Dairy products, polymorphisms in the vitamin D receptor gene and colorectal adenoma recurrence
Vitamin D receptor (VDR) activation inhibits proliferation and angiogenesis in the colorectal epithelium, and inhibits metastasis of colorectal tumors. Polymorphisms in the VDR gene alter receptor cellular levels and functioning, and may confer altered susceptibility to colorectal neoplasia. We aimed to investigate the influence of VDR polymorphisms and dietary factors impacting on vitamin D metabolism on colorectal adenoma (CRA) recurrence. Data on dietary intakes of calcium, vitamin D and dairy products were collected from 853 participants in the United Kingdom Colorectal Adenoma Prevention trial, a randomized trial of aspirin and folate for CRA recurrence prevention. The VDR Cdx2, FokI, BsmI, ApaI and TaqI polymorphisms were genotyped in 546 participants with available DNA, and gene–diet interaction analyses performed in 480. Dairy product intake was inversely related to CRA recurrence risk independent of calcium and vitamin D [relative risk (RR) = 0.64; 95% confidence intervals (CIs): 0.47–0.88, for subjects in the highest compared to lowest intake tertile, ptrend = 0.005]. Milk accounted for 60% of dairy product intake, and on analysis of milk and nonmilk dairy products separately recurrence risk in individuals in the highest tertile of milk intake was half that of lowest tertile individuals (RR = 0.52; 95% CI: 0.38–0.72, ptrend = 3.2 × 10−5), whereas nonmilk dairy products did not influence recurrence. VDR polymorphism genotypes and haplotypes did not directly alter recurrence risk, but the reduction in risk associated with high dairy product intake was confined to individuals with ApaI aA/AA genotype (pinteraction = 0.02). These findings indicate dairy products, and in particular milk, have chemopreventive activity against CRA recurrence. © 2008 Wiley-Liss, Inc.
Colorectal cancer (CRC) is the third commonest cancer worldwide after lung and breast cancer, and two-thirds of CRC occurs in developed countries.1 Up to 90% of CRCs develop from colorectal adenomas (CRAs) and colonoscopic removal of CRA reduces the incidence of CRC.2, 3 In some individuals, the recurrence rate for CRA is as high as 40–50% after 3 years,4 CRA recurrence is a surrogate marker for CRC risk and prevention of CRA recurrence is a useful endpoint for evaluating chemopreventive agents.5
Activation of vitamin D receptors (VDRs) present in many cell types, including colorectal epithelial cells, can inhibit proliferation, angiogenesis and invasiveness, and studies in animal models indicate that vitamin D supplementation can lower the incidence and metastatic potential of intestinal tumors.6–9 Ecological studies correlating sunlight exposure with cancer incidence provided the first evidence that vitamin D levels are inversely related to CRC development in humans.10 Subsequent epidemiological studies have reported lower risks of CRA and CRC in individuals with high intakes and circulating levels of vitamin D.11–14 Vitamin D plays a crucial role in calcium metabolism, and there is experimental,15, 16 epidemiologic17–19 and clinical trial20, 21 data indicating that high calcium intake may also lower colorectal neoplasia risk. Furthermore, there is evidence that vitamin D and calcium may act together to reduce CRA recurrence.22 Dairy products are an important source of calcium, and in the United States, milk is fortified with vitamin D. High dairy product intake is associated with reduced incidence of both CRA23 and CRC,24, 25 and milk consumption, which is the main component of dairy product intake for most individuals, was concluded to result in “probable decreased risk” of CRC in the recent World Cancer Research Fund/American Institute for Cancer Research report into food, nutrition, physical activity and the prevention of cancer.26
The VDR is responsible for mediating the effects of vitamin D, and functioning of the VDR has been shown to be influenced by polymorphisms in the VDR gene.27 The VDR Cdx2 G>A promoter variant maps to a binding site for an intestinal-specific transcription factor and confers altered transcriptional activity.28 The FokI translational start codon polymorphism has structural consequences for the VDR which result in a less transcriptionally potent protein.29, 30 In addition, 2 intronic polymorphisms, BsmI and ApaI, and the TaqI polymorphism in exon 9 have been reported to alter VDR expression, probably through linkage disequilibrium with variants in the 3′UTR altering VDR-mRNA stability.31
Epidemiological studies have reported associations between VDR polymorphism genotype and risk of both CRA32–36 and CRC,37–39 and in some studies, the associations were modified by dietary vitamin D and calcium intakes.32–36, 38, 40 Although not a universal finding,34, 39 carriers of variant FokI alleles have been reported to have an increased risk of CRC40 and advanced CRA32 which is enhanced in individuals with low calcium intake. The homozygous variant BsmI genotype has been found to be associated with a modest protective influence in CRC38 but not in CRA.32, 33, 36 No association between the FokI and TaqI variants and CRA recurrence was reported in one study.22 To date there have been no published reports on the relationship between ApaI and Cdx2 genotypes and colorectal neoplasia risk.
We aimed to test the hypothesis that VDR polymorphism genotype influences risk of CRA recurrence by genotyping individuals participating in a randomized intervention trial for CRA recurrence prevention. By examining the data on intakes of vitamin D, calcium and dairy products, we also investigated the influence of dietary factors on CRA recurrence, and potential gene–diet interactions.
Material and methods
The United Kingdom Colorectal Adenoma Prevention (UKCAP) trial is a recently completed multicenter, double-blind, randomized controlled trial of aspirin and folate for the prevention of CRA recurrence.41 Eligible subjects had one or more histologically confirmed CRA removed at full colonoscopy in the 6 months prior to enrollment, and were not taking regular aspirin, non-aspirin NSAID or prescribed folate supplements. Participants were randomized to aspirin alone (300 mg daily), folate alone (500 μg daily), both aspirin and folate or double placebo. The primary endpoint was the histologically confirmed recurrence of CRA or CRC.
Recurrence was ascertained at follow-up colonoscopy scheduled for 3 years after entry colonoscopy, or performed earlier if symptoms dictated. If follow-up colonoscopy was performed prior to the three-year time point and CRA or CRC was found then the subject left the trial; if no lesion was found then trial medication was continued and further colonoscopy undertaken at the three-year time point. Suspected recurrences found at follow-up colonoscopy were reviewed by the same histopathology department as the original trial entry specimen. Histopathology was performed at local hospitals without central review.
Between 1997 and 2001, 945 subjects were recruited, of which 942 were eligible and were randomized. Information on results of follow-up colonoscopy was available for 853 participants, and blood samples for extraction of germline DNA were collected from 546. Not all subjects from the original trial could be included in this molecular subprotocol as some could not be contacted, and others did not consent to DNA analysis. All subjects included in the genotyping analyses were of Caucasian ethnicity.
Compliance with trial medication was assessed at four-month intervals by research nurses during either telephone interviews or home visits. Participants were asked directly about compliance and potential side effects, and trial tablets were counted. At initial recruitment and follow-up visits, participants were asked to use acetominophen for pain relief where necessary and to avoid the use of NSAIDs. In terms of compliance, 74% of subjects continued in the aspirin arm (aspirin or aspirin-placebo) until follow-up colonoscopy, and in over 94% of these there was ≥85% compliance with prescribed tablets.
Informed consent for the study was obtained from all participants and the study was carried out with ethical review board approval in accordance with the tenets of the declaration of Helsinki.
Face-to-face interviews were conducted by dedicated research staff to gather information on lifestyle and medical factors including ethnicity, demographic information, past medical history, detailed family history data, and multivitamin and individual vitamin supplement use. All subjects also completed a semiquantitative food frequency questionnaire (FFQ) at the time of entry into the trial which addressed their dietary intake in the 5 years prior to recruitment. The FFQ employed was an adaptation of the Willett FFQ42 and has been previously validated in the UK population.43, 44 Subjects were asked how often, on average, they had consumed each of ∼130 food items during the past year. Serving sizes were specified for each food item, and there were 9 possible frequency responses: never or less than once per month; 1–3 times per month; once per week; 2–4 times per week; 5–6 times per week; once per day; 2–3 times per day; 4–5 times per day; and 6 or more times per day. The FFQ also included questions on the specific brands of breakfast cereal eaten, and addition of salt to food whilst cooking or at the table. To remove subjects whose diet was inadequately captured by the FFQ, and incorrectly completed FFQs, those with the highest and lowest 2.5% of energy intakes or questionnaires with more than 10 empty data lines were excluded. After applying these criteria, 716 (84%) and 480 (88%) subjects from the main trial and the genotyped subgroup remained for inclusion in the dietary analyses. Calcium and vitamin D intakes were determined using a nutrient conversion database, and adjusted for total energy intake using the linear residual regression method of Willett and Stampfer.45 Calcium and vitamin D intake from supplements was not energy adjusted. The number of servings of dairy products consumed per day was also calculated, with 1 serving being equivalent to 285 ml (0.5 pint) milk (including whole, semiskimmed, skimmed and dried milk), or 1 medium-sized serving of yoghurt, cheese, cottage or soft cheese, dairy dessert, cream or ice cream.
Constitutional DNA was extracted from EDTA venous blood samples using a standard salt extraction procedure, and quantified by PicoGreen (Invitrogen, Paisley, UK). VDR Cdx2, FokI, BsmI, ApaI and TaqI genotypes were generated using Taqman technology implemented on an ABI 7900HT sequence detection system (Applied Biosystems, Foster City). The assays were validated using control samples of known homozygote wild-type, heterozygote and homozygote variant genotype generated by direct sequencing. A random 10% of samples were repeated with complete concordance. Taqman genotype call rates were >95% for all polymorphisms and any samples that failed were genotyped by direct sequencing and thus genotype data was complete for all variants. Details of PCR primer sequences and reaction conditions are available on request.
The χ2 and t tests were used to compare baseline variables between the genotyped and nongenotyped subgroups of the UKCAP trial population. Genotype frequencies were tested for departure from Hardy–Weinberg equilibrium (HWE) using the χ2 test. Recurrence was defined as recurrence of one or more CRA, or CRC occurrence anytime after randomization. The relationship between genotype, or intakes of calcium, vitamin D or dairy products and risk of recurrence was assessed by means of relative risks (RRs) and 95% confidence intervals (CIs) calculated using Poisson regression with robust error variance, adjusting for age, sex, aspirin and folate intervention, and interval between entry and follow-up colonoscopy, because these variables were found to influence recurrence risk in the main UKCAP trial.46, 47 Analyses of the influence of dairy products were also adjusted for total energy intake. Tertile cut-points based on the distribution in nonrecurrence subjects in the main UKCAP trial were used to designate low (lowest tertile) medium (middle tertile) and high (highest tertile) dietary intakes, and the same cut points were used for analyses involving the genotyped subgroup. Tertiles for calcium and vitamin D intakes were calculated for intake from food only, and for intake from food and supplements combined. To test for linear trend, we modeled the tertile of nutrient intake as a continuous variable in the Poisson regression model, in which each tertile was assigned its median value.
We then considered genotypes in combination with nutrient intakes. Effect modification was evaluated by stratification on the variable of interest, and RRs within each stratum were compared. Interactions between variables with respect to CRA recurrence risk were explored using likelihood ratio testing comparing models with and without a multiplicative term for the 2 variables.
We also performed haplotype analysis for the VDR polymorphisms using genotype data from all subjects to infer chromosomal phase. Haplotypes were analyzed using the program PHASE,48 which implements the expectation-maximization algorithm to estimate haplotype frequencies and individual haplotypes. Possible haplotypes were tested for association with recurrence risk by assigning the most likely pair of haplotypes for each subject as predicted by the expectation-maximization algorithm.
The UKCAP study had ∼80% power to detect a main effect of a dietary variable grouped into tertiles of exposure, assuming a RR of recurrence of 0.65 for the highest versus the lowest tertile of exposure. The analysis of the genotyped subgroup had over 90% and ∼70% power to detect the main effect of genotype in dominant and recessive models, respectively, assuming a variant allele frequency of 0.5 and a RR of 0.5 associated with presence of the variant allele. With the same assumptions as earlier, power for detecting an interaction between a dominantly acting genetic variant and a dietary variable was 30%.
Power to detect main effects and interactions was calculated using the statistical programs POWER Version 1.30 (Epicenter Software) and Power Version 2.2.4 (National Cancer Institute, Bethesda), respectively. All other statistical analyses were undertaken using STATA, version 8.2 (Stata Corporation, College Station, TX), with the exception of the haplotype analyses above. All statistical tests were two-sided, and a p value less than 0.05 was considered significant.
Of the 546 genotyped subjects, 130 (23.8%) had ≥1 CRA, and 7 (1.3%) CRC detected at follow-up colonoscopy. Seventy individuals (12.8%) had advanced colorectal neoplasia, defined as CRAs with villous or tubulovillous features, size ≥1 cm, severe dysplasia or invasive carcinoma. In the main trial a reduced CRA recurrence risk was observed in subjects who received aspirin (RR = 0.81; 95%CI: 0.65–1.02),41 and in the genotyped subgroup a smaller recurrence risk reduction was also seen (RR = 0.92; 95%CI: 0.70–1.22). Folate intervention did not influence recurrence risk in either the main trial or the genotyped subgroup.
Subjects with DNA available for genotyping were more likely to be female compared to the nongenotyped subgroup (p = 0.03; Table I), but there were no significant differences in age, intervention group, interval between entry and follow-up colonoscopy, and CRA recurrence. Thus the genotyped individuals were a representative subgroup of the total trial population. Intake of calcium was higher in genotyped individuals, particularly when intake from both food and supplements was considered (p = 0.04), and vitamin D intake was nonsignificantly lower, whilst dairy product intake was similar. Analysis of dietary exposures in individuals within the 4 intervention arms of the trial indicated no significant differences.
Table I. Comparison of Baseline Characteristics of the Total UKCAP Trial Population and Patients With and Without DNA Available for Genotyping
|Age1,2||57.5 (9.3)||57.3 (9.3)||57.9|
|Sex|| || || |
| Male||477 (56%)||290 (53%)||189 (62%)|
| Female||376 (44%)||256 (47%)||118 (38%)|
|Intervention|| || || |
| Folate alone||215 (25.2%)||144 (26.4%)||71 (23%)|
| Aspirin alone||217 (25.4%)||131 (24.0%)||86 (28%)|
| Folate and aspirin||217 (25.4%)||135 (24.7%)||82 (27%)|
| Double placebo||204 (23.9%)||136 (24.9%)||68 (22%)|
|Colonoscopy interval3|| || || |
| Mean (months)||40.3||40.7||39.8|
| Std. dev. (months)||12.2||11.7||12.9|
| Range (months)||6–79||6–74||9–79|
|Outcome4|| || || |
| Adenoma||210 (24.6%)||130 (23.8%)||80 (26.1%)|
| Carcinoma||10 (1.2%)||7 (1.3%)||3 (1.0%)|
| Advanced neoplasia||104 (12.2%)||70 (12.8%)||34 (11.1%)|
|Number included in dietary analyses||717 (84%)||480 (88%)||237 (77%)|
|Nutrient intake excluding supplements2|| || || |
| Calcium (mg/day)||1,510 (560)||1,540 (570)||1,455 (555)|
| Vitamin D (mcg/day)||5.7 (4.0)||5.5 (3.4)||6.1 (5.0)|
|Nutrient intake including supplements2|| || || |
| Calcium (mg/day)||1,570 (560)||1,600 (600)||1,505 (582)|
| Vitamin D (mcg/day)||7.0 (5.4)||6.9 (5.3)||7.1 (5.7)|
|Dairy product intake (servings/day)2,5||2.0 (0.9)||2.1 (0.9)||2.0 (0.9)|
Intakes of calcium and vitamin D, either from food alone or from food and supplements combined, did not significantly influence CRA recurrence risk (Table II). Dairy product intake, however, was inversely associated with recurrence risk in both the main trial and the genotyped subgroup (RR = 0.64, 95%CI: 0.47–0.88, ptrend = 0.005, and RR = 0.60, 95%CI: 0.42–0.87, ptrend = 0.008, respectively, for individuals in the highest tertile of intake compared to the lowest tertile). When only advanced neoplasia recurrence was considered the inverse relationship with dairy product intake was maintained, although because of the smaller number of endpoints the statistical significance was reduced (RR = 0.61, 95%CI: 0.37–0.96, ptrend = 0.03, and RR = 0.57, 95%CI: 0.33–0.98, ptrend = 0.05, respectively). Adjustment for calcium and vitamin D intake did not modify the relationship between dairy product intake and recurrence risk. Milk accounted for ∼60% of dairy product intake, and when intakes of milk and nonmilk dairy products were considered separately the individuals in the highest tertile of milk intake had an ∼50% reduced risk of recurrence (ptrend = 3.6 × 10−5, Table II), whereas nonmilk dairy product intake did not influence recurrence risk (ptrend = 0.90). Since some subjects reported milk intakes of exactly 1.0 and 1.5 servings per day these tertile cutpoints did not divide the nonrecurrence subjects into equal tertiles. Further analyses indicated that the inverse association between milk intake and recurrence was robust to other methods of classifying milk intake including above and below median (ptrend = 4.0 × 10−6), quartiles (ptrend = 0.9 × 10−4) and quintiles (ptrend = 1.5 × 10−4), and was maintained if milk intake was included as a continuous variable in the model (RR = 0.68, 95%CI: 0.56–0.83, p = 1.6 × 10−4, for a one-serving increment in daily intake). Vitamin B2 and B12 are components of dairy products which could potentially account for the protective influence on CRA recurrence; however, intake of these vitamins did not significantly influence recurrence or modify the protective effects of dairy products. Data on smoking and family history of CRC were available for individuals who underwent genotyping, neither of which significantly influenced recurrence risk (RR = 1.07, 95%CI: 0.79–1.45 and RR = 1.24, 95%CI: 0.91–1.69, for smoking (ever vs. never) and family history, respectively) or modified the influence of dairy products.
Table II. Risk of Colorectal Neoplasia Recurrence According to Intakes of Calcium, Vitamin D and Dairy Products in the UKCAP Trial and in Patients Included in the Genotyping Analysis
|Calcium from food||Low||64/176||Ref||39/115||Ref|
| ||Medium||67/176||0.97 (0.73–1.28)||45/113||0.95 (0.67–1.34)|
| ||High||57/176||0.90 (0.67–1.21)||38/130||0.81 (0.57–1.16)|
|Calcium from food and supplements||Low||61/176||Ref||37/116||Ref|
| ||Medium||67/176||1.00 (0.75–1.33)||43/109||0.99 (0.69–1.41)|
| ||High||60/176||1.00 (0.75–1.35)||42/133||0.95 (0.66–1.36)|
|Vitamin D from food||Low||66/176||Ref||46/122||Ref|
| ||Medium||63/176||0.90 (0.67–1.19)||41/111||0.97 (0.69–1.36)|
| ||High||59/176||0.85 (0.64–1.14)||35/125||0.73 (0.51–1.04)|
|Vitamin D from food and supplements||Low||68/176||Ref||45/123||Ref|
| ||Medium||57/176||0.81 (0.61–1.08)||38/115||0.87 (0.62–1.21)|
| ||High||63/176||0.96 (0.72–1.27)||39/120||0.94 (0.67–1.34)|
|Total dairy products||Low||79/176||Ref||53/119||Ref|
| ||Medium||63/176||0.81 (0.61–1.06)||37/120||0.73 (0.52–1.02)|
| ||High||46/176||0.64 (0.47–0.88)*||32/119||0.60 (0.42–0.87)*|
| ||Medium||72/185||0.84 (0.65–1.09)||49/130||0.81 (0.59–1.11)|
| ||High||43/193||0.52 (0.38–0.72)**||28/134||0.49 (0.34–0.72)***|
|Nonmilk dairy products||Low||64/176||Ref||43/118||Ref|
| ||Medium||64/176||1.03 (0.77–1.37)||41/116||0.98 (0.69–1.39)|
| ||High||60/176||1.03 (0.76–1.38)||38/124||0.96 (0.66–1.40)|
Allele frequencies for all VDR polymorphisms were similar to previous reports in Caucasian populations.49, 50 Genotype frequencies for the Cdx2 and ApaI variants showed evidence of departure from HWE (p = 0.01 for both), whilst frequencies for the FokI, BsmI and TaqI variants were in HWE (p = 0.60, 0.12 and 0.22, respectively). When individuals with heterozygous and homozygous variant genotype for each polymorphism were compared to wild-type individuals there were no significant associations between genotype and CRA recurrence risk (Table III).
Table III. Risk of Any Colorectal Neoplasia Recurrence and Advanced Colorectal Neoplasia Recurrence According to Vitamin D Receptor Genotypes
| ||GA||38/131||0.82 (0.60–1.12)||19/150||0.78 (0.48–1.28)|
| ||AA||13/26||1.05 (0.66–1.67)||8/31||1.20 (0.62–2.32)|
| ||Ff||70/181||1.16 (0.87–1.55)||32/219||0.88 (0.57–1.36)|
| ||ff||16/65||0.92 (0.58–1.46)||8/73||0.81 (0.40–1.65)|
| ||bB||64/215||0.86 (0.64–1.17)||37/242||1.02 (0.64–1.62)|
| ||BB||23/55||1.07 (0.73–1.57)||9/69||0.86 (0.44–1.68)|
| ||aA||73/229||0.97 (0.70–1.33)||38/264||1.07 (0.65–1.75)|
| ||AA||29/87||0.95 (0.63–1.42)||15/101||1.05 (0.56–1.96)|
| ||tT||63/211||0.87 (0.64–1.17)||36/238||0.99 (0.63–1.57)|
| ||TT||23/54||1.09 (0.74–1.59)||9/68||0.85 (0.44–1.68)|
The BsmI and TaqI polymorphisms were in almost complete linkage disequilibrium (pairwise correlation coefficient r2 = 0.97), and more moderate linkage disequilibrium was also seen between the BsmI and ApaI variants (r2 = 0.63). The expectation-maximization algorithm generated 3 common (frequency > 1%) BsmI ApaI TaqI haplotypes: bAt, BaT and bat, with frequencies of 0.49, 0.39 and 0.11, respectively. Using the most frequent bAT haplotype as the reference group neither of the other common haplotypes was associated with CRA recurrence (RR = 1.01, 95%CI: 0.82–1.24, and RR = 1.07, 95%CI: 0.79–1.45, for the BaT and bat haplotypes, respectively). There was no evidence of linkage disequilibrium between the Cdx2, FokI and BsmI polymorphisms (r2 values all >0.02) and so haplotype effects were not considered; however, the joint effects of these 3 polymorphisms were analyzed on the basis of their putative functional effects.49 Individuals with Cdx2 GG, FokI ff and BsmI bb genotypes, which are associated with lower VDR transcriptional activity, lower VDR mRNA levels or lower circulating vitamin D levels, were assigned as having zero “protective” alleles, whilst those with Cdx2 AA, FokI FF and BsmI BB genotypes were assigned 6 “protective” alleles. No significant associations were observed when subjects were grouped according to the number of “protective” alleles carried and RRs were calculated for groups with 2–3 and 4–6 protective alleles using the group with 0 or 1 protective allele as the reference group.
We also examined the effects of genotypes and haplotypes following stratification by dietary intakes of calcium, vitamin D and dairy products. A significant interaction between ApaI genotype and dairy product intake was observed (pint = 0.02), with the reduction in recurrence risk associated with high dairy product intake being confined to individuals with heterozygous or homozygous AA genotype (Table IV). Although there was no significant interaction between milk intake and ApaI genotype, the protective effect of milk again tended to be confined to individuals with heterozygous or homozygous AA genotype, and the low power of our study to detect interactions means that an interaction with milk cannot be excluded. ApaI genotype did not interact with calcium (p = 0.45) or vitamin D intake (p = 0.57), and there were no significant interactions between genotypes for the other VDR polymorphisms and dietary intake in determining recurrence risk, or between aspirin or folate intervention and VDR genotypes or dietary intakes.
Table IV. Association of Colorectal Neoplasia Recurrence Risk and Dairy Product or Milk Intake Stratified by VDR ApaI Genotype
|Dairy products||aa||Recurrence/no recurrence RR3 (95%CI)||10/34 Ref||10/25 1.16 (0.61–2.20)||11/22 1.31 (0.69–2.48)|| |
| ||aA/AA||Recurrence/no recurrence RR3 (95%CI)||42/80 1.61 (0.95–2.73)||30/102 1.01 (0.58–1.77)||19/95 0.72 (0.39–1.31)||0.02|
|Milk||aa||Recurrence/no recurrence RR3 (95%CI)||8/23 Ref||14/30 1.21 (0.62–2.36)||9/28 0.86 (0.43–1.72)|| |
| ||aA/AA||Recurrence/no recurrence RR3 (95%CI)||37/71 1.45 (0.81–2.60)||35/100 1.04 (0.57–1.89)||19/106 0.60 (0.31–1.16)||0.18|
Our study is the first to report the influence of dairy product intake on risk of CRA recurrence. Our finding of an inverse relationship is similar to that reported in case–control and cohort studies investigating dairy product intake and CRA23 and CRC incidence.24, 25 In a pooled analysis of five cohort studies, high compared to low total dairy product intake was associated with an odds ratio for CRC risk of 0.62 (95%CI: 0.52–0.74).51 In the only randomized study on the effects of dairy products on colonic neoplasia, individuals with a past history of CRA who received low fat dairy product supplementation exhibited significant differences in intermediate biomarkers of colonic cancer risk, including proliferative activity of colonic epithelial cells and markers of normal cellular differentiation, compared to those who continued their conventional diets.52 Our observation that milk intake was inversely related to recurrence risk whilst nonmilk dairy products were not also mirrors results from epidemiologic studies. In a pooled analysis of 10 CRC cohort studies, milk intake was significantly inversely related to CRC risk whilst nonmilk dairy products including cheese, yoghurt, cream and ice cream were not.53 Similar to previous reports, in our study, milk was the main component of total dairy product intake, accounting for 60% on average.53 Intakes of other individual nonmilk dairy products in our study were low, which meant we had limited power to test for influences of these items, and thus it is not possible to determine whether the protective effects of dairy products are entirely limited to milk.
In the United States, milk is fortified with vitamin D, and thus dairy products are the main dietary sources of both calcium and vitamin D, and this may account for the inverse association between dairy product intake and colorectal neoplasia.22, 54 Data from both a pooled analysis of CRC cohort studies and a randomized trial of calcium supplementation for CRA recurrence prevention indicate that the protective effects of high calcium intake are seen only in individuals who also have a high intake or high circulating levels of vitamin D.22, 53 Conversely, the benefits of high vitamin D levels are restricted to individuals with high calcium intake. In the United States, dairy products are an important source of both calcium and vitamin D, and may therefore exert a greater influence than micronutrient considered in isolation. In the United Kingdom, however, milk is not fortified with vitamin D and the main dietary sources are oily fish and fortified margarines and cereals. Calcium in milk is highly bioavailable, which may make milk appear to be associated with CRC risk independent of total calcium intake.53 Alternatively, the protective effects of dairy products may be due to other components, such as conjugated linoleic acid and lactoferrin, which inhibit colonic carcinogenesis in animal models,55, 56 or the milk protein casein, which has been reported to have antimutagenic activity on the digestive tract.57
Randomized trials of calcium supplementation indicate a modest but significant protective effect against CRA recurrence (pooled RR from 3 trials combined 0.80, 95%CI: 0.68–0.93).58 In our study, calcium intake did not influence CRA recurrence, and there was no evidence of an interaction between calcium and vitamin D intake, and no influence of calcium was observed when only individuals in the high tertile of vitamin D intake were considered. Daily doses of between 1,200 and 2,000 mg of calcium supplements were used in the randomized trials, however, and it is possible that the differences in calcium intakes between individuals in the low and high tertiles of intake in our study (mean difference 900 mg) were insufficient to impact on recurrence. The relatively high mean calcium intake (1,130 mg) observed in individuals in the lowest tertile of calcium intake in our study may also have diminished the likelihood of observing an inverse relationship between calcium and recurrence. An interaction between calcium and aspirin in preventing recurrence of advanced adenomas has also been suggested,59 with the greatest benefits occurring in individuals exposed to both chemopreventive agents, but we found no evidence for this interaction in our study. The absence of an effect of dietary intake of vitamin D on CRA recurrence risk in our study corroborates results from a previous study.60 Most people in the United Kingdom obtain the majority of their vitamin D from exposure to sunlight, and in our study subjects, no data were available for sunlight exposure and we did not measure plasma vitamin D levels directly.
As with all studies of this nature, our study was dependent on the ability of the FFQ employed to reliably capture participants' dietary habits. The accuracy of FFQs varies between different micronutrients, and it is possible that this variation could account for the absence of a protective association for high calcium and vitamin D intakes compared to that observed for dairy products. The Willet FFQ has been validated in the US population for both females42 and males,61 where the correlation coefficient between the estimation of energy-adjusted calcium intake for the FFQ and the average of two 1-week diet records was 0.42–0.53. The FFQ used in our study was a version of the Willet FFQ adapted for the UK population, which has been validated in female subjects living in the Cambridge region.43, 44 In this analysis the correlation coefficient for calcium intake between the FFQ and a 16-day weighted food record was 0.50, and 46% of individuals were classified into the same quartile of calcium intake by the 2 dietary assessment tools.43 A tendency for overestimation of milk consumption by the FFQ was also noted, whilst estimates of vitamin D intake were not compared. Inaccuracies in assessment of dietary intake will tend to bias estimates of associations toward the null, and as such may account for the failure of our study to detect an association of CRA recurrence with dietary calcium and vitamin D intakes, but are unlikely to account for the significant inverse association observed for diary product intake.
In our study, genotype for VDR polymorphisms did not directly influence recurrence risk. This may reflect a genuine lack of influence in the setting of CRA recurrence, but may also have been due to low statistical power to detect small effects. Common polymorphisms are most likely to confer only modestly altered RRs of disease,62 typically in the order of 1.2–1.5, and although our study had >90% power to detect a common variant conferring a twofold increase in risk, the power to detect variants associated with lower risks was restricted. It is also possible that such polymorphisms exert their influences on colorectal carcinogenesis over longer time periods than the mean 41-month period between colonoscopies in this study. The absence of a main effect of genotype, however, does not preclude the existence of significant genotypic influences in subgroups of the population defined by exposure to relevant environmental factors that also impact on vitamin D metabolism, and in this regard our finding of a significant interaction between ApaI genotype and dairy product intake in determining recurrence risk is noteworthy. This polymorphism has not been previously genotyped in large-scale epidemiological studies of CRA or CRC risk, and because it is not in complete linkage disequilibrium with the BsmI and TaqI variants it may be a better marker for a putative functional variant elsewhere in the region. As such, the ApaI polymorphism warrants further investigation in cohort studies of CRA and CRC in which careful collection of dietary data has also been performed. It should also be noted that the Cdx2 and ApaI variants were not in HWE, indicating possible population stratification, although departure from HWE for VDR polymorphisms was also reported in a previous CRA recurrence prevention trial,22 and any genotypic influences on risk of adenoma development will lead to deviations from expected genotype frequencies in the recurrence setting.
Although dairy products appear to have protective effects against colorectal neoplasia, there is evidence to suggest that they may have an opposing influence in prostate cancer.63 A meta-analysis of twelve prospective studies investigating dairy product intake and prostate cancer risk reported a weak positive association in men with the highest intakes.63 However, a recent analysis of a US cohort of over 29,000 men, including 1,910 with prostate cancer, reported an association with dairy product and dietary calcium intake for nonaggressive disease only, and not for aggressive prostate cancer.64
Aspirin and calcium are currently the only agents that have shown consistent chemopreventive activity against colorectal neoplasia in randomized trials; however, not all individuals benefit from these interventions and some experience side effects.58, 65, 66 On the basis of the inverse relationship observed in our study and similar associations reported in epidemiological studies, we believe that dairy products, and more specifically milk, may have value as chemopreventive agents, making a randomized trial for CRA recurrence prevention an attractive proposition.
R.A.H. is a recipient of Cancer Research UK Clinical Research Training Fellowship.