This article was downloaded by: [East China Normal University] On: 21 March 2014, At: 18:55 Publisher: Routledge Informa Ltd Registered in England and Wales Registered Number: 1072954 Registered office: Mortimer House, 37-41 Mortimer Street, London W1T 3JH, UK

Nutrition and Cancer Publication details, including instructions for authors and subscription information: http://www.tandfonline.com/loi/hnuc20

Fruit and Vegetables Consumption and Gastric Cancer: A Systematic Review and Meta-Analysis of Cohort Studies Nuno Lunet , Antonio Lacerda-Vieira & Henrique Barros Published online: 18 Nov 2009.

To cite this article: Nuno Lunet , Antonio Lacerda-Vieira & Henrique Barros (2005) Fruit and Vegetables Consumption and Gastric Cancer: A Systematic Review and Meta-Analysis of Cohort Studies, Nutrition and Cancer, 53:1, 1-10, DOI: 10.1207/ s15327914nc5301_1 To link to this article: http://dx.doi.org/10.1207/s15327914nc5301_1

PLEASE SCROLL DOWN FOR ARTICLE Taylor & Francis makes every effort to ensure the accuracy of all the information (the “Content”) contained in the publications on our platform. However, Taylor & Francis, our agents, and our licensors make no representations or warranties whatsoever as to the accuracy, completeness, or suitability for any purpose of the Content. Any opinions and views expressed in this publication are the opinions and views of the authors, and are not the views of or endorsed by Taylor & Francis. The accuracy of the Content should not be relied upon and should be independently verified with primary sources of information. Taylor and Francis shall not be liable for any losses, actions, claims, proceedings, demands, costs, expenses, damages, and other liabilities whatsoever or howsoever caused arising directly or indirectly in connection with, in relation to or arising out of the use of the Content. This article may be used for research, teaching, and private study purposes. Any substantial or systematic reproduction, redistribution, reselling, loan, sub-licensing, systematic supply, or distribution in any form to anyone is expressly forbidden. Terms & Conditions of access and use can be found at http:// www.tandfonline.com/page/terms-and-conditions

NUTRITION AND CANCER, 53(1), 1–10 Copyright © 2005, Lawrence Erlbaum Associates, Inc.

REVIEWS

Fruit and Vegetables Consumption and Gastric Cancer: A Systematic Review and Meta-Analysis of Cohort Studies

Downloaded by [East China Normal University] at 18:55 21 March 2014

Nuno Lunet, António Lacerda-Vieira, and Henrique Barros

Abstract: Fruit and vegetable intake is widely recognized as protective for gastric cancer occurrence but prospective research challenged this belief. To evaluate the influence of design options in such results we did a meta-analysis of relevant published cohort studies identified from inception to 2004 in PubMed, EMBASE®, and LILACS. Random-effects meta-analysis, stratification, and meta-regression were used to pool effects and to analyze the association with type of outcome event and length of follow-up independent of other study characteristics. An inverse association was observed between fruit intake and gastric cancer incidence (relative risk, RR = 0.82; 95% confidence interval, CI = 0.73–0.93) and stronger for follow-up periods of ≥10 yr (RR = 0.66; 95% CI = 0.52–0.83) but not when the study outcome was death (RR = 1.08; 95% CI = 0.86–1.35). For vegetables, the RR was 0.88 (95% CI = 0.69–1.13) using all incidence studies and 0.71 (95% CI = 0.53–0.94) when considering only those with the longer follow-up. The association observed between vegetable intake and gastric cancer mortality was 1.05 (95% CI = 0.89–1.25). Other study characteristics assessed added no significant contribution to explain heterogeneity. This meta-analysis showed that design options might play a key role in the observed magnitude or the direction of the association between fruit and vegetable intake and gastric cancer.

Introduction A continued international decline in gastric cancer mortality, especially for cancers of the intestinal histological type, is one of the most remarkable events in cancer epidemiology (1). However, despite the long-lasting decreasing trend in gastric cancer incidence and mortality, the absolute number of incident cases is increasing, and gastric cancer remains worldwide one of the most frequent malignancies, with 930,000 new cases and 700,000 deaths estimated for 2002 (2).

Chronic atrophic gastritis and intestinal metaplasia are considered precursor lesions of intestinal type gastric cancer, the progression to cancer being modulated by environmental exposures, including dietary (3). Fruit and vegetable intake has long been associated with a decreased risk of gastric cancer (4,5), but prospective studies tend to provide less-convincing evidence than case-control investigations (6), partly explained by limited ranges and imprecise measurements of dietary intake within each cohort and recall or selection biases in retrospective studies. However, additional variability is present among prospective studies. The nature and magnitude of the effects of dietary factors in different phases of gastric carcinogenesis are poorly understood, but it is plausible that the protective effect of fruit and vegetables differs with timing of exposure. It is therefore possible to hypothesize that the duration of follow-up or the type of selected outcome event might result in heterogeneous cancer risk estimates. This led us to evaluate such specific methodological aspects by conducting a systematic review of the cohort studies that estimated the effect of fruit and vegetable consumption on gastric cancer.

Methods Published cohort studies presenting results on the association between fruit or vegetable consumption and gastric cancer were identified through PubMed (http://www.ncbi.nlm. nih.gov/entrez/), EMBASE®, and LILACS (http://bases. bvs.br). Searches covered the period from 1966 to 2004 under the terms {[(stomach OR gastric OR cardia) AND cancer] OR stomach cancer OR gastric cancer} AND (nutrition OR diet OR lifestyle OR fruit OR vegetable) AND (cohort analysis OR prospective study OR cohort) without language restrictions. Studies evaluating cohorts of patients with specific pathologies (for example, a previous cancer or a chronic

All authors are affiliated with the Department of Hygiene and Epidemiology, University of Porto Medical School, 4200–319 Porto, Portugal.

Downloaded by [East China Normal University] at 18:55 21 March 2014

disease or gastrectomized patients) were not eligible for review. Additionally, we performed a hand search of our personal database and the reference lists provided by the papers previously identified. In the meta-analysis we included the studies evaluating fruit or vegetable groups classified as “all” or “total.” Exposures presented as raw vegetables, cooked vegetables, green-yellow vegetables, other vegetables, green vegetables, citrus fruits, or other fruit were not taken as equivalent to “all” or “total” and thus were not considered. Studies that reported “fresh vegetables” or “fresh fruit” were included under the hypothesis that fresh vegetables or fruit accounts for a very high proportion of the total consumption (6). All studies fulfilling these selection criteria were considered for our primary analysis. Two reviewers (NL and ALV) extracted data on relative risk (RR) estimates for the association between gastric cancer and the consumption of fruit and/or vegetables considering two exposure levels (lowest vs. highest) as well as study characteristics, following a previously defined protocol. Discrepancies in the evaluation of the articles were resolved by consensus involving a third researcher (HB). When studies had results published for different follow-up times, we opted for the longer follow-up reported. When a study provided several risk estimates, we opted for the one adjusted for the largest number of possible confounding variables. From publications providing separate results for males and females, gender-specific estimates were considered in the final analysis as if obtained from different studies. When studies presented results including and excluding events in the initial phase of follow-up, we opted for the data with all participants, as it was the most frequently presented in the original papers. Each study was characterized according to country of origin, follow-up duration, age of participants at enrollment, method of diet measurement, and control of confounding (Table 1). To estimate the difference between extreme categories of intake, we assigned each class the midpoint of the range. Because the upper category was always open, we assigned it the amplitude of the preceding stratum. When dietary consumption was provided in grams, we considered that a standard serving contained 100 g of fruit or vegetables. For adjusted RRs comparing extreme categories of exposure presented without precision estimates, we considered the adjusted RR estimate and the variance obtained from the crude data. Combined risk estimates and 95% confidence intervals (CIs) were computed with STATA® version 8.0 using a random-effects method. We quantitatively assessed heterogeneity using the I2 statistic (7). Stratified analysis according to study characteristics and meta-regression were performed to explore heterogeneity. Because of the small number of studies available to perform meta-regression, each independent variable was added separately into the model that always included follow-up time and outcome event. 2

Results We identified and screened 763 potentially relevant references (including overlapping publications across the 3 databases); 580 were excluded on the basis of title and abstract (mostly clinical studies, designs other than cohort or animal research) and 183 were retrieved for more detailed assessment. One hundred sixty-two were excluded after further evaluation of the full articles because they were reviews, had study designs other than cohort, or no results were presented on the association between fruit or vegetable consumption and gastric cancer. Twenty-one references were included in the systematic review. Scanning reference lists and personal databases identified two additional articles. Of these 23 publications of cohort studies reporting on the association between fruit and/or vegetable intake and gastric cancer we excluded 4 articles (8–11) that used material presented in more detail elsewhere (12–14), and we selected the results describing subjects not exposed to radiation (15) between two reports from the same study (15,16) that evaluated gastric cancer mortality in atomic bomb survivors. One study was an ecological analysis of 16 cohorts (17) and therefore was excluded. Inoue et al. (13) presented results comparing the same set of cases with subjects with and without atrophic gastritis and we opted to include only the latter. From the remaining 17 references (described in Table 1), 9 studies did not provide results for vegetables as we defined the exposure (13,15,18,20,22,23,26,28,30) and were consequently excluded from the specific analysis. In those studies, nonsignificant negative associations were found for the highest consumption of green-yellow (13,18,28), yellow/orange (26), or raw vegetables (13,15,22,30), but two mortality studies reported no association or a nonsignificant increased risk with green-yellow (20) or other vegetables (20,28). Similarly, four studies (18,23,24,26) did not fulfill our inclusion criteria for fruit consumption. One showed a negative association between gastric cancer and combined consumption of fruit and vegetables (23), another (24) found no significant association between citrus fruit/juice intake and gastric cancer mortality, and two (18,26) did not assess fruit intake. From the remaining 14 publications, we extracted data on RR estimates for the association between gastric cancer and the consumption of fruit (13 articles) or vegetables (8 articles). Seven studies evaluated incidence of new cases and seven evaluated mortality. The median follow-up time was 11.1 yr, ranging from 5 to 26 yr. Most studies were from Asian countries (n = 9), and only one was from Europe (n = 1). Only three studies stated that a validated food-frequency questionnaire was used and most (n = 9) relied on self-administered questionnaires. The difference between extreme categories of exposure was similar for fruit (median, 7.2; interquartile range, 5–9.45) and vegetables (median, 7.5; interquartile range, 7–11.5). Nine studies provided results adjusted for smoking and only five controlled for socioeconomic confounding using information on social status or education. Nutrition and Cancer 2005

Vol. 53, No. 1

Table 1. Studies Identified for Systematic Reviewa Ref. Year of Publication Country

Downloaded by [East China Normal University] at 18:55 21 March 2014

18 1986 Japan 19 1991 United States

Outcome (no. of cases) Follow-up (yr) Mortality (n = 5,247) 17 (maximum) 14.5 (average) Mortality (n = 75) 16.3 (average) 20 (maximum)

Age at Enrollment (yr)

Evaluation of Exposure Validation of Method Items Evaluated

≥40

Interview, FFQ (items, NS) NS Green-yellow vegetables

≥35

Mailed FFQ (35 items) NS Fruit, vegetables, cruciferous vegetables

20 1992 Japan

Mortality (n = 57) 6 (maximum) 5.7 (average)

Male, ≥40 Female, ≥30

Interview, mailed FFQ (25 items) NS Green-yellow vegetables, other vegetables, fruit

21 1994 China

Incidence (n = 539) 5 (maximum)

40–69

Interview, FFQ (9 items) Not validated Fresh vegetables, fresh fruits, pickled vegetables

12 1995 United States

Incidence (n = 250) 26 (maximum) 22.2 (average)

Born 1900–1919 Examined 1965–1968

Interview 24-h dietary recall Vegetables, fruits

13 1996 Japan

Incidence (n = 69) 8 (maximum) 6.0 (average)

≤39 to ≥80

Self-administered FFQ (10 items) NS Raw vegetables, green-yellow vegetables, pickled vegetables, fruit

Exposure to Fruit and Vegetables (“highest” vs. “lowest” category) Results 4 groups (≥4 times/wk vs. ≤3 times/yr) Green-yellow vegetables: men, RR = 0.66; women, RR = 0.66 Quartiles (NS) Fruit: RR = 1.5 (95% CI = 0.75–2.93); vegetables: RR = 0.9 (95% CI = 0.48–1.78) 3 groups (daily vs. ≤1–2 times/mo) Fruit: RR = 1.92 (95% CI = 1.03–3.59); green-yellow vegetables: RR = 1.54 (95% CI = 0.77–3.11); other vegetables: RR = 1.15 (95% CI = 0.59–2.27) 3 groups Fresh vegetables (≥60 vs. ≤30 times/mo) 2 groups: Fresh fruit ≥1 vs. 0 time/mo) Fresh vegetables: RR = 1.1 (95% CI = 0.8–1.4); fresh fruit: RR = 0.9 (95% CI = 0.8–1.1) 4 groups Vegetables (≥80 vs. 0 g/day) Fruits (≥301 vs. 0 g/day) Vegetables: RR = 0.6 (95% CI = 0.3–0.9); fruit: RR = 0.6 (95% CI = 0.4–1.0) 3 groups (daily vs. rare) Fruit: RR = 0.55 (95% CI = 0.22–1.35); raw vegetables: RR = 0.67 (95% CI = 0.29–1.57); green-yellow vegetables: RR = 0.74 (95% CI = 0.17–3.2)

Control of Confounding Age, stratified by sex

Year of birth, smoking

Age, sex

Observations Reasons for Exclusion (when applicable) Excluded: all vegetables and/or all fruits not evaluated Only men (white American men, largely of Scandinavian and German descent) All vegetables not evaluated

Smoking, cancer history in first-degree relatives; controls matched with index case on sex and age (±1 yr)

Nested Case-control

Age

Only men (Hawaiian men of Japanese ancestry) Nested Case-control

Age, sex

All vegetables not evaluated

(continued)

3

4

Table 1 (Continued)

Nutrition and Cancer 2005

Downloaded by [East China Normal University] at 18:55 21 March 2014

Ref. Year of Publication Country

Outcome (no. of cases) Follow-up (yr)

Age at Enrollment (yr)

Evaluation of Exposure Validation of Method Items Evaluated

Exposure to Fruit and Vegetables (“highest” vs. “lowest” category) Results

14 1998 The Netherlands

Incidence (n = 282) 6.3 (maximum) 5.5 (average)

55–69

Mailed FFQ (150 items) Validated against a 9-day diet record Total vegetables and fruit, total vegetables, prepared vegetables, raw vegetables, brassicas, prepared leafy vegetables, raw leafy vegetables, total fruit, citrus fruit, apples, pears

Quintiles Median intake (g/day): fruit (325 vs. 46); vegetables (286 vs. 103) Fruit: RR = 0.97 (95% CI = 0.64–1.48); vegetables: RR = 0.86 (95% CI = 0.58–1.26)

22 1998 United States

Incidence (n = 108) 16.5 (maximum) 14.8 (average)

≥18

Interview, FFQ (13 items) NS Fresh fruit, raw vegetables, pickled vegetables

23 1998 Sweden

Incidence (n = 116) 25.5 (maximum) 21.0 (average)

Born 1886–1925 Evaluated 1967

Mailed dietary habits questionnaire (23 items) NS Fruit and vegetables

24 2001 United States

Mortality (n = 1,349) 14 (maximum)

NS

Mailed FFQ (32 items) NS Citrus fruit/juice, plant foods, vegetables

25 2002 Japan

Mortality (n = 379) 9 (maximum) 7.3 (average)

40–79

Self-administered FFQ (32 items) NS Vegetables, fruit

2 groups (≥7 vs. 0–6 times/wk) Fresh fruit: men RR = 0.6 (95% CI = 0.3–1.0), women RR = 0.7 (95% CI = 0.4–1.4); raw vegetables: men, RR = 0.9 (95% CI = 0.5–1.5), women, RR = 0.7 (95% CI = 0.4–1.4) 4-point scale (high part of diet vs. very little/no part of diet) Fruit and vegetables: RR = 5.53 (95% CI = 1.67–18.31) Tertiles Vegetables (days/wk): men (≥13 vs.