Intakes of Dietary Folate and Other B Vitamins Are Associated with Risks of Esophageal Adenocarcinoma, Barrett's Esophagus, and Reflux Esophagitis 

The Journal of Nutrition, Volume 143, Issue 12, December 2013, Pages 1966–1973, https://doi.org/10.3945/jn.113.174664

Over the past 30 y, dramatic increases in the incidence of esophageal adenocarcinoma (EAC)⁶ have been observed in many Western populations (1, 2). Barrett's esophagus (BE), in which the normal stratified squamous epithelium of the distal esophagus is replaced by specialized intestinal metaplasia, is a recognized precursor of these cancers (3). Because BE can occur as a complication in those with reflux esophagitis (RE), the progression from inflammation through metaplasia to adenocarcinoma has been proposed as a model of the carcinogenic process (4). Although an understanding of the factors involved at different stages of this sequence is crucial for the development of prevention strategies, relatively little is known about the etiology of BE or RE.

Higher vegetable intake has been associated with reduced esophageal cancer (5). The association appears to be stronger for adenocarcinomas than squamous cell carcinomas (6). An inverse association between BE and vegetable intake has been reported in a small number of studies (7). Vegetables, particularly green leafy vegetables, are a rich source of the B vitamin folate. Folate and its synthetic form, folic acid, are key in one-carbon metabolism, disruption of which can interfere with DNA synthesis, repair, and methylation (8). An accumulation of evidence suggests that low-folate status might have a role in carcinogenesis through various mechanisms, including by stimulating aberrant DNA methylation, leading to inappropriate activation of oncogenes (8), or by inducing uracil misincorporation during DNA repair and synthesis, leading to DNA strand breaks, chromosome damage and, eventually, malignant transformation (9).

In addition to folate, efficient one-carbon metabolism requires riboflavin, vitamin B-6, vitamin B-12, and methionine. Vitamin B-12 is a cofactor in the methylation of homocysteine, which is biosynthesized from methionine, vitamin B-6 is required for the conversion of homocysteine to cystathionine, and riboflavin influences the activity of the methylenetetrahydrofolate reductase enzyme (10). These factors, therefore, might influence disease risk either on their own account or by acting together with folate. The available evidence on dietary intakes of folate, vitamin B-12, vitamin B-6, and methionine, independently or together as overall dietary methyl status, is limited for EAC (11–16). Only one study appears to have investigated associations with BE (16) and none to our knowledge have considered RE. Moreover, alcohol and smoking adversely affect folate metabolism, absorption, and/or utilization (17, 18). Alcohol and smoking therefore have the potential to modify relations between folate intake and risk of EAC and its precursors (and vice versa), but no studies to our knowledge have investigated this.

Within the context of an all-Ireland population-based, case-control study [the Factors Influencing the Barrett's Adenocarcinoma Relationship (FINBAR) study], we investigated whether intakes of dietary folate and related B vitamins were associated with risks of EAC, BE, and RE. We further investigated whether folate intake interacted with alcohol or smoking to influence disease risks.

Full details of participant eligibility and recruitment are given elsewhere (19). Recruitment to the FINBAR study commenced on the island of Ireland in March 2002 and continued until July 2005. Three groups of participants were recruited from Northern Ireland and the Republic of Ireland: patients with EAC, patients with long-segment BE, and population controls. In addition, during September 2004 to July 2005, a fourth group, patients with RE, was recruited from Northern Ireland only.

All participants were aged ≤85 y and Caucasian. Eligible EAC cases comprised individuals with an incident, histologically confirmed, first primary invasive tumor located in the esophagus or the gastro-esophageal junction; in situ cancers were not included. In Northern Ireland, cases were identified from electronic records from all pathology laboratories within the province. In the Republic of Ireland, cases were identified from the main hospitals involved in diagnosis and treatment of esophageal cancer. Cases' medical records and histopathological reports were reviewed by a clinical panel to confirm tumor location. Eligible BE cases were individuals with long-length BE at endoscopy (≥3 cm of typical Barrett's mucosa) and presence of specialized intestinal metaplasia confirmed by histological examination. Patients with dysplasia were excluded. RE patients had macroscopically visible erosive esophagitis at endoscopy, which was defined as evidence of mucosal breaks or erosions within the esophagus (grades 2–4 in the Savary Miller classification/Hetzel-Dent classification or grades B, C, or D in the Los Angeles classification). Those with macroscopic or histological evidence of infection, documented dysmotility, or gastric outlet obstruction were excluded. Eligible controls were adults without a history of esophageal or other gastro-intestinal cancer or a known diagnosis of BE. In Northern Ireland, controls were selected at random from the province-wide General Practice Master Index, a province-wide database of all persons registered with a general practitioner. In the Republic of Ireland, controls were selected at random from 4 general practices in Dublin and Cork representing the urban/rural distribution of cases. Controls and BE and RE cases were frequency-matched to EAC cases by 5-y age-band and gender.

Ethical approval was obtained from Queen's University Hospital Research Ethics Committee (Belfast), Cork Teaching Hospitals Clinical Research Ethics Committee, and the Research Ethics Committee Board of St James's Hospital (Dublin). All participants gave written informed consent.

Participants underwent a structured interview with trained interviewers. Information was collected on a range of sociodemographic and lifestyle factors, including self-reported weight at the reference date, history of tobacco use (e.g., age started smoking, age stopped smoking, number of cigarettes smoked per day), and alcohol intake (number of units of alcohol consumed per week at the reference date, with one unit equivalent to one glass of wine, one-half a glass of beer or lager, or one public house-measure of spirits). The interview reference date was 5 y prior to the interview. Dietary information was collected using the validated European Prospective Investigation into Cancer FFQ for Norfolk, England (20), which had been modified to incorporate foods (such as soda bread) reported as commonly eaten in the North/South Ireland Food Consumption Survey (21). Because diets on the island of Ireland and in England are rather similar, fairly little modification was needed. The modified questionnaire collected information on habitual intake of 101 food items and included specific questions on the brands of breakfast cereal eaten. Information was also collected on the broad classes of dietary supplements used by participants. The interviewers measured participants' height and waist and hip size.

Dietary data were converted into estimated nutrient intakes using Tinuviel Software UK (Q-Builder V1.0 program), computerized UK food composition tables (22) that take folic acid fortification of cereals into account. Twenty-three participants were excluded either for providing insufficient dietary information to compute nutrient intakes (n = 22) or for having completely implausible total energy intake (n = 1).

The primary analyses were based on intake from dietary sources alone, because we did not have detailed composition information on dietary supplements. Dietary intakes of folate, vitamin B-12, vitamin B-6, and riboflavin were adjusted for total energy using the nutrient residual method (23). Participants were grouped into intake quartiles based on the distribution of all participants, with the baseline group comprising the lowest quartile. For alcohol, reported units were converted into grams of alcohol per week and participants categorized into 4 groups: nonconsumers and intake tertiles among consumers. For tobacco exposure, participants were categorized by smoking status at the reference point (never, ex, current) and pack-years of smoking. BMI was computed from self-reported weight (kg) at the reference point and height (m) measured by the interviewer, and waist:hip ratio from measurements taken by the interviewer. Social class for men and single women was based on self-reported main occupation 5 y prior to interview or last occupation before retirement; for women who were married, cohabiting, or widowed, this was based on the occupation of their husbands/partners.

For descriptive purposes, participants' characteristics were compared between the 4 groups using chi-square tests. Between-group comparison of means of total energy, folate, vitamin B-12, vitamin B-6, and riboflavin intake were conducted by ANOVA and F-tests. Logistic regression was used to calculate adjusted and multivariate ORs with 95% CIs for each group of cases compared with controls. Adjusted ORs were adjusted for age, sex, and total energy. Multivariate ORs were further adjusted for a range of potential confounders, identified both from the literature and empirically from the data. To fit the best model for each lesion, separate models were built for EAC, BE, and RE. Factors that made a significant contribution (likelihood ratio test P < 0.1) were retained in the models. In addition to age, sex, and total energy, the final model for EAC included years of full-time education, body BMI, and alcohol intake; the BE model included social class, waist:hip ratio, history of hernia, and history of gallstones; and the RE model included years in full-time education, waist:hip ratio, history of hernia, and smoking status. The final models were not adjusted for frequent gastro-esophageal reflux, as this could be viewed as being on the biological pathway. When we repeated our analysis adjusting for gastro-esophageal reflux, this slightly attenuated the results for EAC and BE but did not affect those for RE (data not shown). When we repeated the analyses for RE using only controls from Northern Ireland, the adjusted and multivariate risk estimates for folate, vitamin B-12, and riboflavin were generally further from unity and those for vitamin B-6 were slightly closer to unity (data not shown). Linear trend tests were used to investigate dose-response patterns in risk across dietary quartiles 1–4.

In sensitivity analyses, we investigated whether the main effect of folate was modified by dietary supplement use. Participants who reported taking B vitamins or multi-vitamins were classified as having taken folic acid-containing supplements and the analysis was repeated excluding those who took such supplements.

Effect modification was explored by computing ORs for combinations of folate and other B vitamins, alcohol, and smoking. In these analyses, participants were classified as having “low” (quartiles 1 and 2) or “high” (quartiles 3 and 4) intake. For alcohol, the “low” group comprised nonconsumers plus those in the lowest intake tertile. The test for interaction was the change in deviance (−2*log likelihood) between a main effects model and one including an interaction (cross-product) term. Because of the limited sample size, the analyses of effect modification were considered exploratory.

All models presented had adequate fit according to the Hosmer and Lemeshow test (24). Statistical analyses were conducted using STATA version 9 (StataCorp).

A total of 227 cases with EAC (64% of those eligible), 224 with BE (82%), 230 with RE (69%), and 260 controls (42%) took part. After exclusions due to insufficient or implausible dietary data, the numbers included in the analyses were EAC, 223; BE, 220; RE, 219; and controls, 256. The majority of participants were male (82–84% across groups) (Table 1). The mean age at interview ranged from 61 to 64 y. A higher proportion of those with EAC or BE were in the lower social classes (58–59%) compared with controls (47%) and RE patients (48%; P = 0.02). Almost one-third of those with EAC had self-reported height and weight in the reference period corresponding to a BMI of 30 kg/m² or more compared with 18% of those with BE, 21% with RE, and 19% of controls (P < 0.01). Only 19% of controls reported having very frequent symptoms of heartburn and/or reflux compared with 27% with BE, 40% with RE, and 48% with EAC (P < 0.01). One-third of EAC cases were current smokers, higher than in other groups (BE, 23%; RE, 22%; controls, 18%; P < 0.01).

In analyses adjusted for age, sex, and total energy intake, EAC risk was significantly reduced, by almost 50%, in those with dietary folate intake in the 2 highest quartiles compared with the lowest intake quartile (Table 2). When adjustment was made for other confounders, these risk estimates were slightly attenuated and were borderline significantly different from unity (Q3 vs. Q1 OR: 0.54, 95% CI: 0.30, 0.97; Q4 vs. Q1 OR: 0.56, 95% CI: 0.31, 1.00), although the test for linear trend remained significant (P-trend < 0.01). There were strong inverse trends in risk of BE (multivariate P-trend < 0.01) and RE (multivariate P-trend < 0.01) with increasing folate intake. In multivariate analyses, the risks of both lesions were reduced by at least 60% in those in the highest compared with the lowest intake quartiles (BE, Q4 vs. Q1 = 0.40, 95% CI: 0.21, 0.75; RE, Q4 vs. Q1 = 0.34, 95% CI: 0.18, 0.64). When the analysis was repeated excluding those who used folic acid-containing supplements, the results were unchanged (data not shown).

For vitamin B-6 intake, risks of all 3 lesions decreased significantly with increasing intake (Table 2); tests for trend were highly significant and the risk estimates in the highest intake quartiles were 50–70% lower than in the lowest intake quartiles. EAC risk increased with increasing vitamin B-12 intake (P-trend < 0.01), with the risk for the highest intake quartile almost 4-fold higher than that for the lowest intake (OR: 3.87, 95% CI: 2.06, 7.29). Risk of BE also significantly rose with increasing vitamin B-12 intake (P-trend = 0.02), but the effect estimates were less pronounced than those for EAC. There was no notable association between vitamin B-12 and RE. There was an inverse relation between riboflavin intake and RE, with risk reduced by between 45% and 65% in quartiles 2–4 (P-trend < 0.01). There were no associations between riboflavin and EAC or BE.

There were no significant interactions between dietary folate intake and intakes of vitamin B-12, vitamin B-6, or riboflavin for EAC, BE, or RE (data not shown).

There was a suggestion that folate modified the relation between smoking and EAC, with the test for interaction borderline significant (P-interaction = 0.05) (Table 3). Risk was increased in current smokers compared with those who never smoked, but this relation was much more pronounced among those with low folate intake. Compared with never smokers with high folate intake, the group who were current smokers and had low folate had an 8-fold higher risk (OR: 8.15, 95% CI: 3.61, 18.40), whereas the risk for current smokers with high folate intake was raised only 2-fold (OR: 2.34, 95% CI: 1.08, 5.05). For BE, there was no relation with smoking among those with high folate intake, but the combination of low folate intake and current smoking was associated with a significantly increased risk (OR: 2.93, 95% CI: 1.24, 6.92). There was no interaction between smoking and folate in relation to RE.

As regards alcohol consumption, the low-alcohol, low-folate group had significantly increased risks of EAC, BE, and RE compared with those with low alcohol and high folate intakes, but none of the tests for interaction were significant (Table 3).

Previous studies of folate intake and EAC have reported an inverse relation (11–16). A meta-analysis of 3 studies estimated an OR for the highest compared with the lowest dietary intake quantile of 0.49 (95% CI: 0.35, 0.67) (25). Our OR for the highest-intake quartile was similar (0.56) and a strong linear trend was evident. The one previous study of BE found a nonsignificantly lower risk in the 3 upper quartiles of food folate intake but only for 98 cases with dysplasia (16). Another small study described lower blood folate concentrations in patients with esophageal dysplasia than in cytologically normal individuals (26). Our observation of a similar association across all 3 case groups suggests that, if folate has a role in esophageal carcinogenesis, the effect likely occurs at an early stage.

Major dietary sources of folate include green leafy vegetables, some fruits, legumes, and liver. There may be substantial error in assessment of dietary folate intakes (17), resulting in misclassification, probably nondifferential. Thus, our results may underestimate the true effect of folate. We collected information only on broad classes of dietary supplements used (e.g., multivitamins) and could not estimate total folate intake. However, survey data suggest supplemental folic acid probably contributed relatively little to total folate in older, predominantly male, individuals in Ireland during the study period (27, 28). Few study participants took folic acid-containing supplements (controls, 8%; EAC, 6%; BE, 5%; RE, 8%) and among controls, use did not vary across dietary folate intake quartiles. Unsurprisingly, the associations with folate were unchanged when the analysis was limited to supplement nonusers.

Three previous EAC studies were from the United States (11, 13, 14), where folate intakes and supplement use are fairly high (29, 30). In contrast, the lack of mandatory folic acid fortification of foods, our study, and population surveys (21) suggest that overall folate intake in Ireland is not high and the proportion of intake consumed as folic acid (which is more bioavailable) is low. That we found inverse associations between food folate and esophageal lesions in this setting suggests that, if there is a threshold of folate intake required to obtain a protective effect, it is relatively low.

Whereas an inverse relation between vitamin B-6 intake and EAC has been previously described (11–14, 16), the observed similar associations for BE and RE have not. Dietary sources of vitamin B-6 include meat (pork, poultry), fish, cereals, eggs, and vegetables (31). The overlap between these and folate sources was evident in Spearman rank correlation coefficients (lowest, BE = 0.65; highest, EAC = 0.78; all P values < 0.001). This may explain the similarity in results for the 2 nutrients.

For vitamin B-12, it was intriguing that a positive relation was found and that this was much stronger for EAC than BE (and there was no association with RE). At face value, this pattern suggests this nutrient may be implicated in progression of esophageal lesions. Given the specific role of vitamin B-12 in folate metabolism, we might speculate that the amino acids methionine and homocysteine, related enzymes (such as 5-methyltetrahydrofolate), and/or the DNA methylation cycle generally could be relevant in progression. Vitamin B-12 intakes in our study substantially exceeded recommended amounts (32) and were somewhat higher than average daily intakes among North/South Ireland Food Consumption Survey participants (5.4 μg/d; 662 men, 18–64 y, 1997–1999) (21). The main vitamin B-12 dietary sources are meat, fish, poultry, eggs, and milk (31). In FINBAR, eggs and total fish intakes were unrelated to EAC, but higher intakes of fresh red meat and dairy were associated with increased risk of EAC, but not of BE or RE (33, 34). Thus, our results may be partly explained by the presence of vitamin B-12 in red meat and milk products, or the reported associations with red meat and dairy might be due to the presence of vitamin B-12. Alternatively, the observed association may result from the presence of other nutrients [such as fat, which has been positively associated with these lesions (34)] in vitamin B-12–containing foods. Further research is warranted to better understand inter-relations between vitamin B-12 and related foods and nutrients and esophageal neoplasia etiology.

Rich dietary sources of riboflavin include milk, eggs, lean meats, and green vegetables (31). Some previous studies report an inverse relation between riboflavin and EAC (11–14). Like our study, an Australian study reported no association with EAC but moderate (albeit nonsignificant) reduced risk of BE with higher intake (16). We also found a strong inverse relation with RE. If confirmed elsewhere, this suggests riboflavin may be implicated in the development of early esophageal lesions.

Smoking has been convincingly associated with EAC risk in the FINBAR study (35), but its role in lesions earlier in the pathway is uncertain (36, 37). Smoking adversely affects folate metabolism, depletes vitamin B-6 availability, and interferes with vitamin B-12 metabolism (18), suggesting smokers probably have increased requirements for folate or one-carbon nutrients. Our novel finding that folate intake may modify the smoking-EAC relation, conferring particularly increased risk on smokers with low folate intakes, is consistent with these biological effects, as is the raised risk of BE associated with smoking among those with low folate intake. These exploratory results clearly merit further investigation.

Aspects of alcohol consumption appear related to EAC, BE, and RE risks (38–40). Based on the well-documented effects of alcohol on folate status (17), individuals with high alcohol and low folate intakes might be expected to have increased disease risk. Although for all 3 lesions these individuals had modestly higher risks than those with low alcohol and high folate intakes, risk estimates and tests for interaction were not significant. Even if there was a true interaction between these factors, we may have failed to observed it due to lack of statistical power and misclassification of alcohol intake; the “no alcohol intake” category related to the reference point and included some individuals who previously consumed alcohol. Intriguingly, for EAC, BE, and RE, risk was highest in people with low folate and low alcohol intakes, further suggesting that relations between alcohol and these lesions are complex.

To our knowledge, FINBAR is the first population-based comparison of EAC, BE, and RE risk factors within the same study; it therefore has the potential to improve our understanding of the inflammation-metaplasia-adenocarcinoma sequence by revealing the particular stage at which etiological factors exert their effects. Because diet is potentially modifiable, if one-carbon nutrients were confirmed to be involved at particular stages in this sequence, this could inform development of prevention strategies.

Clinical notes of EAC cases were reviewed to determine tumor site and minimize chances of including gastric cardia tumors. A rapid case ascertainment process helped ensure a relatively high participation rate (62%) for EAC, which has poor prognosis. Reasons for nonparticipation included: death (n = 49), permission to contact refused by clinician (n = 49), and patient refusal (n = 37). The 42% response rate among controls raises the possibility that participation bias could explain the results. The North/South Ireland Food Consumption Survey reported a mean unadjusted daily folate intake of 332 μg/d (SD = 128) from 7-d food records completed by 662 men <65 y during 1997–1999 (21). Although mean (energy-adjusted) daily dietary folate intake among controls was 391 μg/d (SD = 78), folate intake estimated from the European Prospective Investigation into Cancer FFQ (which was the basis of our questionnaire) tends to be systematically higher than that from 7-d diet diaries (41). This suggests that folate intake among controls was not untypical. Moreover, for the other nutrients investigated, directions of association were not always consistent across the 3 lesions, suggesting that the results are not due to bias. The lack of biomarker data may be considered a limitation, as might the lack of information on methionine intake from the FFQ and the possibility of bias in dietary recall. The nutrients analyzed are present in multiple foods and coexist in the diet with other nutrients and non-nutrients. Collinearity among intakes of micronutrients, macronutrients, and foods makes fully accommodating effects of multiple dietary variables in statistical models difficult, if not impossible (42). Thus, we cannot be entirely certain that the observed associations are specific to the nutrients investigated; however, this limitation applies generally to studies investigating associations between aspects of diet and disease.

If folate is etiologically implicated, it is likely that localized rather than systemic folate depletion is important. However, little is known about folate status in esophageal tissue and, although it has been postulated that localized folate deficiencies in esophageal mucosa may occur, the available evidence is indirect (43).

Low folate status could influence carcinogenesis by stimulating aberrant DNA methylation (8), or inducing uracil misincorporation during DNA repair and synthesis (9), and both general mechanisms are implicated in esophageal carcinogenesis. The multi-step development of BE and progression to EAC is characterized by widespread alterations in DNA methylation and epigenetic changes (44), basal DNA damage is higher in BE patients than controls (45), the percentage of DNA damaged cells correlates positively with esophageal pathology (46), and variations in genes that encode DNA repair proteins have been linked to EAC (47). However, there are as yet no direct links between folate, or folate metabolism, and these processes in esophageal carcinogenesis. The single available study reported no interactions between dietary folate and vitamin B-6 intakes and p53 mutations, p53 protein overexpression, or p53 mutations at CpG sites (48), but included only 54 EAC cases. Further research is therefore needed to clarify the precise mechanisms by which folate or one-carbon nutrients might affect esophageal disease processes.

In conclusion, this study supports the hypothesis that folate and other dietary methyl-group factors are implicated in the etiology of EAC and its precursors, BE and RE. Pooled analyses of individual-level data from multiple population-based studies could help further clarify the impact of folate, one-carbon nutrients more generally, and significant cofactors (including smoking and alcohol) in the development of EAC and progression through the inflammation-metaplasia-adenocarcinoma sequence.

The authors thank Siobhan Reynolds, Majella Gallagher, Carol Anderson, and Martin McAnaespie for subject recruitment, Damian McManus for help with classifying the tumor sites, and Helen Mulholland for helpful comments on an earlier draft of the manuscript. The FINBAR study group members are L.J. Murray, L.A. Anderson (Queen's University Belfast); B.T. Johnston, R.G.P. Watson, J. McGuigan, H.R. Ferguson (Belfast Health and Social Care Trust, Belfast, Country Antrim, UK); S.J. Murphy (St Vincent's Hospital, Dublin, Ireland); J.V. Reynolds (St James' Hospital, Dublin, Ireland); and H. Comber (National Cancer Registry Ireland). L.J.M. and L.A.A. designed the FINBAR study; L.A.A. conducted the study; L.J.M. supervised the study; M.M.C. provided dietary assessment expertise; L.S. had the idea for this analysis; A.-E.C. and L.S. analyzed data; and L.S. wrote the paper and has primary responsibility for final content. All authors read and approved the final manuscript.