Associations between dietary folate intake and risks of esophageal, gastric and pancreatic cancers: an overall and dose-response meta-analysis

There are still some controversies on the association between dietary folate intake and the risk of upper gastrointestinal cancers including esophageal, gastric and pancreatic cancers. Hence, a comprehensive meta-analysis on all available literatures was performed to clarify the relationship between dietary folate intake and risks of upper gastrointestinal cancers. An electric search was performed up to December 12th, 2016 within the PubMed, MEDLINE AND EMBASE databases. Ultimately, a total of 46 studies which evaluated the association between folate intake and risks of upper gastrointestinal cancers were included. According to the data from included studies, the pooled results showed significant association between folate intake and esophageal (OR = 0.545, 95%CI = 0.432-0.658), gastric (OR=0.762, 95%CI=0.648-0.876) and pancreatic (OR=0.731, 95%CI=0.555-0.907) cancers. Linearity dose-response analysis indicated that with 100μg/day increment in dietary folate intake, the risk of esophageal, gastric and pancreatic cancers would decrease by 9%, 1.5% and 6%, respectively. These findings indicated that higher level of dietary folate intake could help for preventing upper gastrointestinal cancers including esophageal, gastric and pancreatic cancers.


INTRODUCTION
Folate, also named vitamin B9, is a naturally occurring nutrient and is found in many foods including fruits, vegetables legumes, cereals, and liver. Human can't produce folate de novo and need to uptake folate from dietary intake. Evidences implicated deficient folate is related to increased risks of many cancers [1].
Folate plays an important role in the process of DNA synthesis, repair, and methylation, and was hypothesized to decrease risks of gastrointestinal cancers. The main carcinogenesis mechanisms of folate are inducing DNA strand breaks by causing uracil mis-incorporation into DNA and changing levels of DNA methylation [2]. These aberrant changes may result in potential alterations of critical proto-oncogene and tumor suppressor gene expressions [3]. Animal experiments referring mice and dogs suggested that high levels of folate intake affected DNA methylation and eventually decreased the risks of gastric cancer [4,5]. In addition, the polymorphisms of genes in folate metabolizing pathway may modulate the susceptibility of several cancers.

Meta-Analysis
with the increased risks of prostate [6] and breast [7] cancers, but decreased the risks of colorectal [8] and cervical [9] cancers. Two previous meta-analysis have estimated the associations of folate intakes and risks of esophageal, gastric and pancreatic carcinomas and indicated that increased folate intakes were associated with decreased risks of esophageal and pancreatic cancers [10,11]. However, the results of these studies about the relationship between folate intake and gastric cancer risk remained inconsistent. Larsson et al. indicated that increased folate intake were associated with decreased risks of cardia and non-cardia gastric cancers [11]. Basing on more studies, another systematic review showed no relationship between dietary folate intake and risks of gastric cancers [10]. Therefore, to clarify the associations between folate intake and upper gastrointestinal cancers and evaluate the dose-response relationship between them, we performed an overall meta-analysis based on current observational studies.
To assess the relationship between the risk of esophageal cancer and dietary folate intake, total 19 studies including 2036 patients and 7086 controls were collected. The forest plot is shown in Figure 2A. Significant heterogeneity (p<0.001, I 2 = 73.7%) between these studies suggested that a random effect model was selected. The pooled results showed that dietary folate intake comparing highest levels vs. lowest levels was associated with the decreased risk of esophageal cancer (odds ratio (OR) = 0.545, 95% confidence interval (CI) = 0.432-0.658, Table 2). Table 2 showed the results of specific subgroup analysis based on study designs, countries, histological type and folate intake measurement. All these results were similar in subgroup analysis suggested that folate intake were comprehensive associated with reduced risk of esophageal cancer.
As shown in Figure 3A, the linearity test of doseresponse analysis suggested that with increased 100 μg/ day folate intake from diet, the risk of esophageal cancer decreased 9% degree (OR=0.91, 95%CI=0.88-0.94). The non-linearity test (p<0.001) indicated that the lowest risk of esophageal cancer was at the dose of 405 μg/day (OR=0.69, 95%CI=0.57-0.83). After the dose of folate intake > 405 μg/day, the risk of esophageal cancer would increase after the fall.
As shown in Figure 2B, 5 cohort studies and 16 casecontrol studies were collected to analyze the association between dietary folate intake and risk of gastric cancer. The comprehensive pooled relative risk (RR) indicated a significant association between increased folate intake and decreased risk of gastric cancer (OR=0.762, 95%CI=0.648-0.876, Table 2). There was a significant heterogeneity (p<0.001, I 2 =67.6%) which suggested a further subgroup analysis.   As shown in Figure 3B, non-linearity (p=0.20) dose-response analysis indicated no relationship between folate intake from diet and risk of gastric cancer. However, a linearity relationship (p=0.03) was found and suggested that 1.5% decrease of gastric cancer for each 100 μg/day increase of dietary folate intake (OR=0.985, 95%CI=0.972-0.998).
As shown in Figure 2C, 7 cohort studies and 5 casecontrol studies were collected to analyze the association between dietary folate intake and risk of pancreatic cancer. The comprehensive pooled RR indicated a significant association between increased folate intake and decreased risk of pancreatic cancer (OR=0.731, 95%CI=0.555-0.907, Table 2). There was a significant heterogeneity (p<0.001, I 2 =69.0%) which suggested a further subgroup analysis. As shown in Figure 3C, the linearity test of doseresponse analysis suggested that with increased 100 μg/

Sensitivity analysis and publication bias
One included study of this meta-analysis was omitted each time to evaluate the stability of pooled results. The results remained similar when any result was removed from the pooled results in this meta-analysis. Begg's test and Egger's test were used to evaluate the publication bias, the results were summarized in Table  2. There were significant publication biases in the results which evaluate the associations between folate intake and esophageal cancer (Egger'

DISCUSSION
Folate is a water-soluble B vitamin and is found in many foods including fruits, vegetables legumes, cereals, and liver. Human can't produce folate de novo and need to uptake folate from dietary intake [1,59]. Folate plays an important role in the process of DNA synthesis, repair, and methylation, and was hypothesized to decrease risks of gastrointestinal cancers. Two main mechanisms of folate deficiency leads to carcinogenesis: (1) by leading complete convention of dUMP to dTMP, which makes mis-incorporation of uracil into DNA and induces breaks and mutations of chromosome; and/or (2) inducing alternations in expression of critical proto-oncogenes and tumor suppressor genes by causing aberrant methylated  level of DNA [2,3]. In addition, the polymorphisms of 5,10-methylenetetrahydrofolate reductase, a critical junction protein in folate metabolizing pathway by leading folate metabolites to DNA methylation pathway and away from the DNA synthesis pathway, can regulate the susceptibilities of several cancers [60][61][62].
Our meta-analysis found that increased folate intake was associated with reduced risks of upper gastrointestinal cancers including esophageal, gastric and pancreatic cancers. The dose-response further certified their relationship. Subgroup analysis indicated that the comprehensive inverse associations between dietary folate intake and esophageal cancer. Our data suggested different relationships between dietary folate intake and cancer risks in country, study design, disease type, measurement and sex subgroup analysis of gastric and pancreatic cancers.
The results of this meta-analysis showed that increased dietary folate intake significant decreased risk of esophageal cancer. These results are similar to previous study [10,11,63]. In the subgroup analysis based on country, histological type, study design and dietary measurement, our results suggested an inverse association between dietary folate intake and risks of esophageal cancer in all subgroups. Interesting, we observed a higher OR which suggested a weaker link between folate intake and esophageal cancer in supplement and diet subgroup than in diet subgroup. These results suggested an extra folate supplement is not needed in diet for preventing esophageal cancer. The results of dose-response analysis also indicated that with the folate intake > 450 μg/day, the risk of esophageal cancer would increase weakly comparing with the lowest OR, which suggested that a redundant and supplementary folate is not necessary. Zhang et al. found that the risk ration of breast cancer decreased when the dose of folate was low. However, with the folate dose increasing, a positive association was found between folate intake and breast cancer risk [7].
Different from previous studies [10,11], our results showed a significant association between increased dietary folate intake and reduced gastric cancer risk. Although non-linearity model of dose-response analysis suggested no statistically significant association between folate intake and risk of gastric cancer, linearity model indicated a different result (p=0.03) which certificated our comprehensive pooled OR. Meta-analysis of genetic polymorphisms demonstrated that folate deficiency was associated with increased risk of gastric cancer [11,64,65]. Folate supplement can reverse methylation deficiency, stop global hypomethylation and prevent gastric carcinogenesis in hypergastrinemic transgenic mice [5]. Subgroup analysis indicated an inverse association between dietary folate intake and gastric cancer risk in case-control studies, but no association in cohort studies. A possible reason is that only 5 cohort studies were included in this analysis. Small number of studies and effects of multiple factors may affect recall bias and selection bias and restrict the precision of the last results. Similar to previous studies [11], our data showed a significant inverse association between folate intake and GCA or non-GCA, and a weak inverse link between folate intake and other gastric cancer. These results suggested that dietary folate intake plays different roles in different gastric cancers. In the subgroup analysis based on country, we observed an inverse association between folate intake and gastric cancer only in USA, but not in other countries. In addition, in the subgroup analysis based on measurement, our results showed an inverse association between folate intake coming from diet and risk of gastric cancer. However, no association between folate intake coming from diet and supplement and risk of gastric cancer was found. These results also suggested that an extra folate supplement is not needed in diet for preventing gastric cancer. And the excessive intake of folate may be a risk for gastric cancer since the highest values of 95%CI > 1.00. Different from previous estimate, serum evaluating suggested an increased risk of gastric cancer with high serum concentration. One possible explanation is that since the number of included studies about serum detection of folate and gastric cancer risk is too small, which provide insufficient statistical power to evaluate the risk. Animal experiments suggested a dual role of folate in cancer carcinogenesis: prevention or promotion, depending on the stage of cell transformation at the time of intervention and the dose of folate supplement [66,67]. Significant decreased risks of gastric cancers were observed both in men and women with folate intake increased.

Overall and subgroup analysis
Results of previous meta-analysis about folate intake and pancreatic cancer risk were inconsistent. Bao et al. found folate intake was not associated with overall risk of pancreatic cancer using only prospective cohort studies [68]. However, other studies considered increased folate intake was associated with decreased pancreatic cancer risk [11,69]. Our comprehensive meta-analysis found an inverse association between dietary folate intake and pancreatic cancer risk. Dose-response analysis indicated that a 100 μg/day increment in dietary folate intake was associated with a 6% risk decreasing for pancreatic cancer. Results of subgroup analysis based on country showed an inverse association between folate intake and pancreatic cancer risk in European. However, this association was not found in American and other countries. These results suggested that geographic variation or dietary habit may play an important role in the association. Subgroup analysis by sex indicated that women had higher pancreatic risk with low folate intake when compared with men. Similar to esophageal and gastric cancers, our data showed that an extra folate supplement is not needed in diet for preventing pancreatic cancer.
There are several limitations to current meta-analysis. First, the included studies about esophageal, gastric and pancreatic cancer have few cohort studies which may make influence on the actual result. Since, dose-response analysis didn't separate cohort and case-control studies. Second, subgroup analysis based on measurement only included diet, diet and supplement and serum. Total folate intake and other folate intake measurements were not evaluated for lack of related studies. Third, significant heterogeneity were detected between the studies included in quantitative synthesis. Through further subgroup analysis, we still can't find all the origin of heterogeneity. Forth, this meta-analysis used pooled results for lacking of individual data, which prevents us from finishing a more precise analysis. Last, some subgroup analysis which included small number of studies may not represent objective and exact results. Hence, our results should be treated as exploratory and with caution.
In conclusion, results of current meta-analysis indicated that higher level of dietary folate intake could help for preventing upper gastrointestinal cancers including esophageal, gastric and pancreatic cancers. Dose-response analysis indicated that with 100μg/day increment in dietary folate intake, the risk of esophageal, gastric and pancreatic cancers would decrease by 9%, 1.5% and 6%, respectively. In addition, our analysis indicated that more well-designed studies about associations between esophageal, gastric and pancreatic cancers and folate intake are necessary for further accurately evaluating subgroup analysis based on country, measurement, histological type and sex.

Literature search
A systematically search was performed up to May 2 th , 2017 by two reviewers (H. Z. and Y. Z.) within Pubmed, MEDLINE AND EMBASE, using the terms "folate, folic acid or vitamin B9", "esophageal, oesophagus, gastric, stomach, or pancreatic" and "cancer, neoplasm or carcinoma". In addition, we reviewed the reference lists from original reports and manually selected for other available publications. No language restrictions were imposed in the searching process.

Study selection
The studies were included with the following inclusion criteria: (i) the experimental design was a case-control or cohort study; (ii) studies reported the associations of esophageal, gastric, or pancreatic cancer risk with dietary folate intake from diet, dietary folate intake from diet and supplement and serum levels of folate; (iii) RR, hazard ration (HR) or OR with 95% CI was reported to estimate the relative risk of the highest folate intake vs. lowest folate intake; (iv) patient with disease was identified by histological diagnosis; (v) for dose-response analysis, the number of cases and participants and eligible dose concentration must be provided. The selected studies were only limited in using dietary folate intake as only measurement standard. The most recent study was included for duplicate publications.

Data extraction
The following information was selected independently by two authors (H. Z. and Y. Z.) according to the criteria listed previously: the first author's name, publication year, country, study design, total sample size, sex, number of cases, number of controls, lowest folate level, highest folate level, difference between highest and lowest folate levels, measurement, range of exposure, histological type (ESCC, EAC, gastric cardiac adenocarcinoma (GAC); non-GAC), risk estimates and 95%CI for evaluating the highest folate levels vs. lowest folate levels. Adjusted rations were chosen in preference to the rations with the highest number of adjusted variables. For the studies which the reference groups were not the lowest dose concentration, the EXCEL macro document (RRest9) was used for the reference group transforming and data re-calculating according to the instructions [70]. All controversial questions were resolved by asking a third author.

Statistical analysis
The association of folate intake with esophageal, gastric and pancreatic cancers were examined by the pooled risk estimates (RR or OR) with 95%CI. The heterogeneity test was detected with I 2 statistic. Cut-off points of I 2 value for low, moderate and high degrees of heterogeneity were 25%, 50% and 75%, respectively. A fixed effect model was chosen when heterogeneity was negligible, otherwise, the random effects model was chosen [71]. Sensitivity analysis was investigated to assess robust of pooled results by omitting one study each time. The publication bias was determined by the Begg rank correlation test and Egger's linear regression test [72]. P<0.05 was considered statistically significant, and all p-values were two-sided. The trim-and-fill method was used to re-calculate the publication bias when the P values of Begg test or Egger test >0.05. The new pooled results (RR or OR) were compared with the original results. The results were considered as steady if the new pooled results are similar to the original results. At last, we conducted a dose-response meta-analysis using the correlated natural logs of the RRs or ORs with their standard error (SE) across all folate intake categories [73]. To derive the doseresponse curve, restricted cubic splines with four knots at the 5%, 35%, 65% and 95% percentiles of the distribution were used to assess for potential curvilinear relations. All data in this meta-analysis were performed using Stata 12.0 (StataCorp LP, College Station, TX, USA).

Author contributions
H. Z. performed search, Y. Z. and W. L. prepared tables and figures, C. T. wrote the manuscript and performed power calculation. All authors reviewed the manuscript.