Annals of Oncology original articles

Annals of Oncology 24: 1079–1087, 2013

doi:10.1093/annonc/mds601
Published online 4 December 2012

Cruciferous vegetables intake and the risk of colorectal

cancer: a meta-analysis of observational studies
Q. J. Wu1,2, Y. Yang1,2, E. Vogtmann3, J. Wang2, L. H. Han2, H. L. Li1,2 & Y. B. Xiang1,2*
1
State Key Laboratory of Oncogene and Related Genes; 2Department of Epidemiology, Shanghai Cancer Institute, Renji Hospital, Shanghai Jiaotong University
School of Medicine, Shanghai, P. R. China; 3Department of Epidemiology, University of Alabama at Birmingham, Birmingham, USA

Received 24 August 2012; revised 6 October 2012; accepted 9 October 2012

Background: Epidemiological studies have reported inconsistent associations between cruciferous vegetable (CV)
intake and colorectal cancer (CRC) risk. To our knowledge, a comprehensive and quantitative assessment of the
association between CV intake and CRC has not been reported.
Methods: Relevant articles were identified by searching MEDLINE. We pooled the relative risks (RR) from individual
studies using a random-effect model and carried out heterogeneity and publication bias analyses.
Results: Twenty-four case–control and 11 prospective studies were included in our analysis. When all studies were
pooled, we yielded a significantly inverse association between CV (RR: 0.82; 95% confidence interval 0.75–0.90) intake
and CRC risk. Specific analysis for cabbage and broccoli yielded similar result. When separately analyzed, case–control
studies of CV intake yield similar results, and the results from the prospective studies showed borderline statistical
significance. Moreover, significant inverse associations were also observed in colon cancer and its distal subsite both
among prospective and case–control studies.
Conclusions: Findings from this meta-analysis provide evidence that high intake of CV was inversely associated with
the risk of CRC and colon cancer in humans. Further analysis on other specific CV, food preparation methods, stratified
results by anatomic cancer site, and subsite of colon cancer should be extended in future study.
Key words: colorectal cancer, cruciferous vegetables, dietary, epidemiology, meta-analysis

introduction enzymes (e.g. glutathione S-transferases, GST) by a variety of

CV results in a favorable metabolic profile for the elimination
Colorectal cancer (CRC) is the third most commonly diagnosed of certain chemical carcinogens [5, 6]. On the other hand, CV
cancer worldwide in males and the second in females, with over as a good source of dietary fiber which can prevent CRC by
1.2 million new cases diagnosed in 2008 accounting for 9.7% of several plausible mechanisms, including increased fecal bulk
all incident cancers [1]. Epidemiological and experimental and dilution of carcinogens in the colonic lumen, reduced
studies have provided strong evidence that smoking, physical transit time, and bacterial fermentation of fiber to short-chain
inactivity, overweight and obesity can cause CRC [2]. Despite fatty acids [7, 8]. In an updated report from the World Cancer
the role of diet as a contributing factor in CRC development, Research Fund and the American Institute for Cancer Research
convincing evidence only demonstrated that intakes of alcohol (WCRF/AICR), dietary fiber intake has been upgraded as a
(among males) and red and processed meat are considered to convincing protective dietary factor for CRC [9].
be dietary causes of CRC [3]. A recent meta-analysis which included 19 prospective
Cruciferous vegetables (CV) contain a variety of bioactive studies reported that there is a weak but statistically significant
components such as folate, vitamin C, tocopherols, and inverse association between vegetable intake and CRC
carotenoids [4]. However, the most frequently attributable [summary relative risk (RR): 0.91; 95% confidence interval (CI)
anticancer constituent of CV is glucosinolates (GLS), the 0.86–0.96] [10]. Nevertheless, the epidemiological results of
precursors of isothiocyanates (ITCs) as well as indole-3- whether intake of CV may protect against CRC are still
carbinol (I3C), which may contribute to reduce risk of CRC. controversial. Some epidemiological reviews published in the
Evidence from animal studies has indicated that the joint mid-1990s [11, 12] indicated an inverse association between
induction of phase I (i.e. cytochrome P450s) and phase II CV intake and CRC, but some cohort studies have cast doubt
on the strength of the inverse association between CV intake
*Correspondence to: Prof. Y. B. Xiang, State Key Laboratory of Oncogene and Related and CRC risk [13–16]. Therefore, we carried out a meta-
Genes & Department of Epidemiology, Shanghai Cancer Institute, Renji Hospital, analysis on all prospective and case–control studies published
Shanghai Jiaotong University School of Medicine, No. 25, Lane 2200, Xie Tu Road,
Shanghai 200032, P. R. China. Tel: +86-21-64437002; Fax: +86-21-64046550;
up to April 2012 to evaluate the relationship between CV
E-mail: ybxiang@shsci.org intake and CRC.

© The Author 2012. Published by Oxford University Press on behalf of the European Society for Medical Oncology.
All rights reserved. For permissions, please email: journals.permissions@oup.com.

Licensed to Natalia Sabrina Ortiz - naati.ortiz@hotmail.com

 original articles Annals of Oncology

methods Laird method) [24] was reported, which assumes that the studies included
in the meta-analysis had varying effect sizes. Otherwise, the summary
literature search estimate based on the fixed-effect model (the inverse variance method) [25]
We carried out a comprehensive literature search up to April 2012 using was reported, which assumes that the studies included in the meta-analysis
the MEDLINE (PubMed), with restrictions to the studies of humans and had the same effect size. Summary estimates were calculated for CV,
articles published in English using the following search key words and cabbage, and broccoli intakes. Heterogeneity between subgroups was
medical subject heading terms: (Brassicaceae OR Brassica OR cruciferous evaluated by meta-regression in our analysis. Subgroup analyses were
vegetables OR broccoli OR cabbage OR cauliflower OR Brussels sprouts carried out based on study quality, study design ( prospective versus case–
OR mustard plants OR sauerkraut OR cole slaw OR collards OR bok choy control studies), type of control subjects for case–control ( population-
OR turnip greens OR vegetables) AND (colorectal OR colorectum OR based versus hospital-based controls), geographic location (Europe,
colon OR rectal OR rectum) AND (cancer OR neoplasm OR carcinoma America, and Asia), gender (males versus females), anatomic site of CRC
OR tumor). Furthermore, we also searched the reference lists of all (colon versus rectum cancer), and cancer subsite of colon ( proximal versus
included studies. We followed standard criteria for conducting and distal colon cancer). Moreover, we carried out sensitivity analyses excluding
reporting meta-analyses [17]. one study at a time to explore whether the results were strongly influenced
by a specific study.
Publication bias was evaluated via Egger’s linear regression [26], Begg’s
study selection criteria
rank correlation methods [27] and funnel plots. A P-value <0.05 for
Published studies were included if they (i) used a case–control or Egger’s or Begg’s tests was considered representative of significant statistical
prospective study design; (ii) evaluated the association between CV intake publication bias. Statistical analyses were carried out with Stata (version
and CRC risk; (iii) presented odds ratio (OR), RR, or hazard ratio (HR) 11.0; StataCorp, College Station, TX), and all statistical tests were two-
estimates with 95% CI, standard errors (SE), or data necessary to calculate sided.
these. When multiple publications from the same study were available, we
used the publication with the largest number of cases and most applicable
information. results
literature search
data abstraction and quality assessment
We identified 1147 potentially relevant articles from our search
For each eligible study, two investigators (Q-JW and YY) independently
of the PubMed databases. Of these, 1096 articles were excluded
carried out the eligibility evaluation, data abstraction, and quality
assessment; disagreements were resolved by consensus. Data abstracted
after the first screening based on abstracts or titles, leaving 51
from each study included are as follows: the first author’s last name, year of articles for full-text review. Handsearching of the bibliographic
publication, study region and design, study sample size (number of cases references of these articles identified two additional articles
and controls or cohort size), age range or the mean age of studies, duration [28, 29], of a total of 53 articles for full-text review. Figure 1
years of follow-up for cohort studies, measures and types of CV and intake shows a flow diagram, identifying the relevant studies. After
categories, study-specific adjusted ORs or RRs with their 95% CIs for the exclusion, the remaining 35 articles [13–16, 28–58] were
highest versus lowest category of CV intake (if multiple estimates were included in the systematic review and meta-analysis.
available, we abstracted the estimate that adjusted for the most covariates),
and factors matched by or adjusted for in the design or data analysis. study characteristics and quality assessment
To assess the study quality, a 10-star system on the basis of the
Characteristics of the 35 included articles are shown in
Newcastle–Ottawa Scale [18–21] was used in this meta-analysis. The full
supplementary Table S1, available at Annals of Oncology
score was 10 and the high-quality study was defined as a study with quality
scores ≥7.
online. All included articles, which included 24 275 cases and
1 295 063 subjects, were published between 1978 and 2012,
consisting of 11 prospective studies (10 cohort studies [13–16,
statistical analysis
29, 31–35], one nested case–control study [30]) and 24 case–
The study-specific adjusted RRs were used as the common measure of control studies [28, 36–58] in which 2 articles [38, 39] using
association across studies. Because the absolute risk of CRC is low in the same control population reported the results of colon and
human, the ORs in case–control studies should approximate the RRs or rectum cancer separately. Hence we pooled the results of these
HRs; therefore, we reported all results as RRs for simplicity. Some studies
two studies. Of the 11 prospective studies, eight were
presented individual risk estimates according to the different types of CV
conducted in the United States [13, 14, 16, 29, 31–34], one
and did not report the effect of total CV intake. In this situation, the study-
each in China [30], Finland [15], and The Netherlands [35].
specific effect size in overall analysis was calculated by pooling the risk
Sample sizes ranged from 17 633 [29] to 488 043 [13], and the
estimates of the various CV types, using the inverse-variance method [22].
For studies that reported results separately for males and females or
number of CRC cases varied from 145 [29] to 2972 [13].
proximal and distal colon or colon and rectal cancer, but not combined, we
Of the 24 case–control studies, 6 were conducted in
pooled the results using a fixed-effect model to obtain an overall combined the United States [40, 47, 52, 55, 56, 58], 3 in Australia
estimate before combining with the rest of the studies [8, 10]. [36, 49, 57], 2 each in Canada [38, 39], Singapore [44, 54],
The possible heterogeneity in results across studies was examined by Italy [28, 51], Japan [42, 48], and the UK [41, 46], one each
using the Cochran Q and I 2 statistics [23]. For the Q statistic, a P-value in China [43], Spain [53], India [37], Russia [50], and
<0.1 was considered to be representative of statistically significant Uruguay [45]. The number of patients enrolled in these studies
heterogeneity. I 2 represents the proportion of total variation contributed by ranged from 43 [52] to 3575 [28], and the number of control
between-study variation [23]. When substantial heterogeneity was detected, subjects varied from 41 [52] to 31 782 [48]. Control subjects
the summary estimate based on the random-effect model (DerSimonian– were drawn from the general population in 12 studies [36, 38–

 | Wu et al. Volume 24 | No. 4 | April 2013

Licensed to Natalia Sabrina Ortiz - naati.ortiz@hotmail.com

 Annals of Oncology original articles

Figure 1. Selection of studies for inclusion in meta-analysis

41, 43, 44, 46, 49, 55–57], hospitals in 8 studies [28, 37, 42, 45, moderate heterogeneity (Q = 11.07, P = 0.05, I 2 = 54.8%), and
47, 48, 51, 54], or both in 3 studies [50, 53, 58]. no publication bias was found by the Egger’s test (P = 0.164) or
Study-specific quality scores are summarized in with Begg’s test (P = 0.260).
supplementary Tables S2 and S3, available at Annals of
Oncology online. The range of quality scores was from 4 to 9;
subgroup and sensitivity analyses
the median score was 7. The median scores of prospective and
case–control studies were 7 and 6, separately. High-quality In subgroup analyses of CV intake and CRC, all strata showed
studies (i.e. those studies that had 7 awarded stars) included 10 inverse associations but some associations were not statistically
prospective [13–15, 29–35] and 11 case–control studies [28, 36, significant (Table 1). Similar results were observed for the
38, 39, 42–46, 49, 57]. intake of cabbage. Although the summary results of broccoli
intake were not significant, there were inverse associations in
all the strata and some showed statistically significance
cruciferous vegetables
(Table 1). However, considering study design is the source of
In a random-effect pooled analysis of these studies, high-CV heterogeneity of CV intake and CRC (P-value for meta-
intake (comparing the highest with the lowest category) was regression = 0.025), we separated the studies by study design
associated with a reduced CRC risk (RR: 0.82; 95% CI 0.75– and carried out the subgroup analysis (supplementary
0.90) (Table 1). Statistically significant heterogeneity was Table S4, available at Annals of Oncology online).
observed in the study results (Q = 98.57, P < 0.001, I 2 = 66.5%). In a sensitivity analysis of CV intake and CRC risk, we
There was no indication of publication bias with Egger’s test sequentially removed one study at a time and re-analyzed the
(P = 0.08) or with Begg’s test (P = 0.218). data. The 36 study-specific RRs ranged from a low of 0.81
(95% CI 0.74–0.88) after omitting the study by Pietinen et al.
cabbage [15] to a high of 0.84 (95% CI 0.78–0.92) after omitting the
One cohort [16] and eight case–control studies [37, 42, 48–50, study by Kune et al. [57], but were in general similar.
52, 56, 58] investigated the association between cabbage intake Meanwhile, we removed six studies [40, 51, 53, 55, 57, 58]
and CRC risk. In a fixed-effect pooled analysis of these studies, in which RRs and 95% CI were not reported but calculated
high cabbage intake (comparing the highest with the lowest from raw data. The result from this analysis (RR: 0.88; 95%
category) was associated with a reduced risk of CRC (RR: 0.76; CI 0.81–0.95) was similar. Like CV intake, similar sensitivity
95% CI 0.60–0.97) (Table 1, Figure 2). There was moderate analyses for cabbage and broccoli did not significantly change
heterogeneity among the nine studies (Q = 19.20, P = 0.234, the results (data not shown). Furthermore, we chose not using
I 2 = 58.3%), and no publication bias was found by the Egger’s the inverse-variance method to pool the risk estimates of the
test (P = 0.196) or with Begg’s test (P = 0.917). various CV types—the result also was robust (RR: 0.85; 95%
CI 0.77–0.93).
broccoli
Three cohort studies [16, 31, 34] and three case–control studies
[42, 46, 49] investigated the association between broccoli intake
discussion
and CRC risk. The summary RR was 0.82 (95% CI 0.65–1.02), To our knowledge, this is the first meta-analysis evaluating an
borderline significant, for high broccoli intake (comparing the association between CV intake and CRC. In general, the results
highest with the lowest category) (Table 1, Figure 3), with of this meta-analysis suggested that intake of CV may reduce

Volume 24 | No. 4 | April 2013 doi:10.1093/annonc/mds601 | 

Licensed to Natalia Sabrina Ortiz - naati.ortiz@hotmail.com

 original articles Annals of Oncology

Table 1. Summary risk estimates of the association between cruciferous vegetable consumption and colorectal cancer risk

Number of studies Summary RR (95% CI) Q-statistic I2 P*h P**

h
Value (%)
Overall studies
CV 35 0.82 (0.75–0.90) 98.57 66.5 <0.001 –
Cabbage 9 0.76 (0.60–0.97) 19.20 58.3 0.014 –
Broccoli 6 0.82 (0.65–1.02) 11.07 54.8 0.050 –
Subgroup analyses for CV
High-quality studies (scores ≥7) 20 0.88 (0.80–0.97) 48.57 60.9 <0.001 0.046
Study design
Prospective studies 11 0.93 (0.87–1.00) 15.40 35.1 0.118 0.025
Case–control studies 23 0.74 (0.65–0.84) 74.37 70.4 <0.001
Type of control subjects
Population based 11 0.79 (0.66–0.93) 35.18 71.6 <0.001 0.487
Hospital based 9 0.71 (0.56–0.89) 30.76 74.0 <0.001
Geographic location
Europe 10 0.89 (0.76–1.04) 24.23 62.8 0.004 0.822
America 14 0.81 (0.72–0.92) 34.51 62.3 0.001
Asia 7 0.82 (0.64–1.05) 16.86 64.4 0.010
Gender
Male 16 0.80 (0.69–0.93) 54.75 72.6 <0.001 0.546
Female 16 0.87 (0.77–0.98) 29.26 48.7 0.015
Anatomic cancer site
Colon 16 0.78 (0.69–0.89) 37.06 59.5 0.001 0.225
Rectum 9 0.91 (0.74–1.13) 21.21 62.3 0.007
Cancer subsite of colon
Proximal 6 0.80 (0.67–0.95) 8.66 42.3 0.123 0.709
Distal 6 0.74 (0.56–0.98) 13.84 63.9 0.017
Adjustment for confounders
Body mass index
Yes 14 0.90 (0.85–0.95) 16.58 21.6 0.219 0.020
No 19 0.70 (0.59–0.83) 72.10 75.0 <0.001
Alcohol
Yes 15 0.89 (0.84–0.94) 21.07 33.5 0.100 0.061
No 19 0.73 (0.61–0.86) 73.77 75.6 <0.001
Physical activity
Yes 13 0.90 (0.85–0.96) 15.62 23.2 0.209 0.060
No 21 0.72 (0.59–0.88) 63.05 76.2 <0.001
Total energy intake
Yes 16 0.90 (0.85–0.95) 16.14 7.0 0.373 0.014
No 18 0.71 (0.59–0.85) 75.87 77.6 <0.001
Meat intake
Yes 9 0.92 (0.85–0.99) 7.63 0 0.470 0.290
No 25 0.78 (0.69–0.89) 86.33 72.2 <0.001
Family history of CRC/adenomatous polyposis
Yes 13 0.93 (0.87–0.99) 15.76 23.9 0.202 0.020
No 21 0.73 (0.64–0.84) 73.43 72.8 <0.001
Smoking status
Yes 14 0.91 (0.86–0.97) 18.58 30.0 0.137 0.019
No 20 0.72 (0.62–0.84) 72.68 73.9 <0.001
Subgroup analyses for cabbage
High-quality studies (scores ≥7) 2 0.88 (0.55–1.39) 0 0 0.935 0.633
Study design
Prospective studies 1 0.96 (0.54–1.72) N/A N/A N/A 0.557
Case–control studies 8 0.74 (0.57–0.97) 18.98 63.1 0.008
Subgroup analyses for broccoli
High-quality studies (scores ≥7) 5 0.86 (0.76–0.97) 10.55 43.1 0.103 0.576

Continued

 | Wu et al. Volume 24 | No. 4 | April 2013

Licensed to Natalia Sabrina Ortiz - naati.ortiz@hotmail.com

 Annals of Oncology original articles
Table 1. Continued

Number of studies Summary RR (95% CI) Q-statistic I2 P*h P**

h
Value (%)
Study design
Prospective studies 3 0.91 (0.80–1.03) 1.41 0 0.494 0.359
Case–control studies 3 0.60 (0.32–1.13) 7.32 72.7 0.026

CI, confidence interval; CV, cruciferous vegetables; N/A, not available; RR, relative risk.
*P-value for heterogeneity within each subgroup.
**P-value for heterogeneity between subgroups with meta-regression analysis.

Figure 2. Forest plot (random-effect model) of cabbage consumption and colorectal cancer risk. Squares indicate study-specific relative risks (size of the
square reflects the study-specific statistical weight); horizontal lines indicate 95% CIs; the diamond indicates the summary relative risk estimate with its 95%
CI. CI, confidence interval; F, females; M, males; RR, relative risk.

Figure 3. Forest plot (random-effect model) of broccoli consumption and colorectal cancer risk. Squares indicate study-specific relative risks (size of the
square reflects the study-specific statistical weight); horizontal lines indicate 95% CIs; the diamond indicates the summary relative risk estimate with its 95%
CI. CI, confidence interval; F, females; RR, relative risk.

Volume 24 | No. 4 | April 2013 doi:10.1093/annonc/mds601 | 

Licensed to Natalia Sabrina Ortiz - naati.ortiz@hotmail.com

 original articles Annals of Oncology

the risk of CRC in humans (comparing the highest with the the subgroup analyses of CV intake and CRC risk by study
lowest category). Specific analysis for cabbage yielded similar quality, we observed that the magnitude of risk reduction
inverse association but the result of broccoli showed borderline reported in high-quality studies was weaker than that reported
significance. Our findings from the case–control studies in the overall analysis (18% versus 12%), which indicated that
suggested a reduction between CV intake and CRC risk, but the association may have been changed by poor study
the results from the prospective studies showed borderline methodologies. Likewise, in the subgroup analyses by type of
statistical significance (Table 1, Figures 4 and 5). Additionally, control subjects, the protective effect in hospital-based control
we also observed significant inverse associations in colon subjects was stronger than that in population-based ones
cancer and its distal cancer subsite among both prospective (Table 1), which might mean hospital-based case–control
and case–control studies (supplementary Table S4, available at studies more inclined to selection bias [19].
Annals of Oncology online). For the subgroup analysis of CV intake and CRC risk by
The inverse association between CV intake and risk of CRC anatomic CRC site and subsite of colon cancer, we observed
is biologically plausible. CV is a good source of GLS which can a statistically significant inverse association between CV
be hydrolyzed by the plant enzyme myrosinase into intake and colon cancer and its proximal and distal subsites.
biologically active compounds (ITCs and I3C). Nonetheless, These results are consistent with a recent meta-analysis of the
the anti-carcinogenic actions of GLS are commonly attributed relation between fruit and vegetable and CRC [10] and a
to ITCs. The CRC cancer-protective effects of CV and their large cohort study in The Netherlands [35]. However, the
GLS breakdown products likely involve complex interactions of result of a study in 2007 that has pooled 14 prospective
several mechanisms: the modulation of xenobiotic- cohorts showed that associations with colon cancer risk are
metabolizing enzymes to protect against chemically induced not significant [ pooled multivariate RR: 0.99 (95% CI 0.93–
tumors [59], antioxidant effects, and the induction of apoptosis 1.06) for the highest tertile versus lowest] for CV [63].
and cell-cycle arrest [60, 61]. Furthermore, evidence from Compared with the number of studies on colon cancer, fewer
animal models of colon cancer also showed that CV has studies conducted analyses of rectal cancer (16 versus 9),
generally been shown to inhibit chemical carcinogenesis [60, which may diminish our statistical power to detect an
62]. On the other hand, CV are good sources of fiber which association. Thus, more prospective and well-designed studies
can prevent CRC by several plausible mechanisms, including should focus on this.
increased fecal bulk and dilution of carcinogens in the colonic As with all meta-analyses, several limitations must be
lumen, reduced transit time, and bacterial fermentation of fiber addressed. First, as the observational nature of the data, it is
to short-chain fatty acids [7, 8]. possible that the observed significant inverse association
Compared with retrospective studies, prospective studies are between CV intakes and CRC risk could be due to unmeasured
less susceptible to bias (e.g. recall bias, selection bias) due to or residual confounding. Higher intake of CV may be
their nature. Furthermore, case–control studies had a lower associated with other health behaviors (e.g. higher levels of PA,
median quality score than prospective studies (6 versus 7). In lower intakes of alcohol, and red meat). However, many of the

Figure 4. Forest plot (fixed-effect model) of cruciferous vegetables consumption and colorectal cancer risk in prospective studies. Squares indicate study-
specific relative risks (size of the square reflects the study-specific statistical weight); horizontal lines indicate 95% CIs; the diamond indicates the summary
relative risk estimate with its 95% CI. CI, confidence interval; F, females; M, males; RR, relative risk.

 | Wu et al. Volume 24 | No. 4 | April 2013

Licensed to Natalia Sabrina Ortiz - naati.ortiz@hotmail.com

 Annals of Oncology original articles

Figure 5. Forest plot (random-effect model) of cruciferous vegetables consumption and colorectal cancer risk in case–control studies. Squares indicate
study-specific relative risks (size of the square reflects the study-specific statistical weight); horizontal lines indicate 95% CIs; the diamond indicates the
summary relative risk estimate with its 95% CI. CI, confidence interval; M, males; RR, relative risk.

studies adjusted for known confounding factors (e.g. BMI, sufficient power to detect the putative association between CV
alcohol consumption, PA). In addition, the results generally intake and CRC, although compared with CV, fewer studies
showed inverse association in the subgroup analyses when we and cases were included in the subgroup analyses of cabbage
stratified the results according to adjustment for confounding and broccoli. The relatively large number of included studies
factors or other study characteristics. Second, because most also allowed us to conduct subgroup analyses according to a
studies used FFQ to assess CV intake, our results may have number of study characteristics to identify potential sources of
been influenced by misclassification. On the other hand, heterogeneity.
exposure levels of ITCs can be further affected by different food In summary, this meta-analysis suggested that high intake of
preparation methods [64]. Boiling CV results in a 30%–60% CV can decrease risk of CRC and colon cancer. More in-depth
loss of intact GLS due to thermal degradation and leaching studies are warranted to report more detailed results, including
[65]. However, none of the studies separates the CV intake by other specific vegetables within the CV family, stratified results
cooking methods. Therefore, future epidemiological studies by anatomic cancer site, subsite of colon cancer, food
should consider whether the inverse association of CV will be preparation methods, or adjustment for potential confounders.
affected by the food preparation methods. Third, significant
heterogeneity and possible publication bias must be
considered. There was significant heterogeneity for all studies
acknowledgements
combined (Q = 98.57, P < 0.001, I 2 = 66.5%) in the pooled Q-JW and Y-BX designed research; Q-JW, YY and Y-BX
analysis of CV intake. This could be explained by many conducted research; Q-JW, YY analyzed data; Q-JW wrote the
factors, mainly the design of this study (Table 1). Publication draft; all authors read, reviewed and approved the final
bias can be a problem in meta-analyses of published studies manuscript. Y-BX had primary responsibility for final content.
but we found no statistical evidence of publication bias in this
meta-analysis. Last but not least, due to different methods used
to report CV intake among studies and because we did not funding
attempt to contact the authors for additional information, we This work was supported by the fund of the State Key Project
failed to carry out a dose–response analysis between CV intake Specialized for Infectious Diseases of China (2008ZX10002-015
and CRC. and 2012ZX10002008-002). EV was also supported by the
The strengths of this meta-analysis include a large sample Fogarty International Clinical Research Scholars and Fellows
size with 24 275 cases and 1 295 063 subjects which provided Support Center at the Vanderbilt Institute for Global Health,

Volume 24 | No. 4 | April 2013 doi:10.1093/annonc/mds601 | 

Licensed to Natalia Sabrina Ortiz - naati.ortiz@hotmail.com

 original articles Annals of Oncology

funded by the Fogarty International Center, NIH, through an 20. Larsson SC, Orsini N, Wolk A. Vitamin B6 and risk of colorectal cancer: a meta-
R24 Training Grant (grant number R24TW007988), and the analysis of prospective studies. JAMA 2010; 303: 1077–1083.
Cancer Prevention and Control Training Program at the 21. Oh SW, Myung SK, Park JY et al.. Aspirin use and risk for lung cancer: a meta-
analysis. Ann Oncol 2011; 22: 2456–2465.
University of Alabama at Birmingham funded through the
22. Zhou Y, Zhuang W, Hu W et al.. Consumption of large amounts of Allium
National Institutes of Health (5R25 CA047888).
vegetables reduces risk for gastric cancer in a meta-analysis. Gastroenterology
2011; 141: 80–89.
23. Higgins JP, Thompson SG. Quantifying heterogeneity in a meta-analysis. Stat
disclosure Med 2002; 21: 1539–1558.
The authors have declared no conflicts of interest. 24. DerSimonian R, Laird N. Meta-analysis in clinical trials. Control Clin Trials 1986;
7: 177–188.
25. Woolf B. On estimating the relation between blood group and disease. Ann Hum
Genet 1955; 19: 251–253.
references 26. Egger M, Davey SG, Schneider M et al. Bias in meta-analysis detected by
a simple, graphical test. BMJ 1997; 315: 629–634.
1. Jemal A, Bray F, Center MM et al.. Global cancer statistics. CA Cancer J Clin
2011; 61: 69–90. 27. Begg CB, Mazumdar M. Operating characteristics of a rank correlation test for
publication bias. Biometrics 1994; 50: 1088–1101.
2. Giovannucci E, Kana W. Cancer of the colon and rectum. In Schottenfeld D,
Fraumeni JJ, eds, Cancer Epidemiology and Prevention, 3rd edition. New York: 28. Bosetti C, Filomeno M, Riso P et al.. Cruciferous vegetables and cancer risk in
Oxford University Press, 2006; 809–829. a network of case-control studies. Ann Oncol 2012; 23: 2198–2203.
3. World Cancer Research Fund/American Institute for Cancer Research. Food, 29. Hsing AW, McLaughlin JK, Chow WH et al.. Risk factors for colorectal cancer
Nutrition, Physical Activity and the Prevention of Cancer: A Global Perspective. in a prospective study among U.S. white men. Int J Cancer 1998; 77:
Washington, DC: American Institute for Cancer Research, 2007. 549–553.
4. Kurilich AC, Tsau GJ, Brown A et al.. Carotene, tocopherol, and ascorbate 30. Yang G, Gao YT, Shu XO et al.. Isothiocyanate exposure, glutathione S-
contents in subspecies of Brassica oleracea. J Agric Food Chem 1999; 47: transferase polymorphisms, and colorectal cancer risk. Am J Clin Nutr 2010; 91:
1576–1581. 704–711.
5. Conaway CC, Yang YM, Chung FL. Isothiocyanates as cancer chemopreventive 31. Nomura AM, Wilkens LR, Murphy SP et al.. Association of vegetable, fruit, and
agents: their biological activities and metabolism in rodents and humans. Curr grain intakes with colorectal cancer: the Multiethnic Cohort Study. Am J Clin Nutr
Drug Metab 2002; 3: 233–255. 2008; 88: 730–737.
6. Hecht SS. Inhibition of carcinogenesis by isothiocyanates. Drug Metab Rev 2000; 32. Lin J, Zhang SM, Cook NR et al.. Dietary intakes of fruit, vegetables, and fiber,
32: 395–411. and risk of colorectal cancer in a prospective cohort of women (United States).
Cancer Causes Control 2005; 16: 225–233.
7. Lipkin M, Reddy B, Newmark H et al.. Dietary factors in human colorectal
cancer. Annu Rev Nutr 1999; 19: 545–586. 33. McCullough ML, Robertson AS, Chao A et al.. A prospective study of whole
grains, fruits, vegetables and colon cancer risk. Cancer Causes Control 2003;
8. Aune D, Chan DS, Lau R et al.. Dietary fibre, whole grains, and risk of colorectal
14: 959–970.
cancer: systematic review and dose-response meta-analysis of prospective
studies. BMJ 2011; 343: d6617. 34. Flood A, Velie EM, Chaterjee N et al.. Fruit and vegetable intakes and the risk of
colorectal cancer in the Breast Cancer Detection Demonstration Project follow-up
9. World Cancer Research Fund/American Institute for Cancer Research. http://
cohort. Am J Clin Nutr 2002; 75: 936–943.
www.dietandcancerreport.org/cup/current_progress/colorectal_cancer.php
(1 July 2012, date last accessed). 35. Voorrips LE, Goldbohm RA, van Poppel G et al.. Vegetable and fruit consumption
and risks of colon and rectal cancer in a prospective cohort study: The
10. Aune D, Lau R, Chan DS et al.. Nonlinear reduction in risk for colorectal cancer
Netherlands Cohort Study on Diet and Cancer. Am J Epidemiol 2000; 152:
by fruit and vegetable intake based on meta-analysis of prospective studies.
1081–1092.
Gastroenterology 2011; 141: 106–118.
36. Annema N, Heyworth JS, McNaughton SA et al.. Fruit and vegetable
11. Steinmetz KA, Potter JD. Vegetables, fruit, and cancer prevention: a review.
consumption and the risk of proximal colon, distal colon, and rectal cancers in a
J Am Diet Assoc 1996; 96: 1027–1039.
case-control study in Western Australia. J Am Diet Assoc 2011; 111:
12. Verhoeven DT, Goldbohm RA, van Poppel G et al.. Epidemiological studies on
1479–1490.
brassica vegetables and cancer risk. Cancer Epidemiol Biomarkers Prev 1996;
37. Ganesh B, Talole SD, Dikshit R. A case-control study on diet and colorectal
5: 733–748.
cancer from Mumbai, India. Cancer Epidemiol 2009; 33: 189–193.
13. Park Y, Subar AF, Kipnis V et al.. Fruit and vegetable intakes and risk of
38. Hu J, Mery L, Desmeules M et al.. Diet and vitamin or mineral supplementation
colorectal cancer in the NIH-AARP diet and health study. Am J Epidemiol 2007;
and risk of rectal cancer in Canada. Acta Oncol 2007; 46: 342–354.
166: 170–180.
39. Hu J, Morrison H, Mery L et al.. Diet and vitamin or mineral supplementation
14. Michels KB, Edward G, Joshipura KJ et al.. Prospective study of fruit and
and risk of colon cancer by subsite in Canada. Eur J Cancer Prev 2007; 16:
vegetable consumption and incidence of colon and rectal cancers. J Natl Cancer
275–291.
Inst 2000; 92: 1740–1752.
40. Slattery ML, Curtin K, Sweeney C et al.. Diet and lifestyle factor associations with
15. Pietinen P, Malila N, Virtanen M et al.. Diet and risk of colorectal cancer in a
CpG island methylator phenotype and BRAF mutations in colon cancer. Int J
cohort of Finnish men. Cancer Causes Control 1999; 10: 387–396.
Cancer 2007; 120: 656–663.
16. Steinmetz KA, Kushi LH, Bostick RM et al.. Vegetables, fruit, and colon cancer
41. Turner F, Smith G, Sachse C et al.. Vegetable, fruit and meat consumption and
in the Iowa Women’s Health Study. Am J Epidemiol 1994; 139: 1–15.
potential risk modifying genes in relation to colorectal cancer. Int J Cancer 2004;
17. Stroup DF, Berlin JA, Morton SC et al.. Meta-analysis of observational studies
112: 259–264.
in epidemiology: a proposal for reporting. Meta-analysis Of Observational Studies
42. Hara M, Hanaoka T, Kobayashi M et al.. Cruciferous vegetables, mushrooms, and
in Epidemiology (MOOSE) group. JAMA 2000; 283: 2008–2012.
gastrointestinal cancer risks in a multicenter, hospital-based case-control study
18. Wells GA, Shea B, O’Connell D et al.. The Newcastle-Ottawa Scale (NOS) for
in Japan. Nutr Cancer 2003; 46: 138–147.
assessing the quality of nonrandomised studies in meta-analyses. Available from:
43. Chiu BC, Ji BT, Dai Q et al.. Dietary factors and risk of colon cancer in
http://www.ohri.ca/programs/clinical_epidemiology/oxford.asp.
Shanghai, China. Cancer Epidemiol Biomarkers Prev 2003; 12: 201–208.
19. Yang WS, Va P, Wong MY et al.. Soy intake is associated with lower lung cancer
44. Seow A, Quah SR, Nyam D et al.. Food groups and the risk of colorectal
risk: results from a meta-analysis of epidemiologic studies. Am J Clin Nutr 2011;
carcinoma in an Asian population. Cancer 2002; 95: 2390–2396.
94: 1575–1583.

 | Wu et al. Volume 24 | No. 4 | April 2013

Licensed to Natalia Sabrina Ortiz - naati.ortiz@hotmail.com

 Annals of Oncology original articles
45. Deneo-Pellegrini H, Boffetta P, De Stefani E et al.. Plant foods and differences 56. Young TB, Wolf DA. Case-control study of proximal and distal colon cancer and
between colon and rectal cancers. Eur J Cancer Prev 2002; 11: 369–375. diet in Wisconsin. Int J Cancer 1988; 42: 167–175.
46. Evans RC, Fear S, Ashby D et al.. Diet and colorectal cancer: an 57. Kune S, Kune GA, Watson LF. Case-control study of dietary etiological factors:
investigation of the lectin/galactose hypothesis. Gastroenterology 2002; 122: the Melbourne Colorectal Cancer Study. Nutr Cancer 1987; 9: 21–42.
1784–1792. 58. Graham S, Dayal H, Swanson M et al.. Diet in the epidemiology of cancer of
47. Freedman AN, Michalek AM, Marshall JR et al.. Familial and nutritional risk the colon and rectum. J Natl Cancer Inst 1978; 61: 709–714.
factors for p53 overexpression in colorectal cancer. Cancer Epidemiol Biomarkers 59. Steinkellner H, Rabot S, Freywald C et al.. Effects of cruciferous vegetables
Prev 1996; 5: 285–291. and their constituents on drug metabolizing enzymes involved in the
48. Inoue M, Tajima K, Hirose K et al.. Subsite-specific risk factors for colorectal bioactivation of DNA-reactive dietary carcinogens. Mutat Res 2001; 480–481:
cancer: a hospital-based case-control study in Japan. Cancer Causes Control 285–297.
1995; 6: 14–22. 60. Smith TK, Mithen R, Johnson IT. Effects of Brassica vegetable juice on the
49. Steinmetz KA, Potter JD. Food-group consumption and colon cancer in the induction of apoptosis and aberrant crypt foci in rat colonic mucosal crypts in
Adelaide Case-Control Study. I. Vegetables and fruit. Int J Cancer 1993; 53: vivo. Carcinogenesis 2003; 24: 491–495.
711–719. 61. Smith TK, Lund EK, Clarke RG et al.. Effects of Brussels sprout juice on the cell
50. Zaridze D, Filipchenko V, Kustov V et al.. Diet and colorectal cancer: results cycle and adhesion of human colorectal carcinoma cells (HT29) in vitro. J Agric
of two case-control studies in Russia. Eur J Cancer 1992; 29A: 112–115. Food Chem 2005; 53: 3895–3901.
51. Bidoli E, Franceschi S, Talamini R et al.. Food consumption and cancer of the 62. Uhl M, Kassie F, Rabot S et al.. Effect of common Brassica vegetables
colon and rectum in north-eastern Italy. Int J Cancer 1992; 50: 223–229. (Brussels sprouts and red cabbage) on the development of preneoplastic
52. Wohlleb JC, Hunter CF, Blass B et al.. Aromatic amine acetyltransferase as a lesions induced by 2-amino-3-methylimidazo[4,5-f ]quinoline (IQ) in liver and
marker for colorectal cancer: environmental and demographic associations. colon of Fischer 344 rats. J Chromatogr B Analyt Technol Biomed Life Sci
Int J Cancer 1990; 46: 22–30. 2004; 802: 225–230.
53. Benito E, Obrador A, Stiggelbout A et al.. A population-based case-control study 63. Koushik A, Hunter DJ, Spiegelman D et al.. Fruits, vegetables, and colon cancer
of colorectal cancer in Majorca. I. Dietary factors. Int J Cancer 1990; 45: risk in a pooled analysis of 14 cohort studies. J Natl Cancer Inst 2007;
69–76. 99: 1471–1483.
54. Lee HP, Gourley L, Duffy SW et al.. Colorectal cancer and diet in an Asian 64. Higdon JV, Delage B, Williams DE et al.. Cruciferous vegetables and human
population—a case-control study among Singapore Chinese. Int J Cancer 1989; cancer risk: epidemiologic evidence and mechanistic basis. Pharmacol Res
43: 1007–1016. 2007; 55: 224–236.
55. West DW, Slattery ML, Robison LM et al.. Dietary intake and colon cancer: sex- 65. Zhang Y. Cancer-preventive isothiocyanates: measurement of human exposure
and anatomic site-specific associations. Am J Epidemiol 1989; 130: 883–894. and mechanism of action. Mutat Res 2004; 555: 173–190.

Annals of Oncology 24: 1087–1093, 2013

doi:10.1093/annonc/mds587
Published online 21 November 2012

Influence of imatinib interruption and rechallenge on

the residual disease in patients with advanced GIST:
results of the BFR14 prospective French Sarcoma
Group randomised, phase III trial
A. Patrikidou1, S. Chabaud2, I. Ray-Coquard2, B. N. Bui3, A. Adenis4, M. Rios5, F. Bertucci6,
F. Duffaud7, C. Chevreau8, D. Cupissol9, J. Domont1, D. Pérol2, J. Y. Blay10 & A. Le Cesne1,*,
for the French Sarcoma Group
1
Department of Medicine, Institut Gustave Roussy, Villejuif; 2Biostatistics Department, Centre Léon Bérard, Lyon; 3Department of Medical Oncology, Institut Bergonié,
Bordeaux; 4Department of Medical Oncology Centre Oscar Lambret, Lille; 5Department of Medical Oncology, Centre Alexis Vautrin, Nancy; 6Department of Medical
Oncology, Institut Paoli Calmettes, Marseille; 7Department of Medicine, Hopital La Timone, Marseille; 8Department of Medicine, Institut Claudius Regaud, Toulouse;
9
Department of Medical Oncology, Centre Val d’Aurelle, Montpellier; 10INSERM U590 & Université Claude Bernard & Centre Léon Bérard, Lyon, France

Received 17 June 2012; revised 27 September 2012; accepted 28 September 2012

Background: We previously demonstrated that interruption of imatinib mesylate (IM) in responding patients ( pts) with
advanced gastrointestinal stromal tumours (GISTs) results in rapid reprogression. The impact of interruption on residual

*Correspondence to: Dr A. Le Cesne, Department of Medical Oncology, Institut Gustave

Roussy, 114 rue Edward Vaillant, Villejuif 94805, France. Tel: +33-1-42-11-43-16;
Fax: +33-1-42-11-52-19; E-mail: lecesne@igr.fr

© The Author 2012. Published by Oxford University Press on behalf of the European Society for Medical Oncology.
All rights reserved. For permissions, please email: journals.permissions@oup.com.

Licensed to Natalia Sabrina Ortiz - naati.ortiz@hotmail.com