An update of the WCRF/AICR systematic literature review on esophageal and gastric cancers and citrus fruits intake

Purpose The 2007 World Cancer Research Fund/American Institute for Cancer Research expert report concluded that foods containing vitamin C probably protect against esophageal cancer and fruits probably protect against gastric cancer. Most of the previous evidence was from case–control studies, which may be affected by recall and selection biases. More recently, several cohort studies have examined these associations. We conducted a systematic literature review of prospective studies on citrus fruits intake and risk of esophageal and gastric cancers. Methods PubMed was searched for studies published until 1 March 2016. We calculated summary relative risks and 95 % confidence intervals (95 % CI) using random-effects models. Results With each 100 g/day increase of citrus fruits intake, a marginally significant decreased risk of esophageal cancer was observed (summary RR 0.86, 95 % CI 0.74–1.00, 1,057 cases, six studies). The associations were similar for squamous cell carcinoma (RR 0.87, 95 % CI 0.69–1.08, three studies) and esophageal adenocarcinoma (RR 0.93, 95 % CI 0.78–1.11, three studies). For gastric cancer, the nonsignificant inverse association was observed for gastric cardia cancer (RR 0.75, 95 % CI 0.55–1.01, three studies), but not for gastric non-cardia cancer (RR 1.02, 95 % CI 0.90–1.16, four studies). Consistent summary inverse associations were observed when comparing the highest with lowest intake, with statistically significant associations for esophageal (RR 0.77, 95 % CI 0.64–0.91, seven studies) and gastric cardia cancers (RR 0.62, 95 % CI 0.39–0.99, three studies). Conclusions Citrus fruits may decrease the risk of esophageal and gastric cardia cancers, but further studies are needed. Electronic supplementary material The online version of this article (doi:10.1007/s10552-016-0755-0) contains supplementary material, which is available to authorized users.


Introduction
Esophageal and gastric cancers are the eight and the fifth most common cancers worldwide, respectively [1]. Esophageal cancer accounted for 456,000 new cancer cases in 2012 [1]-it is the sixth most common cause of cancer mortality, with 400,000 deaths in 2012 reflecting its poor prognosis, and has a 5-year survival rate of 15-25 % [2]. Squamous cell carcinoma (SCC) is the predominant histological type of esophageal cancer worldwide but in USA, UK, Australia, and some Western European countries, and the incidence of esophageal adenocarcinomas now exceeds that of SCC [3,4]. Gastric cancer is more common in lowand middle-income countries, and although incidence rates are declining in most parts of the world, almost one million new cases occurred worldwide in 2012 [1]. The incidence of cancers of the gastric cardia has remained stable or Electronic supplementary material The online version of this article (doi:10.1007/s10552-016-0755-0) contains supplementary material, which is available to authorized users. increased at least in Western countries. Gastric cancer is usually diagnosed at advanced stages. This makes the disease the third leading cause of cancer death globally, with an estimated 723,000 deaths in 2012 [1].
Tobacco use is a risk factor for esophageal and gastric cancers. Alcohol and tobacco use are the main risk factors for esophageal SCC [5]. Due to close anatomical proximity and similar etiology, esophageal adenocarcinomas and cancers of the gastric cardia have other risk factors in common, including obesity and gastro-esophageal reflux disease [5,6]. Helicobacter Pylori infection is the major risk factor for non-cardia gastric cancer. Approximately, 80 % of non-cardia gastric cancers are attributable to Helicobacter Pylori infection. Despite the possibility of preventing non-cardia gastric cancer by treating H. Pylori infection, there are concerns with possible adverse consequences of the antibiotic treatment, such as development of antibiotic resistance and alterations of the intestinal microbiota [7]. There is no effective screening for early detection of these cancers.
Diet may also play a role on the development of esophageal and gastric cancers. In 2007, the World Cancer Research Fund/American Institute for Cancer Research (WCRF/AICR) Second Expert Report concluded that there was evidence that high total intake of salt probably increases the risk of gastric cancer, and that vegetables and fruits intake probably protects against esophageal and gastric cancers [8]. With respect to fruit intake, recent meta-analyses of cohort studies reported significant inverse associations with gastric cancer [9] and esophageal SCC [10] but not with adenocarcinomas of the esophagus [11].
Citrus fruits are rich in vitamin C, and foods containing vitamin C were judged probably to protect against esophageal cancer in the WCRF/AICR Second Expert Report [8]. Much of the previous evidence on citrus fruits was based on case-control studies. More recently, a publication from an integrated network of case-control studies [12], conducted in Italy and Switzerland, reported a significantly inverse association between citrus fruits intake and risk of esophageal cancer.
A recent meta-analysis of cohort studies reported nonsignificant inverse association between citrus fruits intake and the risk of gastric cancer for the comparison of the highest versus the lowest intakes [9]. However, there is no recent meta-analysis of cohort studies on citrus fruits intake and risk of esophageal cancer or subtypes of esophageal and gastric cancers. As part of the WCRF/AICR Continuous Update Project (CUP) [13], we conducted a systematic literature review and meta-analysis of cohort studies to investigate the association between citrus fruits intake and the risk of esophageal cancer, adenocarcinomas and squamous cell carcinomas, and total gastric, cardia, and non-cardia gastric cancers.

Search strategy
All cohort studies identified in the systematic literature review for the WCRF/AICR Second Expert Report [8] were indexed in PubMed. Therefore, we updated the search using the same search strategy in PubMed for studies published until 1st March 2016. Searches for esophageal and gastric cancers were carried out separately following protocols that can be accessed at http://www.wcrf.org/int/ research-we-fund/continuous-update-project-cup. In addition, reference lists of relevant reviews identified in the search and of the studies included in the meta-analysis were screened for any further publications.

Study selection
The following inclusion criteria were applied for studies included in this meta-analysis: (a) cohort, nested casecontrol or case-cohort design; (b) reported estimates of the relative risk (hazard ratio, odds ratio, or risk ratio) with confidence intervals (CI); (c) reported quantifiable measure of citrus fruits intake. If several publications using the same study population were identified, the one with the largest number of cases was selected.

Data extraction
The following data were extracted from each study: the first author's last name, publication year, country in which the study was conducted, study name, follow-up period, sample size, sex, age, number of cases, dietary assessment method (type, number of food items, validation), exposure, frequency or amount of intake, associated RR and corresponding 95 % CI, and adjustment variables. The search and data extraction for the systematic literature reviews of esophageal cancer and gastric cancer prior to January 2006 was conducted by the WCRF/AICR Second Expert Report teams at the Pennsylvania State University and the University of Leeds, respectively [8]. The search and data extraction from January 2006 to 1 March 2016 was conducted by the CUP team at Imperial College London. All extracted data are stored in the CUP database [13].

Statistical analyses
We conducted dose-response meta-analyses and summarized the associations for the highest compared to the lowest citrus fruits intake reported in the studies using random-effects models [14].
When not provided in the publications, the linear doseresponse trends were derived from the natural logs of the RRs and CIs across categories of citrus fruits intake, using the method by Greenland and Longnecker [15]. For this method, the distribution of person-years, cases, RRs, and CIs for at least three categories is required. When not available, person-years per quantile were estimated by dividing total person-years by the number of quantiles. Means or medians of intake were assigned to each category, and when a study reported only the range of intake per category, the midpoint was estimated. For open-ended uppermost or lowermost intake categories, we computed the midpoint by assigning the width to match the nearest category. When intake was reported per unit of energy intake [16,17], we estimated the absolute intake per quantile using the mean energy intake of the whole study population provided in the paper. When intake was reported in times or servings per day or per week, we used a standard portion size of 80 g to convert frequency to grams (http://www.wcrf.org/sites/default/files/protocol_ oesophageal_cancer.pdf). The dose-response was expressed for an increment of 100 g/day of citrus fruits. We used the multivariable adjusted RR from each study. The EPIC study [18,19] reported calibrated relative risk estimates to account for possible diet measurement error, and we used these calibrated risk estimates for the linear dose-response meta-analysis.
We first estimated summary RR for all esophageal and gastric cancers, respectively. For these analyses, the RRs for men and women were combined using fixed-effect meta-analysis before pooling. When RRs were reported by cancer subtypes only, we estimated the combined RR of gastric cardia and non-cardia or esophageal adenocarcinoma and squamous cell carcinoma using Hamling's method [20]. The meta-analyses were also conducted by sex and cancer type, for which we combined the RRs of esophageal adenocarcinoma and gastric cardia cancers using fixed-effect models. The extent of heterogeneity in the meta-analyses was assessed using Cochran Q test and I 2 statistics, with low and high heterogeneity extent indicated by I 2 values below 30 % or substantially higher than 50 % [21].
Subgroup analyses were conducted to assess possible sources of heterogeneity, as well as study quality. The predefined factors to explore were sex, outcome type, geographic location, duration of follow-up, number of cases, publication year, and adjustment for confounders including smoking, alcohol intake and adiposity (as measured by BMI), when the number of studies allowed it.
Publication bias was assessed with Egger's test [22] and visually by using funnel plot. All analyses were conducted using Stata version 12 software (Stata Corp, College Station, TX).
Main study characteristics are shown in Table 1. Citrus fruits intake was assessed using food frequency questionnaires. The definition of citrus fruits exposure varied slightly across the studies; in three studies, it included citrus fruits juice [26,27,29] (Table 1).
Gastric or esophageal cancers were primary outcomes in all but two studies [24,25] that reported on multiple cancer sites.
Five studies were conducted in Asia [24,25,27,28], two in Europe [18,19,26], and two in North America [16,17,29] (Table 1). All studies were included men and women apart from Yamaji et al., 2008 [27] which was only included men (Table 1). A summary of the results of metaanalyses by cancer type is presented in Table 2.
Only two studies in men [24,27], one incidence and one on mortality from esophageal cancer and one study on esophageal cancer mortality in women [24] were available. There is not enough data to examine the association of citrus fruits and esophageal cancer risk by sex (Table 3).
In subgroup analysis (all esophageal cancers), no differences emerged across study characteristics, including adjustment factors (Table 3). There is some suggestion that more adjusted studies tend to report stronger associations, but the number of studies is low. A positive not significant association was observed in the only study [24] that did not adjust for tobacco, smoking, and alcohol intake in which the outcome was mortality for esophageal cancer. When this study was omitted from the analysis, the summary RR for an increase of 100 g/day of citrus fruits intake was 0.85 (95 % CI 0.73-0.99) with no heterogeneity.

Highest versus lowest analysis
No. of studies   [24] b Minimally adjusted study for age and study area [24] and another study which did not include alcohol intake in the final model but tested that it did not confound the association [29]  In subgroup analyses by cancer type, inverse association was observed for cancers of the gastric cardia, but not for non-cardia gastric cancers. Three studies [17,19,26] investigated the association between citrus fruits intake and gastric cardia cancer risk with a total of 555 cases among 972,149 participants. The summary RR for 100 g/day increment was 0.75 (95 % CI 0.55-1.01), with moderate [21] heterogeneity (I 2 = 55 %, P heterogeneity = 0.11) (Fig. 2a), and it was 0.62 (95 % CI 0.39-0.99) comparing the highest with lowest intake, with high heterogeneity (I 2 = 67 %, P heterogeneity = 0.05) (Fig. 2b).
It was not possible to formally explore the source of heterogeneity in the analyses on cardia gastric cancer. Visual inspection of the forest plot shows that heterogeneity is driven by the American NIH-AARP study [17] that reported no association of citrus fruits with cardia gastric cancer. The reasons for the different results are unclear. The NIH-AARP study [17] categorized intake by servings/1,000 kcal, whereas the two other studies [19,26] reported in continuous increments in g/day.
Summary risk estimates observed in subgroup analyses for all gastric cancers were mostly similar to that in the overall analysis, with exceptions in some subgroups where a positive association was observed. Estimates of risk were below 1 in studies adjusted for smoking and alcohol and BMI but not in the unadjusted studies (Table 3). Significant associations were observed in subgroup analyses for all gastric cancers among European studies [19,26] and studies with 500-\1,000 cases [19,25,26]. Inverse not significant associations were observed in men but not in women (2 studies) [24,29]. There was no evidence of small study effects such as publication bias (p = 0.25).

Interaction with smoking
One study reported on the interaction of smoking status and citrus fruits intake in relation to esophageal or gastric cancers. In the EPIC study [19], the inverse association of citrus fruits for gastric cancer was restricted to current smokers and not observed in never or former smokers (p for interaction =0.07). Other studies in the review explored the interaction of smoking and intake of total fruits and vegetables, or fruits. In general, no significant interactions with smoking were observed. In the NIH-AARP study, the risk estimates of adenocarcinoma and SCC for total fruits and vegetable intakes appeared similar in smokers, non-smokers, and current smokers [17]. In the study in Japanese men, esophageal SCC risk was inversely associated with total fruits or vegetables intake in never, current and former smokers [27]. In the Netherlands Cohort Study, slightly greater inverse associations of fruit intake with SCC and adenocarcinomas of esophagus and gastric cardia cancer were reported in smokers that in never smokers, but the interaction was not significant (p for interaction = 0.25; 0.15; and 0.49, respectively) [26]. In a Chinese study in men and women, a significant reduction in risk of distal gastric cancer from increased fruit intake was significant among ever smokers and inverse but nonsignificant in never smokers, but the interaction by smoking status was not statistically significant (p for interaction = 0.27) [28].

Discussion
In these meta-analyses of cohort studies, citrus fruits intake was marginally associated with reduced risks of esophageal and gastric cardia cancers. No association with non-cardia gastric cancers was observed. Similar results were observed for adenocarcinomas and SCC of esophagus.
Citrus fruits are rich in vitamin C that could influence cancer risk by scavenging reactive oxygen species, protecting mucosal tissues from the damaging effects of oxidative stress, and inhibiting nitrosamine formation in the stomach [30]. The results of this meta-analysis are consistent with the inverse association of prediagnostic plasma vitamin C concentration and risk of gastric cardia cancer (215 cases) observed in the EPIC study [31] and in a study in a high-risk Chinese population (467 cases) [32]. In the EPIC study, the associations were more pronounced for gastric cardia than non-cardia cancer, although the associations were not statistically significant when stratified by subtype. Further evidence is provided by the Shandong Intervention Trial of vitamin supplementation (vitamin C, E and selenium), in which supplemented individuals had a lower risk of esophageal and gastric cancers [33] and in a meta-analysis of 20 randomized controlled trials of antioxidant supplementation (vitamins A, C, E, and selenium) inverse but not significant lower risk of gastrointestinal cancers was observed [34]. In the NIH-AARP, use of vitamin C supplements was associated with reduced risk of gastric non-cardia adenocarcinomas, but no association was observed with multivitamin supplements use that usually contains vitamin C [35]. Finally, in recent metaanalyses, total fruit intake was associated with significantly lower risk of gastric cancer [9] and esophageal squamous cell carcinoma [10].
In addition to high vitamin C content, citrus fruits contain a wide range of bioactive compounds such as citrus flavonoids, carotenoids, and limonoids. Experimental studies have demonstrated that these bioactive components may protect DNA, regulate cell growth, and induce apoptosis [36][37][38].
The main limitation of this meta-analysis is the small number and limited power of published studies on citrus fruits intake, esophageal and gastric cancer risks, and the unexplained heterogeneity of the inverse association of citrus fruits intake and gastric cancer cardia in the three studies identified [17,19,26].
The observed inverse associations could be due to residual confounding by smoking. In the EPIC study [19], the inverse association of citrus fruits for gastric cancer was restricted to current smokers and not observed in never or former smokers (p for interaction =0.07). However, other studies included in the meta-analysis [16,17,19,26] reported no evidence of interaction of effect modification by smoking status. In the NIH-AARP, the association between fruit and vegetable intake with ESCC [16] was similar in the limited number of non-drinkers and nonsmokers; in a study on gastric cancer [17], there was no evidence of effect modification by cigarette smoking status or alcohol drinking; in the NLCS study, the risk estimates for total fruit intake and risk of all types of gastric and esophageal cancers were further below 1 in current smokers compared to never and former smokers, but the interaction was not significant (p for interaction [0.15) [26]. On the other hand, smokers tend to eat less fruits and vegetables [39,40], have lower concentration of serum antioxidants [41], and may benefit more from higher citrus fruits intake [27].
Measurement error of diet may have attenuated the risk estimates. Only the EPIC cohort corrected for dietary measurement error [18,19]. When non-calibrated risk estimates from the EPIC cohort were used in the sensitivity analysis, the association became significant for gastric cardia cancer (RR 0.75; 95 % CI 0.57-0.99), and the risk estimates did not change for esophageal cancer and remained similar for all gastric cancers (RR 0.96; 95 % CI 0.88-1.05) and gastric non-cardia cancer (RR 1.04; 95 % CI 0.94-1.16). Strengths of this meta-analysis include the prospective design of the included studies, which are less prone to bias than other observational studies, detailed dose-response and categorical meta-analyses, and the increased statistical power to detect modest but statistically significant inverse associations.

Conclusions
In conclusion, there is evidence from cohort studies that citrus fruits may decrease the risk of esophageal and cardia gastric cancers, but the data are still limited.