Markers of Vitamin D Exposure and Esophageal Cancer Risk: A Systematic Review and Meta-analysis

It is estimated that 456,000 new esophageal cancer cases and 400,000 deaths occur annually in the world (1). Esophageal cancer is the sixth most common cause of cancer-related death worldwide, largely due to a particularly poor prognosis: 5-year survival rates are barely 10% in Europe (2, 3). Esophageal cancer has a distinctive epidemiologic pattern according to its most common histologic subtypes: adenocarcinoma and squamous cell carcinoma (SCC). Differing patterns of incidence suggest differential risk factors that may influence these cancer subtypes.

Adenocarcinoma affects the lower third of the esophagus and is thought to arise due to repetitive gastro-esophageal reflux causing alterations to the native squamous epithelium that can lead to Barrett's esophagus and cancer. Western regions have witnessed a rapid increase in esophageal adenocarcinoma incidence (4): a threefold increase has been observed since the 1970s (5). This increase has been associated with lifestyle factors, including obesity and tobacco smoking (6–8).

In contrast, incidence rates of SCC, which typically affects the upper esophagus, appear to be declining in some Western countries (9, 10). However, SCC remains the predominant esophageal cancer type in developing countries and is endemic in parts of Asia or the “esophageal cancer belt” stretching from Northern Iran to North central China (11). SCC can be largely attributed to consumption of alcohol, hot mate, pickled vegetables, opium, tobacco smoking, or chewing of nass (12–14).

Adequate vitamin D status has been linked with reduced risks of colorectal, breast, and other cancers (15–21). The tentatively causal relationship is supported by an abundance of in vitro evidence that has demonstrated several effects of vitamin D on the “hallmarks” of cancer, including regulation of apoptosis, promotion of cell differentiation, and suppression of cell proliferation (19, 22). Synthesis in the skin following exposure to sunshine and dietary intake are the main sources of vitamin D. Very few foods naturally contain vitamin D, so supplements constitute the most important dietary source (23). Once vitamin D is synthetized or ingested, it is hydroxylated in the liver to form 25-hydroxyvitamin D [25(OH)D], the main circulating form of vitamin D and best predictor of vitamin D status (24). After a second hydroxylation reaction, the active form 1,25-dihydroxyvitamin D [1,25(OH)₂D] is created. 1,25(OH)₂D can bind to the vitamin D receptor (VDR) and this complex has the ability to exert downstream biologic effects. Therefore, it is hypothesized that it is not only the availability of vitamin D but also the availability and structure of VDR that determine molecular actions.

The role of vitamin D in occurrence of rarer cancers is less clear; in particular, conflicting findings have been reported for the risk of esophageal cancer (25). Because vitamin D status is easily modifiable, understanding the role of vitamin D for cancer occurrence is highly relevant for making informed decisions about primary prevention. The aim of this systematic review and meta-analyses is to provide a comprehensive summary of the published literature on the risk of esophageal cancer and precursor lesions in relation to vitamin D exposures: [25(OH)D], vitamin D intake, UVB radiation, vitamin D–related genetic variation, and VDR expression. To our knowledge, this is the first systematic review on this topic.

The bibliographic databases Ovid MEDLINE (US National Library of Medicine, Bethesda, Maryland), EMBASE (Reed Elsevier PLC, Amsterdam, Netherlands), and Web of Science (Thomson Reuters, Times Square, New York) were searched from inception to September 8, 2015, for literature related to vitamin D or related exposures and esophageal neoplasia risk.

The search strategy identified studies that contained at least one keyword or Medical Subject Heading (MeSH) term from each of the following exposures: (i) vitamin D, cholecalciferol, ergocalciferol, [25(OH)D], vitamin D receptor(s), or calcitriol receptor(s), or any of these terms combined with SNP(s) or genetic polymorphism(s), or sun exposure, ultraviolet, UVB, solar radiation, sunlight, latitude or geographic variation, combined with terms for the outcomes: (ii) Barrett's esophagus, esophageal cancer, adenocarcinoma, SCC, tumor(s), or neoplasm(s). The search strategy also incorporated limits to studies conducted on humans; however, no language restrictions were specified. Review articles were excluded but checked for references. The systematic review protocol is registered on PROSPERO database 2014:CRD42014007630 (26) and in compliance with MOOSE guidelines (27).

Titles and abstracts were independently examined by two of three reviewers (L. Zgaga, F. O'Sullivan, and H.G. Coleman) to assess eligibility for the review using “PICO” criteria:

• Participants: Individuals of any age who have received a diagnosis of cancer or premalignant conditions of the esophagus (and corresponding control populations) were included in the review.

• Intervention: Assessment of vitamin D status, UVB exposure, vitamin D intake (from foods and/or supplements), VDR expression, and vitamin D–related genetic polymorphisms of the study participants.

• Comparators: Comparisons will be made between vitamin D status and vitamin D–related exposures outlined above with individuals who have not received a diagnosis of cancer or premalignant conditions of the esophagus.

• Outcome: Risk of esophageal cancer, including the histologic subtypes adenocarcinoma and/or SCC and/or premalignant lesions of the esophagus, Barrett's esophagus, and squamous dysplasia.

With regard to study design, observational (case–control, retrospective and prospective cohorts, cross-sectional) and interventional studies were included in the review; ecologic studies and case reports were not eligible for inclusion.

The reviewers initially screened titles and abstracts to remove obviously irrelevant articles and screened full text articles independently to identify studies for inclusion in the systematic review. Discrepancies were resolved by discussion with a fourth reviewer (M.M. Cantwell). Reference lists of included articles were also searched for other relevant studies. Methodologic quality for case–control and cohort studies was evaluated using the Newcastle–Ottawa Scale (28). For the cross-sectional study, we used an adapted version of the Newcastle–Ottawa Scale (29).

A standardized data collection protocol was used for gathering data: apart from results, study authors, publication year, residence of participants, proportion of Caucasians, age and gender distribution, study design, number of cases and controls, measurement method or definition of vitamin D exposure and outcome examined, details of the adjustment for confounders, and other variables were recorded. Corresponding authors were contacted for extra study details to enable evaluation and/or analysis if these were not reported in the articles, such as OR values for each of the SNPs investigated (typically only significant associations were reported), or covariates used in the analysis.

The associations between esophageal lesion risk and vitamin D exposures were summarized in meta-analyses by comparing risk in the highest to the lowest reported category of exposure (the lowest exposure level was the reference). If the original article used the highest category as the reference, the ORs were inverted or recalculated.

ORs and their corresponding 95% confidence intervals (CI) adjusted for the maximum number of confounding variables were extracted from published reports. In some studies, relative risk (RR) estimates were used, whereas adjusted HRs were extracted from cohort studies. These measures were used in the meta-analysis as given, because the HR, OR, and RR are approximate to one another when event rates are small, as is the case with esophageal cancer (30). Random-effects models were used to calculate pooled OR estimates. We used forest plots to show study-specific risk estimates and to present summary ORs where a minimum of two studies were published for: (i) per esophageal lesion subtype (adenocarcinoma, SCC, squamous dysplasia, and Barrett's esophagus) and (ii) for esophageal cancer overall (adenocarcinoma and SCC). Although stratified analysis by gender, ethnicity, and geographic location was planned, lack of studies precluded this.

The I² statistic was calculated to quantify the degree of heterogeneity between studies: larger I² values indicate greater heterogeneity (31). Risk of publication and selection bias was evaluated by checking for asymmetry in the funnel plots of the study OR against the standard error of the logarithm of the OR (32). Analysis was conducted using R software and the “metaphor” package (33).

Flowchart for study selection is shown in Fig. 1. Following initial screening of 690 titles and abstracts (n = 475 after removing duplicates), and then 45 full text articles, we identified 15 articles (34–48) that examined relationship between vitamin D exposures and esophageal neoplasms. These publications related to risk of esophageal cancer or precursor lesions, and: [25(OH)D] concentration (n = 3), vitamin D intake (n = 4), UVB radiation (n = 1), and/or vitamin D–related genetic variants or molecular expression (n = 7), as outlined in Table 1. Further specific limitations of the original study designs are outlined in Supplementary Table S1.

Only two studies investigated the role of [25(OH)D] in SCC occurrence (39, 40). In the meta-analysis, we found a nonsignificantly increased SCC risk when comparing the high versus low levels of circulating [25(OH)D]; OR, 1.20 (95% CI, 0.77–1.63). Esophageal adenocarcinoma risk was investigated in a single nested case–control study (39), and squamous dysplasia risk in a single cross-sectional study (38). We found an increased risk of esophageal cancer overall (adenocarcinoma and SCC) when comparing high versus low levels of [25(OH)D] level in the meta-analysis (OR, 1.39; 95% CI, 1.03–1.74; Fig. 2).

Four studies have reported on the association between vitamin D intake from food (supplement use not considered) and esophageal neoplastic lesion risk: two studies examined risk of adenocarcinoma (35, 37), three examined risk of SCC (34–36), and a single study examined risk of Barrett's esophagus (37). A nonsignificantly increased risk was found in the meta-analysis for adenocarcinoma (OR, 1.45; 95% CI, 0.65–2.24) and nonsignificantly decreased risk for SCC (OR, 0.80; 95% CI, 0.48–1.12; Fig. 3) with higher vitamin D intakes. No association was observed overall between vitamin D intake and risk of cancer (OR, 1.03, 95% CI, 0.65–1.42). No associations with vitamin D and Barrett's esophagus were found in an Ireland-based study (37).

Only a single study examined the relationship between esophageal cancer and UVB (47). This study found decreased risk of esophageal adenocarcinoma (OR, 0.49; 95% CI, 0.31–0.79) and esophago-gastric junction adenocarcinoma (OR, 0.52; 95% CI, 0.33–0.81) in individuals with higher lifetime mean daily UV radiation exposure, but not with SCC (OR, 0.95; 95% CI, 0.57–1.59). For meta-analysis across cancer types from this study, see Supplementary Fig. S1.

Risk of esophageal neoplasia was investigated in relation to VDR polymorphisms (5 studies), VDR expression (single study), and vitamin D level–related genetic variation (single study).

Haplotype rs2238135/rs1989969 was examined using the same cohort in relation to SCC and also Barrett's esophagus and adenocarcinoma. We found a decreased cancer risk in G/T haplotype carriers in meta-analysis of unadjusted findings (OR, 0.45; 95% CI, 0.00–0.91; Fig. 4).

A suggestive association was also found between esophageal cancer risk and variant rs2107301 T versus G in a meta-analysis of two studies adjusted OR estimates (OR, 0.66; 95% CI, 0.28–1.05; Figs. 4 and 5). No association was found between TaqI or FokI and esophageal neoplasia (OR, 1.31; 95% CI, 0.41–2.20 and OR, 1.03; 95% CI, 0.72–1.33), respectively.

A single study did not find any differences in VDR expression between Barrett's esophagus, adenocarcinoma, or normal mucosa samples analyzed, although this investigation was restricted to only six biopsies samples per disease state (42).

One Chinese case–control study, assessed 12 SNPs that were shown to modify vitamin D status in relation to risk of SCC. In this relatively large study comprising about 4,000 participants (1,942 cases), no significant associations were found between any of these SNPs individually or their genetic score and the risk of SCC (48).

In this systematic review, we attempt to summarize all available evidence to give the most comprehensive overview of the associations between vitamin D exposures and esophageal neoplastic lesions to date. Our effort has been limited by the scarcity and quality of published studies, and the use of different vitamin D exposures and outcomes, which makes interpretation and comparisons difficult.

Although we observed increased esophageal lesion risk associated with higher [25(OH)D] concentration in the meta-analysis, at this time, we are reluctant to suggest that higher [25(OH)D] increases the risk of esophageal cancer. The small number of published studies, the limitations of their designs (Supplementary Table S1), and possibility of population-specific effects raise concerns. Nonetheless, current evidence exposes a possibility that population subgroups may exist where risk of esophageal cancer is increased with higher [25(OH)D] concentration. All three studies contained a large proportion of the Han Chinese population, of which two were in Linxian, China. The Linxian region in China has among the highest rates of esophageal SCC in the world (38–40). It is not clear whether any region-specific environmental or genetic exposures (or their interactions) drive these high rates. Some authors suggest that vitamin D may be increasing cancer risk in this population by affecting the metabolism of polycyclic aromatic hydrocarbons (high use of coal in the region leads to high exposure to these toxins; ref. 38). Hence, the majority of evidence to date comes from studies conducted in China, where distribution of [25(OH)D] and etiology of esophageal lesions are likely to be different to that of the Western populations (49). The notion that the effect of vitamin D on esophageal cancer may vary by ethnicity has implications for vitamin D supplementation recommendations aimed at increasing [25(OH)D] level. In conclusion, further studies in different populations are needed and population-specific effects of [25(OH)D] cannot be excluded at this time.

It is worth mentioning the study of Giovannucci and colleagues that measured predicted [25(OH)D] concentration in a large cohort (N = 47,800), by modeling multiple factors that influence vitamin D status, such as UVB, diet, supplements, skin pigmentation, and body mass index (BMI). A prediction equation was then developed and it was found that higher levels of predicted [25(OH)D] were associated with a decreased risk of esophageal cancer (RR, 0.37; 95% CI, 0.17–0.80; ref. 50); however, factors used to predict [25(OH)D] could affect cancer risk independent from their association to vitamin D status, for example, BMI.

We found a nonsignificantly increased risk of adenocarcinoma and nonsignificantly decreased risk of SCC for higher dietary vitamin D intakes—this is contradictory to nonsignificantly increased risk of SCC found for higher [25(OH)D] levels. As subtypes of esophageal cancer seem to have different etiology, it is plausible to hypothesize that vitamin D could have a beneficial effect on one while having no effect or a detrimental effect on another subtype. Overall, only four studies investigated the relationship between vitamin D intake and neoplastic lesions of the esophagus. Notably, none of these studies collected information on vitamin D supplementations, which make a major contribution to vitamin D status for individuals choosing to take these (23). Widespread vitamin D supplementation is not currently advocated; however, with increasing prevalence of use, it is important for contemporary studies to assess this as part of vitamin D intake data.

A single article reported on measures of lifetime UV radiation exposure and esophageal cancer risk in an Australian population–based case–control study (47). Individuals with the highest tertile of mean lifetime daily UV radiation exposure had a reduced risk of esophageal adenocarcinoma and esophago-gastric junctional tumors. In addition, an inverse association was also found between the number of nevi (another marker of sun exposure) and adenocarcinoma in the same study. This is in contrast to no association observed in a single [25(OH)D] study and contradictory to the increased adenocarcinoma risk found for higher dietary vitamin D intake. Similarly, no association was found between SCC and UVB, but a suggestive positive association was observed in [25(OH)D] studies. The inconsistency between these results may be due to the underlying population differences which we are not able to address at this time due to the lack of published studies.

It is worth mentioning that multiple ecological studies examined UVB radiation exposure and esophageal cancer risk (these were ineligible for inclusion in our review due to study design). However, they found a significantly lower esophageal cancer risk and mortality in regions with higher UVB irradiance (47, 51, 52). Boscoe and colleagues used satellite-measured solar UVB levels and found a reduced esophageal cancer risk (RR, 0.79; 95% CI, 0.75–0.83) and lower mortality rates (RR, 0.74, 95% CI, 0.71–0.76) when looking at solar UVB exposure in Southern versus Northern United States. Similar correlations with incidence and mortality were observed with latitude, with an index of UVB intensity in France (53) and with mortality in China (54).

The advantage of studying genetic polymorphisms is that the exposure is constant and present throughout life. However, genetic effects are typically small and very large cohorts are needed to achieve sufficient power for effect detection. The majority of studies included in this review had small sample sizes (median number of cases was 141).

There has been some evidence to suggest different polymorphisms in the VDR gene (and consequential variations in the VDR protein) can modify activity of vitamin D–VDR complex (55). For example, rs11568820 was found to directly influence transcriptional activity due to its location in the VDR promoter region (56), and rs10735810 has been shown to affect the translational start site of 1,25-dihydroxycholecalciferol (57). Therefore, VDR polymorphisms or altered expression can potentially lead to the modification of cancer risk and survival (58–60). To date, polymorphisms in VDR gene have been linked to risk of cancers, including prostate (61), breast (62), skin and colorectal (63, 64), and high VDR expression has been linked to increased survival in prostate and breast cancers (65–67).

In the meta-analysis, an association was found for VDR haplotype rs2238135/rs1989969 G/T and esophageal neoplasia risk; while there was suggestive association for variant rs2107301. In accordance with this, previous studies have also reported a reduced risk in prostate cancer for individuals who have the more common G allele, when compared with those with the rarer C allele for variant rs2238135 (OR, 0.51; 95% CI, 0.29–0.85; ref. 68). Contrastingly however, this article also found an increased risk of prostate cancer with variant rs2107301 T versus C (OR, 2.47; 95% CI, 1.52–4.0; ref. 68), whereas Anic and colleagues found no association between rs2107301 and risk of glioma (69).

It had previously been suggested that apoptosis mechanisms were important for the transformation of Barrett's esophagus into adenocarcinoma and that expression of VDR in the esophagus was important in regulating apoptosis (70). It has further been shown that high expression of VDR can impact disease progression and overall survival of prostate, colon, and breast cancers (65, 67, 71). In a very small study by De Gottardi and colleagues, no difference in VDR expression was observed in esophageal biopsies from patients with normal mucosa, Barrett's esophagus, or adenocarcinoma (42). Trowbridge and colleagues have also published a series of articles on VDR expression in esophageal tissue, but these were not eligible for inclusion in our review due to the lack of risk estimates presented, or inability to calculate these. In their studies, a change in VDR expression has been noted in columnar metaplasia but not in native squamous epithelium of the esophagus (72, 73). This suggests that vitamin D does not have an opportunity to bind locally in the esophagus and therefore exert any biologic effects, unless the cell lining has undergone the metaplastic transition and may explain discrepancies in results by histologic subtype.

Finally, Wang and colleagues (2015) found no associations between 12 genetic variants associated with vitamin D status and risk of SSC in their Chinese case–control study. However, it may be inappropriate to assess these SNPs that were shown to modify vitamin D status in GWAS in individuals of European ancestry (74) in the Chinese population (48).

In summary, because of the small sample size of most studies and an overall scarcity of published articles, evidence available at this time is deficient and meaningful conclusions cannot be made. However, results presented here do suggest vitamin D–related genetic variation is worthy of further examination, in larger, adequately powered studies.

This is the first systematic review that has examined the relationship between vitamin D and related exposures and esophageal neoplasia risk. A strong point of this review is that all major environmental and genetic vitamin D–related exposures have been considered. Dietary vitamin D intake is known to correlate only weakly with [25(OH)D], the best biomarker for exposure to vitamin D. This is probably due to the fact that dietary sources of vitamin D are scarce, and accurate assessment of vitamin D intake over time is difficult. Therefore, the associations observed may in fact reflect the effect of additional exposures other than vitamin D.

The major limitation relates to the published information; namely, only a small number of studies that suffer from various methodologic limitations and typically include small sample sizes were available. It cannot be excluded that reported findings arose due to unmeasured or residual confounding, as the level of adjustment varied across retrieved studies. Moreover, we noted relatively large heterogeneity in meta-analyses and our capacity to detect publication bias is limited (Supplementary Figs. S2–S5) because meta-analyses were based on a small number of studies (75).

Future studies should be sufficiently powered and aim to measure 25(OH)D (ideally at multiple time points), collect accurate data on vitamin D intake (in particular, information on vitamin D supplementation), attempt to approximate individual UVB exposure, and assess genetic factors relevant for vitamin D metabolism, to provide comprehensive evidence. Information on important confounders should be collected and included. Associations should ideally be examined in a cohort (particularly for vitamin D intake and UVB) or case–control (genetic factors) studies and designed carefully to minimize the possibility of confounding and reverse causation. Different esophageal neoplastic lesions should be examined separately due to the known differences in their etiology.

This is the first systematic review which has examined the relationship between esophageal neoplasia risk and vitamin D exposures; we present the most comprehensive overview of available evidence to date by including all major personal, environmental, and genetic factors related to vitamin D. While vitamin D has generally been shown to be protective for most other cancers, we found that higher [25(OH)D] concentration was associated with an increased risk of esophageal cancer, in predominantly Chinese populations. Interestingly, albeit nonsignificantly, dietary vitamin D intake was associated with a decreased risk of SCC, but an increased risk of adenocarcinoma. One study reported higher lifetime UVB exposure was associated with a decreased risk of adenocarcinoma. There is some evidence to suggest VDR polymorphisms modify the risk; however, no consistent associations were detected. Hence, results are strikingly inconsistent and we are unable to make any firm conclusions with respect to the role of vitamin D in esophageal neoplasia at this time.

Because vitamin D deficiency is common, pressure exists to promote vitamin D supplementation. Our findings have implications for the guidelines on supplementation and population-wide interventions that are currently being revisited and debated in many countries, because results suggest that population subgroups may exist where attempts to increase [25(OH)D] concentration could be harmful. Issues like this need to be considered and harmful effect clarified before interventions are put in place.

Therefore, it is critical to examine the suggested detrimental effects of vitamin D on health in well-designed adequately powered studies, before public health measures aimed at increasing [25(OH)D] are introduced.

No potential conflicts of interest were disclosed.

F. O'Sullivan is funded via FP7-PEOPLE-2013-CIG SOGVID, project number 631041. H.G. Coleman is supported by a Cancer Research UK Population Research Postdoctoral Fellowship (A15333).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.