Modification of the inverse association between dietary vitamin D intake and colorectal cancer risk by a FokI variant supports a chemoprotective action of Vitamin D intake mediated through VDR binding

Abstract

Vitamin D has anticarcinogenic properties and might influence colorectal cancer (CRC) risk, but the epidemiological evidence is inconsistent. Many mechanisms of action for vitamin D have been proposed, with some of them initiating via its binding to the vitamin D receptor (VDR). Using a large Scottish case–control study, we investigated (i) main associations between CRC, vitamin D and calcium dietary intake and 4 VDR single nucleotide polymorphisms (rs10735810, rs1544410, rs11568820, rs7975232) and (ii) interaction associations between the VDR variants, vitamin D and calcium intakes. Inverse and dose-dependent associations were found between CRC risk, dietary [Odds ratio (OR) = 0.77, 95% confidence intervals (CI) 0.63, 0.92, p-trend = 0.012] and total vitamin D (OR = 0.80, 95% CI 0.65, 0.98, p-trend = 0.014) intake in multivariable-adjusted logistic regression models, whereas neither calcium intake nor any of the VDR variants were associated with CRC. Additionally, we observed statistically significant interactions (case–control, case-only designs) between vitamin D and calcium intake and rs10735810 (p-interaction 0.02, 0.006, respectively). We conducted meta-analyses of cohort, case–control and serum studies that also showed an inverse association between dietary vitamin D intake and CRC (serum studies: combined OR = 0.70, 95% CI 0.56, 0.87). The evidence of interaction we report here further supports the inverse association between vitamin D mediated through binding to the VDR. © 2008 Wiley-Liss, Inc.

Vitamin D can be ingested or synthesised in the skin from inactive precursors through the action of UV sunlight. Its active form, 1α,25(OH)₂D₃ is produced after 2 hydroxylation steps in the liver and kidneys.1 The recommended dietary intake of vitamin D for people over 50 years old is 10 μg per day. Prevalence of vitamin D deficiency in Scotland might be high because of high latitude (with skin being unable to make vitamin D effectively in winter months) and routine vitamin D and calcium supplementation for the housebound (>65 years old) is recommended.2

Apart from the regulation of calcium blood concentration and absorption, vitamin D may affect colorectal cancer (CRC) risk via its binding to the vitamin D receptor (VDR)1 influencing cell proliferation, differentiation, apoptosis and angiogenesis3, 4 or affecting insulin resistance.5 Among the VDR (chromosome 12q) variants is a poly-A repeat at the 3′ untranslated region of the gene, which is in linkage disequilibrium with 4 restriction fragment length polymorphisms (RFLPs) of unknown functional effects known as BsmI (exon 8, rs1544410), ApaI (exon 8, rs7975232), TaqI (exon 9, rs731236)6 and Tru9I.7 An RFLP (FokI, rs10735810) at the first potential start site of the gene (ATG to ACG) results in a long version of the VDR protein (T [“f”] allele) or a protein shortened by 3 amino acids (C [“F”] allele).6, 8

Results from case–control9–18 and cohort studies3, 19–30 examining the associations between dietary vitamin D intake and CRC are inconclusive (web-Tables I and II). In addition, a randomised clinical trial investigating the effects of daily calcium and vitamin D supplementation for 7 years showed no effect on CRC incidence among postmenopausal women.31 However, results from serum studies31–37 are more consistent, indicating an inverse association with CRC (web-Table III). Some case–control studies have investigated the associations between VDR variants and CRC5, 18, 38–44 or adenomas7, 45–50 (web-Tables IV and V) with a few of them showing a positive association with the variant allele of FokI (rs10735810).38, 39, 42 Few studies have also performed stratified and interaction analyses by vitamin D and/or calcium analysis suggesting that the effect of VDR variants might depend on the intake of these nutrients.7, 18, 39, 45–48, 50

The aim of this large case–control study was to evaluate the associations between CRC vitamin D and/or calcium and to investigate whether any association is mediated via the VDR pathway. We examined: (i) the main associations of CRC with vitamin D intake, calcium intake and 4 single nucleotide polymorphisms (SNPs) of VDR [FokI (rs10735810), BsmI (rs1544410), rs11568820 and ApaI (rs7975232)] in the whole sample and after sex, age, site of cancer, smoking, calcium and vitamin D intake stratification; (ii) the interaction associations between the VDR SNPs and vitamin D and the VDR SNPs and calcium intake.

Material and methods

Study population

We studied cases and controls from a case–control study of CRC (Study of Colorectal Cancer in Scotland, SOCCS). Ethical approval was obtained from the MultiCentre Research Ethics committee for Scotland (MREC) and relevant Local Research Ethics committees, and all participants provided written informed consent. We aimed to recruit all incident cases (1999–2006) of adenocarcinoma of colorectum presenting to surgical units in Scotland (16–79 years old). Exclusions were patient death before ascertainment, patient too ill to participate, recurrent cases, or patient unable to give informed consent due to learning difficulties or other medical conditions. We recruited about 40% of all incident cases in Scotland over the study period. During the same period, controls were drawn randomly from a population-based register (community health index) and invited to participate. Participation rates among those approached were ∼58% for cases and an estimated 57% for controls. Questionnaire completion was sufficient for analysis in 82% of cases and 97% of controls recruited. More than 99% of the study participants were white Caucasian (see Ref. 51 for further recruitment details). Genetic analysis was performed on 3,005 cases and 3,072 controls for FokI (rs10735810) and BsmI (rs1544410) and on 2,015 cases and 2,071 controls for ApaI (rs7975232) and rs11568820. In the dietary analysis, 2,070 cases and 2,793 controls were included and gene-environment interaction analysis was performed on 1,864 cases and 2,591 controls for FokI (rs10735810) and BsmI (rs1544410) and on 1,396 cases and 1,826 controls for ApaI (rs7975232) and rs11568820.

Lifestyle and dietary data

Subjects completed 2 questionnaires: one with lifestyle and cancer information reporting their status 1 year prior diagnosis or recruitment51 and one semi-quantitative food frequency questionnaire (Scottish Collaborative Group FFQ, Version 6.41; http://www.foodfrequency.org). Its main characteristics (http://www.foodfrequency.org) and its validity for ranking macro- and micro-nutrients in younger adults52 have been previously described. Particularly, in a sample of young individuals correlation coefficients between FFQ and weighted dietary record measurements were 0.51 (men) and 0.39 (women) for vitamin D and 0.52 (men) and 0.78 (women) for calcium.52 However, this validity study was carried out in younger subjects than the subjects included in the current study and we cannot be certain of the degree of validity in this older age group. Participants were also asked to give full details of dietary supplement taken. Frequencies of consumption of the specified measures of each food were converted into nutrients using an in-house calculation programme (McCance and Widdowson's the composition of foods. 5th ed). Nutrient information on supplements was collected from the manufacturer's product information.

Genotyping data

Genotyping was undertaken in 2 phases as part of an array-based candidate gene approach, using the Illumina Infinium I Custom array platform and performed by Illumina (San Diego). In phase I, 2 VDR gene variants [FokI (rs10735810) and BsmI (rs1544410)] of 1,012 patients and 1,012 controls (<55 years old) were genotyped, whereas in phase II, 4 VDR gene variants [FokI (rs10735810), BsmI (rs1544410), rs11568820 and ApaI (rs7975232)] of 2,013 patients and 2,071 controls (21–83 years old) were genotyped. DNA samples were accurately quantified by Pico-Green™ and quality controlled prior to dispatch to San Diego. Case and control DNA samples were stored, genotyped and analysed in the same way. In addition, to avoid potential systematic batch-to-batch variation or bias, samples were anonymised as to affection status and were randomly distributed within plates. Data were subject to Illumina quality control procedures and genotypes were discarded if call rates were less than 99.5%.

Statistical analysis

The statistical package used was STATA version 10.0 (Stata Corp, College Station, Tex). Spearman rank correlation coefficients were calculated to test the correlation between vitamin D and calcium intake. The Pearson χ² test and the t-test were used to test the difference between cases and controls in terms of categorical and continuous confounding variables.

Logistic regression models (crude analysis, model I: adjusted for age, sex, deprivation score) were used to estimate the strength of association between CRC risk and the 4 SNPs [FokI (rs10735810), BsmI (rs1544410), rs11568820 and ApaI (rs7975232)]. The deprivation score (Carstairs and Morris Index) concerns material deprivation derived from 4 census variables and includes 7 scores (1 least deprived, 7 most deprived) and it is considered a proxy for socioeconomic status.53 Odds ratios (ORs) and 95% confidence intervals (CI) were calculated after sex, age (≤55 years old, >55 years old), cancer site (colon cancer, rectal cancer), smoking (non and former smokers, current smokers), vitamin D intake (low, high intake) and calcium intake (low, high intake) stratification. In addition, the associations between CRC and residually54 energy adjusted vitamin D and calcium intake (dietary and total intake) were tested in 2 logistic regression models (model I, model II) in the whole sample and after sex, age, cancer site, smoking, vitamin D and calcium dietary intake stratification. Participants were divided into quintiles based on the combined distributions of cases and controls. Model I was corrected for age, sex and Carstairs Deprivation Index. Model II was further corrected for energy (MJ/day, continuously, included as a covariate to reduce random error55), fibre intake (g/day, energy adjusted), smoking (non-smoker, former smoker and current smoker), body mass index (BMI, kg/m², continuous), regular Non Steroidal Anti-Inflammatory Drug (NSAID) intake (yes vs. no), family history of cancer, physical activity (hours of cycling and other sports activities, 4 groups). Interaction associations were estimated by using 2 interaction analysis designs that are compatible (case–control and case-only). In the case–control design, both the genetic (i.e., one of the VDR SNPs) and the environmental (i.e., vitamin D intake) factors were assessed for all the cases and controls and interaction was tested by fitting an interactive model and its nested multiplicative one. In the case-only design, the genetic and environmental factors are assessed only in the cases (independence between genotypes and environment exposure is assumed). ORs in a case-only study are interpreted as a synergy index on a multiplicative scale.56 For both designs, the reference category used was homozygotes of the wild-type allele of a low vitamin D or calcium dietary intake.

Meta-analysis of published studies

We identified published cohort and case–control studies (keywords: Colorectal Neoplasms [MeSH], adenomas, Vitamin D [MeSH], vitamin D receptor, VDR) searching MEDLINE. References from these publications were also examined to identify previous studies. The inclusion criteria were (i) cohort or case–control studies examining the associations between CRC (primary endpoint) and vitamin D dietary intake (providing at least 3 categories of the exposure) and/or 25(OH)D serum concentration and/or 1α,25(OH)₂D₃ serum concentration and/or VDR polymorphisms; (ii) limited to humans and publications in English published until May 2007; (iii) providing relative risks (RRs) (ORs for the case–control studies) and 95% CI or information allowing us to calculate them. Review Manager software (4.2) was used to perform meta-analyses of 11 published cohort studies and 9 case–control studies (including the current study) to compare high versus low dietary intakes of vitamin D and 7 serum nested case–control studies to compare high versus low serum levels of vitamin D metabolites (25(OH)D). In addition, we performed meta-analyses of 5 studies on CRC and 2 on colorectal adenomas to compare the FokI (rs10735810) genotypes and meta-analyses of 4 studies on CRC and 3 studies on colorectal adenomas to compare the BsmI (rs1544410) genotype. Fixed effect models were adopted when there was no evidence for heterogeneity, which was quantified using a χ²-test and I² score.

Results

The variant allele frequencies in the control sample of 3 SNPs [FokI (rs10735810), ApaI (rs7975232) and rs11568820] were under Hardy-Weinberg equilibrium (p > 0.05), but BsmI (rs1544410) was not (p = 0.0122). There were no significant differences between the cases and the controls in terms of age, sex, calcium intake (energy adjusted), BMI, smoking and area deprivation index (Table I). Control individuals reported a significantly lower total daily energy, higher vitamin D (energy adjusted) and fibre (energy adjusted) intake (p < 0.00005, p = 0.0016, p < 0.0005), a significant higher regular intake of NSAIDs (p < 0.0005) and of supplements (p = 0.004), a more active lifestyle (p = 0.031) and had a lower family history risk of CRC (p < 0.00005) than cases (Table I). In Tables II and III, the distributions of dietary, demographic, lifestyle and other risk factors for different levels of vitamin D and calcium intake among the controls are presented.

Table I. Demographic Characteristics and Lifestyle Factors for the Study Population

Variables	Cases1 (n = 2,070)	Controls1 (n = 2,793)	p-value2
1 Mean values and in parentheses standard deviations for continuous variables; number of subjects and in parenthesis percentages for categorical variables. 2 p-values from the Pearson χ^2 for categorical variables; from t-test for continuous variables. 3 Energy adjusted. 4 Smokers were defined as individuals who have smoked at least 1 cigarette per day. 5 Frequent use was defined as an intake of at least 4 days per week for at least 1 month. 6 Intake of supplements with vitamin D. 7 Locally based deprivation index (Carstairs deprivation index) based on the 2001 Census data; 7 categories ranging from very low deprivation (DEPCAT = 1) to very high deprivation (DEPCAT = 7).
Dietary variables
Energy intake (MJ/day)	11.27 (4.37)	10.66 (3.95)	<0.00005
Vitamin D intake3 (μg/day)	4.35 (2.52)	4.60 (2.79)	0.0016
Calcium intake3 (g/day)	1.11 (0.26)	1.11 (0.28)	0.53
Fibre intake3 (g/day)	21.38 (5.85)	22.23 (6.18)	<0.00005
Demographic variables
Age (years)	62.0 (10.78)	62.4 (10.52)	0.19
Sex
Men	1,185 (57.3)	1,591 (57.0)
Women	885 (42.7)	1,202 (43.0)	0.84
Lifestyle and other variables
BMI (kg/m^2)	26.64 (4.42)	26.74 (4.63)	0.45
Family history of cancer
Low	1,616 (82.8)	2,709 (98.9)
Moderate/high	335 (17.2)	30 (1.1)	<0.00005
Smoking
No	873 (42.7)	1,200 (43.4)
Former	829 (40.6)	1,059 (38.3)
Current4	342 (16.7)	508 (18.3)	0.17
Frequent NSAIDs intake
No	1,450 (70.5)	1,764 (63.8)
Yes5	607 (29.5)	1,001(36.2)	<0.0005
Physical activity (sport activities and cycling)
0 hr/day	1,139 (58.0)	1,456 (54.1)
0–3.5 hr/day	486 (24.8)	701 (26.0)
3.5–7 hr/day	204 (10.4)	339 (12.6)
>7 hr/day	135 (6.9)	197 (7.3)	0.031
Supplement intake6
No	1,595 (77.1)	2,051 (73.4)
Yes	475 (22.9)	742 (26.6)	0.004
Site of cancer
Colon cancer	1,217
Rectal cancer	874
Deprivation7
1	194 (9.4)	260 (9.3)
2	436 (21.1)	573 (20.5)
3	535 (25.9)	762 (27.3)
4	489 (23.6)	648 (23.2)
5	220 (10.6)	295 (10.6)
6	140 (6.8)	178 (6.4)
7	55 (2.7)	76 (2.7)	0.96

Table II. Distribution of Dietary, Lifestyle and Demographic Risk Factors for Different Levels of Vitamin D Intake (Vitamin D Quintiles) Among Controls

Variables1	Vitamin D intake (μg/day)
	<2.51	2.51–3.39	3.40–4.40	4.41–5.99	≥6.00	p-value2
1 Mean values and in parentheses standard deviations for continuous variables; number of subjects and in parenthesis percentages for categorical variables. 2 p-values calculated from ANOVA test (continuous variables) or chi-square test (categorical variables). 3 Energy adjusted. 4 Smokers were defined as individuals who have smoked at least 1 cigarette per day. 5 Frequent use was defined as an intake of at least 4 days per week for at least 1 month. 6 Intake of supplements with vitamin D. 7 Locally based deprivation index (Carstairs deprivation index) based on the 2001 Census data; 7 categories ranging from very low deprivation (DEPCAT = 1) to very high deprivation (DEPCAT = 7).
Dietary variables
Energy intake (MJ/day)	10.3 (5.20)	10.9 (4.07)	10.6 (3.32)	10.6 (3.50)	10.8 (3.41)	0.10
Calcium intake3 (g/day)	1.06 (0.33)	1.11 (0.27)	1.14 (0.27)	1.12 (0.25)	1.12 (0.26)	<0.00005
Fibre intake3 (g/day)	22.4 (6.99)	22.1 (6.17)	21.6 (5.77)	22.3 (5.76)	22.8 (6.10)	0.04
Demographic variables
Age (years)	60.7 (11.0)	61.2 (10.4)	62.8 (10.8)	63.4 (10.2)	63.6 (9.9)	<0.00005
Sex
Men	302 (55.9)	317 (58.0)	316 (57.9)	326 (57.5)	330 (55.7)
Women	238 (44.1)	230 (42.0)	230 (42.1)	241 (42.5)	263 (44.3)	0.89
Lifestyle and other variables
BMI (kg/m^2)	26.9 (4.97)	26.8 (4.52)	26.6 (4.63)	27.0 (4.55)	26.4 (4.46)	0.25
Family history of cancer
Low	522 (99.1)	529 (98.7)	530 (98.7)	548 (99.1)	580 (99.0)
Moderate/high	5 (0.9)	7 (1.3)	7 (1.3)	5 (0.9)	6 (1.0)	0.95
Smoking
No	231 (43.3)	247 (45.7)	227 (42.1)	243 (43.2)	252 (42.6)
Former	187 (35.0)	183 (33.9)	224 (41.6)	217 (38.5)	248 (42.0)
Current4	116 (21.7)	110 (20.4)	88 (16.3)	103 (18.3)	91 (15.4)	0.03
Frequent NSAIDs intake
No	339 (63.6)	350 (64.7)	352 (65.2)	335 (59.7)	388 (65.8)
Yes5	194 (36.4)	191 (35.3)	188 (34.8)	226 (40.3)	202 (34.2)	0.22
Physical activity (sport activities and cycling)
0 hr/day	304 (58.7)	295 (55.5)	284 (53.5)	281 (51.9)	292 (51.0)
0–3.5 hr/day	121 (23.3)	140 (26.4)	146 (27.5)	147 (27.2)	147 (25.7)
3.5–7 hr/day	62 (12.0)	52 (9.8)	69 (13.0)	75 (19.9)	81 (14.2)
>7 hr/day	31 (6.0)	44 (8.3)	32 (6.0)	38 (7.0)	52 (9.1)	0.15
Supplement intake6
No	414 (76.7)	405 (74.0)	407 (74.5)	412 (72.7)	413 (69.7)
Yes	126 (23.3)	142 (26.0)	139 (25.5)	155 (27.3)	180 (30.3)	0.10
Deprivation7
1	41 (7.6)	51 (9.3)	48 (8.8)	50 (8.8)	70 (11.8)
2	104 (19.2)	90 (16.6)	117 (21.4)	133 (23.5)	129 (21.7)
3	144 (26.7)	153 (28.0)	148 (27.1)	144 (25.4)	173 (29.2)
4	128 (23.7)	128 (23.4)	132 (24.2)	134 (23.6)	126 (21.3)
5	67 (12.4)	73 (13.4)	58 (10.6)	53 (9.3)	44 (7.4)
6	40 (7.4)	37 (6.8)	32 (5.9)	36 (6.4)	33 (5.6)
7	16 (3.0)	14 (2.5)	11 (2.0)	17 (3.0)	18 (3.0)	0.12

Table III. Distribution of Dietary, Lifestyle and Demographic Risk Factors for Different Levels of Calcium Intake (Calcium Quintiles) Among Controls

Variables1	Calcium intake (g/day)
	<0.89	0.89–1.03	1.03–1.15	1.15–1.32	≥1.32	p-value2
1 Mean values and in parentheses standard deviations for continuous variables; number of subjects and in parenthesis percentages for categorical variables. 2 p-values calculated from ANOVA test (continuous variables) or chi-square test (categorical variables). 3 Energy adjusted. 4 Smokers were defined as individuals who have smoked at least 1 cigarette per day. 5 Frequent use was defined as an intake of at least 4 days per week for at least 1 month. 6 Intake of supplements with vitamin D. 7 Locally based deprivation index (Carstairs deprivation index) based on the 2001 Census data; 7 categories ranging from very low deprivation (DEPCAT = 1) to very high deprivation (DEPCAT = 7).
Dietary variables
Energy intake (MJ/day)	9.96 (5.62)	10.6 (3.92)	11.0 (3.49)	11.1 (3.15)	10.6 (2.82)	<0.00005
Vitamin D intake3 (μg/day)	4.31 (3.11)	4.66 (2.65)	4.54 (2.38)	4.80 (2.87)	4.66 (2.87)	0.05
Fibre intake3 (g/day)	20.8 (6.14)	21.7 (5.52)	22.3 (5.48)	23.3 (6.76)	23.2 (6.49)	<0.00005
Demographic variables
Age (years)	61.6 (10.7)	62.8 (10.7)	63.3 (10.3)	61.9 (11.2)	62.4 (10.2)	0.05
Sex
Men	345 (61.1)	367 (64.2)	313 (60.0)	305 (53.7)	261 (46.1)
Women	220 (38.9)	205 (35.8)	209 (40.0)	263 (46.3)	305 (53.9)	<0.0005
Lifestyle and other variables
BMI (kg/m^2)	26.7 (4.94)	26.9 (4.39)	27.0 (4.81)	26.9 (4.33)	26.2 (4.62)	0.05
Family history of cancer
Low	546 (98.9)	560 (99.5)	503 (98.8)	549 (98.6)	551 (98.8)
Moderate/high	6 (1.1)	3 (0.5)	6 (1.2)	8 (1.4)	7 (1.2)	0.66
Smoking
No	235 (42.0)	213 (37.5)	229 (44.4)	256 (45.6)	267 (47.6)
Former	191 (34.1)	238 (41.9)	193 (37.4)	225 (40.0)	212 (37.8)
Current4	134 (23.9)	117 (20.6)	94 (18.2)	82 (14.4)	82 (14.6)	<0.0005
Frequent NSAIDs intake
No	353 (63.0)	358 (63.4)	324 (62.9)	370 (65.7)	359 (63.9)
Yes5	207 (37.0)	207 (36.6)	191 (37.1)	193 (34.3)	203 (36.1)	0.87
Physical activity (sport activities and cycling)
0 hr/day	331 (61.1)	297 (53.7)	288 (57.6)	263 (47.9)	277 (50.5)
0–3.5 hr/day	126 (23.3)	147 (26.6)	121 (24.2)	160 (29.2)	147 (26.8)
3.5–7 hr/day	56 (10.3)	58 (10.5)	60 (12.0)	82 (14.9)	83 (15.1)
>7 hr/day	29 (5.3)	51 (9.2)	31 (6.2)	44 (8.0)	42 (7.6)	0.001
Supplement intake6
No	426 (75.4)	425 (74.3)	397 (76.1)	405 (71.3)	398 (70.3)
Yes	139 (24.6)	147 (25.7)	125 (23.9)	163 (28.7)	168 (29.7)	0.12
Deprivation7
1	52 (9.2)	49 (8.6)	42 (8.1)	56 (9.9)	61 (10.8)
2	83 (14.7)	123 (21.5)	100 (19.2)	137 (24.1)	130 (23.0)
3	145 (25.6)	170 (29.7)	159 (30.5)	158 (27.8)	130 (23.0)
4	153 (27.1)	116 (20.3)	109 (20.9)	119 (20.9)	151 (26.7)
5	66 (11.7)	64 (11.2)	61 (11.7)	58 (10.2)	46 (8.1)
6	49 (8.7)	36 (6.3)	30 (5.8)	27 (4.8)	36 (6.3)
7	17 (3.0)	14 (2.4)	20 (3.8)	13 (2.3)	12 (2.1)	0.001

The 5 main food sources of vitamin D were fried oily fish (22.9% of total dietary intake), smoked oily fish (10.0%), grilled, poached, baked or pickled oily fish (6.8%), tuna (tinned or fresh) (5.4%) and mince or meat sauce, e.g., Bolognese (4.5%). 742 (26.6%) controls and 475 (22.9%) cases have been taking vitamin D supplements and the typical dose was 5 μg/day. The 5 main food sources for calcium were semi-skimmed milk (18.0%), full fat hard cheese (8.7%), full fat milk (5.8%), low fat yoghurt (4.5%) and skimmed milk (3.6%). 163 (5.8%) controls and 99 (4.8%) cases have been taking calcium supplements and the typical dose was 200 mg/day. Spearman's correlation coefficient for correlation between vitamin D and calcium intake (energy adjusted) was 0.07 (p value < 0.00005).

Intakes of dietary and of total vitamin D showed statistically significant inverse associations with CRC risk in both models I and II (model I: p for trend 0.014, 0.001; model II: p for trend 0.012, 0.014; respectively) with a 20–23% reduction in risk for those of high versus those of low intake (model I: OR = 0.81, 95% CI 0.67, 0.97; OR = 0.77, 95% CI 0.64, 0.92; model II: OR = 0.77, 95% CI 0.63, 0.94, OR = 0.80, 95% CI 0.65, 0.98; respectively) (Table IV). Intakes of dietary and of total calcium were not associated with CRC risk in either model I or II (model I: p for trend 0.90, 0.77; model II: p for trend 0.86, 0.62, respectively) (Table IV).

Table IV. Logistic Regression Models of the Effect of Vitamin D and Calcium Intake from Dietary Sources and of Total (Diet and Supplement) Intake on Colorectal Cancer in the Whole Population; Logistic Regression Models of the Effect of four Single Nucleotide Polymorphisms on Colorectal Cancer in the Whole Sample and after Vitamin D and Calcium Dietary Intake Stratification

1 Adjusted for energy (residual method), age, sex and deprivation score. 2 Adjusted for energy (residual method), energy (included as a covariate) age, sex, deprivation score, fibre intake (energy adjusted), family history of cancer, body mass index, smoking, (non-smoker, current smoker, former smoker), NSAIDs intake (yes/no), physical activity. 3 Adjusted for age, sex and deprivation area score.
Dietary data	Cases	Controls	Model I1		Model II2
			OR	95% CI	OR	95% CI
Vitamin D intake (μg/day)
Dietary intake
<2.51	433	540	1.00		1.00
2.51–3.39	426	547	0.98	0.82, 1.17	0.95	0.77, 1.16
3.40–4.40	426	546	0.98	0.82, 1.18	0.94	0.77, 1.15
4.41–5.99	406	567	0.90	0.75, 1.08	0.89	0.73, 1.10
≥6.00	379	593	0.81	0.67, 0.97	0.77	0.63, 0.94
p for trend (quintiles)			0.014		0.012
p for trend (continuous)			0.003		0.001
Total intake
<2.76	436	537	1.00		1.00
2.76–3.92	440	533	1.03	0.86, 1.23	0.96	0.79, 1.18
3.93–5.54	422	550	0.95	0.80, 1.14	0.93	0.76, 1.14
5.54–8.30	401	572	0.87	0.73, 1.04	0.84	0.69, 1.03
≥8.31	371	601	0.77	0.64, 0.92	0.80	0.65, 0.98
p for trend (quintiles)			0.001		0.014
p for trend (continuous)			<0.0005		0.004
Calcium intake mg/day
Dietary intake
<887.6	408	565	1.00		1.00
887.6–1026.1	401	572	0.97	0.81, 1.17	0.92	0.75, 1.12
1026.2–1153.2	450	522	1.20	1.00, 1.44	1.06	0.86, 1.29
1153.3–1319.4	405	568	0.99	0.83, 1.19	0.96	0.79, 1.18
≥1319.5	406	566	1.00	0.84, 1.20	0.96	0.78, 1.17
p for trend (quintiles)			0.90		0.86
p for trend (continuous)			0.59		0.61
Total intake
<892.4	409	564	1.00		1.00
892.4–1035.8	407	407	0.99	0.83, 1.19	0.91	0.74, 1.11
1035.9–1163.8	440	532	1.15	0.96, 1.37	1.01	0.82, 1.23
1163.9–1341.0	423	550	1.06	0.89, 1.27	1.02	0.83, 1.24
≥1341.1	391	581	0.93	0.78, 1.12	0.89	0.72, 1.09
p for trend (quintiles)			0.77		0.62
p for trend (continuous)			0.41		0.45
Genetic data	Cases	Controls	Crude analysis		Model I3
			OR	95% CI	OR	95% CI
rs10735810 (FokI; Ex4+4T>C)
Whole sample
CC (FF)	1,051	1,148	1.00		1.00
CT (Ff)	1,450	1,413	1.12	1.00, 1.25	1.12	1.00, 1.25
TT (ff)	439	477	1.01	0.86, 1.17	1.02	0.87, 1.19
			p for trend 0.48		p for trend 0.42
Vitamin D
<3.86 μg/day
CC (FF)	372	447	1.00		1.00
CT (Ff)	457	595	0.92	0.77, 1.11	0.93	0.77, 1.11
TT (ff)	147	212	0.83	0.65, 1.07	0.84	0.65, 1.08
			p for trend 0.15		p for trend 0.16
≥3.86 μg/day
CC (FF)	314	530	1.00		1.00
CT (Ff)	432	609	1.20	0.99, 1.44	1.19	0.99, 1.44
TT (ff)	142	198	1.21	0.94, 1.56	1.21	0.93, 1.56
			p for trend 0.07		p for trend 0.08
			p for interaction 0.02		p for interaction 0.02
Calcium
<1.1 g/day
CC (FF)	363	500	1.00		1.00
CT (Ff)	425	574	1.02	0.85, 1.23	1.01	0.84, 1.22
TT (ff)	130	210	0.85	0.66, 1.10	0.85	0.66, 1.10
			p for trend 0.36		p for trend 0.33
≥1.1 g/day
CC (FF)	323	477	1.00		1.00
CT (Ff)	464	630	1.09	0.90, 1.31	1.09	0.90, 1.31
TT (ff)	159	200	1.17	0.91, 1.51	1.17	0.91, 1.51
			p for trend 0.19		p for trend 0.20
			p for interaction 0.006		p for interaction 0.006
rs1544410 (BsmI; Intron 8, 60890G>A)
Whole sample
AA (BB)	469	476	1.00		1.00
AG (Bb)	1,431	1,536	0.95	0.82, 1.09	0.96	0.82, 1.11
GG (bb)	1,084	1,026	1.07	0.92, 1.25	1.08	0.92, 1.25
			p for trend 0.16		p for trend 0.16
Vitamin D
<3.86 μg/day
AA (BB)	147	184	1.00		1.00
AG (Bb)	472	650	0.91	0.71, 1.16	0.91	0.71, 1.16
GG (bb)	357	420	1.06	0.82, 1.38	1.05	0.81, 1.37
			p for trend 0.35		p for trend 0.39
≥3.86 μg/day
AA (BB)	150	209	1.00		1.00
AG (Bb)	432	666	0.90	0.71, 1.15	0.90	0.71, 1.15
GG (bb)	306	462	0.92	0.72, 1.19	0.92	0.72, 1.19
			p for trend 0.65		p for trend 0.66
			p for interaction 0.68		p for interaction 0.70
Calcium
<1.1 g/day
AA (BB)	135	198	1.00		1.00
AG (Bb)	445	633	1.03	0.80, 1.32	1.03	0.80, 1.33
GG (bb)	338	453	1.09	0.84, 1.42	1.10	0.85, 1.43
			p for trend 0.45		p for trend 0.41
≥1.1 g/day
AA (BB)	162	195	1.00		1.00
AG (Bb)	459	683	0.81	0.64, 1.03	0.80	0.63, 1.02
GG (bb)	325	429	0.91	0.71, 1.17	0.90	0.70, 1.16
			p for trend 0.82		p for trend 0.77
			p for interaction 0.44		p for interaction 0.45
rs7975232 (ApaI, IVS10-49G>T)
Whole sample
TT (AA)	527	571	1.00		1.00
TG (Aa)	1,030	1,050	1.06	0.92, 1.23	1.07	0.93, 1.24
GG (aa)	439	416	1.14	0.96, 1.37	1.14	0.96, 1.37
			p for trend 0.14		p for trend 0.14
Vitamin D
<3.86 μg/day
TT (AA)	185	230	1.00		1.00
TG (Aa)	355	430	1.03	0.81, 1.30	1.04	0.82, 1.32
GG (aa)	170	170	1.24	0.93, 1.66	1.22	0.92, 1.64
			p for trend 0.15		p for trend 0.18
≥3.86 μg/day
TT (AA)	183	280	1.00		1.00
TG (Aa)	353	506	1.07	0.85, 1.34	1.07	0.85, 1.34
GG (aa)	150	210	1.09	0.83, 1.45	1.10	0.83, 1.46
			p for trend 0.52		p for trend 0.48
			p for interaction 0.32		p for interaction 0.35
Calcium
<1.1 g/day
TT (AA)	170	243	1.00		1.00
TG (Aa)	355	467	1.09	0.86, 1.38	1.09	0.86, 1.39
GG (aa)	152	191	1.14	0.85, 1.52	1.13	0.85, 1.52
			p for trend 0.38		p for trend 0.39
≥1.1 g/day
TT (AA)	198	267	1.00		1.00
TG (Aa)	353	469	1.01	0.81, 1.28	1.02	0.81, 1.28
GG (aa)	168	189	1.20	0.91, 1.58	1.21	0.91, 1.59
			p for trend 0.22		p for trend 0.21
			p for interaction 0.49		p for interaction 0.50
rs11568820 (-29648A>G)
Whole sample
GG	1,226	1,308	1.00		1.00
AG	678	643	1.12	0.98, 1.29	1.13	0.99, 1.29
AA	92	86	1.14	0.84, 1.55	1.15	0.85, 1.57
			p for trend 0.08		p for trend 0.06
Vitamin D
<3.86 μg/day
GG	456	548	1.00		1.00
AG	220	241	1.10	0.88, 1.37	1.13	0.91, 1.41
AA	34	41	1.00	0.62, 1.60	1.00	0.62, 1.61
			p for trend 0.58		p for trend 0.47
≥3.86 μg/day
GG	406	634	1.00		1.00
AG	249	326	1.19	0.97, 1.47	1.19	0.97, 1.47
AA	31	36	1.34	0.82, 2.21	1.33	0.81, 1.18
			p for trend 0.06		p for trend 0.06
			p for interaction 0.75		p for interaction 0.74
Calcium
<1.1 g/day
GG	405	577	1.00		1.00
AG	234	279	1.19	0.96, 1.48	1.21	0.97, 1.50
AA	38	45	1.20	0.77, 1.89	1.16	0.73, 1.82
			p for trend 0.11		p for trend 0.12
≥1.1 g/day
GG	457	605	1.00		1.00
AG	235	288	1.08	0.87, 1.33	1.10	0.89, 1.35
AA	27	32	1.12	0.66, 1.89	1.14	0.67, 1.94
			p for trend 0.44		p for trend 0.36
			p for interaction 0.98		p for interaction 0.98

ORs, 95% CIs and p-values for trend for CRC risk were estimated as before for groups stratified by sex, age, cancer site, smoking (no or former, current smokers), calcium intake (low, high intake) and vitamin D intake (low, high intake) for models I and II (data not shown). Associations between vitamin D and CRC were non-significantly stronger among males than among females (model II, high vs. low quintile of total intake OR = 0.73, 95% CI 0.56, 0.97, p-trend = 0.023; OR = 0.91, 95% CI 0.67, 1.24, p-trend = 0.33; respectively); among those aged >55 years old than among those aged ≤55 years old (model II, high vs. low quintile of total intake, OR = 0.75, 95% CI 0.59, 0.96, p-trend = 0.011; OR = 1.05, 95% CI 0.70, 1.57, p-trend = 0.57; respectively) and among non or former smokers than among smokers (model II, high vs. low quintile of total intake, OR = 0.77, 95% CI 0.62, 0.97, p-trend = 0.016; OR = 0.95, 95% CI 0.55, 1.62, p-trend = 0.53; respectively). For model I and II analysis, calcium (dietary and total) was not associated with CRC after sex, age, smoking and vitamin D intakes stratification. In contrast, there was a non-statistically significant inverse association with rectal cancer but not with colon cancer (model II, high vs. low quintile of total intake, OR = 0.73, 95% CI 0.55, 0.96, p-trend = 0.09; OR = 1.06, 95% CI 0.83, 1.37, p-trend = 0.73; respectively). Finally, joint vitamin D and calcium intake was not significantly associated with CRC (Model I p_interaction = 0.26. Model II p_interaction = 0.17; data not shown).

We further investigated the associations between CRC, vitamin D and calcium after genotype stratification. The associations between vitamin D and CRC after FokI (rs10735810) genotype stratification were as follows: high versus low total vitamin D intake for FF: OR = 0.67, 95% CI 0.51, 0.89, p-trend = 0.002; for Ff: OR = 0.87, 95% CI 0.67, 1.11, p-trend = 0.18; for ff: OR = 0.91, 95% CI 0.60, 1.40, p-trend = 0.71 (data not shown). In addition, calcium intake was not associated with CRC for FF (CC) and Ff (CT) individuals, but it was positively associated for the ff (TT) individuals (high vs. low total calcium intake for FF: OR = 0.85, 95% CI 0.64, 1.14, p-trend = 0.56; for Ff: OR = 0.93, 95% CI 0.73, 1.19, p-trend = 0.58; for ff: OR = 1.70, 95% CI 1.09, 2.67, p-trend = 0.020).

None of the variants were significantly associated with CRC in the whole sample and there were no differences between the crude analysis and Model I (adjusted for age, sex and deprivation area score) (Table IV). There were no significant differences in the associations between CRC and the 4 VDR SNPs after sex, age and cancer site stratification (data not shown). Although after age stratification, the GG genotype of the ApaI (rs7975232) SNP was associated with a significant increase risk of CRC (OR = 2.37, 95% CI 1.28, 4.40, p-trend = 0.012) for the individuals younger than 55 years old (data not shown).

Associations between the 4 VDR SNPs and CRC were examined after vitamin D and calcium stratification. In addition, interaction relationships between VDR SNPs and vitamin D and calcium were investigated. FokI (rs10735810) interacted significantly with vitamin D and calcium dietary intake (case–control design: p for interaction 0.02 and 0.006 respectively) (Table IV, Fig. 1). Case-only analysis confirmed these findings showing a significant interaction between high calcium dietary intake and the variant genotype of FokI (rs10735810) (OR_interaction (95% CI), p for interaction: 1.34 (1.01, 1.77), 0.038; data not shown). Stratified analysis showed a non-significant, slightly lower risk of the variant genotype for the individuals of low vitamin D or low calcium dietary intake (crude analysis: OR = 0.83, 95% CI 0.65, 1.07, p-trend = 0.15; OR = 0.85, 95% CI 0.65, 1.10, p-trend = 0.36; respectively) and an increase risk of the variant genotype for the individuals of high vitamin D or calcium dietary intake (crude analysis: OR = 1.21, 95% CI 0.94, 1.56, p-trend = 0.07; OR = 1.17, 95% CI 0.91, 1.51, p-trend = 0.19; respectively) (Table IV). None of the other 3 SNPs interacted significantly with either vitamin D or calcium dietary intake (Table IV).

Figure 1. (a) Effect of FokI (rs10735810) SNP and vitamin D dietary intake on colorectal cancer risk; reference group wild-type individuals (FF) with a low vitamin D intake (p for interaction 0.02). (b) Effect of FokI (rs10735810) SNP and calcium dietary intake on colorectal cancer risk; reference group wild-type individuals (FF) with a low calcium intake (p for interaction 0.006).

The combined effect of high intake of vitamin D on CRC in the data set of 11 cohort studies and 9 case–control studies showed a weak significant inverse association for the cohort studies and a weak non-significant inverse association for the case–control studies (combined RR = 0.91, 95% CI 0.84, 1.00; combined OR = 0.90, 95% CI 0.80, 1.02; respectively) (web-Figs. 1 and 2). The combined effect of high 25(OH)D in the blood on CRC in the data set of 7 nested case–control studies showed a significant inverse association with CRC (combined OR = 0.70, 95% CI 0.56, 0.87) (web-Fig. 3). The combined effect of the FokI (rs10735810) ff (TT) genotype versus the FF (CC) genotype was not significant in either CRC or colorectal adenomas studies (web-Figs. 4 and 6). In contrast, BsmI (rs1544410) GG genotype was associated with an increased CRC risk after combining the results from 4 studies (combined OR = 1.18, 95% CI 1.04, 1.33), but there was no association with colorectal adenomas (web-Figs. 5 and 7).

Discussion

In this large case–control study none of the 4 examined SNPs was associated with CRC. However, FokI (rs10735810), a SNP that affects VDR function, showed a statistically significant interaction with vitamin D and calcium dietary intake. Individuals homozygous for the variant and who had a high dietary intake of vitamin D or calcium had a higher risk compared with those homozygous for the wild type with a high dietary intake of vitamin D and calcium. The observed interaction was not due to recall bias, because a case-only analysis confirmed these findings showing a significant interaction between high calcium dietary intake and the variant genotype of FokI (rs10735810).

The F (C) allele of FokI (rs10735810) has been found to result in a 3 amino acid shorter version of the VDR protein that is more efficient in binding vitamin D than the longer version coded by the f (T) allele. Therefore, higher vitamin D or calcium intake might enhance its activity.6 Both vitamin D and calcium interact biologically with VDR and it has been suggested that they act together in their anticarcinogenic properties, with their effects being mainly at the earlier stages of carcinogenesis (adenomas).46 Ingles et al.47 showed that the f (T) allele was inversely associated with large colorectal adenomas (>1 cm in diameter; more likely to progress to adenocarcinomas) among individuals with low vitamin D and calcium intake and concluded that the association between VDR variants and colorectal adenoma risk are modified by vitamin D and calcium intake, findings which are in accordance with our results.

Additionally, in our population calcium intake (either dietary or total) was not associated with CRC. In contrast, dietary and total vitamin D intakes were inversely associated with CRC and there was a dose-response relationship. These associations persisted after controlling for several confounding factors (Model II). We also observed an inverse association between CRC and supplementary intake of vitamin D (for 2.5–5 μg/day vs. 0 μg/day supplementary intake OR = 0.83, 95% CI 0.72, 0.96; for more than 5 μg/day vs. 0 μg/day supplementary intake: OR = 0.90, 95% CI 0.69, 1.16; p-trend = 0.035). The main sources of vitamin D in our study were fish and fish products, which are the main sources of the ω3 poly-unsaturated fatty acids (ω3-PUFAs). Because of the common food sources, these 2 nutrients were highly correlated with each other (r = 0.82, p-value < 0.00005) and it is very difficult to know whether the inverse association with CRC is driven by vitamin D and/or by ω3-PUFAs. However, taken into consideration the results from the meta-analyses (particularly of the nested case–controls that measured vitamin D metabolite levels in the serum; web-Fig. 3) and the interaction between vitamin D and the VDR variant, we think that it is reasonable to believe that vitamin D is inversely associated with CRC.

We further investigated the associations between CRC, vitamin D and calcium after genotype stratification to test whether their associations are modified according to the particular genotype. We observed that the inverse association between vitamin D and CRC was more profound for individuals of the FF (CC) FokI (rs10735810) genotype than for individuals of the Ff (CT) or ff (TT) genotypes. In consistence with these findings, Peters etal.50 reported that the inverse association between 25(OH)D and colorectal adenomas was much stronger among FokI (rs10735810) FF (CC) individuals than among the ff (TT), though there was no significant evidence of interaction. For FokI (rs10735810), calcium intake was not associated with CRC for FF (CC) and Ff (CT) individuals, but it was positively associated for the ff (TT) individuals. The effect modification by the FokI (rs10735810) on the association between CRC and vitamin D and between CRC and calcium intake supports a possible causal role in colorectal carcinogenesis for the vitamin D.

We found weak and mainly non-significant inverse associations between vitamin D intake and CRC after combining results from cohort and case–control studies (web-Figs. 1 and 2). These inconsistent and weak associations might be due to the fact that the studies included did not capture total vitamin D intake (dietary, supplementary intake and skin production) coupled to the measurement error in dietary measures of vitamin D intake. The combined effect of studies measuring serum (plasma) vitamin D (25(OH)D) was stronger and significant (web-Fig. 3). A recent clinical trial of vitamin D and calcium supplementation for 7 years in post-menopausal women, though did not show any association with CRC.31 However, a large proportion of women assigned to vitamin D/calcium supplementation and assigned to placebo were also taking supplements on their own and the authors suggested that this may have limited their ability to affect the rates of CRC further. In addition, this finding might be due to insufficient time for vitamin D to affect carcinogenesis or it might be possible that vitamin D and/or calcium might affect colorectal carcinogenesis via the VDR at the early adenoma stages. In addition, there was no association between FokI (rs10735810), CRC and colorectal adenomas after combining previous studies. However, the variant genotype of BsmI rs1544410 (GG) was found to be significantly associated with CRC. We did not replicate this finding, possibly because BsmI (rs1544410) was not in Hardy Weinberg equilibrium in our study.

The strengths of our study include a very large sample size, use of a validated FFQ,52 the identification of vitamin D intake from dietary supplements and data on 4 VDR SNPs. General limitations of our study have been previously described.51 Briefly, under- representation of cases which were very ill when at presentation might limit external validity of results. Validity studies on nutrient estimates of this FFQ were carried out in younger subjects and we cannot be certain of the degree of validity in this older age group.52, 57 However, any measurement error would most likely be non-differential and thus underestimate true relationships. Limitations of observational studies employing food frequency questionnaires include misclassification bias due to imprecise measures of dietary intake and residual confounding after attempts to control for confounders and of case–control studies in particular recall and selection bias. However, we attempted to limit these problems by careful adjustment, adoption of identical study procedures in cases and controls, use of a food frequency questionnaire which had been validated,52, 58 use of images of portion sizes and careful instructions to improve accuracy of reporting diet and adoption of a recall period 1 year before diagnosis or recruitment date to reduce recall bias. In addition, we have no reason to believe that there would be any differences in the vitamin D and calcium intake between the participants and non-participants.

In this analysis, we have only considered dietary and supplementary vitamin D intake, because we did not have information regarding its skin production by UV sunlight. Therefore, subjects might have been misclassified due to the lack of measuring sunshine-produced vitamin D. However, a UVB irradiation threshold of 20 mJ/cm² is required to induce the vitamin D3 skin production, and apparently this threshold is not reached for countries above latitude 40° during the winter months (Scotland's latitude 55°).59 Therefore the sun exposure in Scotland, especially during the winter months is relatively low and this will probably make diet a more important contributor. Finally, given the multiple interactions examined, we cannot rule out the possibility that the reported interaction between FokI (rs10735810), vitamin D and calcium might be due to chance.

In summary, we report the results of the largest case–control study to date (in terms of cases) investigating the association between CRC and 4 VDR SNPs, dietary and total vitamin D and dietary and total calcium intake, as well as the interaction relationships between the SNPs and these nutrients. Our findings of an inverse association between vitamin D intake and CRC risk with a dose response relationship in addition to the OR effect modification by the FokI (rs10735810) support the interpretation of a possible preventive role for vitamin D in colorectal carcinogenesis. In addition, meta-analysis of nested case–control serum studies showed a significant inverse association with CRC. The fact that meta-analyses of cohort and case–control studies did not provide clear evidence of an inverse association, might be due to the lack of the individual studies to measure skin-produced vitamin D. The observed statistically significant interactions between high vitamin D and high calcium intake and the VDR SNP FokI (rs10735810) that has a demonstrated functional effect on the VDR protein, suggest that vitamin D and calcium act synergistically and that their effects are mediated via the VDR.

Acknowledgements

The authors thank and acknowledge the contribution and support of Mrs. R. Bisset, Mrs. M. Edwards, Mrs. L. McGoohan and Mrs. G. Barr for recruitment supervision and data management. Ms. Theodoratou was also supported by a studentship from the Greek State Scholarship Foundation.