% CI 0.140.92) but not in women (1.14, 95% CI
0.225.76).423 The statistically significant protective effect
for men had dissipated at the 10-year follow-up.424 (Also see
Dorant 1996 0.55 (0.350.88)
Gonzalez 2006 0.31 (0.0111.69) chapter
Summary estimate 0.55 (0.350.87) Allium vegetables are high in flavonols and organosulphur
Case control
compounds. They also, particularly garlic, have antibiotic
Haenszel 1972 0.49 (0.330.72) properties. Although this may act directly against H pylori
Trichopoulos 1985 0.23 (0.160.35) (a known cause of stomach cancers), a study in humans has
You 1988 0.32 (0.170.59)
Buiatti 1989 0.85 (0.721.00)
not shown this effect.425 It is also possible that antibacteri-
Hansson 1993 0.79 (0.471.34) al effects of garlic might inhibit the secondary colonisation
Ji 1998 Men 0.66 (0.431.02) of the stomach after H pylori-induced atrophy. At present,
Ji 1998 Women 0.69 (0.381.24)
Gao 1999 0.00 (0.000.02)
there is no evidence to support or refute this mechanism. An
De Stefani 2001 0.29 (0.110.81) animal study provides evidence that dietary garlic can reduce
Munoz 2001 0.70 (0.610.80) the severity of H pylori-associated gastritis.426
Takezaki 2001 1.45 (0.742.82)
Sipetic 2003 0.40 (0.200.80)
Lissowka 2004 0.86 (0.671.10) The evidence, though not copious and mostly from
Nan 2005 0.49 (0.280.84)
case-control studies, is consistent, with a dose-
Zickute 2005 0.91 (0.711.16)
Summary estimate 0.59 (0.470.74) response relationship. There is evidence for plausible
mechanisms. Allium vegetables probably protect
against stomach cancer.
0.5 1 2
Relative risk, per 100 g/day
Two cohort studies361 362 and six case-control studies427-435
investigated garlic consumption.
Both cohort studies reported non-significant decreased risk
when comparing the highest with the lowest intake groups,
on 14 studies, giving a summary effect estimate of 0.59 (95% with effect estimates of 0.77 (95% CI 0.511.16)361 and 0.68
CI 0.470.74) per 100 g/day, with high heterogeneity (95% CI 0.461.01) (figure 4.2.15).362
(figure 4.2.14).109 129 152 162 171 178 182 187 191 194 232 247 270 416 All six case-control studies showed decreased risk for the
Both ecological studies reported statistically significant highest consumers of garlic,427-435 which was statistically sig-
decreased risk with increased consumption.202 208 nificant in three (figure 4.2.16).431 432
A statistically significant dose-response relationship is There is considerable preclinical evidence with model car-
apparent from cohort and case-control data. cinogens and transplantable tumours that supports an anti-
cancer effect of garlic and some of its allyl sulphur
Garlic components. Animal studies demonstrate that allyl sulphides
The single cohort study, which was specific to supplemen- effectively inhibit colon tumour formation and also can
tary garlic, showed a non-significant increased risk when inhibit cell growth in the laboratory.436-439
comparing garlic supplement use versus no supplement use
(1.29, 95% CI 0.622.67).413
Fifteen of the case-control studies showed decreased risk
when comparing highest to lowest intake groups,109 129 182 184 Figure 4.2.15 Garlic and colon cancer; cohort studies
195 232 246 247 251 418 420-422
which was statistically significant in
seven.129 182 232 246 247 418 420 422 One study showed a non- Relative risk (95% CI)
significant increased risk.262 Meta-analysis was possible on
five studies, giving a summary effect estimate of 0.41
Giovannucci 1994 Men 0.77 (0.511.16)
(95% CI 0.230.73) per serving/day.109 129 182 232 421
One ecological study showed statistically significant Steinmetz 1994 Women 0.68 (0.461.01)

decreased risk with increased intake208; the other showed no

significant association.203
0.2 0.5 1 2 5
Intervention study Relative risk, highest vs lowest exposure category
The double-blind, randomised trial had an intervention dura-
tion of 3 years, and a 5- and 10-year follow-up, and more
than 5000 participants, all of whom had been identified as
being at increased risk of stomach cancer. The intervention

P A R T 2 E V I D E N C E A N D J U D G E M E N T S Fruits
Figure 4.2.16 Garlic consumption and colorectal cancer;
Mouth, pharynx, and larynx
case-control studies
One cohort study,447 35 case-control studies21 22 24-26 28 30-33 35
Relative risk (95% CI)
36 39-50 59-61 63 64 67 69 72 74 448-450
and 2 ecological studies52 68
investigated fruits and mouth, pharynx, and larynx cancers;
Huetal 1991 0.21 (0.040.98) and 1 cohort study,66 23 case-control studies23 26-29 31 33 34 37-
Le Marchand 1997 Men 0.80 (0.541.19)
39 41-43 45-47 50 63 65 75 451 452
and 1 ecological study52 investigated
Le Marchand 1997 Women 0.90 (0.591.37) citrus fruits. In addition, 1 cohort study54 and 6 case-control
Franceschi 1998 0.90 (0.801.01)
studies33 39 45 55-57investigated non-starchy vegetables and
Levi 1999 0.39 (0.210.71)
fruits in combination (also see evidence on non-starchy
Kamel 2002 0.83 (0.441.55)
vegetables, chapter

0.2 0.5 1 2 5 General fruits

Relative risk, highest vs lowest exposure category
The single cohort study, which adjusted for smoking,
showed a non-significant decreased risk for the highest when
compared to the lowest intake groups, with an effect esti-
mate of 0.82 (95% CI 0.641.04) (figure 4.2.17).447
Most (32) of the case-control studies reported decreased
risk associated with higher intake of fruits,24-26 28 30-33 35 36 39-
48 50 59-61 63 64 69 72 74 448 450
which was statistically significant
The evidence, though not copious and mostly from in 17.26 31 32 35 39-43 46-48 50 63 64 69 72 448 No study reported sta-
case-control studies, is consistent, with a dose- tistically significant increased risk. Meta-analysis was possi-
response relationship. There is evidence for plausible ble on 7 studies (all of which adjusted for smoking), giving
mechanisms. Garlic probably protects against a summary effect estimate of 0.72 (95% CI 0.590.87) per
colorectal cancer. 100 g/day, with high heterogeneity (figure 4.2.17).30 39 42 44
45 69 72
Heterogeneity came from the varying size, not direc-
The Panel is aware that since the conclusion of the SLR, one tion, of effect.
case-control study78 has been published. This new information One ecological study showed a weak inverse correlation
does not change the Panel judgement (see box 3.8). between fruits and oral cancer.68 The other observed inverse
correlations among women for fruit and both oral and laryn-
In addition to this judgement, data on garlic have geal cancers and positive correlations among men for the
contributed to the evidence base for allium vegetables and same two sites.52
stomach cancer (also see chapter 7.5). Carrots
Cervix Figure 4.2.17 Fruits and mouth, pharynx, and larynx
cancer; cohort and case-control studies
Five case-control studies440-444 and one ecological study445
investigated carrots and cervical cancer.
Relative risk (95% CI)
Case-control studies were consistent in showing reduced
risk for the highest levels of consumption, which was sta- Cohort
tistically significant in three.440-442 All studies used hospital- Chyou 1995 0.82 (0.641.04)

based controls and none adjusted for human papilloma virus Summary estimate 0.82 (0.641.04)

status. The single ecological study showed non-significant

Case control
increased risk with high intake of carrots.445
La Vecchia 1991 0.59 (0.410.86)
Some carotenoids, including beta-carotene and alpha- De Stefani 1994 0.58 (0.301.12)
carotene, which are found at high levels in carrots, are pre- De Stefani 2000 0.66 (0.520.85)
cursors of vitamin A. They also have properties independent Bosetti 2002 0.92 (0.880.97)

of their pro-vitamin A activity. Carotenoids are recognised Marchioni 2003 0.96 (0.861.08)

antioxidants and low blood levels of dietary antioxidants are Gaudet 2004 0.90 (0.741.10)
Kapil 2005 0.17 (0.100.31)
associated with human papilloma virus persistence.446
Summary estimate 0.72 (0.590.87)

The evidence, from case-control studies only, is sparse

but consistent. There is limited evidence suggesting 0.1 0.5 0.75 1 1.5

that carrots protect against cervical cancer. Relative risk, per 100 g/day

Data on carrots have contributed to the evidence base for

non-starchy vegetables and mouth, pharynx, and larynx can-
cers (chapter 7.1) and lung cancer (chapter 7.4). Also see

C H A P T E R 4 F O O D S A N D D R I N K S

Figure 4.2.18 Citrus fruits and mouth, pharynx, and Figure 4.2.20 Citrus fruits and mouth, pharynx, and larynx
larynx cancer; case-control studies cancer; case-control studies: dose response

Relative risk (95% CI)

Franco 1989
Franco 1989 0.72 (0.560.93)
Zheng 1993 0.45 (0.310.66) Zheng 1993
Levi 1998 0.88 (0.820.95)
Levi 1998
De Stefani 2000 0.27 (0.140.52)
Bosetti 2002 0.91 (0.850.97) De Stefani 2000
Pisa 2002 0.71 (0.510.99)
Bosetti 2002
Gaudet 2004 0.89 (0.731.10)
Summary estimate 0.76 (0.660.87) Pisa 2002

Gaudet 2004

0.5 0.75 1 1.5 2

Relative risk, per 50 g/day

0 100 200 300

Citrus fruits (g/day)

Figure 4.2.19 Fruits and mouth, pharynx, and larynx

cancer; case-control studies: dose response

LaVecchia 1991
came from the varying size, not direction, of effect.
De Stefani 1994
A dose-response relationship was apparent from case-
De Stefani 2000 control but not cohort data for both general and citrus fruits
Bosetti 2002 (figures 4.2.19 and 4.2.20). There is some suggestion that
Marchioni 2003
the greatest effect appears to be with the first increment.
That is, some fruit consumption confers a protective effect
Gaudet 2004
compared to none. However, it is not clear that the effect
Kapil 2005 continues in a linear fashion with increased doses.
One ecological study found no significant association
between citrus fruit consumption and cancer mortality in
men or women.52
0 200 400 600 800 1000
Studies that reported on combined intake of non-starchy
Fruits (g/day)
vegetables and fruits showed evidence of an association with
decreased risk (see chapter
The general mechanisms through which fruits could plau-
sibly protect against mouth, pharynx, and larynx cancer are
outlined below.

The evidence, including on fruit subtypes, though

Citrus fruits mostly from case-control studies, is consistent, with a
The single cohort study, which was specific to oranges and dose-response relationship. There is evidence for
was adjusted for smoking, showed a non-significant plausible mechanisms. Fruits probably protect against
decreased risk for the highest when compared to the lowest mouth, pharynx, and larynx cancers.
intake groups, with an effect estimate of 0.50 (95% CI
0.301.00), with a p value for trend of 0.03.66 This risk esti- The Panel is aware that since the conclusion of the SLR, two
mate was for cancers of the upper aerodigestive tract. cohort studies76 77 and one case-control study79 have been pub-
Twenty-two of the case-control studies showed decreased lished. This new information does not change the Panel judge-
risk associated with higher intake of fruits,23 27-29 31 33 34 37-39 ment (see box 3.8).
41-43 45-47 50 63 65 75 451 452
which was statistically significant in
13.23 27-29 31 33 37 39 42 43 47 50 75 The 23rd study showed no effect Oesophagus
on risk.26 Meta-analysis was possible on 7 studies (all of Four cohort studies,80 82 83 447 36 case-control studies22 40 60
which adjusted for smoking), giving a summary effect esti- 84 86 87 89 94-96 98-100 102 104 108-110 112-115 125-129 134-136 138 453-456
mate of 0.76 (95% CI 0.660.87) per 50 g/day, with high 7 ecological studies52 68 116 118 119 234 457 458 investigated fruits
heterogeneity (figure 4.2.18).23 29 37 39 42 45 65 Heterogeneity and oesophageal cancer; 1 cohort study,66 16 case-control

P A R T 2 E V I D E N C E A N D J U D G E M E N T S

Figure 4.2.21 Fruits and oesophageal cancer; cohort Figure 4.2.22 Fruits and oesophageal cancer;
and case-control studies case-control studies

Relative risk (95% CI) Relative risk (95% CI)

Cohort Jun-lao 1989 1.52 (0.474.91)

Yu 1993 0.99 (0.851.15) Hanaoke 1994 0.30 (0.130.67)
Chyou 1995 0.65 (0.391.08)
Castelletto 1994 0.33 (0.130.80)
Guo 1999 0.90 (0.771.06)
Gao 1999 0.38 (0.0120.03)
Tran 2005 0.80 (0.700.91)
De Stefani 2000 0.53 (0.420.67)

Case control Wolfgarten 2001 0.40 (0.270.59)

Notani 1987 0.99 (0.631.57) Sharp 2001 0.84 (0.740.96)
Victoria 1987 0.66 (0.500.88) De Stefani 2005 0.59 (0.490.71)
Nakachi 1988 Women 0.23 (0.120.45)
Summary estimate 0.56 (0.420.74)
Nakachi 1988 Men 0.31 (0.220.44)
Brown 1988 0.50 (0.290.87)
Jun-lao 1989 1.00 (0.821.22)
0.2 0.5 0.75 1 1.5
De Stefani 1990 0.33 (0.210.52)
Negri 1991 0.30 (0.210.42) Relative risk, per 100 g/day
Tavani 1993 0.40 (0.190.85)
Tavani 1994 0.30 (0.110.85)
Castelletto 1994 0.70 (0.311.57)
Hanaoka 1994 0.50 (0.181.39)
Srivastava 1995 3.15 (1.297.72)
Gimeno 1995 0.55 (0.330.94)
Zhang 1997 0.40 (0.160.98)
De Stefani 1999 0.30 (0.170.52)
Wang 1999 0.51 (0.270.98) Figure 4.2.23 Citrus fruits and oesophageal cancer;
Gao 1999 0.75 (0.361.56) cohort and case-control studies
Cheng 2000 0.08 (0.010.56)
Nayar 2000 0.96 (0.452.05)
Relative risk (95% CI)
De Stefani 2000 0.18 (0.090.37)
Takezaki 2000 0.70 (0.520.94)
Phukan 2001 0.30 (0.042.17)
Kjaerheim 1998 0.50 (0.270.91)
Terry 2001 0.60 (0.360.99)
Wolfgarten 2001 0.32 (0.120.89)
Case control
Takezaki 2001 0.91 (0.481.73)
Brown 1988 0.50 (0.290.87)
Sharp 2001 0.64 (0.251.65)
Cheng 1992 0.10 (0.040.26)
Zhang 2001 0.64 (0.420.97)
Castelletto 1994 1.60 (0.813.15)
Onuk 2002 7.10 (3.2115.73)
Cheng 1995 Non-smokers 0.39 (0.160.97)
Lik 2003 0.08 (0.060.11)
Cheng 1995 Never drinkers 0.59 (0.231.52)
Hung 2004 0.60 (0.400.90)
Zhang 1997 0.70 (0.291.71)
Yang 2005 0.42 (0.190.91)
Launoy 1998 0.54 (0.330.89)
De Stefani 0.21 (0.100.44)
Levi 2000 0.22 (0.090.54)
Bosetti 2000 0.42 (0.250.71)
Terry 2000 0.90 (0.501.61)

0.2 0.5 1 2 5 Chen 2002 0.48 (0.211.10)

De Stefani 2005 0.28 (0.150.54)
Relative risk, highest vs lowest exposure category

0.2 0.5 1 2 5
Relative risk, highest vs lowest exposure category

studies,33 85 86 88 90 97 105 111 113 124 125 128 130 132 133 136 459 and 1 CI 0.81.1; adjusted for smoking),80 and 0.99 (95% CI
ecological study52 investigated citrus fruits. 0.851.15; not adjusted for smoking).82
Thirty-two of the case-control studies reported reduced
General fruits risk for the highest intake groups when compared to the low-
All of the cohort studies reported reduced risk with higher est (figure 4.2.21),22 40 60 84 86 87 89 95 96 98 99 102 104 108-110 113-115
intakes of fruit,80 82 83 447 which was statistically significant 125-129 134-136 138 453-456
which was statistically significant in
in two.83 447 One study reported a statistically significant 24.40 60 84 86 87 89 95 96 102 104 110 113-115 127 134-136 138 454-456 One
dose-response relationship, with a risk estimate of 0.68 study reported statistically significant increased risk,100 one
(95% CI 0.530.88) per 100 g/day after adjustment for reported no effect on risk,112 and one reported no statistically
smoking.447 The other three reported risks for the highest significant association.94 Meta-analysis was possible on eight
intake groups relative to the lowest, with risk estimates of studies, giving a summary effect estimate of 0.56 (95% CI
0.8 (95% CI 0.70.9; not adjusted for smoking),83 0.9 (95% 0.420.74) per 100 g/day, with high heterogeneity (figure

C H A P T E R 4 F O O D S A N D D R I N K S

published. This new information does not change the Panel

Figure 4.2.24 Citrus fruits and oesophageal cancer;
judgement (see box 3.8).
case-control studies

Relative risk (95% CI) Lung

Twenty-five cohort studies,214 216 282-300 337 339 360 461-467 32
Cheng 1992 0.46 (0.390.55) case-control studies,261 303-306 308-318 320-322 324 326-328 330 331 343 346
Castelletto 1994 1.22 (0.791.89) 349 350 352 355 357 358 468-472
and 7 ecological studies52 116 332-334
Cheng 1995 Never drinkers 0.65 (0.480.89) 473 474
investigated fruits and lung cancer.
Cheng 1995 Non-smokers 0.57 (0.430.75)
Levi 2000 0.75 (0.670.84)
Twenty of the cohort studies showed decreased risk for the
Terry 2000 0.97 (0.841.13) highest intake groups when compared to the lowest,214 216 282-
De Stefani 2005 0.62 (0.490.78)
289 291-294 296 297 299 300 337 461-467
which was statistically signifi-
Summary estimate 0.70 (0.560.88) cant in four.216 289 292 300 461 464 Four studies showed
non-significant increased risk290 295 339 360 and the other report-
0.2 0.5 0.75 1 1.5 2 ed no statistically significant association.298 Meta-analysis was
Relative risk, per 50 g/day possible on 14 cohort studies, giving a summary effect esti-
mate of 0.94 (95% CI 0.900.97) per 80 g serving/day, with
low heterogeneity (figure 4.2.25). All but one of these stud-
ies adjusted for smoking.462
Pooled analysis from 8 cohort studies (over 430 000 par-
ticipants, followed up for 6 to 16 years, more than 3200 lung

4.2.22). Heterogeneity may be partially explained by differ-

ential adjustment for confounders between studies.
All seven ecological studies reported reduced risk with Figure 4.2.25 Fruits and lung cancer; cohort and
case-control studies
increased intake,52 68 116 118 119 234 457 458 which was statistically
significant in one.68 458
Relative risk (95% CI)

Citrus fruits Cohort

The single cohort study, which was specific to oranges and Alavanja 2004 (pesticide applicators) 0.89 (0.591.35)
Alavanja 2004 (applicator spouses) 0.65 (0.281.51)
was adjusted for smoking, showed a non-significant Breslow 2000 0.92 (0.711.18)
decreased risk for the highest when compared to the lowest Feskanich 2000 (HPFS) Men 1.05 (0.941.17)

intake groups, with an effect estimate of 0.50 (95% CI Feskanich 2000 (NHS) Women 0.94 (0.871.02)
Fraser 1991 0.47 (0.320.69)
0.301.00), with a p value for trend of 0.03.66 This risk esti- Fu 1997 0.94 (0.671.32)
mate was for cancers of the upper aerodigestive tract; 22 out Holick 2002 0.94 (0.891.00)
Jansen 2004 0.84 (0.651.09)
of 71 cases were oesophageal cancers. Miller 2002 0.96 (0.861.07)
Fifteen of the case-control studies reported decreased risk Olson 2002 0.91 (0.860.96)
for the highest intake groups when compared to the lowest,33 Shibata 1992 0.99 (0.871.14)
Skuladottir 2004 0.95 (0.861.05)
85 86 88 90 97 105 111 113 124 125 130 132 133 136 459
which was statisti- Takezaki 2003 0.63 (0.241.63)
cally significant in 10 (figure 4.2.23).33 85 86 88 97 105 113 132 133 Vorrips 2000 0.92 (0.860.98)
136 459
The other study reported a non-significant increased Summary estimate 0.94 (0.900.97)

risk.128 Meta-analysis was possible on six studies, giving a Case control

summary effect estimate of 0.70 (95% CI 0.560.88) per Axelsson 1996 0.77 (0.461.28)

50 g/day, with high heterogeneity (figure 4.2.24).33 97 128 130 Brennan 2000 1.08 (0.522.28)
De Stefani 2002 0.84 (0.661.06)
132 133
Four of these studies adjusted for smoking.33 97 128 132 Gao 1993 0.45 (0.300.67)
Heterogeneity may be partially explained by differential Hu 2002 1.04 (0.851.27)

adjustment for confounders between studies. Ko 1997 1.00 (0.541.84)

Kreuzer 2002 0.79 (0.451.39)
The single ecological study reported a non-significant Lagiou 2004 0.79 (0.601.04)
increased risk.52 Pawlega 1997 0.01 (0.000.24)

The general mechanisms through which fruits could plau- Rachtan 2002
Raunoi-Ravina 2002
0.49 (0.320.75)
1.19 (0.741.91)
sibly protect against cancer are outlined below. De Stefani 1999 0.67 (0.540.82)
Suzuki 1994 1.33 (0.563.18)
Swanson 1997 0.91 (0.761.09)
The evidence, including on fruit subtypes, though Summary estimate 0.80 (0.680.94)
mostly from case-control studies, is consistent, with a
dose-response relationship. There is evidence for
0.05 1 2
plausible mechanisms. Fruits probably protect against
Relative risk, per serving/day
oesophageal cancer.

The Panel is aware that since the conclusion of the SLR, one
cohort study140 and two case-control studies143 460 have been

P A R T 2 E V I D E N C E A N D J U D G E M E N T S

Of the seven ecological studies, four reported non-

Figure 4.2.26 Fruits and lung cancer; cohort and
significant decreased risk in areas of higher fruit consump-
case-control studies; dose response
tion,52 332 473 474 one reported no consistent association,334 and
two reported non-significant increased risk.116 333
The general mechanisms through which fruits could plau-
Voorrips 2000
sibly protect against cancer are outlined below. In addition,
Takezaki 2003 flavonoids found in fruit directly inhibit expression of
Skuladottir 2004 CYP1A1 (a cytochrome P450 enzyme that helps to
Oslon 2002 metabolise toxins), resulting in decreased DNA damage.475
Miller 2002 Elevated CYP1A1 activity has been associated with increased
Jansen 2004
risk of lung cancer, primarily in smokers.476 The protective
Holick 2002
association of flavonoids is associated with specific CYP1A1
Fu 1997
Fraser 1991
genotypes, which supports the importance of flavonoids and
Feskanich 2000 (NHS) potentially explains heterogeneous results.476 477
Feskanich 2000 (HPFS)
Men The evidence is ample and consistent. A dose-response
Breslow 2000 relationship is apparent from both cohort and case-
Alavanja 2004 control studies and there is evidence for plausible
(Pesticide applicators)
mechanisms operating in humans. The evidence that
Alavanja 2004
(Applicator spouses) fruits protect against lung cancer is convincing.

Case control The Panel is aware that since the conclusion of the SLR, one
Swanson 1997
case-control study478 has been published. This new information
Suzuki 1994
Stefani 1999
does not change the Panel judgement (see box 3.8).
Ruano-Ravina 2002
Rachtan 2002 Stomach
Pawlega 1997 Sixteen cohort studies,71 80 144-147 149 150 213-217 252-254 414 51
Lagiou 2004 case-control studies,89 109 129 151 154 156 158-163 167-169 174-176 178-180
Kreuzer 2002 182 184-187 189-191 193 195 219 221 222 224-227 229 230 246 255-258 260 261 264
Ko 1997
Hu 2002
270 479-482
and 23 ecological studies52 116 118 119 197 198 200-202 204-
Gao 1993 investigated fruits.
209 234 236-240 483-485

De Stefani 2002 Ten cohort studies reported decreased risk for the highest
Brennan 2000 intake groups when compared to the lowest,71 80 144 146 150 214-
Axelsson 1996 217 253 254
which was statistically significant in one,253 and in
women only in a second study.216 Six studies showed
increased risk,145 147 149 213 214 252 414 which was statistically sig-
0 1 2 3 4 5
nificant in one.213 Meta-analysis was possible on eight stud-
Fruits (servings/day)
ies, giving a summary effect estimate of 0.95 (95% CI
0.891.02) per 100 g/day, with low heterogeneity (figure
One of the cohort studies considered in the meta-analysis
above (EPIC, more than 521 000 participants in over 10
European countries) reported results stratified by H pylori
cancer cases) showed a statistically significant reduced risk for status. The effect estimate for the H pylori-negative group
the highest intake group when compared to the lowest (0.77, was 0.72 (95% CI 0.391.33) and 0.98 (95% CI 0.811.2)
95% CI 0.670.87), with a p value for trend of < 0.001.336 for the positive group.140
Twenty-one case-control studies showed decreased risk for Forty case-control studies showed decreased risk for the
the highest intake groups when compared to the lowest,261 highest intake groups when compared to the lowest,89 109 151
303 305 306 308 309 311 312 315 317 318 320-322 324 327 328 331 343 346 349 350 156 158-160 162 163 167-169 174 176 178-180 184 186 187 189-191 195 219 221 222

which was statistically significant in 7.

355 357 358 468 469 472 261 225 226 229 230 246 256-258 260 261 264 270 479-481
which was statistical-
Three studies reported no effect
309 311 324 327 343 346 357 358 468 472
ly significant in 25.89 109 151 158-160 162 163 167-169 174 176 178 186 187
on risk310 316 330 352 and 8 showed increased risk,304 313 314 326 190 191 221 222 226 229 246 256 261 264 479 481
Seven showed increased
470 471
which was statistically significant in 3.304 326 470 Meta- risk,129 161 182 185 193 224 482 which was statistically significant
analysis was possible on 14 case-control studies, giving a in two.182 193 One study showed non-significant increased risk
summary effect estimate of 0.80 (95% CI 0.680.94) per in men and non-significant decreased risk in women.227 Two
serving/day, with moderate heterogeneity (figures 4.2.25 studies showed no effect on risk154 255 and the remaining one
and 4.2.26). All but 2 of these studies adjusted for smoking, reported that there was no significant association.175 Meta-
and exclusion of these 2 studies did not significantly alter analysis was possible on 26 studies, giving a summary effect
the meta-analysis.316 352 estimate of 0.67 (95% CI 0.590.76) per 100 g/day, with

C H A P T E R 4 F O O D S A N D D R I N K S

showed non-significant increased risk in white men and

Figure 4.2.27 Fruits and stomach cancer; cohort and
Japanese men and women, and non-significant decreased
case-control studies
risk in white women.483
Relative risk (95% CI)
The stomach is a particularly unusual chemical environ-
ment and it is possible that, in addition to the general mech-
Cohort anisms described below, specific mechanisms apply, for
Chyou 1990 0.95 (0.841.07)
Botterweck 1998 0.92 (0.831.02)
instance, in relation to nitrosamine formation.486 It is also
Galanis 1998 0.68 (0.510.92) plausible that bioactive constituents in fruit would protect
Fujino 2002 Men 1.01 (0.901.13) against H pylori-induced damage, particularly inflammation,
Fujino 2002 Women 1.12 (0.851.49)
Kobayashi 2002 0.75 (0.551.00)
which is implicated in the development of stomach cancers.
Ngoan 2002 0.94 (0.281.19)
Khan 2004 Men 1.14 (0.284.70)
The evidence is ample and more consistent with a
Gonzalez 2006 1.04 (0.911.19)
Summary estimate 0.95 (0.891.02)
dose-response relationship for case-control studies
than for cohorts. There is evidence for plausible
Case control
mechanisms. Fruits probably protect against stomach
Jedrychowski 1981 0.71 (0.530.95)
Jedrychowski 1986 0.98 (0.661.47) cancer.
You 1988 0.61 (0.480.79)
Burr 1989 Men
Burr 1989 Women
0.53 (0.320.89)
0.39 (0.150.97)
The Panel is aware that since the conclusion of the SLR, three
Coggon 1989 0.48 (0.073.45) case-control studies487-489have been published. This new infor-
De Stefani 1990 0.42 (0.300.61) mation does not change the Panel judgement (see box 3.8).
Kato 1990 Men 0.89 (0.651.22)
Kato 1990 Women 0.85 (0.491.50)
Lee 1990 1.05 (0.791.39) Nasopharynx
Wu-Williams 1990 Men 0.64 (0.371.11) Six case-control studies investigated general fruits and
Hoshiyama 1992 0.54 (0.410.70)
Memik 1992 0.57 (0.360.90)
nasopharyngeal cancers274 275 281 490-492; a further five case-
Cornee 1995 0.75 (0.670.99) control studies investigated citrus fruits.273 278-281 Preserved
De Stefani 1998 0.49 (0.420.58) fruits were excluded from all categories.
Ji 1998 Men 0.49 (0.370.80)
Ji 1998 Women 0.55 (0.370.80)
Of the six case-control studies that investigated general
Gao 1999 5.46 (0.3617.98) fruits, four reported decreased risk for the highest intake
Huang 1999 0.93 (0.801.08)
groups when compared to the lowest,275 281 491 492 which was
Ward 1999 0.90 (0.551.47)
Mathew 2000 0.90 (0.551.47) statistically significant in two.275 491 The other two studies
De Stefani 2001 0.65 (0.530.79) reported that there was no significant effect on risk, with-
Takezaki 2001 0.60 (0.370.97)
out further detail.274 490 All five of the case-control studies
Nishimoto 2002 0.74 (0.621.06)
Lee 2003 0.36 (0.180.73) that investigated citrus fruits reported decreased risk for the
Sipetic 2003 0.17 (0.070.32) highest intake groups when compared to the lowest,273 278-
Suh 2003
Lissowska 2004
0.80 (0.690.92)
0.65 (0.470.91)
four of which were statistically significant.273 278-280
Boccia 2995 2.06 (1.103.84) Preserved fruits were excluded as they introduced sub-
Summary estimate 0.67 (0.590.76) stantial heterogeneity.
The general mechanisms through which fruits could plau-
0.5 1 2
sibly protect against nasopharyngeal cancer are outlined
Relative risk, per 100 g/day
below. In addition, it is possible that active constituents in
fruit could act directly on Epstein-Barr virus infection.493

The evidence, from case-control studies only, is sparse.

There is limited evidence suggesting that fruits protect
against nasopharyngeal cancer.

high heterogeneity (figure 4.2.27). Pancreas

A dose-response relationship is apparent from case-control Six cohort studies,214 216 252 494-496 16 case-control studies,219
but not cohort data. There is statistically significant hetero- 497-511
and 8 ecological studies52 197 485 512-515 investigated
geneity between study types. fruits and pancreatic cancer.
Eighteen ecological studies showed decreased risk with All six cohort studies showed decreased risk for the high-
increased intake of fruits,116 118 197 200 201 204-208 234 237-239 484 485 est intake groups when compared to the lowest,214 216 252 494-
which was statistically significant in eight.204-208 237 Four stud- 496
which was statistically significant in one.496 Meta-analysis
ies showed increased risk with increased intake,52 118 119 202 was possible on three cohort studies, giving a summary effect
239 240
which was statistically significant in one.240 Two stud- estimate of 0.92 (95% CI 0.811.04) per 100 g/day, with no
ies showed non-significant decreased risk in women and heterogeneity.216 494 495
non-significant increased risk in men209 236; one study Eleven case-control studies showed decreased risk for the
showed non-significant decreased risk in men and non- highest intake groups when compared to the lowest,219 497 498
significant increased risk in women198; and one study 500 501 503-509 511
which was statistically significant in four,503

P A R T 2 E V I D E N C E A N D J U D G E M E N T S

504 508 511

and in men but not women in a fifth study,506 and sex, a statistically significant decreased risk was apparent in
in women but not men in a sixth.501 One study reported a sta- women (0.81, 95% CI 0.850.98 per serving/day based on
tistically significant increased risk for men and a statistically five studies), with low heterogeneity.
significant decreased risk for women.510 No other study Because of the abundant prospective data from cohort
reported statistically significant increased risk. Meta-analysis studies, case-control studies were not summarised.
was possible on eight case-control studies, giving a summa- The mechanism for this sex difference is unknown. There
ry effect estimate of 0.89 (95% CI 0.820.98) per 100 g/day, is speculation the mechanism could be related to the (part-
with high heterogeneity.497 498 502 503 505 506 508 510 Heterogeneity ly understood) explanation for protective effects observed in
could be partly explained by proxy reporting, poor study qual- postmenopausal women provided with hormone replace-
ity, and varying adjustment for known confounders. ment therapy. Another possibility is that the result could be
A dose-response relationship is apparent from case-control, artifactual if men are poorer at reporting their diets than
but not cohort data. women.
Ecological studies show no consistent association.52 197 485 The general mechanisms through which fruits could plau-
sibly protect against colorectal cancer are outlined below.
The general mechanisms through which fruits could plau-
sibly protect against pancreatic cancer are outlined below. There is a substantial amount of evidence but it is
The evidence is inconsistent. There is limited evidence inconsistent. There is limited evidence suggesting that
suggesting that fruits protect against pancreatic cancer. fruits protect against colorectal cancer.

The Panel is aware that since the conclusion of the SLR, one The Panel is aware that since the conclusion of the SLR, one
cohort study516 has been published. This new information does cohort530 and five case-control studies 261 380 531-533 have been
not change the Panel judgement (see box 3.8). published. This new information does not change the Panel
judgement (see box 3.8).
Liver cancer
One cohort study216 517 and five