12506
Received 4 June 2013; accepted 3 September 2013; Accepted Article online 3 December 2013
Abstract
Aim The available evidence was reviewed to compare tivity of CT colonography for the detection of polyps
the effectiveness of CT colonography with that of colo- < 6 mm in diameter was low and heterogeneous,
noscopy for colorectal cancer (CRC) screening. although it was higher for polyps > 10 mm. The main
factors contributing to a greater sensitivity of CT colo-
Method An electronic search was conducted using
nography were the inclusion of only populations with an
PubMed, EMBASE, the Cochrane Library and Centre
average CRC risk and colonic insufflation with CO2. The
for Reviews and Dissemination databases, from incep-
incidence of adverse effects was very low for both tests.
tion to July 2009. Studies were included if investiga-
tions used CT colonography for CRC screening in Conclusion CT colonography has high specificity but
asymptomatic populations. Studies were excluded if heterogeneous sensitivity, although in most cases it is
investigations were conducted for the diagnosis of CRC not as sensitive or specific as conventional colonoscopy.
or in elderly, high-risk or symptomatic populations. CT colonography could therefore be useful as a screen-
ing test for populations with an average risk of CRC.
Results Of the 213 references identified, nine studies
were included. The specificity of CT colonography in Keywords Colonography, computed tomographic,
screening for CRC was high, although it decreased with colonoscopy, early detection of cancer, colorectal neo-
decreasing diameter of polyp to be detected. The sensi- plasms
O82 Colorectal Disease ª 2013 The Association of Coloproctology of Great Britain and Ireland. 16, O82–O89
J. E. Martın-L
opez et al. Accuracy of CT colonography and colonoscopy in CRC diagnosis
databases (CRD), Clinical Evidence, UptoDate, Emer- reviewer when there was any doubt about their eligibil-
gency Care Research Institute, Drug Effectiveness ity.
Review Project, Hayes, Technology Evaluation Center When information given in titles/abstracts suggested
and Web of Knowledge and the Global Health Library. that the study included patients who underwent CT
The search strategies combined MeSH (medical subject colonography for the screening of colorectal polyps and
headings), Emtree terms and text words to define the cancer with subsequent colonoscopy and assessed the
population, index test, gold standard and outcome test for accuracy and safety, full paper articles were
(PICO format: patient/population, intervention, com- retrieved for further assessment. If there was doubt
parison and outcome) (Table 1) with restrictions related regarding inclusion from the title and abstract, the full
to study design (comparative and prospective studies). article was obtained for clarification. Stringent inclu-
There were no language restrictions and searches were sion/exclusion criteria were applied to the full paper to
carried out from 1994 (when CT colonography was first obtain the final set of included studies. Inclusion criteria
described) up to January 2012. Additionally, the refer- were prospective or randomized trials with populations
ence lists from the final selection of articles were without abdominal symptoms of average age ≥ 50 years
checked manually to identify other relevant papers. who underwent CT colonography and colonoscopy
Titles and abstracts of all retrieved articles were scanned within 3 months for CRC screening. Eligible studies
for inclusion by one reviewer with reference to a second had to report the detection of colorectal polyps,
Table 1 Population characteristics, CRC risk, age, allocation and selection criteria.
Average age
Study (years) Risk stratification* Allocation Selection criteria
Graser et al. [18] 60.5 Average Consecutive Asymptomatic. Age > 50 years
Johnson et al. [16] 65.0 Average Consecutive Asymptomatic. Age ≥ 40 years
Johnson et al. [12] 58.3 Average (89%), NR Asymptomatic. Age ≥ 50 years. Colonoscopy
moderate (10%) programmed
and high (1%)
Juchems et al. [13] 62.6 Average NR Asymptomatic. Age 55–75 years. No personal
history of colon adenomas and/or CRC in
first-degree relatives before 60 years
David and Kim [19] 57.6 Average (95%) Consecutive Age ≥ 50 years or ≥ 40 years with familial CRC
and high (5%) history. Average CRC risk
Macari et al. [17] 55.0 Average Consecutive Asymptomatic. Age ≥ 50 years. FOBT negative
results. No familial CRC history. No personal
history of colorectal polyps or sigmoidoscopy
Pickhardt et al. [14] 57.8 Average (97.4%) Consecutive Age 50–79 years with average CRC risk.
and moderate (2,6%) Age 40–79 years with familial CRC history
Regge et al. [20] 60.0 Moderate (63%) and Relative contacts Group 1: first-degree relatives of patients with
high (37%) advanced colorectal neoplasia diagnosed before
the age of 60 years and aged 40–65 years
(family history group)
Group 2: individuals entered a colonoscopy
surveillance programme after endoscopic removal
of colorectal adenomas and were aged 18–70
years (postpolypectomy group)
Group 3: individuals with positive results from
FOBT who were aged 59–69 years and
participating in a CRC screening programme
(FOBT positive group)
Vogt et al. [15] 58.0 Moderate Consecutive Average CRC risk referred for colonoscopy
because of nonspecific abdominal symptoms
(chronic constipation or chronic abdominal pain)
*Colorrectal cancer risk stratification according to American Cancer Society guidelines [3].
FOBT, faecal occult blood test; CRC, colorectal cancer; NR, not referred.
Colorectal Disease ª 2013 The Association of Coloproctology of Great Britain and Ireland. 16, O82–O89 O83
Accuracy of CT colonography and colonoscopy in CRC diagnosis J. E. Martın-L
opez et al.
advanced neoplasia and CRC and had to include true- professional who interpreted the images, use of sec-
positive, true-negative, false-positive and false-negative ond-look colonoscopy, determination of size on CT
values. colonography, positive test criteria and histopathologi-
Studies were included if the reference standard test cal confirmation. The number and location of tumours
used to define the true disease status was histological detected by CT colonography were documented.
diagnosis and/or colonoscopy for verification. Studies Histologically proved cancers were considered on CT
with an excessively high cancer prevalence because of a colonography if they were detected prior to colonos-
priori patient selection, e.g. previous endoscopy, were copy, surgery and/or pathological confirmation. Results
excluded, as well as studies that included individuals at colonography were considered a true-positive result if
with a confirmed diagnosis of CRC. Systematic assess- they were identified at initial blinded endoscopy before
ment of the quality and documentation of the selected the unblinding of CT colonography results. The sensi-
articles was performed. To grade the quality of the tivity of colonoscopy could only be assessed for studies
study the quality assessment (QUADAS) tool was used where this condition was applied. CT colonography and
with special focus on the time interval between tests. colonoscopy were considered as independent technolo-
The technical quality of the CT colonography technique gies for analysis. For the analysis per patient and per
was assessed with reference to the European procedure lesion, 2 9 2 contingency tables were constructed to
guideline [7] (Table 2) and with particular reference to allow the calculation of sensitivity and specificity. Data
patient position during the test, the use of spasmolytic extracted were as follows: polyp size 5–7, 8–10 and
drugs, bowel insufflation with CO2 or the use of intra- ≥ 10 mm, based on the associated potential CRC risk.
venous contrast. Data extraction was performed using The analysis for polyps < 5 mm was not considered in
predefined data extraction forms. For each primary this work because most included studies did not report
study the following data were abstracted: on it.
1 Study population: mean or median age with age
range, method of allocation whether consecutive or
Statistical analysis
not, CRC risk, country, inclusion/exclusion criteria
and the indication for CT colonography. REVIEW MANAGER 5.0 was used to produce a methodo-
2 Imaging features: bowel preparation, dietary restric- logical quality summary table and forest plots of sensi-
tions, tagging and bowel distension, use of spasmo- tivity and specificity for each test as well as to present
lytic drugs and type of CT system and CT meta-analytic results in receiver operating characteristic
parameters (radiation dose), the positioning of the (ROC) space. Data presented on a patient and lesion
patient, the use of intravenous contrast medium dur- basis were considered separately in the analysis. Patient-
ing CT colonography and complications. based data were used as the basis for the main analysis
3 Colonoscopy: type of bowel preparation, use of drugs, and for the investigation of heterogeneity. Per patient
clinical experience of professional who interpreted the and per lesion sensitivity and specificity values for CT
images, dietary restrictions and complications. colonography and colonoscopy for colorectal polyps
4 Imaging and diagnostic criteria: experience of the radi- and cancer were summarized using the software META-
ologist reading the CT colonography, the grade of the DISC (http://www.hrc.es/incestigacion/metadisc_htm).
CT Per Per
Study Patients Lesions colonography Colonoscopy patient lesion
O84 Colorectal Disease ª 2013 The Association of Coloproctology of Great Britain and Ireland. 16, O82–O89
J. E. Martın-L
opez et al. Accuracy of CT colonography and colonoscopy in CRC diagnosis
Heterogeneity between primary studies was assessed nonpertinent articles were excluded on the basis of their
by using the I2 statistic, so that I2 values equal to 25, title and abstract, 26 studies were considered for
50 and 75% were assumed to represent low, moderate inclusion. From these 26 potential full text articles, 17
and high heterogeneity. The aspects that could have dif- articles, including three meta-analyses [8–10], were
fered between test procedures under investigation were excluded owing to the inclusion of asymptomatic popu-
eligible for sensitivity analysis, including patient posi- lations [10]. This left nine articles [11–18,20] for inclu-
tion, routine use of intravenous contrast material, bowel sion in this systematic review and meta-analysis with a
insufflation or spasmolytic drugs. Other analyses were total of 5640 asymptomatic patients aged between 57
conducted to investigate factors that may have been and 65 years (mean 59.4) and with different CRC risks
sources of heterogeneity related to population selection (71.5% average, 21.5% moderate and 7% high)
for faecal occult blood test (FOBT) results and CRC (Table 1).
risk of the population. Table 2 shows the number of studies and patients/
lesions in the main groups for analysis. All participants
received CT colonography and colonoscopy on the
Results
same day. In most cases the reference standard was a
The initial search yielded 260 articles (Fig. 1), and after combination of histology and colonoscopy results; three
duplicates were excluded 230 articles remained. After studies [11–13] used only histology and two [14,15]
Identification
Articles
Articles screened
excluded
(n = 230)
(n = 204)
Studies included in
qualitative synthesis
(meta-analysis)(n = 9)
Colorectal Disease ª 2013 The Association of Coloproctology of Great Britain and Ireland. 16, O82–O89 O85
Accuracy of CT colonography and colonoscopy in CRC diagnosis J. E. Martın-L
opez et al.
Sensitivity
phate preparation [14,17] and no specific colon cleans-
ing directions were given in the remaining study [20]. 0.5
Five studies used oral tagging [12,13,14,16,20] and 0.4
one used an intravenous contrast medium [19] . The
use of tagging was not specified in three studies 0.3
[15,17,18]. Methods used for bowel distension varied. 0.2
Five studies used CO2 insufflation [11,12,18,20] com-
bined with butylscopolamine bromide [13,20] or gluca- 0.1
gon hydrochloride [12] as a spasmolytic drug. In two 0
[18,19] no spasmolytic drug was administered. In the 1 0.9 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0
Sensitivity
remaining studies, manual insufflation with room air
Legend
was used [13–17]. Both supine and prone images were Colonoscopy CT colonography
acquired in seven of the nine studies, while in the
remaining two studies [11,15] only one position was Figure 3 Comparative patient-based data for the sensitivity
used. In five studies [12,13,15–17] the images were and specificity of CT colonography (red) and accuracy of colo-
interpreted by experienced radiologists, and in one the noscopy (black). ○,Colonoscopy; ♢, CT colonography.
only reader had 5 years experience in CT colonographic
data interpretation [17]. The size of the polyp was
determined by the pathologist [12,13,16–20] , elec- nography (n = 4) was 80.3% (95% CI 77.7–82.8%) and
tronic callipers [14] or consensus between readers based the specificity (n = 3) was 73.2% (95% CI 67.7–78.1%).
on visual criteria [15]. The estimated sensitivity of the pooled CT colonog-
The estimated sensitivity of pooled CT colonography raphy (n = 2) was 60.9% (95% CI 53.4–68.0%) and
(n = 4) and colonoscopy (n = 3) was 66.8% [95% confi- 88.7% for colonoscopy (n = 2) (95% CI 83.3–92.8%).
dence interval (CI) 62.7–70.8%] and 92.5% (95% CI The estimated specificity of pooled CT colonography
89.0–95.2%) (Fig. 2). Figure 3 shows the ROC analysis was 25.2% (95% CI 17.5–34.4%) and 90.3% (95% CI
of the sensitivity and specificity for CT colonography 82.9–95.2%) for colonoscopy (n = 2). Tables 3 and 4
and colonoscopy. Heterogeneity between studies (I2) show the accuracy of CT colonography and colonos-
was 40.2%. The estimated pooled sensitivity of CT colo- copy by patient and by lesion for different polyp sizes.
Colonoscopy
O86 Colorectal Disease ª 2013 The Association of Coloproctology of Great Britain and Ireland. 16, O82–O89
J. E. Martın-L
opez et al. Accuracy of CT colonography and colonoscopy in CRC diagnosis
Colorectal Disease ª 2013 The Association of Coloproctology of Great Britain and Ireland. 16, O82–O89 O87
Accuracy of CT colonography and colonoscopy in CRC diagnosis J. E. Martın-L
opez et al.
Sensitivity Specificity
Numbers in brackets in the left-most column are the number of studies included in the analysis.
FOBT, faecal occult blood test; CRC, colorectal cancer.
CT Colonography sensitivity copy for CRC were 96.1% (n = 49) and 94.7%
according to different factors analysed
(n = 25). There are a few explanations available for the
CT Colonography global
sensitivity differences between the greater sensitivity of CT colo-
100 nography than colonoscopy. The first is that inclusion
90
80 of symptomatic populations could be overestimating the
70
60 accuracy of CT colonography. The second could be the
50 use of populations with a different CRC risk profile.
40
30 The main methodological limitation found in primary
20
10 studies was the lack of information on clinical data.
0
g n nt on ug 2 This aspect is crucial, because in practice clinical data
in tio e iti dr O
pl ul
a ag s C
are available for radiologists, who might be underestimat-
am p st po y tic ith
s po tra nt ol w
ing the significance of the findings. The analysed studies
BT ris
k
co
n tie m io
n
FO Pa as at
s sp fl used two different methods of blinding. The first was
R
C
n ou of n su
e
C ve se e li segmental blinding in which the endocopist knew the
ag tra U w result of CT colonography after the examination of every
er i n Bo
Av of
se colonic segment. The other method was so-called opti-
U
mized blinding in which the endocopist looked at the
Figure 4 Sensitivity analysis. Factors influencing the accuracy colonoscopic images for discordance between the tests.
of CT colonography (FOBT, faecal occult blood test; CRC, The second method was considerably better given that it
colorectal cancer).
has a lower risk of bias and real blinding was impossible
for ethical reasons, although neither method could detect
sensitivity of 86.7% for lesions between 5 and 7 mm all lesions [21]. The sensitivity of CT colonography
and 92.9% for those > 10 mm. The maximum value for might have been underestimated in the studies because a
colonoscopy was for lesions between 8 and 10 mm colonoscopy is an imperfect reference.
(sensitivity 99.2%, specificity 91.3%). Although the spec- We included populations with different levels of risk,
ificity of CT colonography is higher than for colonos- which resulted in some homogeneity. It was neverthe-
copy, sensitivity is greater for colonoscopy. For the less reasonable to assume that the results could be
detection of CRC the pooled sensitivity of CT colonog- applied to the screening population. The small number
raphy (96%) was slightly higher than that for colonos- of included studies is another important limitation,
copy (91%). which renders the results of subgroup analysis uncer-
Our results are similar to previously published data. tain. Despite the large number of systematic reviews
A meta-analysis of 47 prospective studies of symptom- and meta-analyses on the accuracy of CT colonography,
atic patients [8] found a similar overall CT colonogra- no studies of an asymptomatic population were found.
phy sensitivity of 69% and specificity of 83%. Another One of the strengths of the present work is the likely
meta-analysis [10] of six prospective studies with popu- inclusion of most of the relevant literature. It comprised
lations with at least 50 asymptomatic individuals of five of the studies quoted in the most recent previous
average risk also found similar results for polyps with a systematic review [10] with an additional four included
sensitivity and specificity of 83.3 and 87.9% for polyps for the first time (which compared the diagnostic accu-
≥ 10 mm. In a recent study of symptomatic populations racy and safety of both tests). A second strength is that
[9], the sensitivities of CT colonography and colonos- only data on participants with an average risk were used
O88 Colorectal Disease ª 2013 The Association of Coloproctology of Great Britain and Ireland. 16, O82–O89
J. E. Martın-L
opez et al. Accuracy of CT colonography and colonoscopy in CRC diagnosis
to assess the ability to identify all polyps as well as tion of polyps and colorectal tumors: a systematic review
CRC. Thus the study provides an estimate of the diag- and meta-analysis. Digestion 2009; 80: 1–17.
nostic value of CT colonography and colonoscopy over- 9 Pickhardt PJ, Hassan C, Halligan S et al. Colorectal cancer:
all and for subgroups according to lesion size in a CT colonography and colonoscopy for detection–systematic
review and meta-analysis [Review]. Radiology 2011; 259:
screening setting.
393–405.
In conclusion, the pooled sensitivity of CT colonog-
10 de Haan MC, van Gelder RE, Graser A et al. Diagnostic
raphy for polyps appears to be lower than that of colo- value of CT-colonography as compared to colonoscopy in
noscopy, except in a specific subgroup with polyps an asymptomatic screening population: a meta-analysis. Eur
between 8 and 10 mm and according to analysis per Radiol 2011; 21: 1747–63.
lesion. The pooled specificity of CT colonography was 11 Kim YS, Kim N, Kim SH et al. The efficacy of intravenous
higher than that of colonoscopy according to analysis contrast-enhanced 16-raw multidetector CT colonography
per patient, but not to analysis per lesion. The main for detecting patients with colorectal polyps in an asymp-
factors contributing to the increase in the sensitivity of tomatic population in Korea. J Clin Gastroenterol 2008;
CT colonography were restricted to populations with 42: 791–8.
an average risk of CRC and for colonic insufflation with 12 Johnson CD, Chen MH, Toledano AY et al. Accuracy of
CT colonography for detection of large adenomas and can-
CO2. The incidence of adverse events was very low.
cers. N Engl J Med 2008; 359: 1207–17.
13 Juchems M, Rompp A, Kestler HA et al. A prospective
References comparison of video colonoscopy and CT colonography in
asymptomatic patients screened for colorectal cancer. Dtsch
1 Shin HR, Bray F, Forman D, Mathers C, Parkin DM, Med Wochenschr 2010; 135: 1656–61.
Ferlay J. GLOBOCAN 2008 v1.2, Cancer Incidence and 14 Pickhardt PJ, Choi JR, Hwang I et al. Computed tomo-
Mortality Worldwide: IARC CancerBase No. 10 [Internet]. graphic virtual colonoscopy to screen for colorectal neoplasia
Lyon, France: International Agency for Research on in asymptomatic adults. N Engl J Med 2003; 349: 2191–200.
Cancer; 2010. Available from: http://globocan.iarc.fr 15 Vogt C, Cohnen M, Beck A et al. Detection of colorectal
(accessed 31 January 2012). polyps by multislice CT colonography with ultra-low-dose
2 Bond JH. Clinical evidence for the adenoma-carcinoma technique: comparison with high-resolution videocolonos-
sequence, and the management of patients with colorectal copy. Gastrointest Endosc 2004; 60: 201–9.
adenomas. Semin Gastrointest Dis 2000; 11: 176–84. 16 Johnson CD, Fletcher JG, MacCarty RL et al. Effect of
3 Byers T, Levin B, Rothenberger D, Dodd GD, Smith RA. slice thickness and primary 2D versus 3D virtual dissection
American Cancer Society guidelines for screening and sur- on colorectal lesion detection at CT colonography in 452
veillance for early detection of colorectal polyps and cancer: asymptomatic adults. AJR Am J Roentgenol 2007; 189:
update 1997. American Cancer Society Detection and 672–80.
Treatment Advisory Group on Colorectal Cancer. CA 17 Macari M, Bini EJ, Jacobs SL et al. Colorectal polyps and
Cancer J Clin 1997; 47: 154–60. cancers in asymptomatic average-risk patients: evaluation
4 Uradomo L, Steele A, Greenwald BD, Raufman JP, Aslinia with CT colonography. Radiology 2004; 230: 629–36.
F. Quality assessment of colonoscopic cecal intubation: an 18 Graser A, Stieber P, Nagel D et al. Comparison of CT col-
analisis of 6 years of continuous practice at a university hos- onography, colonoscopy, sigmoidoscopy and faecal occult
pital. Am J Gastroenterol 2006; 101: 721–31. blood tests for the detection of advanced adenoma in an
5 Winawer SJ, Zauber AG, Ho MN et al. Prevention of average risk population. Gut 2009; 58: 241–8.
colorectal cancer by colonoscopic polypectomy. The 19 Kim DH, Pickhardt PJ, Taylor AJ et al. CT colonography
National Polyp Study Workgroup. N Engl J Med 1993; versus colonoscopy for the detection of advanced neoplasia.
329: 1977–81. N Engl J Med 2007; 357: 1403–12.
6 McFarland EG, Levin B, Lieberman DA et al. Revised col- 20 Regge D, Laudi C, Galatola G et al. Diagnostic accuracy of
orectal screening guidelines: joint effort of the American computed tomographic colonography for the detection of
Cancer Society, U.S. Multisociety Task Force on Colorectal advanced neoplasia in individuals at increased risk of colo-
Cancer, and American College of Radiology. Radiology rectal cancer. JAMA 2009; 301: 2453–61.
2008; 248: 717–20. 21 van Rijn JC, Reitsma JB, Stoker J, Bossuyt PM, van
7 Taylor SA, Laghi A, Lefere P, Halligan S, Stoker J. Euro- Deventer SJ, Dekker E. Polyp miss rate determined by tan-
pean Society of Gastrointestinal and Abdominal Radiology dem colonoscopy: a systematic review. Am J Gastroenterol
(ESGAR): consensus statement on CT colonography. Eur 2006; 101: 343–50.
Radiol 2007; 17: 575–9.
8 Chaparro M, Gisbert JP, Del CL, Cantero J, Mate J. Accu-
racy of computed tomographic colonography for the detec-
Colorectal Disease ª 2013 The Association of Coloproctology of Great Britain and Ireland. 16, O82–O89 O89