ABSTRACT National surveys conducted since 1982 were used to assess maternal and child obesity in Latin
American and Caribbean countries and in U.S. residents of Mexican descent. Obesity in women, a body mass
index (BMI) ¢30 kg/m2, was 3% in Haiti, 8–10% in eight Latin American countries and 29% in Mexican Americans.
Median BMI for Latin American women were near or above the 50th percentile of the general U.S. population;
values exceeded the 75th percentile in the case of Mexican Americans. The prevalence of overweight (ú1 SD
above mean weight-for-height) in children 1–5 y of age ranged from 6% in Haiti to 24% in Peru among 13
countries. Overweight occurred in 24% of Mexican-American children. Prevalences of overweight in children and
of obesity in women were greater in urban areas and in households of higher socioeconomic status. Overweight
in children increased with higher maternal education; however, in some countries, obesity in women decreased
with higher education. No general pattern of change over time was observed in eight countries in overweight in
children. Obesity in women increased in the three countries with such data and in Mexican-American women and
children. There was a tendency for greater national incomes to be associated with greater obesity levels in women
and with lower levels of stunting in children. Levels of obesity in the region indicate a public health concern,
particularly among women, considering that studies have identified mortality and morbidity risks associated with
obesity in adults. J. Nutr. 128: 1464–1473, 1998.
KEY WORDS: • overweight • obesity • Latin America and Caribbean • Mexican Americans
Undernutrition remains the nutrition problem of greatest creases in levels of fatness and obesity. In addition to diet,
concern in developing countries. However, there is increased reduced levels of physical activity, increased use of alcohol
recognition, particularly in Latin America, that dietary pat- and tobacco and increased stress, particularly in the rapidly
terns and lifestyles are changing dramatically, and that as a growing cities of developing countries, are among the determi-
result, chronic and degenerative diseases are becoming im- nants of obesity and other chronic and degenerative diseases.
portant public health concerns (Monteiro et al. 1995, Popkin Obesity is increasingly recognized as a growing problem
1994, Sinha 1995, Walker 1995). These changes are occurring in Latin America. Using national survey data from Brazil,
at a time when undernutrition levels, although reduced over researchers have shown that obesity increased between 1974
the years, still afflict large sectors of the population (Adminis- and 1989 in adults (Monteiro et al. 1995, Sichieri et al. 1994)
trative Committee on Coordination/Sub-Committee on Nutri- but not in children (Monteiro et al. 1995). Defining obesity
tion 1992 and 1997). as a body mass index (BMI)5 ¢30 kg/m2, prevalences increased
Popkin (1994) notes that authors have emphasized the de- from 2.5 to 4.8% in men and from 6.9 to 11.7% in women
mographic (i.e., from high fertility and high childhood mortal- (Sichieri et al. 1994). Obesity increased with family income
ity to low fertility and low childhood mortality) and epidemio- in children (Monteiro et al. 1995) and adults (Monteiro et al.
logic (i.e., shift in causes of mortality from infections to 1995, Sichieri et al. 1994). However, the relationship between
chronic diseases) changes occurring in many developing coun- adult obesity and income was nonlinear; at low and middle
tries while neglecting what he calls the concurrent ‘‘nutrition income levels, obesity increased with income but at high levels
transition.’’ The latter includes adoption of ‘‘Western’’ diets the pattern was reversed, just as in developed countries (Mon-
(i.e., high in saturated fats, sugar and refined foods) and in- teiro et al. 1995). Results from the 1988 National Nutrition
Survey of Mexico suggest that obesity in women was as com-
1
mon in Mexico as in the United States in the 1970s (Hernan-
Presented in preliminary form at the XI Congreso de la Sociedad Latinoamer-
icana de Nutrición, November 9–15, 1997, Guatemala City, Guatemala.
dez et al. 1996). Approximately 17% of Mexican women aged
2
Supported by the Food and Nutrition Program of the Pan American Health
Organization and by the World Bank.
3 5
The costs of publication of this article were defrayed in part by the payment Abbreviations used: BMI, body mass index; DHS, Demographic and Health
of page charges. This article must therefore be hereby marked ‘‘advertisement’’ Surveys; GNP, gross national product; NCHS, National Center for Health Statis-
in accordance with 18 USC section 1734 solely to indicate this fact. tics; NHANES, National Health and Nutrition Examination Survey; PI, poverty
4
To whom correspondence should be addressed. index; SES, socioeconomic status.
1464
12–49 y exceeded the 85th percentile, or a BMI of Ç27 kg/ contain height and weight information for children õ5 y of age,
m2, of the distribution found in women measured in the 1976– sometimes for women of reproductive age (15–49 y) and almost never
80 Second National Health and Nutrition Examination Sur- for men. Most surveys used were collected by the Demographic and
vey (NHANES). The Ministry of Public Health of Costa Rica Health Surveys (DHS/IRD, 1992) program, which has assisted coun-
tries since 1984 in conducting national surveys on fertility, family
reported increases at the national level in the prevalence of planning, and maternal and child health (Institute for Resource De-
obesity (¢25 BMI) between 1982 and 1996 of 34.6–45.9% velopment/MACRO International 1990). In these surveys, women
(Ministerio de Salud 1997). National surveys in different Ca- of childbearing age were interviewed or measured with the use of
ribbean countries found that 7–20% of males and 22–48% of standard survey instruments; the areas of data collection were family
females ú15 y of age exceeded 120% of reference weight for planning knowledge, attitudes and practices; maternal and child
height, or a BMI of Ç27 kg/m2 (Sinha 1995). In addition, health; nutritional status of women and their children; and social
serial data from Barbados showed increases in the prevalence and economic background indicators (Institute for Resource Develop-
of obesity from 7 to 16% in males and 33 to 38% in females ment/MACRO International 1990). DHS surveys are available to
between 1969 and 1981 (Sinha 1995). investigators through the World Wide Web [http://www.macroint.-
Results from studies of selected samples complement those com/dhs/].
We used 16 DHS data sets from 9 Latin American and Caribbean
from national surveys. In Costa Rica, 14% of men in a rural countries, data from 8 other national nutrition and health surveys
area exceeded 130% of reference weight for height (i.e., a BMI from 5 Latin American countries, and data from Mexican Americans
of Ç30 kg/m2) compared with 21% in an urban area; in (i.e., persons of Mexican origin living in the United States) who
women, the corresponding values were 33 and 39% (Campos were included in two national nutrition and health surveys from the
et al. 1992). In Santiago, Chile, the prevalence of obesity, United States (Table 1). All 26 data sets included anthropometric
defined as ú120% of reference weight for height, was 20% in data for children but only 14 data sets contained anthropometric data
men and 30% in women (Atalah 1993). Although socioeco- for women (Table 1). Information about Mexican Americans was
nomic status (SES) was unrelated to obesity in men, it was included for comparison because this population, of Latin American
related negatively in women such that levels were Ç10% for origin, has a diet and a lifestyle characteristic of an industrialized
country. Not all national surveys carried out in Latin America since
high SES and 40% for low SES. There was also an interaction 1982 were accessible to us. However, those included constitute the
with stature; the highest prevalence of obesity was found largest compilation ever assembled to assess the extent of obesity in
among low SES women of short stature, nearly 50% (Atalah Latin America and the Caribbean.
1993). In a study of school children from Santiago, the preva- The variables of interest in women were height, weight, age, fam-
lence of obesity, defined as ú120% of reference weight for ily, socioeconomic status, residence (urban or rural) and education.
height, was 8.9% and increased as a function of income (Iva- Only records with complete data for all variables were used. The age
novic et al. 1987). In an urban community from Trinidad, range in women was restricted to between 15 and 49 y, the focus of
Beckles et al. (1985) found that obesity, defined as a BMI virtually all available national surveys. Pregnant women were ex-
¢30, was greater in women than in men and greatest among cluded. Children 12–60 mo of age were selected for study, but some
those of African descent. Obesity was present in 8, 7 and 4% surveys provided data only for children 12–36 mo. Anthropometric
Z-scores were computed for children relative to the WHO/National
in men of African, mixed and Indian ancestry, respectively; Center for Health Statistics (NCHS) reference population (Dibley
in women, the values were 32, 29 and 25%, respectively. et al. 1987) using the Anthropometric Software Package Tutorial
The above results suggest that obesity is a problem in Latin Guide and Handbook (Jordan 1990). Records with height-for-age,
America, particularly in women. However, varying definitions weight-for-age, or weight-for-height Z-scores ú /5 or õ 05 were
of obesity and different age ranges of the subjects studied make excluded from the analyses. Populations õ5000 were designated as
cross-country comparisons difficult. Also, estimates of obesity ‘‘rural’’ and those above as ‘‘urban’’ in the DHS datasets; similar
have been reported for only a few of the many national nutri- criteria were applied to non-DHS datasets. Maternal education was
tion surveys that have been carried out in the region. This coded as low (primary school or less) and high (at least 1 y of second-
omission reflects the fact that most surveys have been used ary schooling). An index of socioeconomic status (SES) was calcu-
lated for individual families in each dataset based on amenities (e.g.,
to provide information about undernutrition in women and electricity or water), possessions (e.g., radio, television or bicycle)
children. The scarcity of information on men and school age and characteristics of the home (e.g., type of floor or roof). The
children also reflects the focus of most nutrition surveys on calculation was particular to each country and thus, the socioeco-
women and young children. nomic index is not valid across countries; rather, it reflects relative
It is important for policy makers to have accurate informa- socioeconomic status within each country. The poverty index in the
tion not only about the level of obesity at the country level, datasets from the United States was generated by NCHS and is a
but about how values vary across subgroups of the population continuous variable based on income, estimates of living costs appro-
as defined on the basis of region, urban/rural residence, levels priate for the area of residence and date of measurement, and family
of education and socioeconomic status. Also useful to policy composition. A poverty index (PI) ¢1.0 implies that the family
makers is information about changes in the extent of obesity should be able to meet its basic needs; a PI õ1 was used as the
definition of low SES.
over time. Obesity in children was defined as weight-for-height Z-scores ú2
The primary objective of this study is to estimate preva- SD as recommended by WHO (1995). In addition, overweight in
lences of obesity in women and children from Latin American children was defined as a Z-score value ú1 SD above the WHO/
countries based on national survey data collected since 1982. NCHS mean weight-for-height. Wasting and stunting were defined
Uniform definitions of overweight and obesity are used to as Z-scores õ2 SD below the reference mean for weight for height and
facilitate comparison across countries. A second objective is height, respectively. By definition, Ç15.9% of cases in the reference
to investigate how obesity varies by area of residence, socioeco- population were overweight and Ç2.3% were obese. Obesity in
nomic level and education. A third objective is to estimate women was defined as a BMI ¢30 kg/m2. Also, estimates were pro-
trends in obesity. vided using the definition of ¢27.3 kg/m2 to facilitate comparison
with the literature and for grades 1, 2 and 3 of obesity as defined by
WHO (1990 and 1995) (1 Å 25.0–29.9, 2 Å 30.0–39.9 and 3
METHODS Å ¢40 kg/m2).
Sample weights were used in all cases where applicable. The analy-
National nutrition surveys from Latin America and the Caribbean ses of relationships between relative weight and SES, education and
collected since 1982 were the focus of study. These surveys typically area of residence focused on obesity (¢30 BMI) in women but on
/ 4W43$$5038 08-04-98 14:06:35 nutra LP: J Nut October
1466 MARTORELL ET AL.
TABLE 1
National nutrition and health surveys included in this study1
overweight (ú1 SD) in children because the prevalence of obesity obesity as a BMI ¢25 kg/m2, as in the WHO classification,
(ú2 SD) in children was low and statistical power consequently poor. led to large numbers of women being classified as obese, spe-
Odds ratios for overweight or obesity (0 Å No, 1 Å Yes) were esti- cifically, over one third of Latin American women and more
mated in multivariate logistic regressions that included area of resi- than one half of Mexican-American women.
dence (0 Å rural, 1 Å urban), SES (0 Å low, 1 Å medium or high)
and education (0 Å none or primary, 1 Å secondary or higher) as
Median BMI values at various ages for each sample are
independent variables. In analyses of overweight in children, gender plotted in Figure 1 relative to the U.S. distribution published
(0 Å male, 1 Å female) and age in months were included as additional by Frisancho (1990), which is based on data from the first
covariates. In analyses of obesity in women, age and age squared were (1971–1973) and second (1976–1980) National Health and
included in the models because these provided a better fit than age Nutrition Examination Surveys (NHANES). Medians for all
alone. Multivariate analyses were restricted to the most recent na- countries, with the exception of Haiti, were near or above the
tional survey when countries had more than one survey. Statistical 50th percentile of the U.S. distribution. Mexican American
significance was defined as P õ 0.05. women had medians that were closer to the 75th percentile,
Trends for overweight in children were estimated for eight Latin whereas medians for Haitian women were closer to the 25th
American countries and for Mexican Americans. In two countries, percentile.
Guatemala and Colombia, children 12–36 mo of age were measured
in the first survey and children 12–60 mo of age were measured in The percentage of obese (BMI ¢30) women is given in
the second survey. Restricting the second survey to 12–36 mo pro- Table 3 by area of residence, SES, and educational level. The
vided similar results to those for 12–60 mo, reflecting that overweight risk of obesity (BMI ¢30) by category of the social variables
prevalences are similar over the 12- to 60-mo range. Unfortunately, is given in Table 4. Results from unadjusted models (see Table
trends in obesity in women could be estimated only for three Latin 4, footnote 1) indicate that obesity was 1.5–4 times more
American countries and for Mexican Americans. common in urban areas except in Brazil, Colombia, Mexico
Finally, analyses were carried out at the country level to explore and the U.S. where obesity rates were similar in urban and
whether overweight and obesity varied by the level of child stunting rural areas. High SES and obesity covaried in all samples ex-
or the gross national product (GNP) per capita in 1992. Information cept among Mexican and Mexican-American women where
about national incomes was obtained from the State of the World’s
Children, 1995 (UNICEF 1995).
SES bore no relationship to obesity. Better education was asso-
ciated with greater obesity in Haiti, Guatemala and Peru, some
of the poorest countries, but there was no relationship in Bo-
RESULTS livia, Colombia, the Dominican Republic, Honduras and in
Obesity among women. Sample sizes, age and anthropo- Mexican Americans; in Brazil and Mexico, there was more
metric characteristics of the women are given in Table 2. obesity in poorly educated women. Multivariate models atten-
Mean ages were very similar across surveys, but body sizes uated the associations, particularly for education. Controlling
varied markedly. Women from Guatemala, Peru and Bolivia for area of residence and SES, there was a tendency for higher
were very short in comparison to women from Haiti and the prevalences of obesity in poorly educated women except in
U.S. Mean BMI were between 23 and 25 kg/m2 for all but Haiti and Guatemala where more education was associated
Mexican Americans who were the heaviest at 27.1 kg/m2 and with increased obesity. Urban residency and high SES were
Haitian women who were the leanest at 21.2 kg/m2. Similarly, strongly and independently related to greater levels of obesity
the extent of obesity (¢30 BMI) was lowest for Haitian (2.6%) in most countries. Only in Mexican Americans were all three
and greatest for Mexican-American women (28.7%). The social variables unrelated to obesity levels.
range in the prevalence of obesity was 5–40% when a BMI Overweight and obesity in children. Sample sizes, age, and
¢27.3 kg/m2 was used as the criterion for obesity. Defining anthropometric characteristics are given for children in Table
/ 4W43$$5038 08-04-98 14:06:35 nutra LP: J Nut October
OBESITY IN LATIN AMERICA 1467
TABLE 2
Age and anthropometric characteristics of Latin American and Caribbean women 15–49 y old
y kg cm (kg/m2) %
Bolivia (1994) 2347 28.6 (7.0) 55.4 (9.3) 151.0 (5.8) 24.3 (3.7) 16.8 7.6 26.2 7.1 0.5
Brazil (1996) 3158 28.2 (6.5) 58.6 (11.1) 156.3 (6.5) 24.0 (4.3) 19.8 9.7 25.0 9.3 0.4
Colombia (1995) 3319 28.3 (6.7) 58.4 (10.1) 154.5 (6.0) 24.5 (4.0) 21.6 9.2 31.4 8.9 0.3
Dominican
Republic (1996) 7356 28.9 (9.6) 59.8 (12.5) 156.8 (6.4) 24.3 (4.9) 23.3 12.1 26.0 11.3 0.8
Guatemala (1995) 4978 28.9 (7.2) 53.1 (10.1) 148.1 (6.3) 24.2 (3.9) 17.9 8.0 26.2 7.5 0.4
Haiti (1994/1995) 1896 30.1 (7.5) 53.2 (9.3) 158.4 (7.0) 21.2 (3.4) 5.2 2.6 8.9 2.6 0.0
Honduras (1996) 885 30.0 (8.2) 54.4 (12.2) 152.1 (6.3) 23.5 (4.7) 18.7 7.8 23.8 7.3 0.5
Mexico (1987) 3681 28.9 (8.7) 57.7 (10.5) 156.2 (5.5) 23.7 (4.3) 20.0 10.4 23.1 9.8 0.6
Peru (1996) 10747 29.5 (6.8) 56.7 (8.8) 150.4 (5.6) 25.1 (3.6) 22.8 9.4 35.5 9.0 0.4
United States2
(1988–1994) 1404 29.4 (10.8) 67.7 (17.1) 158.1 (6.6) 27.1 (6.7) 39.9 28.7 26.5 24.5 4.1
1 Means (SD).
2 Data for Mexican-American women only.
5. Stunting was pronounced in several countries, whereas wast- The percentage of overweight children is given in Table 6
ing was less common. The country with the greatest proportion by area of residence, SES, maternal education and gender. As
of stunted children was Guatemala (56%); the highest propor- in the case of women, the risk of overweight is given by cate-
tion of wasted children was observed in Haiti (8%). The per- gory of each social variable in Table 7. In general, social
centage of overweight children (Z score ú 1 for weight-for- variables were not as strongly related to relative body mass in
height) or obese (Z score ú 2) was highest in Mexican Ameri- children as they were in women. Urban residency, high SES
cans and lowest in Honduras and Haiti. The samples with and greater maternal education were associated with greater
higher rates of obesity than the 2.3% value found in the refer- risk of overweight in children in most countries.
ence population were the U.S. Mexican Americans (7.4%), There was a tendency for obesity to be 10–20% more com-
Peru (4.7%), the Dominican Republic (4.6%), Brazil (4.1%), mon in girls than in boys, but this was significant only in the
Mexico (3.9%) and Paraguay (2.7%). Dominican Republic, Guatemala and Peru (Table 7).
FIGURE 1 Median body mass index of women 15–49 y old relative to the 25th, 50th and 75th percentiles (dashed lines) of the distribution in
the first and second National Health and Nutrition Examination Surveys (Frisancho 1990). The sample for the United States includes data for Mexican-
American children only.
TABLE 3
Percentage of obese [body mass index (BMI) ¢30] women from Latin America and the Caribbean and denominator by area of
residence, socioeconomic status (SES) and educational level1
Trends in obesity and overweight. Repeated surveys for weight is plotted in Figure 2 for countries with data for more
women were available for only four samples. Prevalences of than one survey. Overweight increased in Mexican Americans,
obesity (BMI ¢30) increased in all. In Brazil, values increased but the pattern was mixed for Latin American countries; some
from 7.6 to 9.2% between 1989 and 1996; small increases were reported increases and other decreases.
also observed in Peru between 1992 and 1996, 8.8–9.4%. Relationship with stunting and national incomes. At the
Greater increases occurred between 1991 and 1996 in the national level, there was a significant relationship between the
Dominican Republic, 7.3–12.1%. Among Mexican Ameri- level of overweight in children and per capita GNP (r Å 0.53,
cans, the increase in obesity from 1982–1984 to 1988–1994 P Å 0.05, n Å 14); however, excluding the data point for
was large, from 19.5 to 28.7%. More information is available Mexican Americans (identified as U.S. in the figures), an out-
from repeated surveys for children. The percentage of over- lier in terms of income, reduced the correlation to 0.11
TABLE 4
Odds ratios for obesity [body mass index ¢ 30] (BMI); 0 Å nonobese, 1 Å obese] among women from Latin America and the
Caribbean as a function of area of residence, socioeconomic status (SES) and educational level
TABLE 5
Age and anthropometric characteristics of Latin American children
Bolivia (1994) 12–35 1866 23.3 (6.8) 01.4 (1.4) 01.0 (1.2) 00.1 (1.1) 4.7 33.9 13.1 2.1
Brazil (1996) 12–59 3165 35.4 (13.6) 00.5 (1.3) 00.3 (1.2) 0.1 (1.0) 2.0 11.3 14.7 4.1
Colombia (1995) 12–59 3600 34.8 (13.6) 01.0 (1.1) 00.7 (1.0) 0.0 (0.9) 1.5 17.0 12.2 1.8
Dominican Republic (1996) 12–59 2984 34.9 (13.4) 00.6 (1.2) 00.4 (1.2) 0.1 (1.0) 1.2 11.5 15.3 4.6
El Salvador (1993) 12–59 2887 34.7 (15.5) 01.3 (1.3) 00.9 (1.1) 00.1 (1.0) 1.6 25.9 9.5 1.7
Guatemala (1995) 12–59 6477 34.6 (13.3) 02.1 (1.2) 01.4 (1.0) 00.2 (1.0) 3.5 56.1 10.0 2.0
Haiti (1994/1995) 12–59 2079 35.0 (14.1) 01.5 (1.5) 01.4 (1.2) 00.6 (1.1) 8.2 36.6 5.7 1.4
Honduras (1996) 12.1–60 1329 35.3 (14.4) 01.6 (1.4) 01.2 (1.1) 00.3 (0.9) 1.1 36.9 3.7 1.4
Mexico (1987) 12–59 5286 35.3 (14.8) 01.0 (1.7) 00.8 (1.3) 00.2 (1.3) 6.4 25.5 14.5 3.9
Nicaragua (1993/1994) 12–59 2581 35.4 (15.9) 01.3 (1.5) 00.8 (1.3) 0.0 (1.1) 2.1 28.2 12.1 2.2
Paraguay (1990) 12–59 2824 35.0 (13.4) 00.8 (1.2) 00.3 (1.0) 0.3 (0.8) 0.3 15.0 17.1 2.7
Peru (1996) 12–59 11796 35.5 (13.3) 01.3 (1.3) 00.5 (1.1) 0.4 (0.9) 1.1 29.2 23.9 4.7
Trinidad and Tobago (1987) 12–36 616 24.0 (7.3) 00.2 (1.1) 00.5 (1.2) 00.4 (1.0) 3.7 4.5 8.9 1.9
United States3 (1988–1994) 12–60 1359 36.2 (14.0) 0.1 (1.0) 0.3 (1.2) 0.4 (1.1) 0.4 1.9 23.7 7.4
1 Mean (SD).
2 Wasting defined as õ2 SD below the reference mean for weight for height; stunting defined as õ2 SD below the reference mean for height for
age; overweight defined as ú1 and obese as ú2 SD above the reference mean for weight for height.
3 Data for Mexican-American children only.
(P Å 0.72, n Å 13; Fig. 3). The relationship between percent- between the percentage of children stunted and the percentage
age of children stunted and GNP was negative and significant of children overweight was negative but not significant (r
(r Å 00.53, P Å 0.05, n Å 14), even after excluding Mexican Å 00.45, P Å 0.11, n Å 14); excluding Mexican Americans
Americans (r Å 00.61, P Å 0.03, n Å 13). The relationship gave similar results (r Å 00.29, P Å 0.34, n Å 13).
TABLE 6
Percentage of overweight (ú1 SD) children from Latin America and the Caribbean and denominator by area of residence,
socioeconomic status (SES), maternal education and gender1
Bolivia (1994) 15.3 10.6 10.2 14.1 12.1 14.5 12.3 13.9
983 883 553 1313 1133 733 942 924
Brazil (1996) 15.9 10.8 9.6 15.9 11.9 17.1 13.9 15.5
2366 799 731 2434 1544 1621 1631 1534
Colombia (1995) 13.9 9.4 6.6 13.4 10.9 13.6 11.5 13.0
2281 1319 637 2963 1817 1783 1812 1788
Dominican Republic (1996) 19.3 10.1 6.4 17.1 13.2 18.7 14.1 16.7
1540 1444 689 2295 1984 1000 1539 1445
El Salvador (1993) 10.8 8.4 6.7 12.0 7.8 14.0 9.0 10.1
1365 1522 1348 1539 2118 769 1465 1422
Guatemala (1995) 12.5 8.7 7.6 10.5 9.5 12.8 8.7 11.3
1532 4945 1363 5114 5926 551 3258 3219
Haiti (1994/1995) 6.8 5.2 5.6 5.7 5.5 7.2 5.4 6.0
748 1331 494 1585 1818 261 1048 1031
Honduras (1996) 8.5 3.1 2.0 5.9 4.5 8.2 5.3 4.8
598 731 277 1052 1107 222 692 637
Mexico (1987) 15.7 14.2 11.6 15.2 13.4 17.2 13.9 15.2
905 4381 766 4520 3560 1726 2715 2571
Nicaragua (1993/1994) 14.9 9.2 9.9 13.4 10.9 15.4 11.1 13.2
1295 1286 981 1600 1898 683 1299 1282
Paraguay (1990) 17.9 16.6 15.3 17.7 17.2 17.0 17.0 17.3
1055 1769 827 1997 2237 587 1464 1360
Peru (1996) 28.0 17.3 15.6 26.5 20.0 27.7 22.8 24.9
6315 5481 3591 8205 6497 5299 5891 5905
Trinidad and Tobago (1987) 11.2 7.5 8.4 9.1 8.1 9.9 11.1 7.0
241 375 119 497 322 294 288 328
United States2 (1988–1994) 24.4 22.5 25.6 21.7 26.9 22.4 24.4 23.0
863 496 742 617 404 955 673 686
TABLE 7
Odds ratios for overweight (ú1 SD; 0 Å nonoverweight, 1 Å overweight) among children from Latin America and the Caribbean as
a function of area of residence, socioeconomic status (SES) and maternal education
Maternal Maternal
education education
Area of SES (0 Å none/ Area of SES (0 Å none/
residence (0 Å low, primary, 1 Å residence (0 Å low, primary, 1 Å Gender3
(0 Å rural, 1 Å med/ secondary/ (0 Å rural, 1 Å med/ secondary/ (0 Å male,
Country (year) 1 Å urban) high) higher) 1 Å urban) high) higher) 1 Å female)
FIGURE 2 Trends in overweight (weight-for-height Z-score ú 1.00) among children (ú12 mo). The sample for the United States includes data
for Mexican-American children only.
serial data for several countries, we could not discern a clear to meet the needs of a broad range of income and education
pattern of change in overweight in Latin American children. levels. In countries such as Brazil and Mexico, obesity can no
Finally, we call attention to the need for information from longer be dismissed as a condition of the elite, but viewed
national surveys on school-age children, a group not generally rather as a concern of the disadvantaged.
included, in whom overweight and obesity would be a clearer Relationships with stunting and GNP. The analyses at
concern than among preschool children. country level were limited by low statistical power such that
Obesity and social factors. In their review of the world most relationships examined were not significant, particularly
literature on the relationship between obesity and SES, Sobol after excluding the data for the U.S. In general, the results
and Stunkard (1989) found that the prevalence of obesity in suggest that greater national incomes are associated with
adults and children increased with rising wealth in developing greater obesity levels in women and lower levels of stunting
societies. In developed societies, in contrast, prevalence is in children. Also, there was a tendency for greater levels of
lower with greater income, at least among women. Thus, devel- stunting in children to be associated with less obesity in
oping countries in transition to greater wealth should be found women and in children. These results are consistent with our
along a continuum that begins with a positive relationship initial expectation that increased national wealth would be
between obesity and SES in the poorest countries, to no rela- associated with less undernutrition and greater obesity.
tionship in those with middle incomes, to a negative relation- The results reviewed have important policy implications.
ship in those with the greatest wealth. The pattern of relation- Prevalences of obesity (BMI ¢30) among women were be-
ships between income and obesity described by Monteiro et tween 8 and 10% in all countries except Haiti. These statistics
al. (1995) for different income levels of the Brazilian popula- suggest an already existing public health concern. Although
tion fits this expected pattern (see introduction). the prevalences of overweight and obesity in children ap-
Unfortunately, almost all of the surveys we analyzed lacked proached or exceeded that in the U.S. reference population
measures of income. Furthermore, the socioeconomic indices (Dibley et al. 1987) in a few countries, most countries had
we calculated were specific to the country and relied on relative low values. Because uncertainty remains about the significance
wealth based on amenities, possessions and home characteris- of overweight in children õ5 y of age, it is unclear what level
tics. Educational attainment of women and area of residence of concern to express about the results reported here.
were used as additional social variables. Despite the limitations Tackling the problems of overweight and obesity poses
of our studies, the findings should still be of interest to scholars enormous challenges for Latin American countries. First, many
trying to describe the relationship between SES or income and of these countries still have high levels of undernutrition and
obesity or overweight. We found that, in general, the patterns it is imperative that this issue remain a priority. In addition,
of relationships among children were those expected for poor, Latin American countries are not prepared institutionally to
developing countries. Overweight in children tended to be deal with problems of diet and chronic disease. Professionals
greater in urban areas, in families of higher SES and in house- assigned to diet and chronic disease within ministries of public
holds with higher maternal education. Relationships were health are rare, even in countries where these problems are
much stronger and more complex in women compared with large relative to those of undernutrition. Finally, consensus is
children. Obesity in women was more common in urban areas lacking about how to address diet and health problems in
in the poorest countries and was unrelated to area of residence developed countries (Blackburn and Kanders 1994). Achiev-
in the richest two (i.e., Brazil and Mexico) and in Colombia. ing behavior change that leads to reduced obesity has proven
SES was related positively and significantly to obesity in difficult in the United States and elsewhere; even when success
women in all Latin American and Caribbean countries except occurs, programmatic lessons must be made applicable to each
Mexico. Controlling for SES and residence, obesity was more Latin American country. These are difficult issues that warrant
common in women with higher education only in Haiti and urgent discussion among researchers, public health prac-
Guatemala, two very poor countries; in all other countries, titioners and policy makers.
there was a tendency for more obesity in women with lower
education. On the other hand, no significant relationships ACKNOWLEDGMENTS
were observed between social variables and obesity in Mexi-
can-American women. In the context of the United States, Helpful comments are acknowledged from Wilma Freire and Man-
the Mexican-American population is of low socioeconomic uel Peña of PAHO and from Kenneth Resnicow, Richard Levinson
status and has one of the highest levels of obesity. and Dirk Schroeder of Emory University.
One limitation is that some countries show great variations
in wealth by region, and these analyses do not investigate LITERATURE CITED
regional differences. Brazil and Mexico are perhaps the best
Administrative Committee on Coordination/Sub-Committee on Nutrition (1992)
examples of countries with much intracountry variation. Second Report on the World Nutrition Situation. Volume I, Global and Re-
Grouping countries as single entities, as we did, probably masks gional Results. ACC/SCN, Geneva, Switzerland.
patterns of relationships between social variables and obesity. Administrative Committee on Coordination/Sub-Committee on Nutrition (1997)
We wish to emphasize that obesity in women is less com- Update on the Nutrition Situation 1996. Summary of Results for the Third
Report on the World Nutrition Situation. ACC/SCN, Geneva, Switzerland.
mon, but is not rare, in rural areas and in poor households Atalah, E. (1993) Análisis de la situación nutricional de la población de Santi-
and that there is a strong tendency for obesity to be more ago. Rev. Méd. Chile 121: 819–826.
common in poorly educated women. Because poor and rural Beckles, G.L.A., Miller, G. J., Alexis, S. D., Price, S.G.L., Kirkwood, B. R., Carson,
D. C. & Byam, N.T.A. (1985) Obesity in women in an urban Trinidadian
households represent significant proportions of the total popu- community. Prevalence and associated characteristics. Int. J. Obes. 9: 127–
lation, large numbers of obese women and children are found 135.
among them. For example, on average, about a third of obese Blackburn, G. L. & Kanders, B. S., eds. (1994) Obesity: Pathophysiology, Psy-
chology and Treatment. Chapman & Hall, New York, NY.
women in Latin America come from rural areas. The social Campos, H., Mata, L., Siles, X., Vives, M., Ordovas, J. M. & Schaefer, E. J. (1992)
mapping of obesity should be examined at the country level, Prevalence of cardiovascular risk factors in rural and urban Costa Rica. Circu-
including examination by region, when contemplating possible lation 85: 648–658.
DHS/IRD (1992) Demographic and Health Surveys. Country Publication Se-
programs and policies. Even the limited information presented ries. Institute of Resource Development, Columbia, MD.
here indicates that educational campaigns should be designed Dibley, M. J., Goldsby, J. B., Staehling, N. W. & Trowbridge, F. L. (1987) Devel-
opment of normalized curves for the international growth reference: historical Third National Health and Nutrition Examination Survey, 1988–94. Vital Health
and technical considerations. Am. J. Clin. Nutr. 46: 749–762. Statistics 1: 32 [DHHS Publication (PHS) 94-1308].
Frisancho, A. R. (1990) Anthropometric Standards for the Assessment of Pawson, I. G., Martorell, R. & Mendoza, F. E. (1991) Prevalence of overweight
Growth and Nutritional Status. University of Michigan Press, Ann Arbor, MI. and obesity in U.S. Hispanic populations. Am. J. Clin. Nutr. 53: 1522S–1528S.
Guo, S. S., Roche, A. F., Chumlea, W. C., Gardner, J. D. & Siervogel, R. M. Popkin, B. M. (1994) The nutrition transition in low income countries: an
(1994) The predictive value of childhood body mass index values for over- emerging crisis. Nutr. Rev. 52: 285–298.
weight at age 35 y. Am. J. Clin. Nutr. 59: 810–819. Schroeder, D. G. & Martorell, R. (1998) Fatness and body mass index from
Hernández, B., Peterson, K., Sobol, A., Rivera, J., Sepúlveda, J. & Lezana, M. A. birth to young adulthood in a rural Guatemalan population. In: The Assess-
(1996) Sobrepeso en mujeres de 12 a 49 años y niños menores de cinco ment of Childhood and Adolescent Obesity. Results of an International Obe-
años en México. Salud Pública de México 38: 178–188. sity Task Force Workshop. Am. J. Clin. Nutr. (suppl.) (in press).
Serdula, M. K., Ivery, D., Coates, R. J., Freedman, D. S., Williamson, D. F. & Byers,
Institute of Resource Development/MACRO International (1990) Model ‘‘B’’
T. (1993) Do obese children become obese adults? A review of the litera-
Questionnaire with Commentary for Low Contraceptive Prevalence Countries.
ture. Prev. Med. 22: 167–177.
DHS-II Basic Documentation, no. 2. Institute for Resource Development, Co- Sichieri, R., Coitinho, D. C., MarıB lia, L. M., Recine, E. & Everhart, J. E. (1994)
lumbia, MD. High temporal, geographic, and income variation in body mass index among
Ivanovic, D., Ivanovic, R. & Buitron, C. (1987) Nutritional status, birth weight adults in Brazil. Am. J. Public Health 84: 793–798.
and breast feeding of elementary first grade Chilean students. Nutr. Rep. Int. Sinha, D. P. (1995) Changing patterns of food, nutrition and health in the Carib-
36: 1347–1361. bean. Nutr. Res. 15: 899–938.
Jordan, M. D. (1990) Anthropometric Software Package Tutorial Guide and Sobal, J. & Stunkard, A. J. (1989) Socioeconomic status and obesity: a review
Handbook. The Centers for Disease Control, Center for Health Promotion and of literature. Am. Psych. Assoc. 105: 260–275.
Education, Division of Nutrition, Statistics Branch, Atlanta, GA. Solomon, C. G. & Manson, J. E. (1997) Obesity and mortality: a review of the
Kaplowitz, H., Martorell, R. & Mendoza, F. S. (1989) Fatness and fat distribu- epidemiologic data. Am. J. Clin. Nutr. 66: 1044S–1050S.
tion in Mexican-American children and youths from the Hispanic Health and Stevens, J., Cai, J., Pamuk, E. R., Williamson, D. F., Thun, M. J. & Wood, J. L.
Nutrition Examination Survey. Am. J. Hum. Biol. 1: 631–648. (1998) The effect of age on the association between body-mass index and
Kuczmarski, R. J., Flegal, K. M., Campbell, S. M. & Johnson, C. L. (1994) In- mortality. N. Engl. J. Med. 338: 1–7.
creasing prevalence of overweight among U.S. adults: the National Health Troiano, R. P. & Flegal, K. M. (1998) Overweight children and adolescents: de-
and Nutrition Examination Surveys, 1960 to 1991. J. Am. Med. Assoc. 272: scription, epidemiology and demographics. Pediatrics 101: 497–504.
205–211. UNICEF (1995) The State of the World’s Children. Oxford University Press,
Martorell, R., Mendoza, F. S., Castillo, R. O., Pawson, I. G. & Budge, C. C. (1987) New York, NY.
Short and plump physique of Mexican-American children. Am. J. Phys. An- Walker, A.R.P. (1995) Nutrition-related diseases in Southern Africa: with special
thropol. 73: 475–487. reference to urban African populations in transition. Nutr. Res. 15: 1053–1095.
Whitaker, R. C., Wright, J. A., Pepe, M. S., Seidel, K. D. & Dietz, W. H. (1997)
Ministerio de Salud (1997) Encuesta Nacional de Nutrición, 1996,
Predicting obesity in young adulthood from childhood and parental obesity.
AntropometrıB a. San José, Costa Rica. N. Engl. J. Med. 337: 869–873.
Monteiro, C. A., Mondini, L., Medeiros de Souza, A. L. & Popkin, B. M. (1995) World Health Organization (1990) Diet, Nutrition, and the Prevention of Chronic
The nutrition transition in Brazil. Eur. J. Clin. Nutr. 49: 105–113. Disease: Report of a WHO Study Group. World Health Organization Technical
National Center for Health Statistics (NCHS) (1985) Plan and operation of the Report Series no. 797, pp. 69–72. WHO, Geneva, Switzerland.
Hispanic Health and Nutrition Examination Survey, 1982–84. Vital Health Sta- World Health Organization (1995) Physical Status: The Use and Interpretation
tistics 1: 19 [DHHS Publication (PHS) 85-1321]. of Anthropometry: Report of a WHO Expert Committee. World Health Organi-
National Center for Health Statistics (NCHS) (1994) Plan and operation of the zation Technical Report Series no. 854. WHO, Geneva, Switzerland.