Variety in vegetable and fruit consumption and the risk of gastric and esophageal cancer in the European prospective investigation into cancer and nutrition
Diets high in vegetables and fruits have been suggested to be inversely associated with risk of gastric cancer. However, the evidence of the effect of variety of consumption is limited. We therefore investigated whether consumption of a variety of vegetables and fruit is associated with gastric and esophageal cancer in the European Prospective Investigation into Cancer and Nutrition study. Data on food consumption and follow-up on cancer incidence were available for 452,269 participants from 10 European countries. After a mean follow-up of 8.4 years, 475 cases of gastric and esophageal adenocarcinomas (180 noncardia, 185 cardia, gastric esophageal junction and esophagus, 110 not specified) and 98 esophageal squamous cell carcinomas were observed. Diet Diversity Scores were used to quantify the variety in vegetable and fruit consumption. We used multivariable Cox proportional hazard models to calculate risk ratios. Independent from quantity of consumption, variety in the consumption of vegetables and fruit combined and of fruit consumption alone were statistically significantly inversely associated with the risk of esophageal squamous cell carcinoma (continuous hazard ratio per 2 products increment 0.88; 95% CI 0.79–0.97 and 0.76; 95% CI 0.62–0.94, respectively) with the latter particularly seen in ever smokers. Variety in vegetable and/or fruit consumption was not associated with risk of gastric and esophageal adenocarcinomas. Independent from quantity of consumption, more variety in vegetable and fruit consumption combined and in fruit consumption alone may decrease the risk of esophageal squamous cell carcinoma. However, residual confounding by lifestyle factors cannot be excluded.
Over the past fifty years, incidence and mortality of gastric cancer have decreased worldwide, especially in developed countries, while the incidence of esophageal adenocarcinomas is rising. Gastric cancer is however still the 3rd and esophageal cancer the 7th cancer-related cause of death world-wide.1
A positive association with gastric cancer risk has been observed for consumption of processed meat,2 nitrite,3 low socio-economic status,4 smoking5 and infection with Helicobacter Pylori (H. Pylori).4, 6 Inverse associations with gastric cancer risk have been reported for consumption of fruit and (raw) vegetables,4, 7, 8 dietary fibres9, 10 and physical activity.11
Esophageal adenocarcinoma risk has been positively associated with male gender,12 reflux symptoms,12 obesity,12 smoking12 and the consumption of processed meat.2 Inverse associations were found for the consumption of vegetables and fruit.8
The risk of esophageal squamous cell carcinoma has been positively associated with low socio-economic status, smoking and alcohol consumption,13 whereas an inverse association has been reported for dietary fibres,9 dietary antioxidants and the consumption of vegetables and fruit.14
In 2007, the evidence that diets rich in vegetables and fruits are inversely associated with gastric and esophageal cancer risk was considered probable by the World Cancer Research Fund/American Institute for Cancer Research (WCRF/AICR).15 Within the European Prospective Investigation into Cancer and Nutrition (EPIC) cohort, a suggestion of an inverse association between vegetables and the intestinal type of gastric cancer and adenocarcinoma of the esophagus has been found, which was based on 330 gastric adenocarcinomas and 65 adenocarcinomas of the esophagus.8 A case–control study nested within EPIC showed an inverse association between vitamin C levels and gastric cancer risk.16 Another study from EPIC observed a strong inverse association between quantity of vegetables and fruits and risk of upper aerodigestive squamous cell cancers, with about one quarter arising in the esophagus.17
To our knowledge, no cohort studies have investigated the relationship between the variety in vegetable and fruit consumption and gastric and esophageal cancer risk. Different vegetables and fruits contain different bioactive compounds suggesting that a mix of anti-carcinogenic substances, like antioxidants, may be responsible for the inverse associations observed.18 Only two case–control studies evaluated the influence of variability in the consumption of vegetables and fruit and both suggested an inverse association with both gastric and esophageal cancers.19, 20
The aim of this prospective study was to evaluate the association between the variety in vegetable and fruit consumption, independent from quantity of total vegetable and fruit consumption, and the risk of gastric and esophageal cancer among participants in the EPIC.
Material and Methods
EPIC is an ongoing multicenter cohort study designed to investigate the relation between diet, lifestyle, metabolic and environmental factors and the incidence of chronic diseases. The cohort consists of cohorts of men and women recruited in 23 centers in 10 European countries: Denmark, France, Germany, Greece, Italy, The Netherlands, Norway, Spain, Sweden and the United Kingdom. The populations and methods have been described in full elsewhere.21 In brief, the EPIC cohort consists of 521,468 subjects, mostly aged 35–70 years, recruited in the period 1992–2000 mostly from the general population, with some exceptions in specific geographical areas, i.e., in France, Norway, Utrecht (The Netherlands) and Naples (Italy) only women were recruited. All participants gave written informed consent. The study was approved by the local ethics committee in the participating countries and the Internal Review Board of the International Agency for Research on Cancer.
Diet and lifestyle questionnaires
At baseline, the usual diet reflecting the situation 12 months before enrolment was assessed by country-specific validated questionnaires designed to capture local dietary habits, and to provide high compliance.22 Although the design of the questionnaires was based on the same general format, there were differences between the questionnaires used in several countries. Extensive quantitative dietary questionnaires were used in three centers in northern Italy, The Netherlands and Germany (self-administered), as well as in Greece (interviewer-administered). In France (self-administered), Spain and Ragusa (Italy) interviewer-administered questionnaires similar to the dietary questionnaires, but structured by meals, were used. In order to increase compliance, the centers in Spain and Ragusa performed a face-to-face interview using a computerized program. Semi-quantitative food frequency questionnaires with the same standard portion assigned to all participants were used in Denmark, Norway, Naples (Italy) and Umeå (Sweden). In Malmö (Sweden), a nonquantitative food frequency questionnaire was combined with a 14-day record on hot meals. Details of the food items included in the selected vegetable and fruit subgroups used in the analysis have been reported in full by Agudo et al.23
Lifestyle questionnaires included questions on education, occupation, medical history, lifetime history of tobacco use and physical activity.21, 24
Diet Diversity Scores for vegetable and fruit consumption
Country-specific dietary questionnaires differed in the number of vegetables and fruits included. In order to improve between-country comparability of the scores, we decided to only select vegetable and fruit items that were included in four or more country-specific dietary questionnaires. This included the majority of products.
Based on the baseline dietary questionnaires, four different Diet Diversity Scores (DDS) were calculated, with DDSvegfr (range 0–40) counting the total number of individual vegetable and fruit products eaten at least once in two weeks. DDSvegsub (range 0–8) counted the total number of vegetable subgroups eaten at least once in two weeks. The eight subgroups of vegetables used included: leafy vegetables, fruiting vegetables, root vegetables, cabbages, mushrooms, grain and pod vegetables, onion and garlic and stalk vegetables.25 DDSveg (range 0–26) counted the total number of individual vegetable products eaten at least once in two weeks. DDSfruit (range 0–14) counted the total number of individual fruit products eaten at least once in two weeks. The consumption of fruit included fresh, dried and canned fruits but excluded nuts, seeds and olives (Appendix).
Follow-up and endpoints
Follow-up was based on population-based cancer registries in seven of the participating countries: Denmark, Italy, Netherlands, Spain, Sweden, United Kingdom and Norway. In France, Germany and Greece, a combination of methods was used, including health insurance records, cancer and pathology hospital registries and active follow-up. Mortality data were also collected from registries at the regional or national level.21 Cancer of the stomach and esophagus included cancers coded to C16 and C15 according to the 10th Revision of the International Statistical Classification of Diseases, Injuries and Causes of Death (ICD-10). For the current analysis, participants were followed from study entry until a first primary gastric or esophageal cancer, death, emigration or end of follow-up period.
Gastric and esophageal cancer cases classified as malignant lymphoma (morphology code 9590) and small cell carcinoma (morphology code 8041–8045) or coded as “benign” (5th digit of the morphology code is zero), “uncertain whether benign or malignant” (5th digit of the morphology code is one) or “carcinoma in situ” (5th digit of the morphology code is two) were censored at time of diagnosis. Only cancer cases coded as “malignant” (5th digit of the morphology code is three) were included as cases. We classified cases according to five major histological types, i.e., gastric cardia adenocarcinoma, noncardia gastric adenocarcinoma, adenocarcinoma of the gastric esophageal junction (GEJ), adenocarcinoma of the esophagus (8140) and squamous cell carcinoma of the esophagus (8070). GEJ was defined as the proximal end of the gastric folds (at macroscopy) or by the proximal limit of the gastric oxyntic mucosa (by histology).26 For the purpose of this study, tumors crossing the GEJ and those developing just below it were grouped in a broad group of the “cardia” carcinomas.26
Validation and confirmation of the diagnosis, classification of tumor site and morphology of tumor (according to ICDO2 Classification and Lauren classification) of the gastric and esophageal adenocarcinomas was performed by a panel of pathologists, including a representative from each participating country and a coordinator. The panel reviewed material provided by the centers (original histological slides and/or slides obtained from paraffin blocks, as well as original histopathological reports).26 Validation and confirmation of the esophageal squamous cell carcinomas was not performed.
Cox proportional hazard regression was used to analyze the association between the DDSs and gastric and esophageal cancer. Entry time was defined as age at recruitment and follow-up time as age of diagnosis (cases) or age at censoring (at-risk subjects). All analyses were stratified by age at baseline to control for length of follow-up, and by gender, and center to control for country effects such as follow-up procedures and questionnaire design. Cases diagnosed after censoring date were considered as noncases. Individuals with missing information on (non)dietary habits were excluded from the analysis (n = 814). Also individuals in the upper and lower 1% percentiles of “energy consumption/required energy” of the total EPIC cohort were excluded (n = 459), because of the likelihood of unrealistic reporting of consumption of food-products.
Due to possible differences in etiology, separate analyses were performed for 3 groups of cancers (i) noncardia adenocarcinomas of the stomach, (ii) adenocarcinomas of the esophagus, the GEJ and the gastric cardia and (iii) esophageal squamous cell carcinomas. In addition, gastric adenocarcinomas were subdivided in diffuse and intestinal adenocarcinomas.
The DDS were divided into tertiles (T1–3), according to the distribution observed in the full cohort, with the lowest tertile as reference category. Additionally, we analyzed the effect of DDS continuously per 2 products increment for DDSvegfr, DDSveg, DDSfruit and with increments of 1 subgroup for DDSvegsub.
Individuals with a more varied consumption of vegetables and fruits are more likely to eat more vegetables and fruits, therefore we controlled for the quantity of consumption of vegetables and/or fruit (g/day). Also smoking status (current, former and never), smoking duration (years), lifetime smoking dose (number of cigarettes/day), BMI (kg/m2), alcohol consumption (g/day) and combined red and processed meat consumption (g/day) were taken into account as confounders. In order to improve error correction, estimated energy intake was divided into energy from fat and from other macronutrients, because it is mostly the nonfat components of the diet that contribute to fruit and vegetable intake. All models included energy intake from fat and nonfat sources (kcal/day).27 Indicator values were used for missing values.
We derived probability values for a linear trend from regression models using a continuous variable with values equal to the tertile medians, to take the unequal distances of the tertiles into account.
The potential modifying effect of H. Pylori infection was analyzed in a subset of 1,260 participants, including 88 with noncardia adenocarcinoma, for whom H. Pylori status was known. This subgroup was reported in a previous EPIC study.28 We also tested the highest educational level as proxy for social economic status and physical activity but these variables had no influence on the models.
Analyses were performed on the full cohort, but also without the first two years of follow-up, including 89 cases of gastric and esophageal adenocarcinomas and 25 esophageal squamous cell carcinomas, to exclude potential pre-diagnostic changes in diet in patients with pre-clinical gastric cancer. The cohort was also split in a northern (Germany, The Netherlands, Sweden, Denmark, United Kingdom and Norway) and a southern (France, Italy, Spain and Greece) cohort, because of the differences in food patterns and incidence of H. Pylori infection. For the subgroup of gastric adenocarcinomas, we performed separate analyses for intestinal and diffuse gastric adenocarcinomas, because of differences in etiology. For all subgroups, we performed separate analyses for ever (former and current combined) and never smokers. Interactions (on the multiplicative scale) were tested using the interaction term of vegetable and fruit variety (in tertiles) with smoking (ever vs. never) and cohort (north vs. south). Finally, analyses were performed separately for males and females.
All analyses were performed using SAS version 9.1 (SAS Institute Inc., Cary, NC). A p < 0.05 was considered to be significant.
After a mean follow-up of 8.4 years, 475 gastric and esophageal adenocarcinomas and 98 esophageal squamous cell carcinomas were diagnosed. Of the patients with an adenocarcinoma, 180 were classified as noncardia adenocarcinomas and 185 as adenocarcinomas of the esophagus, GEJ or gastric cardia. In 110 patients, the location of the adenocarcinoma was unknown. Table 1 shows the frequency of adenocarcinomas and squamous cell carcinomas by country.
Table 1. Number of gastric and esophageal cancer cases per country. The EPIC Cohort Study
Selected characteristics across tertiles of variety in vegetable and fruit consumption (DDSvegfr) are shown in Table 2. In general, differences were most apparent for participants in the highest third of DDSvegfr, which were more often women, more often never smokers, more often H. Pylori negative and reported higher education. Additionally, they consumed more vegetables and fruit, less red and processed meat and were more often alcohol consumers.
Table 2. Baseline characteristics by tertiles of the Diet Diversity Score for intake of vegetables and fruits (for quantitative variables mean (SD) or median (range) and for qualitative variables (percentages). The EPIC Cohort Study
Table 3 shows the adjusted hazard ratios (HRs) and 95% confidence intervals (95% CI) separately for (i) noncardia adenocarcinomas, (ii) adenocarcinomas of the esophagus, GEJ and gastric cardia, and (iii) esophageal squamous cell carcinomas. None of the analyses showed a difference between males and females (data not shown).
Table 3. Adjusted1 hazard ratios and 95% confidence intervals (95% CIs) for gastric and esophageal cancer of the four different Diet Diversity Scores (DDS) for the full cohort. The EPIC cohort study
Results on noncardia adenocarcinoma risk suggested a positive association comparing the highest tertile of variety of consumption of vegetable and fruit products (DDSvegfr) with the lowest tertile (HR 1.82; 95% CI 1.02–3.25, p-trend = 0.04). However, the continuous variable did not show a significant association with risk (HR per 2 products increment 1.06; 95% CI 0.98–1.14). Excluding the first two years of follow-up attenuated somewhat the categorical association (T3 vs. T1 HR 1.71; 95% CI 0.89–3.28, p-trend = 0.11). Analyzing by H. Pylori status in a subset of 1,260 participants with known H. Pylori status, further weakened the results in the H. Pylori positive subgroup (DDSvegfr T1 vs. T3 HR 1.67; 95% CI 0.33–8.44, p-trend = 0.61). The H. Pylori negative subgroup did not contain a sufficient number of cases to perform accurate analysis. Stratified analysis by smoking status showed a positive association in ever smokers for the variety of consumption of vegetable and fruit products combined (DDSvegfr T1 vs. T3 HR 2.72; 95% CI 1.31–5.68). In addition, we found a positive association in never smokers for fruit consumption alone (DDSfruit T1 vs. T2 HR 2.08; 95% CI 1.06–4.07). However, the continuous variable did not show a significant association (HR per 2 products increment 1.11; 95% CI 1.00–1.23 and HR 1.07; 95% CI 0.87–1.33 respectively) (data not shown).
There was an apparent difference between diffuse and intestinal gastric adenocarcinomas with a nonsignificant inverse association between variety in vegetable and fruit consumption (DDSvegfr) and risk of intestinal gastric adenocarcinomas (T2 vs. T1 HR 0.56; 95% CI 0.36–0.87; T3 vs. T1 HR 0.72; 95% CI 0.39–1.35; p-trend = 0.26; data not shown) and a nonsignificant positive association of variety in vegetable and fruit consumption and risk of diffuse gastric adenocarcinomas (T2 vs. T1 HR 1.63; 95% CI 1.01–2.63; T3 vs. T1 HR 1.64; 95% CI 0.81–3.30; p-trend = 0.17; data not shown).
None of the DDSs were associated with risk of adenocarcinomas of the esophagus, GEJ and gastric cardia combined (Table 3). In the analysis stratified by smoking status, no associations between the DDS scores and risk were seen in ever and never smokers (data not shown).
Results on esophageal squamous cell carcinoma showed that each increase in consumption of two different products of vegetables and fruits combined statistically significantly decreased the risk of esophageal squamous cell carcinoma by 12% (continuous HR 0.88; 95% CI 0.79–0.97) (Table 3). Excluding the first two years hardly affected the results (continuous HR 0.89; 95% CI 0.79–1.00). In addition, increasing diversity in fruit consumption was statistically significantly and inversely associated with risk of esophageal squamous cell carcinoma both in the categorical (p-trend = 0.04) as in the continuous analysis with a 24% risk reduction for each increase in consumption of two different fruit products (continuous HR 0.76; 95%CI 0.62–0.94). Excluding the first two years hardly affected the strength of this association (continuous HR 0.79; 95% CI 0.62–1.02). When analyzing ever smokers separately for esophageal squamous cell carcinomas, we found a significant inverse association for a varied fruit consumption (continuous HR per 2 products increment 0.74; 95% CI 0.57–0.96, Table 4). In the subgroup of never smokers we did not find a significant association for a varied fruit consumption and esophageal squamous cell carcinoma risk (continuous HR per 2 products increment 1.03; 95% CI 0.67–1.61). However, this subgroup did not contain a sufficient number of cases (n = 21) to perform an accurate analysis.
Table 4. Adjusted1 hazard ratios and 95% confidence intervals (95% CIs) for esophageal squamous cell cancer of the four different Diet Diversity Scores (DDS) in ever and never smokers. The EPIC cohort study
We found no difference in associations between vegetable and fruit consumption and risks of gastric cancer between the northern and southern countries. In addition, the interaction was not statistically significant for noncardia adenocarcinoma risk (p = 0.75) and risk of adenocarcinoma of the esophagus, GEJ and gastric cardia (p = 0.92). For esophageal squamous cell carcinoma a somewhat stronger and statistically significant inverse association was only seen in the northern countries (Northern cohorts continuous HR 0.87; 95% CI 0.78–0.97 vs. Southern cohorts continuous HR 0.96; 95% CI 0.73–1.25). However, subgroups were small especially in southern countries. In addition, the test for interaction was not statistically significant (p = 0.88).
Overall, we found that, independent from quantity of consumption of total vegetable and fruit, more variety in vegetable and fruit consumption and fruit consumption alone were associated with a decreased risk of esophageal squamous cell carcinoma. This association was particularly seen among ever smokers. Variety in vegetable and/or fruit consumption was not associated with risk of gastric and esophageal adenocarcinomas. However, our results did show a nonsignificant inverse association between variety in vegetable and fruit consumption and risk of intestinal gastric adenocarcinomas.
In a previous EPIC analysis based on 330 gastric adenocarcinomas and 65 adenocarcinomas of the esophagus,8 no association was found between total vegetable or fruit consumption and gastric adenocarcinomas risk (vegetable HR 0.91; 95% CI 0.65–1.28 and fruit HR 1.04; 95%CI 0.91–1.20 per 100 g). However, a statistically nonsignificant inverse association between vegetable consumption and esophageal adenocarcinoma risk was observed (HR 0.72; 95% CI 0.32–1.64 per 100 g). Our current results using a larger number of cases suggest that there is however no association between the variety in vegetable and/or fruit consumption and gastric and esophageal adenocarcinoma risks.
DDS have previously been used to describe the variety within diets or food groups in relation to mortality and gastric and esophageal cancer risk. Jansen et al. (2004)29 looked specifically into diversity of vegetable and fruit consumption in relation to cancer risk and found decreased overall cancer risks among the individuals with a higher variety in vegetable consumption. However, an inverse relation between total cancer risk and variety in fruit consumption was only seen after exclusion of the first two years of follow-up.29 One case–control study on gastric cancer19 and one case–control study on esophageal squamous cell carcinoma20 found inverse associations for both cancers for a more varied diet, especially a more varied consumption of vegetables and fruit. Nevertheless, the authors not adjust specifically for the total number of fruit and vegetable servings consumed.
Our results suggest a positive association between variety in vegetable and fruit consumption and risk of gastric noncardia adenocarcinomas, especially in never smokers. This association weakened after excluding the first 2 years of follow-up and in participants positive for H. Pylori which may be explained by chance, especially because the continuous analysis does not support the positive association. For the subanalysis on H. Pylori the sample size did not contain a sufficient number of noncardia cancer cases negative for H. Pylori infection and therefore does not allow definitive conclusions to be made. In addition, H. Pylori infection did not modify the association between the quantity of vegetable and fruit consumption and risk of gastric and esophageal adenocarcinomas in previous studies.8 Large cohort studies with long-term follow-up are needed to definitely conclude whether H. Pylori infection affects the association between variety in vegetable and fruit consumption and risk of gastric noncardia adenocarcinomas. In addition, our results for never smokers were underpowered and might be due to chance and due to relatively small subgroups.
The statistically nonsignificant inverse association between variety of vegetable and fruit consumption and risk of intestinal gastric adenocarcinomas may be in line with the previous result found in the EPIC cohort (intestinal type n = 116, diffuse type n = 120) on quantity of vegetable and fruit intake and risk of intestinal gastric adenocarcinomas (HR 0.66; 95% CI 0.35–1.22 per 100 g increase).8 In addition, our results are comparable to a case–control study by Lunet et al.30 Nevertheless, evidence from cohort studies is lacking and the metabolic pathway and features of this histological type are still unknown. It has been suggested that the intestinal type has a predominately environmental etiology, resulting in the possibility of a more prominent role of vegetable and fruit consumption in this group, whereas the diffuse type is thought to be more genetically determined. H. Pylori infection, however, seems to have similar effects on the risk of both intestinal and diffuse gastric cancer risk.30
Our observation that more variety in vegetable and/or fruit consumption may decrease the risk of esophageal squamous cell carcinomas may well be in line with the strong inverse association observed in EPIC between quantity of vegetables and fruits and risk of upper aerodigestive cancers of which about one quarter were located in the esophagus.17 However, the effect of vegetable and fruit combined may be in part attributed to the effect of fruit alone. A recent EPIC study on lung cancer also found an inverse association between variety in vegetable and fruit consumption and risk of squamous cell carcinomas of the lung.31 Compared to adenocarcinomas, squamous cell carcinomas appear to be more affected by smoking. The association between vegetable and fruit consumption and risk of squamous cell carcinoma may therefore differ between ever and never smokers, because antioxidants in vegetables and fruit may protect against the damage caused by free radicals in cigarette smoke. On the other hand, oxidative stress due to smoking may reach such high levels that antioxidants are not sufficiently capable to protect against oxidative stress.32 In our study, we found a significant inverse association for a varied fruit consumption and esophageal squamous cell carcinoma risk in ever smokers (continuous HR per 2 products increment 0.74; 95% CI 0.57–0.96). In the subgroup of never smokers, we did not find a significant association for a varied fruit consumption and esophageal squamous cell carcinoma risk (continuous HR per 2 products increment 1.03; 95% CI 0.67–1.61). However, the subgroup of never smokers did not contain sufficient cancer cases to yield stable risk estimates. Yet absolute risks for never smokers (T1 0.01%, T2 0.001%, T3 0.01%) seem to be lower compared to ever smokers (T1 0.06%, T2 0.03%, T3 0.01%). It should be noted however that the association in ever smokers may also be caused by residual confounding. Therefore, our suggested evidence of an inverse association between variety in vegetables and/or fruits and the risk of squamous cell cancers of the esophagus in particular in ever smokers does require further research.
Analyses stratified by geographic region showed only minor differences in risk estimates between the northern and the southern cohorts with a suggestion of an inverse association with risk of esophageal squamous cell carcinomas only in the northern cohorts. This difference remained even when separate tertiles were used for northern and southern countries (data not shown). Such findings may well be explained by chance since southern cohorts were having substantially fewer cases.
The main limitation of our study is the small size of the subgroups of cancer types with concomitant loss of power to detect the usually weak diet–cancer associations. The positive association in the categorical analyses between a more varied vegetable and fruit consumption and risk of noncardia adenocarcinomas may be the result of the multiple tests that we performed. Another limitation is that dietary questionnaires may not be adequate enough to validly rank participants according to variety in vegetable and fruit intake. A recent EPIC study concluded that high blood levels of vitamin C and carotenoids were inversely associated with gastric cancer risk, whereas dietary intakes of vitamin C,16 β-carotene and vitamin E33 were not related to risk, suggesting that blood levels of vitamins and antioxidants may lead to more accurate results than questionnaires.34 In addition, blood levels of vitamins measure the actual amount of vitamins absorbed, as absorption may be influenced by other factors, like H. pylori infection.
EPIC is the largest prospective investigation on diet, lifestyle and cancer so far involving 10 European countries.21 Because there is a wide range in dietary and lifestyle habits within EPIC, the study is well suited to examine diet diversity and gastric cancer risk. As the dietary questionnaires used in this project slightly differed between the different EPIC centers, we calculated the DDSs based on fruit and vegetable products included in four or more dietary questionnaires. This makes the DDSs better comparable between countries. The more vegetable and fruit products are included in a questionnaire the more likely individuals may report eating them. And thus, we may have over- or under-estimated the variety in consumption in the few centers that used questionnaires with longer lists of products; however, this concerns a minority of products only. Conversely, we may have underestimated diversity in centers with only few items included in dietary questionnaires such as Norway. Finally, we were able to adjust for quantity of consumption of vegetables and fruits. It should be kept in mind that the individuals with a more varied consumption of vegetable and fruit are, in general, the individuals consuming more vegetables and fruit and these individuals probably share also other lifestyle factors that may be linked to the cancers of interest in this study and, consequently, our findings may be explained by residual confounding.
In conclusion, irrespective of the quantity of total vegetable and fruit consumption, more variety in vegetable and fruit consumption and in fruit consumption alone is inversely associated with the risk of esophageal squamous cell carcinoma; the inverse association for fruits was particularly seen among ever smokers. Variety in vegetables and/or fruits was not associated with risk of gastric and esophageal adenocarcinoma. In participants consuming a diverse diet in vegetables and fruits, a reduced risk of intestinal adenocarcinomas was observed but this association was not statistically significant.
Table A1. Food items included in the Diet Diversity Scores