Reflux oesophagitis and Barrett's oesophagus represent early transitional stages in the progression of normal oesophageal tissue to oesophageal adenocarcinoma, arising from the repetitive reflux of gastric acid and bile. This spectrum of oesophageal disorders is of growing concern in developed nations. Although relatively rare, the widely reported rise in oesophageal adenocarcinoma incidence in the UK has continued into the new millennium(Reference Lepage, Rachet and Jooste1). A systematic analysis of time trends has also confirmed that the prevalence of both gastro-oesophageal reflux disease and oesophagitis has increased since the 1980s, particularly in North America and Europe(Reference El-Serag2). Although Barrett's oesophagus is a designated premalignant condition for oesophageal adenocarcinoma(Reference Shaheen and Richter3), progression along the carcinogenic pathway is selective, with approximately 0·5 % of patients going on to develop this malignancy(Reference Wood and Yang4–Reference Thomas, Abrams and De Caestecker6). Such selective development has generated interest in identifying modifiable risk factors for cancer progression in this small subsample of patients.
Despite certain dietary triggers such as fat, chocolate, mints, onions, citrus fruit and tomatoes being observed for reflux episodes, there is little evidence to suggest that these play a role in the subsequent development of oesophageal adenocarcinoma(Reference Terry, Lagergren and Wolk7). Only a limited number of studies have explored diet as an aetiological agent for these oesophageal lesions, with studies to date largely focusing on fruit and vegetable, macronutrient and antioxidant vitamin intake(Reference El-Serag, Satia and Rabeneck8, Reference Anderson, Watson and Murphy9–Reference Mulholland, Cantwell and Anderson13).
There is considerable evidence in the scientific literature citing a protective role for vitamin D against the development of many cancers(Reference Giovannucci14). Vitamin D has been shown to have several generic anti-carcinogenic effects such as suppressing cell proliferation, promoting cell differentiation and regulating apoptosis(Reference Dusso, Brown and Slatopolsky15). Consistent inverse associations are evident for optimum vitamin D status and colorectal cancer and adenoma risk in particular(Reference Gorham, Garland and Garland16, Reference Wei, Garland and Gorham17), and such findings are fuelling momentum for randomised controlled trials of vitamin D supplementation in cancer prevention. However, the hypothesised protective effects of vitamin D in cancer prevention may not be universal. There have been notable exceptions to this protective relationship though from studies of vitamin D and other digestive tract neoplasms(Reference Stolzenberg-Solomon, Vieth and Azad18). Increased risks of oesophageal squamous cell carcinoma and its precursor, squamous dysplasia of the oesophagus, have been demonstrated for individuals with the highest serum concentrations of 25-hydroxyvitamin D in a Chinese population(Reference Abnet, Chen and Dawsey19, Reference Chen, Dawsey and Qiao20).
Vitamin D and Ca are highly related in physiological terms, with low serum Ca status known to induce the synthesis of the active form of vitamin D from 25-hydroxyvitamin D in the kidney(Reference Davis21). Adequate Ca consumption has been demonstrated to reduce the risk of colorectal cancer, although reports of its relationship with the risk of other cancers have been inconsistent(12). These equivocal findings are suggested to be due to confounding from high fat contents of common food sources, particularly dairy products.
The most recent World Cancer Research Fund global report on food, nutrition, physical activity and the prevention of cancer could not draw any conclusions about the association of vitamin D, Ca or dairy product intake with oesophageal cancer risk due to a lack of available evidence(12). Furthermore, no studies to date have investigated the association between vitamin D, Ca or dairy intake and the risk of the precursor conditions, reflux oesophagitis and Barrett's oesophagus.
The aim of the present investigation was to explore the association between vitamin D, Ca and dairy intake and the risk of reflux oesophagitis, Barrett's oesophagus and oesophageal adenocarcinoma in an all-Ireland case–control study.
Methods
Study design
The Factors INfluencing the Barrett's Adenocarcinoma Relationship (FINBAR) study is an all-Ireland population-based case–control study that was established to investigate the aetiology of reflux oesophagitis, Barrett's oesophagus and oesophageal adenocarcinoma, as described previously(Reference Anderson, Watson and Murphy9, Reference Mulholland, Cantwell and Anderson13, Reference Anderson, Johnston and Watson22). Incident cases of oesophageal adenocarcinoma (n 227) and long-segment Barrett's oesophagus (n 224), and normal population controls (n 260) were recruited throughout the island of Ireland between March 2002 and July 2005. Reflux oesophagitis cases (n 230) were recruited in Northern Ireland only. Barrett's oesophagus, reflux oesophagitis and control subjects were frequency matched within 5-year age and sex strata to the distribution of oesophageal adenocarcinoma patients, up to a maximum age of 85 years. The present study was conducted according to the guidelines laid down in the Declaration of Helsinki, and all procedures involving human subjects/patients were approved by the Research Ethics Committee of the Queen's University Belfast (Northern Ireland); Clinical Research Ethics Committee of Cork Teaching Hospitals; and Research Ethics Committee Board of St James's Hospital (Dublin, Ireland). Written informed consent was obtained from all subjects.
Study participants
Incident cases of histologically confirmed oesophageal adenocarcinoma were identified from hospital clinical records in the Republic of Ireland or via electronic pathology records in Northern Ireland. Patients with endoscopically diagnosed long-segment Barrett's oesophagus ( ≥ 3 cm in length) were recruited if the presence of specialised intestinal metaplasia had been histologically confirmed. Patients with dysplasia were excluded. Reflux oesophagitis patients were included if they had been diagnosed with mucosal erosions in the oesophagus, classified as grades 2–4 or grades B, C or D using the Savary Miller/Hetzel-Dent or Los Angeles methods(Reference Nayar and Vaezi23), respectively, at the time of endoscopy. Population-based controls were adults with no previous history of Barrett's oesophagus, oesophageal or other gastrointestinal cancers. Controls were recruited at random in Northern Ireland via the General Practice Master Index and in the Republic of Ireland from four General Practices representing urban and rural areas (Dublin and Cork). The response rates of eligible participants ranged from 42 % for controls to 69, 82 and 74 % for reflux oesophagitis, Barrett's oesophagus and oesophageal adenocarcinoma cases, respectively.
Data collection
Trained interviewers collected data from study participants using an electronic questionnaire. Dietary intake was assessed using a 101-item FFQ, adapted for the Irish population from a version of the European Prospective Investigation into Cancer and Nutrition FFQ(Reference Day, Oakes and Luben24), by incorporating foods reported as commonly eaten in the North/South Ireland Food Consumption Survey(Reference Harrington, Robson and Kiely25). The FFQ did not specifically ask for information on vitamin D or Ca supplement usage. Mean daily dietary intakes were calculated from the FFQ using Q-Builder (Tinuviel Software, Llanfechell, Anglesey, UK). Participants were asked to recall their dietary habits over the 12-month period 5 years before the interview. BMI 5 years before the interview was calculated using self-reported weight (kg) divided by current height (m2), as measured by the interviewer. Helicobacter pylori infection status was assessed from serum samples using a Western blot assay, as described previously(Reference Anderson, Murphy and Johnston26).
Statistical analysis
A total of twenty-two study participants failed to provide any dietary information, and information on dairy product intake was missing for a further twenty-nine participants, and therefore these individuals were excluded from the present analysis. This equated to twenty-two reflux oesophagitis cases, twelve Barrett's oesophagus cases, nine oesophageal adenocarcinoma cases and eight controls being excluded, leaving 208 reflux oesophagitis patients, 212 Barrett's oesophagus patients, 218 oesophageal adenocarcinoma patients and 252 population controls for consideration in this analysis. The comparison group for all analysis exploring reflux oesophagitis risk was restricted to controls recruited in Northern Ireland, since reflux oesophagitis patients were recruited in Northern Ireland only.
Key characteristics and mean nutrient intakes of cases in relation to controls were examined using independent t tests for continuous variables and χ2 tests for categorical variables. Unconditional multiple logistic regression analysis was utilised to explore the association between vitamin D, Ca and dairy intake and disease risk by tertiles of intake, resulting in OR and corresponding 95 % CI. In order to test for trend, each individual within a particular tertile was assigned the median intake value for that tertile before inclusion in the regression model. In multivariate-adjusted models, confounding variables were age (years), sex, energy intake (kJ/d), smoking status (current/previous/never), education (years), BMI 5 years before the interview (kg/m2), occupation (manual/non-manual), alcohol intake (g/d), regular non-steroidal inflammatory drug use (weekly use for at least 6 months' duration), H. pylori infection (seronegative/seropositive), energy-adjusted glycaemic index intake (units/4184 kJ per d), energy-adjusted saturated fat intake (g/4184 kJ per d) and location (Northern Ireland/Republic of Ireland). Reflux oesophagitis analyses were restricted to Northern Ireland controls only. Covariates were chosen based on previously known risk factors for oesophageal lesions. In a separate multivariate-adjusted regression model, we further tested for regular gastro-oesophageal reflux symptoms (ever/never), since it is debatable whether reflux symptoms may confound or be on the causal pathway between disease risk and the dietary variables of interest. Stratified analyses were conducted according to BMI categories, smoking and H. pylori infection status, which were further adjusted for each other, and the likelihood ratio test was applied to evaluate potential interactions. Statistical analysis was performed using Intercooled STATA version 9.2 (StataCorp, College Station, TX, USA).
Results
Table 1 summarises the key characteristics and dietary variables of interest for all three case groups and their respective controls. Reflux oesophagitis and Barrett's oesophagus cases did not differ from their respective controls by daily energy, vitamin D, Ca or dairy product intakes. In contrast, oesophageal adenocarcinoma cases had significantly higher energy and vitamin D intakes compared with controls, although Ca and dairy intakes were similar between the groups. Vitamin D and Ca intakes represent intakes from foods only and do not include intakes from supplements, which we were unable to assess.
GOR, gastro-oesophageal reflux; NSAID, non-steroidal anti-inflammatory drug.
* Cases compared with Northern Ireland controls only.
† Cases compared with all controls.
‡ Dairy product intake defined as milk, yogurt, Greek yogurt, cheese, processed cheese and cottage cheese.
§ Heartburn/acid reflux symptoms experienced at least once weekly or >50 times/year >5 years before the interview date.
∥ Ever use defined as use at least once weekly for ≥ 6 months' duration.
As illustrated in Table 2, Ca intake was inversely associated with odds of reflux oesophagitis when comparing the highest with the lowest tertile of intake (OR 0·37, 95 % CI 0·17, 0·80; P for trend = 0·01) in age- and energy-adjusted regression analysis. However, no significant associations were identified for Ca, vitamin D or total dairy product intake and reflux oesophagitis risk once the adjusted for potential confounders.
GOR, gastro-oesophageal reflux.
* Multivariate adjustments: age (years); sex; energy intake (kJ/d); smoking status (current/previous/never); BMI 5 years before the interview; education (years); occupation (manual/non-manual); alcohol (g/d); regular non-steroidal anti-inflammatory drug use (ever/never: ever, weekly use for at least 6 months' duration); Helicobacter pylori infection (seropositive/seronegative); energy-adjusted glycaemic index intake (units/4184 kJ per d); energy-adjusted saturated fat intake (g/4184 kJ per d); location (Northern Ireland/Republic of Ireland).
† GOR symptoms experienced at least once weekly or >50 times/year >5 years before the interview date.
‡ Analysis limited to Northern Ireland controls only.
§ Dairy product intake defined as milk, yogurt, Greek yogurt, cheese, processed cheese and cottage cheese.
No significant associations were found between vitamin D or Ca intakes and Barrett's oesophagus risk either before or after multivariate adjustments, as shown in Table 2. There was also little evidence of an association between dairy product consumption and odds of Barrett's oesophagus in multivariate-adjusted regression analysis. Once the analysis was further adjusted for gastro-oesophageal reflux symptoms, however, a significant positive association emerged, indicating a nearly twofold increased risk of Barrett's oesophagus for individuals with the highest tertile of dairy product intake.
A significant direct association was observed between the highest tertile v. the lowest tertile of vitamin D intake and odds of oesophageal adenocarcinoma, even after adjustment for confounders (OR 1·99, 95 % CI 1·03, 3·86; P for trend = 0·02). This was attenuated slightly after adjustment for gastro-oesophageal reflux symptoms, but a significant trend remained (P for trend = 0·03). Consumption of dairy products also tended to be positively associated with oesophageal adenocarcinoma risk, although this did not achieve statistical significance until multivariate-adjusted analyses further accounted for gastro-oesophageal reflux symptoms (OR 1·89, 95 % CI 1·02, 3·50; P for trend = 0·04). No significant relationship was found between dietary Ca and oesophageal adenocarcinoma risk (Table 2).
To investigate the unexpected direct association observed between vitamin D and oesophageal adenocarcinoma risk further, we explored intake of foods commonly consumed in the Irish diet that are rich sources of vitamin D. Oily fish, egg, margarine and total red meat intakes were not significantly associated with oesophageal adenocarcinoma risk and could not explain the association observed (Table 3). Individuals who consumed liver were at heightened odds of oesophageal adenocarcinoma compared with non-consumers after adjusting for potential confounders (OR 1·73, 95 % CI 1·01, 2·97). However, further adjustment for liver intake actually strengthened, rather than attenuated, the positive association between vitamin D and oesophageal adenocarcinoma risk (OR 1·99, 95 % CI 1·03, 3·86) when comparing extreme categories of intake. Furthermore, individuals who consumed liver also had elevated odds of Barrett's oesophagus compared with non-consumers (OR 1·77, 95 % CI 1·06, 2·98), thus failing to demonstrate the specificity to oesophageal adenocarcinoma that was shown by the results of vitamin D. Additional adjustment for total vitamin A intake could not explain the deleterious effect of liver, although it was not possible to investigate the effects of retinyls and pro-vitamin A carotenoids separately. It should be noted that unlike in the USA, milk is not fortified with vitamin D in either Northern Ireland or the Republic of Ireland and was therefore not considered in the vitamin D-related food analysis.
GOR, gastro-oesophageal reflux; tsp, teaspoon.
* Multivariate adjustments: age (years); sex; energy intake (kJ/d); smoking status (current/previous/never); BMI 5 years before the interview (kg/m2); education (years); occupation (manual/non-manual); alcohol (g/d); regular non-steroidal anti-inflammatory drug use (weekly use for at least 6 months' duration); location (Northern Ireland/Republic of Ireland); Helicobacter pylori infection (seropositive/seronegative); energy-adjusted saturated fat intake (g/4184 kJ per d); energy-adjusted glycaemic index intake (units/4184 kJ per d).
† GOR symptoms experienced at least once weekly or >50 times/year >5 years before the interview date.
‡ Total red meat intake: beef+pork+ham+lamb+liver.
When analysis was stratified by BMI categories, as shown in Table 4, a 2·5-fold elevated OR for oesophageal adenocarcinoma was limited to overweight and obese individuals with the highest category of vitamin D intake, and was not observed in normal-weight individuals. The direct association between vitamin D and oesophageal adenocarcinoma risk also appeared to be modified by H. pylori status, and was only evident in individuals with the highest intakes who were H. pylori-positive (OR 3·78, 95 % CI 1·44, 9·91; P for trend = 0·008) and not H. pylori-negative cases (OR 0·98, 95 % CI 0·35, 2·77; P for trend = 0·72). Ever smokers were also at increased odds of oesophageal adenocarcinoma with increasing vitamin D intakes, which appeared to be largely due to previous smoking habits. Notably though, all formal tests for the interaction between vitamin D and these stratified factors failed to reach statistical significance.
* Multivariate adjustments: age (years); sex; energy intake (kJ/d); education (years); occupation (manual/non-manual); alcohol (g/d); regular non-steroidal anti-inflammatory drug use (weekly use for at least 6 months' duration); location (Northern Ireland/Republic of Ireland); energy-adjusted saturated fat intake (g/4184 kJ per d); energy-adjusted glycaemic index intake (units/4184 kJ per d).
† BMI analysis further adjusted for H. pylori infection (seropositive/seronegative) and smoking status (current/previous/never).
‡ H. pylori analysis further adjusted for BMI 5 years before the interview and smoking status (current/previous/never).
§ Smoking status analysis further adjusted for BMI 5 years before the interview and H. pylori infection (seropositive/seronegative).
Discussion
Findings from this population-based case–control study indicate that high vitamin D intakes from foods are associated with elevated oesophageal adenocarcinoma risk but are unlikely to influence earlier stages of the carcinogenic pathway. Ca intake was not related to the development of any of the oesophageal lesions investigated. The association between dairy product intake and Barrett's oesophagus and oesophageal adenocarcinoma risk only became significant once analyses were further adjusted for gastro-oesophageal reflux symptoms. This would suggest that individuals are consuming dairy products to soothe their reflux symptoms, rather than being associated with the aetiology of these conditions.
The present findings that high vitamin D intake increases oesophageal adenocarcinoma risk are stronger than those of a US population-based case–control study, which identified a non-significant 10 % increased risk of oesophageal adenocarcinoma comparing the 75th centile with the 25th centile of intake(Reference Mayne, Risch and Dubrow27). Findings from the same study corroborate ours, in that Ca intake was also unrelated to oesophageal adenocarcinoma risk; however, an increased risk was identified for those individuals with high-fat dairy intake(Reference Mayne, Risch and Dubrow27, Reference Navarro Silvera, Mayne and Risch28). No such connection was observed for low-fat dairy intake, and so the authors have concluded that fat content was responsible for the elevated oesophageal adenocarcinoma risk(Reference Navarro Silvera, Mayne and Risch28). We were unable to disentangle low- and high-fat dairy product intake in the present study; however, all analyses were adjusted for saturated fat intake.
A biologically plausible mechanism for the overall direct association between dietary vitamin D and oesophageal adenocarcinoma risk can only be speculated. Undoubtedly, the evidence for anti-carcinogenic properties of vitamin D in the body(Reference Garland, Garland and Gorham29) far outweighs that describing an opposing role. Specific evidence to explain an increased risk of oesophageal malignancy is also lacking. A feeding study of a tumorigenic diet administered to mice in addition to vitamin D was found to have no significant effects in oesophagus tissue(Reference Koohdani, Baghdadchi and Sasani30). To fully probe the association, we first investigated common sources of vitamin D in the Irish diet to explore whether there was a nefarious effect of a particular food. Only liver was directly related to oesophageal adenocarcinoma risk; however, further adjustment for liver intake only served to strengthen the association observed for vitamin D. It is therefore unlikely that an alternative carcinogenic agent in vitamin D food sources is responsible for the association observed.
We then performed stratified analysis to determine whether the direct association was limited to a certain subgroup of individuals. The increased risk was heightened among overweight individuals, H. pylori-positive individuals and ever, particularly past, smokers, although interactions between these factors and vitamin D intake were not significant. Nevertheless, BMI is a well-established correlate of vitamin D status, with obesity linked with lower circulating 25-hydroxyvitamin D concentrations(Reference Kull, Kallikorm and Lember31–Reference Lagunova, Porojnicu and Lindberg33), which makes the present findings even more surprising. This association is possibly confounded by lower outdoor physical activity levels among overweight individuals. Similar converse results were observed for H. pylori positivity, which is a highly significant protective factor in oesophageal adenocarcinoma development(Reference Anderson, Murphy and Johnston26), and yet we observed a heightened risk among H. pylori-positive individuals with the highest vitamin D intakes. A significant positive correlation between serum vitamin D concentrations and H. pylori status has been reported in a small cross-sectional study of end-stage renal disease patients(Reference Nasri and Baradaran34), suggesting that vitamin D may modulate H. pylori activity, but this hypothesis requires considerable investigation to elucidate any potential pathways. A history of smoking also appeared to increase oesophageal adenocarcinoma risk for high vitamin D consumers. Correspondingly, results from a pooled sample of upper gastrointestinal cancers illustrated a significant reduced risk for never smokers with low vitamin D status(Reference Abnet, Chen and Chow35). Given the lack of statistical significance in interaction tests in the present study though, it is likely that the associations observed in stratified analysis are subject to differential misclassification bias and should be interpreted with caution.
Finally, it may be that nutrient–gene interactions are explaining the associations observed. The vitamin D receptor has been demonstrated to regulate bile acid metabolism in vivo (Reference Nishida, Ozeki and Makishima36), and bile acids are known to induce DNA damage that will ultimately encourage malignant growth in human gastrointestinal tissue(Reference Bernstein, Bernstein and Payne37). Therefore, the association between vitamin D receptor polymorphisms and oesophageal adenocarcinoma risk will provide a useful insight into potential mechanisms involved.
The direct association between dietary vitamin D and oesophageal adenocarcinoma risk is perhaps not entirely surprising though. An international consortium has recently published their findings from a Vitamin D Pooling Project of Rarer Cancers(Reference Helzlsouer38). Combining data of circulating 25-hydroxyvitamin D concentrations, the research team found that high vitamin D status significantly doubled the risk of pancreatic cancer(Reference Stolzenberg-Solomon, Jacobs and Arslan39), while there was no evidence of an association between 25-hydroxyvitamin D concentrations and the combined risk of oesophageal and gastric cancers, which included 104 oesophageal adenocarcinomas(Reference Abnet, Chen and Chow35). The research team deemed that ‘careful consideration’ should be taken when recommending healthy individuals to increase their vitamin D intake(Reference Stolzenberg-Solomon, Jacobs and Arslan39). Importantly, these findings do not negate the proven inverse association between vitamin D and colorectal cancer and adenoma risk(Reference Gorham, Garland and Garland16, Reference Wei, Garland and Gorham17), but suggest that more appropriate targeting of individuals is required for advising an increased intake of vitamin D from foods or conduction of supplementation trials in future.
The present study has several strengths, including its population-based design, which maximises the generalisability of the findings, and a large number of cases included. A wide variety of confounding variables were able to be accounted for in regression analyses, including H. pylori infection, which was not available in the Cohort Consortium Vitamin D Pooling Project(Reference Abnet, Chen and Chow35). However, there is the possibility that some residual confounding may be occurring in these analyses. We do not underestimate the key limitation that this study examines vitamin D intake only and does not account for additional consumption via supplements, nor serum 25-hydroxyvitamin D concentrations, and therefore does not account for cutaneous vitamin D production from sunlight. Indeed, recent data from the USA confirm that dietary vitamin D is only partially correlated with serum 25-hydroxyvitamin D concentrations and therefore does not fully reflect vitamin D status(Reference Millen, Wactawski-Wende and Pettinger40). Unfortunately, it would be inappropriate to assess vitamin D status in the present study due to the possibility of reverse causation, and it would be useful for serum 25-hydroxyvitamin D and oesophageal adenocarcinoma risk to be assessed in future prospective studies. It is possible that dietary recall bias among the oesophageal adenocarcinoma cases or dietary measurement error from the FFQ utilised may be explaining the association observed. The potential for selection bias may also have been introduced by the lower response rate among our population-based controls compared with cases. However, mean vitamin D intakes observed in the present study are in line with those measured in the representative North/South Ireland Food Consumption Survey of 3·7 μg/d(Reference O'Brien, Kiely and Harrington41), as were Ca intakes(Reference Hannon, Kiely and Harrington42). Moreover, supplements were found to contribute to only 6·2 and 11·4 % of vitamin D intakes by men and women in the same nationally representative survey, respectively(Reference O'Brien, Kiely and Harrington41). Vitamin D insufficiency is highly prevalent in healthy Irish adults(Reference O'Sullivan, Nic Suibhne and Cox43), prompting calls for a nationwide fortification or supplementation programme. Yet, it cannot be ignored that certain population subgroups may be at risk of harm if advice to increase vitamin D intake from foods was implemented on a widespread basis. These findings provide specific evidence for a research direction call on vitamin D to identify population groups that may be at risk with vitamin D intake modification. Many Barrett's oesophagus patients remain asymptomatic and therefore undiagnosed(Reference Fitzgerald44), and so any necessary restrictions should perhaps be extended to regular sufferers of gastro-oesophageal reflux symptoms. However, further research is required to confirm the direct association between vitamin D intake and oesophageal adenocarcinoma risk and to elucidate the potential mechanism involved.
In conclusion, intake of vitamin D-containing foods was directly associated with oesophageal adenocarcinoma risk in this Irish population but was not related to the risk of its precursor conditions. Ca and dairy product intakes did appear to contribute to the development of these oesophageal lesions.
Acknowledgements
The authors' contributions were as follows: H. G. M. performed the statistical analysis and drafted the manuscript; L. J. M. was the original grant holder for the FINBAR study; L. A. A. recruited the study participants and collected the study data; M. M. C. provided guidance on the statistical analysis and nutritional knowledge. All authors contributed to the writing and editing of the final manuscript. The authors state that there are no conflicts of interest. The FINBAR study group members include L. J. M., L. A. A. (Queen's University Belfast), B. T. Johnston, R. G. P. Watson, J. McGuigan, H. R. Ferguson (Belfast Health and Social Care Trust, Belfast, County Antrim, UK), S. J. Murphy (St Vincent's Hospital Dublin, Dublin, Ireland), J. V. Reynolds (St James' Hospital, Dublin, Ireland) and H. Comber (National Cancer Registry of Ireland, Cork, Ireland). We appreciate the contributions made by the study participants, their families and all who assisted with the study, particularly the Northern Ireland Cancer Registry and the National Cancer Registry Cork. We also thank Dr Christian Abnet (National Cancer Institute, Bethesda, MD, USA) for his helpful comments on the manuscript. The present study was supported by funding from the Ulster Cancer Foundation, the Northern Ireland R&D office and the Health Research Board.