Skip to main content
PMC home page
American Journal of Epidemiology logoLink to American Journal of Epidemiology
. 2017 Feb 17;185(6):452–464. doi: 10.1093/aje/kww143

Interactions Between Genome-Wide Significant Genetic Variants and Circulating Concentrations of 25-Hydroxyvitamin D in Relation to Prostate Cancer Risk in the National Cancer Institute BPC3

Vasiliki I Dimitrakopoulou *, Ruth C Travis, Irene M Shui, Alison Mondul, Demetrius Albanes, Jarmo Virtamo, Antonio Agudo, Heiner Boeing, H Bas Bueno-de-Mesquita, Marc J Gunter, Mattias Johansson, Kay-Tee Khaw, Kim Overvad, Domenico Palli, Antonia Trichopoulou, Edward Giovannucci, David J Hunter, Sara Lindström, Walter Willett, J Michael Gaziano, Meir Stampfer, Christine Berg, Sonja I Berndt, Amanda Black, Robert N Hoover, Peter Kraft, Timothy J Key, Konstantinos K Tsilidis b
PMCID: PMC5856084  PMID: 28399564

Abstract

Genome-wide association studies (GWAS) have identified over 100 single nucleotide polymorphisms (SNPs) associated with prostate cancer. However, information on the mechanistic basis for some associations is limited. Recent research has been directed towards the potential association of vitamin D concentrations and prostate cancer, but little is known about whether the aforementioned genetic associations are modified by vitamin D. We investigated the associations of 46 GWAS-identified SNPs, circulating concentrations of 25-hydroxyvitamin D (25(OH)D), and prostate cancer (3,811 cases, 511 of whom died from the disease, compared with 2,980 controls—from 5 cohort studies that recruited participants over several periods beginning in the 1980s). We used logistic regression models with data from the National Cancer Institute Breast and Prostate Cancer Cohort Consortium (BPC3) to evaluate interactions on the multiplicative and additive scales. After allowing for multiple testing, none of the SNPs examined was significantly associated with 25(OH)D concentration, and the SNP–prostate cancer associations did not differ by these concentrations. A statistically significant interaction was observed for each of 2 SNPs in the 8q24 region (rs620861 and rs16902094), 25(OH)D concentration, and fatal prostate cancer on both multiplicative and additive scales (P ≤ 0.001). We did not find strong evidence that associations between GWAS-identified SNPs and prostate cancer are modified by circulating concentrations of 25(OH)D. The intriguing interactions between rs620861 and rs16902094, 25(OH)D concentration, and fatal prostate cancer warrant replication.

Keywords: 25-hydroxyvitamin D, BPC3, gene-environment interactions, genome-wide association studies, prostate cancer

Prostate cancer is the most common type of cancer in men and has large clinical heterogeneity, ranging from well-differentiated indolent tumors to aggressive and fatal disease. Genome-wide association studies (GWAS) have identified over 100 genetic variants associated with prostate cancer risk, explaining approximately 30% of the genetic variance of the disease, but have generally failed to identify variants specific for aggressive or fatal disease (1, 2). Some of these single nucleotide polymorphisms (SNPs) are located in intergenic regions, which are still under investigation for their potential functions. A few studies have investigated the interaction of these genetic associations according to other established or suspected risk factors for prostate cancer, including age, ethnicity, family history, body mass index, diabetes, or circulating concentrations of insulin-like growth factors or steroid sex hormones, and have found no strong evidence for multiplicative interaction (3–6). However, little is known about whether these genetic associations are modified by circulating concentrations of 25-hydroxyvitamin D (25(OH)D). There is ample biological evidence of an anticancer role for 25(OH)D, as metabolites of vitamin D control cellular growth and differentiation (7), and administration of vitamin D analogs inhibits the progression of prostate cancer in animal models and in phase II trials (8–10). Two meta-analyses of epidemiologic studies—Yin et al. (11) in 2009 and Gilbert et al. (12) 2011—have observed null associations for circulating 25(OH)D and risk of prostate cancer overall or aggressive prostate cancer, while a more recent meta-analysis (Xu et al. (13) in 2014) observed a statistically significant positive association for all prostate cancer. To further investigate the mechanistic basis for the association of GWAS-identified SNPs and prostate cancer risk, we examined the additive and multiplicative interactions of 46 SNPs and circulating 25(OH)D concentrations in relation to total and fatal prostate cancer risk in the Breast and Prostate Cancer Cohort Consortium (BPC3).

METHODS

Source and study population

BPC3 is a consortium of 9 cohort studies being conducted in the United States and Europe that was established in 2004 to investigate genetic risk factors for breast and prostate cancer (14). The studies recruited patients over various periods beginning in the 1980s. BPC3 includes the following studies: the Alpha-Tocopherol, Beta-Carotene Cancer Prevention Study (ATBC); the American Cancer Society Cancer Prevention Study II; the European Prospective Investigation into Cancer and Nutrition (EPIC); the Health Professionals Follow-up Study (HPFS); the Multi-Ethnic Cohort; the Nurses’ Health Study; the Physicians’ Health Study (PHS); the Prostate, Lung, Colorectal, and Ovarian Cancer Screening Trial (PLCO); and the Women's Health Study. Incident cancer cases were identified through linkage to cancer registries or through self-reports that were confirmed by medical records and/or pathology reports. Detailed information about this consortium and its component studies can be found elsewhere (14). All of these studies have been approved by the institutional review boards or ethics committees of their respective institutions.

The present study used 5 cohorts (ATBC, EPIC, HPFS, PHS, and PLCO) that enrolled male participants who provided genetic and circulating vitamin D data. Men were excluded if they had prevalent cancer at recruitment or if they were not of white European ancestry, resulting in a total number of 3,811 cases (511 of whom were known to have died from the disease) and 2,980 controls.

Genotyping

A total of 47 SNPs were genotyped based on published GWAS for prostate cancer; they were: rs13385191, rs1465618, rs721048, rs10187424, rs12621278, rs2292884, rs2660753, rs7629490, rs6763931, rs10936632, rs17021918, rs7679673, rs2242652, rs12653946, rs2121875, rs130067, rs1983891, rs339331, rs9364554, rs12155172, rs10486567, rs6465657, rs2928679, rs1512268, rs1016343, rs16901979, rs16902094, rs620861, rs6983267, rs4242382, rs1571801, rs10993994, rs7127900, rs12418451, rs10896449, rs10875943, rs902774, rs11649743, rs4430796, rs1859962, rs8102476, rs11672691, rs2735839, rs5759167, rs11704416, rs5945619, rs5919432. PHS and PLCO did not have data for rs10187424, rs6763931, rs10936632, rs2242652, rs2121875, rs130067, rs10875943, and rs5919432. Missing genotypes for all 47 SNPs were imputed by sampling from the observed frequency distribution in men without missing genotypes stratified by case-control status, as previously reported in detail (15). The allele frequencies and results from all statistical analyses in the present study were similar before and after the imputations, and thus only results using the directly genotyped information were presented.

Genotyping was performed using the TaqMan assay (Applied Biosystems, Foster City, California) in 6 genotyping laboratories in 3 countries: Cancer Genomics Research Laboratory at the National Cancer Institute, Harvard T.H. Chan School of Public Health, University of Southern California, German Cancer Research Center, University of Cambridge, and Imperial College London. The median genotyping success rate was 98.7% overall (interquartile range, 97.4%–99.6%; range, 82.4%–100%). Blinded duplicate samples (approximately 5%) were included within each study, and the concordance rate was greater than 99%. All but 1 autosomal SNP (rs1983891) were in Hardy-Weinberg equilibrium (P > 0.001), and that variant was removed from the analysis.

Circulating vitamin D concentrations

Prediagnostic circulating 25(OH)D concentrations were measured in specialist laboratories. All laboratory personnel were blinded to the case-control status of the samples. Detailed information on assay methods and quality control statistics by the participating cohorts can be found elsewhere (16–20).

Blood samples were collected at different time points and assessed in different batches except for PLCO, for which all blood samples were assessed in a single batch. Because vitamin D concentrations are dependent on seasonality, and analysis in different batches normally induces laboratory variation, we created “cohort-, batch-, and season-specific” tertiles using the distribution of 25(OH)D concentrations among the controls. We also used a continuous measure of 25(OH)D concentration that was standardized for cohort, batch, and season using linear regression models as described in Rosner et al. (21).

Statistical analysis

The main aim of the present study was to examine the interactions of genes and circulating vitamin D with respect to total and fatal prostate cancer risk. However, to properly assess interactions on the multiplicative and absolute scales, all possible marginal pairwise associations were also assessed and presented herein. In particular, we estimated the following: 1) the associations of the 46 SNPs with risk of total and fatal prostate cancer, 2) the associations of the 46 SNPs with circulating concentrations of 25(OH)D, and 3) the associations of 25(OH)D concentration with risk of total and fatal prostate cancer. We performed all analyses using cohort-, batch-, and season-specific tertiles of vitamin D concentrations and a Rosner-standardized continuous vitamin D variable.

Logistic regression models with adjustment for age (continuous) at blood draw, cohort, and country (within EPIC) were used to assess the associations between the 46 SNPs and the risk of total and fatal prostate cancer. Odds ratios and their 95% confidence intervals were calculated per carried allele that was associated with an increased risk of prostate cancer in the GWAS literature. Logistic regression models were also used to evaluate the association between 25(OH)D concentrations and risk of total and fatal prostate cancer after adjustment for age at blood draw, year of blood draw, cohort, and country (within EPIC) as well as diabetes, alcohol intake, and body mass index. Geometric means and 95% confidence intervals for the circulating 25(OH)D concentrations were calculated by genotype for each SNP (rare homozygote, heterozygote, common homozygote), using linear regression models for the natural logarithmic transformation of 25(OH)D. Models adjusted for age at blood draw, year of blood draw, case-control status, cohort, and country (within EPIC). The F-distribution was used to assess differences between the 3 geometric means.

Multiplicative and additive interactions of genes and vitamin D in relation to total and fatal prostate cancer risk were examined using several methods. For the multiplicative interactions, we employed both a case-control and a case-only design. The per-allele odds ratios for total and fatal prostate cancer for each SNP were compared across 25(OH)D concentrations using logistic regression models that adjusted for age at blood draw, year of blood draw, cohort, and country (within EPIC). The P values for interaction were calculated using likelihood ratio tests based on per-allele odds ratios. All reported P values were uncorrected for multiple hypotheses testing, but they are interpreted in view of the 46 independent comparisons made, as the examined SNPs were not in linkage disequilibrium. Using the Bonferroni correction and a significance level of 5%, only an uncorrected P value of less than 0.001 would be regarded as statistically significant. As a sensitivity analysis, the false-discovery-rate approach was also used to account for multiple testing, as it may have higher statistical power (22). The false-discovery-rate method gave similar results to the Bonferroni correction (data not shown), and thus results are shown throughout the text only for the latter approach.

A 2-step case-only approach was also used to evaluate multiplicative interactions with greater statistical power. SNPs were dichotomized according to the presence of the risk allele, and logistic regression models were fitted to assess the association between the binary SNPs and 25(OH)D concentrations among only the prostate cancer cases after adjustment for age at blood draw, year of blood draw, cohort, and country (within EPIC). Under the assumption that genes and 25(OH)D concentrations are independent in the overall population, a statistically significant association in this model indicates an interaction association (23). For the observed nominally significant interactions, we reverted to the less powerful case-control analysis to more fully assess interactions while reducing the number of tested comparisons.

As an attempt to further increase the power to detect multiplicative interactions, another sensitivity analysis (conditional logistic regression with counterfactuals) was employed. When the studied genes are independent of the exposure and the Hardy-Weinberg equilibrium holds, which is the case in the present analysis, it has been shown that this method is unbiased and achieves higher statistical power than the conventional analysis (24). This method gave results similar to our primary analysis (data not shown), and thus results are shown throughout the text only for the latter approach.

Additive interactions for cohort-, batch-, and season-specific tertiles and Rosner-standardized continuous 25(OH)D levels with dichotomized SNPs were estimated using the relative excess risk due to interaction (RERI) method, as described in Hosmer and Lemeshow (25). To estimate the 95% confidence intervals of the RERIs, we performed bootstrap sampling with 1,000 samples. Each time, samples were drawn separately for cases and controls to maintain the original numbers. Using the resulting bootstrap sampling distribution and, more specifically, its 2.5th and 97.5th percentiles, we estimated the confidence intervals (26). Although different methods have been suggested for the construction of confidence intervals for RERIs (25, 26), this method has been shown to have the best coverage, especially in cases of asymmetry (26–28).

The cumulative multiplicative and additive interactions for all 46 SNPs were assessed by creating an additive genetic score after summing the number of risk alleles across the 46 SNPs for each participant. We regarded missing genotypes as zero risk alleles and considered the gene score as a continuous variable. All analyses were performed in STATA, version 12 (StataCorp LP, College Station, Texas). We also implemented kernel machine models to better evaluate the aggregate association with all 46 SNPs. Logistic regression kernel machine models (29–31) treat each of the SNPs as a random effect, while the adjusting covariates are incorporated as fixed effects. Kernel machine models consider the joint association with the entire set of SNPs and take into account the between-SNPs correlation, thereby improving the power of the test and reducing the multiple testing burden. Kernel machine analysis was performed using the iSKAT package (https://www.hsph.harvard.edu/xlin/) in R (R Foundation for Statistical Computing, Vienna, Austria) within a Unix environment.

RESULTS

Selected characteristics of the 3,811 prostate cancer cases and 2,980 controls are shown by cohort in Web Table 1 (available at http://aje.oxfordjournals.org/). The cases were, on average, aged 63.2 years at the time of blood draw and 68.6 years at cancer diagnosis, while the controls had an average age of 63.5 years at blood draw. The mean concentrations of 25(OH)D differed by cohort, with smaller values observed in the ATBC Study, which was conducted in Finland.

Twenty-nine SNPs (63%) were nominally statistically significantly associated with risk of any prostate cancer (Web Table 2); the directions of all associations were consistent with previous GWAS findings (2). Seventeen SNPs were not significantly associated with risk in this study, likely due to the smaller sample size compared with the published GWASs. Thirteen SNPs (28%) were nominally statistically significantly associated with risk of fatal prostate cancer (Web Table 3). Individuals in the top tertile of 25(OH)D concentrations had a higher risk of any prostate cancer than did individuals in the bottom tertile (odds ratio (OR) = 1.16, 95% confidence interval (CI): 1.03, 1.31; P for trend = 0.01), whereas no association was observed for risk of fatal prostate cancer (Web Table 4; for the top vs. bottom tertile, OR = 0.96, 95% CI: 0.75, 1.22; P for trend = 0.73). These results did not differ by participating cohort (P for heterogeneity > 0.41).

The distribution of 25(OH)D concentrations by genotype for each of the 46 SNPs is shown in Table 1. Only 4 associations were nominally statistically significant, and none remained significant after adjusting for multiple testing. In order to investigate whether the genetic associations with risk of total and fatal prostate cancer were stronger for specific strata of 25(OH)D concentrations, we evaluated gene-environment interactions on the multiplicative (Tables 2 and 3) and additive scales (Web Tables 5 and 6). Table 2 presents the assessment of multiplicative interactions for total prostate cancer. Analyses included the case-control and the case-only approach using cohort-, batch-, and season-specific tertiles and Rosner-standardized continuous terms for 25(OH)D concentration. None of the multiplicative interactions was statistically significant for any SNP after allowing for multiple testing, and this was also evident when the additive genetic score and the kernel machine genetic score were used. Only one SNP, rs620861, had a marginally statistically significant interaction using the case-only approach. We used a P threshold of 0.01 for the case-control analyses, applying a correction to the standard threshold of 0.05 to account for multiple testing on the basis of finding 5 nominally significant SNPs for 25(OH)D tertiles in the case-only analyses. When interactions on the additive scale were evaluated, again no statistically significant association was found after adjustment for multiple testing (Web Table 5).

Table 1.

Distribution of 25-Hydroxyvitamin D Concentrations (1982–2004) According to the Genotype of 46 Single Nucleotide Polymorphisms Identified in Genome-Wide Association Studies, Breast and Prostate Cancer Cohort Consortium

SNP Chromosome Gene Zygosity and 25(OH)D Concentrationa P Valueb
Rare Homozygote Heterozygote Common Homozygote
GM 95% CI GM 95% CI GM 95% CI
rs13385191 2 C2orf43 22.4 21.5, 23.4 21.8 21.4, 22.2 22.2 21.8, 22.5 0.18
rs1465618 2 THADA 21.8 20.9, 22.9 22.2 21.8, 22.6 21.9 21.7, 22.2 0.59
rs721048 2 EHBP1 22.0 20.9, 13.2 21.9 21.5, 22.3 22.1 21.8, 22.4 0.82
rs10187424 2 GGCX/VAMP8 22.4 20.5, 24.4 21.3 19.6, 23.2 21.7 19.9, 23.6 0.05
rs12621278 2 ITGA6 19.0c 15.4, 23.5 22.3 21.6, 23.0 22.0 21.8, 22.2 0.32
rs2292884 2 MLPH 22.7 21.6, 23.8 22.1 21.7, 22.6 21.9 21.5, 22.3 0.29
rs2660753 3 Unknown 21.8 20.0, 23.7 22.1 21.6, 22.6 22.0 21.8, 22.3 0.94
rs7629490 3 Unknown 21.7 21.0, 22.4 22.0 21.6, 22.4 22.1 21.7, 22.5 0.66
rs6763931 3 ZBTB38 21.5 19.8, 23.4 21.7 20.0, 23.6 21.5 19.7, 23.4 0.71
rs10936632 3 CLDN11/SKIL 21.6 19.9, 23.5 21.7 20.0, 23.6 21.7 20.0, 23.6 0.95
rs17021918 4 PDLIM5 21.9 21.2, 22.6 22.1 21.8, 22.5 22.1 21.7, 22.4 0.80
rs7679673 4 TET2 21.8 21.2, 22.3 22.2 21.8, 22.5 22.0 21.6, 22.4 0.45
rs2242652 5 TERT 19.8 17.3, 22.6 21.1 18.8, 23.7 21.0 18.7, 23.6 0.21
rs12653946 5 IRX4 22.0 21.4, 22.5 21.9 21.6, 22.3 22.2 21.7, 22.6 0.66
rs2121875 5 FGF10 22.1 20.1, 24.2 21.3 19.6, 23.2 22.2 20.4, 24.2 0.01
rs130067 6 CCHCR1 21.6 18.9, 26.7 21.3 18.9, 23.9 20.8 18.5, 23.4 0.22
rs339331 6 RFX6 22.1 21.3, 23.0 22.0 21.7, 22.4 22.0 21.7, 22.3 0.97
rs9364554 6 SLC22A3 22.7 21.9, 23.6 21.9 21.5, 22.2 22.1 21.8, 22.4 0.16
rs12155172 7 SP8 22.0 21.0, 23.0 21.9 21.6, 22.3 22.1 21.8, 22.3 0.91
rs10486567 7 JAZF1 21.2 20.3, 22.1 22.0 21.7, 22.4 22.1 21.8, 22.4 0.18
rs6465657 7 LMTK2 22.0 21.5, 22.4 22.1 21.8, 22.4 22.0 21.6, 22.4 0.87
rs2928679 8 SLC25A37 22.3 21.8, 22.8 21.9 21.6, 22.2 22.1 21.7, 22.5 0.44
rs1512268 8 NKX3.1 22.0 21.5, 22.5 21.9 21.6, 22.2 22.2 21.8, 22.6 0.54
rs1016343 8 Unknown 22.0 21.1, 23.0 22.3 21.9, 22.7 21.9 21.7, 22.2 0.38
rs16901979 8 Unknown 23.0c 18.3, 28.9 21.8 21.0, 22.7 22.1 21.9, 22.3 0.79
rs16902094 8 Unknown 22.2 20.8, 23.6 22.1 21.7, 22.5 22.0 21.7, 22.3 0.92
rs620861 8 Unknown 21.5 20.9, 22.1 22.1 21.8, 22.4 22.0 21.6, 22.4 0.26
rs6983267 8 Unknown 21.9 21.4, 22.3 22.0 21.7, 22.3 22.2 21.8, 22.6 0.53
rs4242382 8 Unknown 24.0 22.2, 25.9 22.0 21.6, 22.5 22.0 21.8, 22.3 0.11
rs1571801 9 DAB21P 22.5 21.7, 23.4 21.9 21.5, 22.3 22.1 21.7, 22.4 0.40
rs10993994 10 MSMB 22.0 21.5, 22.6 22.3 22.0, 22.6 21.8 21.4, 22.1 0.12
rs7127900 11 Unknown 21.9 20.9, 23.0 22.0 21.7, 22.4 22.0 21.7, 22.3 0.96
rs12418451 11 Unknown 22.3 21.6, 23.1 21.9 21.5, 22.2 22.2 21.8, 22.5 0.37
rs10896449 11 Unknown 22.1 21.7, 22.6 22.1 21.8, 22.4 21.9 21.5, 22.4 0.82
rs10875943 12 TUBA1C/PRPH 21.1 18.7, 23.9 21.0 18.6, 23.6 20.9 18.6, 23.5 0.86
rs902774 12 KRT8 22.5 21.1, 23.9 22.1 21.6, 22.5 22.1 21.8, 22.4 0.86
rs11649743 17 HNF1B 21.8 20.6, 23.0 21.8 21.4, 22.2 22.1 21.9, 22.4 0.40
rs4430796 17 HNF1B 22.5 22.0, 23.0 21.9 21.6, 22.2 22.0 21.6, 22.4 0.10
rs1859962 17 Unknown 21.8 21.3, 22.2 22.1 21.8, 22.4 22.2 21.8, 22.7 0.32
rs8102476 19 Unknown 22.1 21.6, 22.6 22.1 21.7, 22.4 21.9 21.5, 22.3 0.75
rs11672691 19 Unknown 22.2 21.2, 23.2 22.1 21.7, 22.5 22.0 21.6, 22.3 0.79
rs2735839 19 KLK2/KLK3 19.9 18.5, 21.4 22.0 21.5, 22.5 22.1 21.8, 22.3 0.02
rs5759167 22 BIL/TTLL1 21.6 21.2, 22.1 22.2 21.8, 22.5 22.1 21.6, 22.5 0.19
rs11704416 22 TNRC6B 21.9 20.8, 23.0 22.2 21.8, 22.7 21.9 21.6, 22.2 0.50
rs5945619 X NUDT11 22.3 22.0, 22.7 21.9 21.6, 22.2 0.06
rs5919432 X AR 20.3 18.0, 22.9 21.1 18.8, 23.8 0.03
Open in a new tab

Abbreviations: CI, confidence interval; GM, geometric mean; SNP, single nucleotide polymorphism.

a From a linear regression analysis between natural log-transformed 25-hydroxyvitamin D concentrations (Rosner-standardized) for cohort, batch, and season and further adjusted for age at blood draw, year of blood draw, case-control status, cohort, and country (within the European Prospective Investigation into Cancer and Nutrition).

b Test of the difference of 3 geometric means using the F distribution. Conventional P values are shown; all P values were nonsignificant after allowance for multiple testing.

c The total sample size in this genotype subgroup was ≤50 observations.

Table 2.

Per-Allele Associations Between 46 Single Nucleotide Polymorphisms Identified in Genome-Wide Association Studies and Risk of Total Prostate Cancer According to 25-Hydroxyvitamin D Concentrations (1982–2004), Breast and Prostate Cancer Cohort Consortium

SNP No. of Cases No. of Controls Tertiles of 25(OH)Da P for Interactionb P for Interactionc P for Interactiond P for Interactione
First Second Third
OR 95% CI OR 95% CI OR 95% CI
rs13385191 3,377 2,556 1.09 0.94, 1.26 1.15 0.99, 1.33 1.01 0.87, 1.16 0.48 0.23 0.04 0.24
rs1465618 3,648 2,816 1.12 0.96, 1.30 1.15 0.99, 1.33 1.06 0.92, 1.23 0.62 0.81 0.33 0.63
rs721048 3,675 2,849 1.14 0.97, 1.34 1.12 0.97, 1.31 1.23 1.05, 1.43 0.50 0.79 0.57 0.57
rs10187424 2,252 1,782 1.11 0.95, 1.31 1.03 0.88, 1.20 1.04 0.89, 1.21 0.53 0.54 0.02 0.10
rs12621278 3,682 2,834 0.98 0.75, 1.30 1.00 0.76, 1.31 1.25 0.95, 1.64 0.26 0.87 0.42 0.33
rs2292884 3,109 2,455 1.12 0.96, 1.30 0.99 0.85, 1.15 1.02 0.88, 1.19 0.47 0.39 0.42 0.65
rs2660753 3,692 2,884 1.06 0.87, 1.28 1.20 1.00, 1.44 1.14 0.94, 1.37 0.51 0.33 0.23 0.14
rs7629490 3,115 2,441 1.15 1.00, 1.32 1.09 0.95, 1.25 1.09 0.95, 1.25 0.62 0.56 0.37 0.23
rs6763931 2,318 1,837 0.98 0.85, 1.14 1.03 0.88, 1.20 1.06 0.91, 1.23 0.52 0.43 0.26 0.23
rs10936632 2,348 1,858 0.96 0.83, 1.12 1.03 0.89, 1.20 1.09 0.94, 1.27 0.26 0.84 0.09 0.40
rs17021918 3,656 2,831 1.07 0.94, 1.21 1.00 0.88, 1.14 1.20 1.06, 1.37 0.18 0.18 0.19 0.34
rs7679673 3,677 2,824 1.15 1.02, 1.30 1.21 1.07, 1.37 1.09 0.96, 1.23 0.54 0.26 0.64 1.00
rs2242652 2,350 1,858 1.11 0.92, 1.33 1.29 1.06, 1.57 1.43 1.19, 1.73 0.05 0.04 0.80 0.24
rs12653946 3,365 2,548 1.10 0.97, 1.26 1.06 0.94, 1.21 0.99 0.87, 1.12 0.25 0.41 0.24 0.25
rs2121875 2,285 1,739 1.06 0.90, 1.26 0.92 0.78, 1.08 1.09 0.93, 1.29 0.81 0.77 0.19 0.18
rs130067 2,340 1,856 1.06 0.87, 1.28 1.20 0.99, 1.45 0.95 0.79, 1.15 0.43 0.50 0.60 0.87
rs339331 3,387 2,564 1.11 0.97, 1.28 0.99 0.85, 1.14 1.07 0.93, 1.23 0.74 0.44 0.38 0.55
rs9364554 3,694 2,876 1.08 0.94, 1.25 1.06 0.92, 1.21 1.07 0.94, 1.22 0.91 0.14 0.38 0.73
rs12155172 3,650 2,830 0.97 0.84, 1.13 1.11 0.96, 1.28 1.05 0.91, 1.21 0.47 0.43 0.75 0.67
rs10486567 3,677 2,873 1.22 1.05, 1.41 1.12 0.97, 1.29 1.19 1.03, 1.37 0.85 0.56 0.08 0.21
rs6465657 3,681 2,868 1.13 1.00, 1.27 1.08 0.96, 1.22 1.02 0.91, 1.15 0.21 0.42 0.77 0.49
rs2928679 3,692 2,828 1.05 0.93, 1.19 1.02 0.91, 1.16 0.98 0.87, 1.11 0.48 0.46 0.94 0.80
rs1512268 3,709 2,851 1.09 0.96, 1.23 1.02 0.91, 1.16 1.11 0.98, 1.25 0.88 0.81 0.32 0.54
rs1016343 3,694 2,875 1.27 1.10, 1.46 1.02 0.88, 1.17 1.23 1.07, 1.42 0.79 0.39 0.62 0.31
rs16901979 3,655 2,852 1.38 1.00, 1.90 1.21 0.87, 1.67 1.87 1.34, 2.60 0.20 0.14 0.42 0.42
rs16902094 3,362 2,621 1.11 0.94, 1.32 1.00 0.85, 1.18 1.19 1.01, 1.41 0.61 0.02 0.23 0.59
rs620861 3,460 2,673 0.97 0.85, 1.11 1.07 0.94, 1.21 1.25 1.10, 1.43 0.01 0.10 0.04 0.06
rs6983267 3,670 2,847 1.20 1.06, 1.35 1.13 1.00, 1.28 1.31 1.16, 1.48 0.32 0.80 0.28 0.62
rs4242382 3,755 2,921 1.33 1.10, 1.62 1.46 1.21, 1.77 1.58 1.31, 1.90 0.27 0.83 0.17 0.48
rs1571801 3,598 2,773 1.02 0.88, 1.17 1.23 1.07, 1.42 1.04 0.91, 1.20 0.89 0.84 0.98 0.65
rs10993994 3,653 2,852 1.20 1.06, 1.36 1.18 1.05, 1.34 1.18 1.04, 1.33 0.91 0.69 0.83 0.47
rs7127900 3,666 2,815 1.12 0.96, 1.30 1.09 0.94, 1.27 1.30 1.12, 1.51 0.16 0.14 0.20 0.35
rs12418451 3,773 2,932 1.10 0.96, 1.26 1.09 0.96, 1.24 1.14 1.01, 1.30 0.66 0.35 0.72 0.10
rs10896449 3,635 2,850 1.22 1.08, 1.37 1.18 1.04, 1.33 1.21 1.08, 1.36 0.99 0.49 1.00 0.59
rs10875943 2,363 1,867 1.02 0.87, 1.20 1.08 0.92, 1.28 0.99 0.85, 1.16 0.81 0.77 0.51 0.86
rs902774 3,574 2,752 1.02 0.86, 1.22 1.16 0.98, 1.37 1.06 0.90, 1.25 0.88 0.71 0.83 0.48
rs11649743 3,661 2,858 1.24 1.06, 1.45 1.19 1.02, 1.39 1.08 0.92, 1.26 0.18 0.12 0.46 0.68
rs4430796 3,596 2,802 1.30 1.14, 1.47 1.29 1.14, 1.46 1.17 1.04, 1.32 0.19 0.44 0.05 0.18
rs1859962 3,701 2,882 1.21 1.07, 1.37 1.37 1.22, 1.55 1.08 0.96, 1.22 0.16 0.99 0.99 0.49
rs8102476 3,622 2,796 0.88 0.78, 0.99 1.13 1.00, 1.28 1.08 0.96, 1.22 0.01 0.07 0.15 0.22
rs11672691 3,338 2,549 1.24 1.06, 1.44 0.96 0.83, 1.12 1.12 0.96, 1.29 0.37 0.10 0.61 0.73
rs2735839 3,614 2,795 1.13 0.95, 1.36 1.20 1.01, 1.43 1.07 0.90, 1.28 0.63 0.47 0.55 0.14
rs5759167 3,661 2,819 1.08 0.96, 1.22 1.14 1.01, 1.28 1.26 1.12, 1.42 0.08 0.08 0.20 0.02
rs11704416 3,354 2,564 1.16 0.99, 1.36 1.15 0.98, 1.34 1.01 0.87, 1.18 0.23 0.17 0.79 0.58
rs5945619 3,703 2,871 1.18 1.08, 1.29 1.04 0.96, 1.14 1.10 1.01, 1.20 0.29 0.23 0.33 0.38
rs5919432 2,336 1,851 1.02 0.89, 1.16 1.00 0.87, 1.14 1.19 1.04, 1.36 0.13 0.19 0.01 0.04
Additive SNP scoref 3,811 2,980 1.05 1.04, 1.06 1.03 1.02, 1.05 1.04 1.03, 1.06 0.50 0.97 0.60 0.63
Kernel machine scoreg 0.21 0.28 <0.001 <0.001
Open in a new tab

Abbreviations: 25(OH)D, 25-hydroxyvitamin D; CI, confidence interval; OR, odds ratio; SNP, single nucleotide polymorphism.

a From a logistic regression model of SNPs and total prostate cancer risk by cohort-, batch-, and season-specific tertiles of 25(OH)D concentration adjusted for age at blood draw, year of blood draw, cohort, and country (within the European Prospective Investigation into Cancer and Nutrition).

bP for interaction was calculated based on the case-control analysis and a variable for cohort-, batch-, and season-specific tertiles of 25(OH)D concentration. Conventional P values are shown; all P values were nonsignificant after allowance for multiple testing (P threshold = 0.001).

cP for interaction was calculated based on the case-control analysis and a continuous 25(OH)D variable (Rosner-standardized) for cohort, batch, and season. Conventional P values are shown; all P values were nonsignificant after allowance for multiple testing (P threshold = 0.001).

dP for interaction was calculated based on the case-only analysis for dichotomized SNPs and cohort-, batch-, and season-specific tertiles of 25(OH)D concentration. Conventional P values are shown; all P values were nonsignificant after allowance for multiple testing. We used a P threshold of 0.01 for the case-control analyses, applying a correction to the standard threshold of 0.05 to account for multiple testing on the basis of finding 5 nominally significant SNPs for 25(OH)D tertiles in the case-only analyses.

eP for interaction was calculated based on the case-only analysis for dichotomized SNPs and a continuous 25(OH)D variable (Rosner-standardized) for cohort, batch, and season. Conventional P values are shown; all P values were nonsignificant after allowance for multiple testing. We used a P threshold of 0.01 for the case-control analyses, applying a correction to the standard threshold of 0.05 to account for multiple testing on the basis of finding 5 nominally significant SNPs in the case-only analyses.

f From a logistic regression model of a continuous additive genetic score (after summing the number of risk alleles across the 46 SNPs for each participant) and total prostate cancer risk. For the case-only analysis, the additive genetic score was dichotomized at the median among controls.

g From a logistic regression kernel machine model across the entire set of 46 SNPs and total prostate cancer risk.

Table 3.

Per-Allele Association Between 46 Single Nucleotide Polymorphisms Identified in Genome-Wide Association Studies and Risk of Fatal Prostate Cancer According to 25-Hydroxyvitamin D Concentrations (1982–2004), Breast and Prostate Cancer Cohort Consortium

SNP No. of Cases No. of Controls Tertiles of 25(OH)Da P for Interactionb P for Interactionc P for Interactiond P for Interactione
First Second Third
OR 95% CI OR 95% CI OR 95% CI
rs13385191 419 2,556 1.07 0.78, 1.46 1.03 0.76, 1.40 0.92 0.66, 1.27 0.50 0.46 0.75 0.88
rs1465618 459 2,816 1.07 0.79, 1.45 1.23 0.91, 1.66 1.11 0.81, 1.51 0.83 0.86 0.62 0.71
rs721048 489 2,849 0.97 0.69, 1.35 0.98 0.71, 1.35 1.34 0.97, 1.85 0.19 0.30 0.58 0.77
rs10187424 307 1,782 1.19 0.87, 1.65 1.24 0.87, 1.75 1.07 0.76, 1.51 0.58 0.87 0.91 0.91
rs12621278 475 2,834 0.84 0.49, 1.45 1.11 0.63, 1.98 1.45 0.77, 2.72 0.21 0.56 0.47 0.008
rs2292884 403 2,455 1.26 0.92, 1.73 0.73 0.51, 1.05 1.13 0.83, 1.56 0.67 0.91 0.75 0.71
rs2660753 486 2,884 0.84 0.54, 1.30 1.08 0.73, 1.61 0.90 0.59, 1.37 0.87 0.18 0.51 0.20
rs7629490 400 2,441 1.10 0.83, 1.47 1.05 0.79, 1.40 0.94 0.70, 1.26 0.45 0.45 0.54 0.75
rs6763931 321 1,837 0.88 0.65, 1.19 0.78 0.56, 1.09 1.36 0.99, 1.86 0.07 0.07 0.34 0.28
rs10936632 315 1,858 0.85 0.61, 1.17 1.26 0.90, 1.77 1.05 0.77, 1.44 0.30 0.89 0.89 0.90
rs17021918 469 2,831 1.01 0.78, 1.31 0.85 0.64, 1.12 1.05 0.81, 1.38 0.85 0.31 0.69 0.55
rs7679673 472 2,824 1.27 0.99, 1.62 1.32 1.01, 1.72 1.29 0.99, 1.68 0.99 0.89 0.96 0.86
rs2242652 324 1,858 0.87 0.61, 1.23 0.97 0.64, 1.46 1.63 1.07, 2.49 0.03 0.42 0.86 0.80
rs12653946 420 2,548 1.33 1.02, 1.75 0.94 0.71, 1.25 1.15 0.88, 1.50 0.41 0.95 0.64 0.79
rs2121875 309 1,739 1.00 0.72, 1.38 1.12 0.78, 1.60 1.27 0.90, 1.78 0.34 0.42 0.37 0.58
rs130067 323 1,856 0.97 0.66, 1.42 0.93 0.60, 1.42 0.80 0.53, 1.19 0.52 0.85 0.83 0.93
rs339331 421 2,564 1.26 0.94, 1.69 0.75 0.55, 1.02 1.03 0.76, 1.39 0.32 0.08 0.28 0.43
rs9364554 489 2,876 1.09 0.83, 1.44 1.03 0.78, 1.37 1.14 0.86, 1.51 0.87 0.60 0.71 0.64
rs12155172 472 2,830 0.86 0.64, 1.17 1.27 0.95, 1.70 1.05 0.78, 1.43 0.35 0.42 0.35 0.72
rs10486567 500 2,873 0.81 0.62, 1.06 1.00 0.75, 1.33 1.09 0.81, 1.47 0.16 0.23 0.59 0.72
rs6465657 478 2,868 1.06 0.83, 1.36 1.03 0.80, 1.33 0.82 0.63, 1.06 0.18 0.24 0.45 0.36
rs2928679 477 2,828 1.12 0.87, 1.45 0.99 0.77, 1.29 0.96 0.74, 1.24 0.32 0.68 0.16 0.58
rs1512268 482 2,851 1.07 0.84, 1.37 0.92 0.71, 1.18 0.92 0.72, 1.19 0.42 0.59 0.35 0.27
rs1016343 487 2,875 1.60 1.21, 2.10 1.17 0.87, 1.57 1.10 0.81, 1.50 0.07 0.06 0.28 0.30
rs16901979 490 2,852 1.06 0.55, 2.06 1.45 0.77, 2.73 1.90 1.01, 3.58 0.20 0.11 0.40 0.17
rs16902094 450 2,621 0.96 0.68, 1.34 1.31 0.95, 1.80 1.60 1.15, 2.23 0.03 0.002 0.05 0.11
rs620861 463 2,673 0.75 0.58, 0.97 1.06 0.81, 1.39 1.48 1.11, 1.97 0.001 0.009 0.001 0.02
rs6983267 498 2,847 1.27 1.00, 1.60 1.19 0.92, 1.52 1.22 0.95, 1.57 0.79 0.94 0.72 0.83
rs4242382 505 2,921 1.08 0.73, 1.59 1.97 1.39, 2.80 1.35 0.95, 1.91 0.44 0.40 0.29 0.17
rs1571801 482 2,773 1.22 0.94, 1.60 0.98 0.73, 1.31 1.17 0.88, 1.55 0.86 0.48 0.45 0.90
rs10993994 494 2,852 0.88 0.68, 1.13 1.24 0.97, 1.59 1.36 1.06, 1.76 0.02 0.03 0.04 0.007
rs7127900 474 2,815 1.11 0.82, 1.50 0.90 0.65, 1.25 1.14 0.83, 1.56 0.86 0.88 0.41 0.66
rs12418451 504 2,932 1.10 0.84, 1.43 0.96 0.73, 1.26 1.21 0.93, 1.58 0.60 0.42 0.34 0.78
rs10896449 493 2,850 1.05 0.83, 1.33 1.04 0.81, 1.34 1.29 1.00, 1.65 0.28 0.17 0.33 0.73
rs10875943 321 1,867 1.13 0.81, 1.58 1.08 0.77, 1.51 1.11 0.81, 1.51 0.98 0.57 0.98 0.75
rs902774 469 2,752 0.91 0.64, 1.29 1.41 1.01, 1.96 1.22 0.87, 1.70 0.23 0.47 0.17 0.55
rs11649743 496 2,858 1.41 1.01, 1.97 1.39 0.99, 1.96 0.93 0.67, 1.30 0.09 0.08 0.64 0.35
rs4430796 489 2,802 1.67 1.28, 2.18 1.43 1.10, 1.85 1.05 0.82, 1.35 0.02 0.10 0.005 0.007
rs1859962 495 2,882 1.32 1.03, 1.69 1.30 1.01, 1.67 1.06 0.83, 1.36 0.26 0.51 0.89 0.45
rs8102476 495 2,796 0.91 0.72, 1.15 1.04 0.81, 1.33 1.14 0.88, 1.46 0.25 0.43 0.29 0.08
rs11672691 418 2,549 1.72 1.22, 2.42 0.80 0.58, 1.10 1.45 1.04, 2.01 0.46 0.52 0.46 0.91
rs2735839 483 2,795 1.21 0.83, 1.74 0.76 0.55, 1.06 0.92 0.63, 1.35 0.31 0.37 0.39 0.15
rs5759167 471 2,819 0.99 0.78, 1.26 1.08 0.84, 1.40 1.30 1.01, 1.69 0.12 0.17 0.11 0.03
rs11704416 422 2,564 1.32 0.98, 1.79 1.17 0.84, 1.65 1.24 0.90, 1.70 0.78 0.91 0.52 0.70
rs5945619 490 2,871 1.22 1.03, 1.46 1.05 0.87, 1.26 1.26 1.06, 1.51 0.77 0.76 0.40 0.58
rs5919432 320 1,851 0.88 0.68, 1.13 0.95 0.71, 1.26 1.27 0.94, 1.71 0.06 0.05 0.26 0.06
Additive SNP scoref 511 2,980 1.03 1.01, 1.05 1.03 1.01, 1.06 1.05 1.03, 1.08 0.22 0.04 0.66 0.95
Kernel machine scoreg 0.005 0.005 <0.001 <0.001
Open in a new tab

Abbreviations: 25(OH)D, 25-hydroxyvitamin D; CI, confidence interval; OR, odds ratio; SNP, single nucleotide polymorphism.

a From a logistic regression model of SNPs and total prostate cancer risk by cohort-, batch-, and season-specific tertiles of 25(OH)D concentrations, adjusted for age at blood draw, year of blood draw, cohort, and country (within the European Prospective Investigation into Cancer and Nutrition).

bP for interaction was calculated based on the case-control analysis and a variable for cohort-, batch-, and season-specific tertiles of 25(OH)D concentrations. Conventional P values are shown; only 1 P value was marginally statistically significant for rs620861 after allowance for multiple testing (P threshold = 0.001).

cP for interaction was calculated based on the case-control analysis and a continuous 25(OH)D variable (Rosner-standardized) for cohort, batch, and season. Conventional P values are shown; all P values were nonsignificant after allowance for multiple testing (P threshold = 0.001).

dP for interaction was calculated based on the case-only analysis for dichotomized SNPs and cohort-, batch-, and season-specific tertiles of 25(OH)D. Conventional P values are shown; only 2 P values were statistically significant (for rs620861 and the kernel machine score) after allowance for multiple testing. We used a P threshold of 0.0125 for the case-control analyses, applying a correction to the standard threshold of 0.05 to account for multiple testing on the basis of finding 4 nominally significant SNPs in the case-only analyses.

eP for interaction was calculated based on the case-only analysis for dichotomized SNPs and a continuous 25(OH)D variable (Rosner-standardized) for cohort, batch, and season. Conventional P values are shown; only 2 P values were statistically significant (for rs620861 and the kernel machine score) after allowance for multiple testing. We used a P threshold of 0.01 for the case-control analyses, applying a correction to the standard threshold of 0.05 to account for multiple testing on the basis of finding 5 nominally significant SNPs in the case-only analyses.

f From a logistic regression model of a continuous additive genetic score (after summing the number of risk alleles across the 46 SNPs for each participant) and fatal prostate cancer risk. For the case-only analysis, the additive genetic score was dichotomized at the median among controls.

g From a logistic regression kernel machine model across the entire set of 46 SNPs and fatal prostate cancer risk.

Multiplicative interactions for fatal prostate cancer are presented in Table 3. After correcting for multiple testing in the case-control approach with a Bonferroni threshold P value of 0.001, or after conducting the 2-step approach in the case-only design, we identified only 1 statistically significant interaction, for rs620861. The per-allele odds ratio for fatal prostate cancer was significantly lower for men in the lowest third of 25(OH)D (OR = 0.75, 95% CI: 0.58, 0.97), null for the second third (OR = 1.06, 95% CI: 0.81, 1.39), and significantly higher for the highest third (OR = 1.48, 95% CI: 1.11, 1.97; P for interaction = 0.001). In other words, the per-tertile association of 25(OH)D concentrations with fatal prostate cancer risk yielded a statistically significantly inverse odds ratio of 0.49 (95% CI: 0.33, 0.73) for men with the TT genotype, whereas the association was not significant for men with the CC genotype (OR = 1.15, 95% CI: 0.94, 1.41). A statistically significant interaction was observed for the kernel machine genetic score but not for the additive genetic score (Table 3).

When interactions on the additive scale were evaluated, nominal statistical significance was evident for 7 SNPs, but only rs620861 (P = 0.00001) and rs16902094 (P = 0.00001) survived the correction for multiple testing (Web Table 6). We observed the following marginal risk estimates for the association of rs620861 (dichotomized) with fatal prostate cancer (OR = 0.24) and the association of the cohort-, batch-, and season-specific tertiles of 25(OH)D (per tertile) with fatal prostate cancer (OR = 0.50), with an odds ratio for interaction of 2.10. Based on these estimates, we calculated a RERI of 0.51 (95% CI: 0.25, 0.70). The positive values indicate that the excess risk of fatal prostate cancer due to the presence of low concentrations of vitamin D is greater for those who have the risk allele (T) for rs620861 than those who do not have it, which agrees with the results from the interaction analysis on the multiplicative scale.

DISCUSSION

In this large pooled analysis, we investigated potential departure from multiplicative and additive interactions for 46 susceptibility SNPs, circulating concentrations of 25(OH)D, and the risk of total and fatal prostate cancer. We observed that the SNP and total prostate cancer associations did not differ by 25(OH)D concentrations on the multiplicative or additive scale after correction for multiple comparisons, although we found evidence of multiplicative and additive interaction between 25(OH)D and each of 2 SNPs in the 8q24 region (rs620861 and rs16902094) with risk of fatal prostate cancer.

The exact biological mechanisms behind such a potential interaction are unclear. The SNPs are located in chromosomal region 8q24, which is considered an intergenic region with pleiotropic associations with several cancers and other diseases (32). The mechanisms by which genetic variation in this region influences the risk of prostate cancer are not yet fully understood (33). Nonetheless, studies have shown that 8q24 physically interacts with the nearby proto-oncogene MYC (33–36), which is a well-defined oncogenic transcription factor and the most frequently amplified protein-coding gene across all cancer types (32). Loci on 8q24 act as tissue-specific regulators (enhancers) of MYC (34) and have been found to interact with the MYC promoter specifically in prostate cancer cell lines (35, 36).

There was no indication in the present analysis of an association of rs620861 or rs16902094 with concentrations of 25(OH)D, nor, to our knowledge, has such an association been mentioned in the literature, which could suggest that the observed interaction may be due to chance. However, there is ample literature on the regulatory role of vitamin D on the MYC gene. Vitamin D signaling can suppress expression of genes regulated by c-MYC, providing a molecular basis for the cancer-preventive actions of vitamin D (37). Furthermore, it has been shown that the active regulator of vitamin D—1,25(OH)2D3—down-regulates c-MYC and its transcription factor E2F, subsequently resulting in reduced growth of several prostate cancer cell lines (38–40). Future studies are needed—first to verify the observed intriguing interactions between 8q24 rs620861 and rs16902094, 25(OH)D concentration, and fatal prostate cancer risk, and second to shed light on the potential underlying biological mechanisms.

Reports of associations between circulating concentrations of vitamin D metabolites and risk of prostate cancer are inconsistent in the epidemiologic literature. A meta-analysis of 25 studies published in 2011 provided little evidence that 25(OH)D concentrations were associated with the risk of total or aggressive prostate cancer (12). Some recent prospective studies have also reported null associations between 25(OH)D and risk of any prostate cancer but inverse associations for aggressive or lethal disease (18, 41). However, other recent prospective studies observed positive associations for total disease and null associations for lethal disease (16, 42, 43) or a statistically significant U-shaped association for total and aggressive disease (44). A meta-analysis published in 2014 observed a statistically significant, 17% elevated risk of any prostate cancer for individuals with higher levels of 25(OH)D, but the investigators did not explore associations by stage and grade of the disease. Potential reasons for the inconsistencies in the literature addressing vitamin D and prostate cancer have been described in detail elsewhere (43); briefly they include the use of different 25(OH)D assays, single (instead of multiple) measurement of 25(OH)D, the different screening practices by country, and the large clinical heterogeneity of prostate cancer. Future studies should include consideration of prostate cancer mortality as the most clinically relevant prostate cancer endpoint (45).

To our knowledge, this is the first study to investigate potential interactions between GWAS-identified SNPs related to prostate cancer risk and circulating levels of 25(OH)D. We evaluated the strength of the evidence for the observed statistically significant interactions based on published guidelines (46). The strength of the literature evidence for the main association of 25(OH)D concentrations with risk of fatal prostate cancer can be considered weak, as few studies have examined fatal prostate cancer as an outcome, whereas the strength of the evidence for the association between the 2 SNPs in 8q24 (rs620861 and rs16902094) and risk of fatal disease is considered moderate; several GWASs have confirmed these findings for total prostate cancer, but results are sparse for fatal disease. In summary, this corresponds to a moderate a priori likelihood for the existence of an interaction. However, the overall strength of the evidence for an interaction is weak, given that replication is currently lacking and the evidence in the present analysis is based on only approximately 500 fatal prostate cancer cases.

These results imply a lack of robust interactions on the multiplicative or additive scales associating 46 prostate cancer susceptibility SNPs and 25(OH)D concentration with risk of either total or fatal prostate cancer. However, lack of statistical interaction does not imply lack of biological (causal) interaction. We cannot exclude the possibility that there may be modest or weak gene–vitamin D interactions that this study had insufficient statistical power to detect. Moreover, we tested only for 2-level interactions in the present study, and higher-order interactions may have been missed, although power to detect such interactions would be lower. BPC3 investigators are in a unique position to explore gene-environment interactions because BPC3 consists of 9 well-established cohort studies (of which 5 were included in this analysis) with prospectively collected blood specimens, high-quality biomarker assays, and genotyping data for thousands of participants. With 3,811 cases and 2,980 controls, this study had more than 80% power to detect a multiplicative interaction association of 1.7, assuming an allele frequency of 30% and a SNP or 25(OH)D main association with total prostate cancer of 1.1, but the power was reduced for fatal prostate cancer. Recently published GWASs have identified several prostate cancer SNPs other than the 46 SNPs studied here. Therefore, more studies with a larger number of participants are needed to reexamine our findings, to study untested GWAS-identified SNPs, and to evaluate the gene and vitamin D interactions for total and fatal prostate cancer in individuals with European ancestry and other ethnicities.

Overall, we did not find strong evidence that associations between GWAS-identified SNPs and prostate cancer are modified by circulating concentrations of 25(OH)D. The intriguing multiplicative interactions between rs620861 and rs16902094, 25(OH)D concentration, and fatal prostate cancer warrant replication.

Supplementary Material

Web Material
Click here for additional data file. (524.4KB, pdf)

ACKNOWLEDGMENTS

Author affiliations: Department of Hygiene and Epidemiology, School of Medicine, University of Ioannina, Ioannina, Greece (Vasiliki I. Dimitrakopoulou, Konstantinos K. Tsilidis); Cancer Epidemiology Unit, Nuffield Department of Population Health, University of Oxford, Oxford, United Kingdom (Ruth C. Travis, Timothy J. Key); Department of Epidemiology, Harvard T.H. Chan School of Public Health, Boston, Massachusetts (Irene M. Shui, J. Michael Gaziano, Meir Stampfer); Division of Public Health Sciences, Fred Hutchinson Cancer Research Center, Seattle, Washington (Irene M. Shui); Department of Epidemiology, School of Public Health, University of Michigan, Ann Arbor, Michigan (Alison Mondul); Division of Cancer Epidemiology and Genetics, National Cancer Institute, Bethesda, Maryland (Demetrius Albanes, Sonja I. Berndt, Amanda Black, Robert N. Hoover); Department of Chronic Disease Prevention, National Institute for Health and Welfare, Helsinki, Finland (Jarmo Virtamo); Unit of Nutrition, Environment and Cancer, Cancer Epidemiology Research Programme, Catalan Institute of Oncology (ICO-IDIBELL), Barcelona, Spain (Antonio Agudo); Department of Epidemiology, German Institute of Human Nutrition (DIfE), Potsdam-Rehbrücke, Germany (Heiner Boeing); Department for Determinants of Chronic Diseases, National Institute for Public Health and the Environment, Bilthoven, The Netherlands (H. Bas Bueno-de-Mesquita); Department of Gastroenterology and Hepatology, University Medical Centre, Utrecht, The Netherlands (H. Bas Bueno-de-Mesquita); Department of Epidemiology and Biostatistics, School of Public Health, Imperial College London, London, United Kingdom (H. Bas Bueno-de-Mesquita, Marc J. Gunter, Konstantinos K. Tsilidis); Department of Social and Preventive Medicine, Faculty of Medicine, University of Malaya, Kuala Lumpur, Malaysia (H. Bas Bueno-de-Mesquita); Genetic Epidemiology Group, Genetics Section, International Agency for Research on Cancer, Lyon, France (Mattias Johansson); Department of Biobank Research, Faculty of Medicine, Umeå University, Umeå, Sweden (Mattias Johansson); Cambridge Institute of Public Health, School of Clinical Medicine, University of Cambridge, Cambridge, United Kingdom (Kay-Tee Khaw); Section for Epidemiology, Department of Public Health, Aarhus University, Aarhus, Denmark (Kim Overvad); Molecular and Nutritional Epidemiology Unit, Cancer Research and Prevention Institute, Florence, Italy (Domenico Palli); Hellenic Health Foundation, Athens, Greece (Antonia Trichopoulou); Department of Hygiene, Epidemiology and Medical Statistics, University of Athens Medical School, Athens, Greece (Antonia Trichopoulou); Department of Nutrition, Harvard T.H. Chan School of Public Health, Boston, Massachusetts (Edward Giovannucci, David J. Hunter, Walter Willett); Department of Epidemiology, Harvard T.H. Chan School of Public Health, Boston, Massachusetts (Edward Giovannucci); Department of Medicine, Harvard Medical School, Boston, Massachusetts (Edward Giovannucci, J. Michael Gaziano); Program in Genetic Epidemiology and Statistical Genetics, Department of Epidemiology, Harvard T.H. Chan School of Public Health, Boston, Massachusetts (David J. Hunter, Sara Lindström, Peter Kraft); Division of Aging, Department of Medicine, Brigham and Women's Hospital, Boston, Massachusetts (J. Michael Gaziano); Early Detection Research Group, Division of Cancer Prevention, National Cancer Institute, Bethesda, Maryland (Christine Berg).

The Breast and Prostate Cancer Cohort Consortium was supported by the National Cancer Institute (U01-CA98233-07 to D.J.H., U01-CA98710-06 to M.J.T., U01-CA98216-06 to E.R. and R.K., U01-CA98758-07 to B.E.H., and the Intramural Research Program of the National Institutes of Health/National Cancer Institute to the Division of Cancer Epidemiology and Genetics). I.M.S. was supported by a Department of Defense Prostate Cancer Research Fellowship. The coordination of EPIC is financially supported by the European Commission (DG-SANCO) and the International Agency for Research on Cancer. The national cohorts are supported by Danish Cancer Society (Denmark); German Cancer Aid, German Cancer Research Center, Federal Ministry of Education and Research, Deutsche Krebshilfe, Deutsches Krebsforschungszentrum, and Federal Ministry of Education and Research (Germany); the Hellenic Health Foundation (Greece); Associazione Italiana per la Ricerca sul Cancro-AIRC-Italy and National Research Council (Italy); Dutch Ministry of Public Health, Welfare and Sports, Netherlands Cancer Registry (NKR), LK Research Funds, Dutch Prevention Funds, Dutch Zorg Onderzoek Nederland, World Cancer Research Fund, and Statistics Netherlands (The Netherlands); Health Research Fund (FIS) PI13/00061 (EPIC-Granada) and PI13/01162 (EPIC-Murcia) and the Regional Governments of Andalucía, Asturias, Basque Country, Murcia, and Navarra as well as ISCIII Health Research Funds RD12/0036/0018 (cofounded by FEDER funds/European Regional Development Fund ERDF) (Spain); Swedish Cancer Society, Swedish Research Council, and County Councils of Skåne and Västerbotten (Sweden); Cancer Research UK (14136 to EPIC-Norfolk; C570/A16491 and C8221/A19170 to EPIC-Oxford) and Medical Research Council (1000143 to EPIC-Norfolk, MR/M012190/1 to EPIC-Oxford) (UK). For information on how to submit an application for gaining access to EPIC data and/or biospecimens, please follow the instructions at http://epic.iarc.fr/access/index.php.

We thank Dr. Eleni-Ioanna Delatola from University College Dublin for programming support in the Unix environment.

Conflict of interest: none declared.

REFERENCES

  • 1. Al Olama AA, Kote-Jarai Z, Berndt SI, et al. . A meta-analysis of 87,040 individuals identifies 23 new susceptibility loci for prostate cancer. Nat Genet. 2014;46(10):1103–1109. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 2. Eeles R, Goh C, Castro E, et al. . The genetic epidemiology of prostate cancer and its clinical implications. Nat Rev Urol. 2014;11(1):18–31. [DOI] [PubMed] [Google Scholar]
  • 3. Lindstrom S, Schumacher F, Siddiq A, et al. . Characterizing associations and SNP-environment interactions for GWAS-identified prostate cancer risk markers—results from BPC3. PLoS One. 2011;6(2):e17142. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 4. Stevens VL, Ahn J, Sun J, et al. . HNF1B and JAZF1 genes, diabetes, and prostate cancer risk. Prostate. 2010;70(6):601–607. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 5. Tsilidis KK, Travis RC, Appleby PN, et al. . Interactions between genome-wide significant genetic variants and circulating concentrations of insulin-like growth factor 1, sex hormones, and binding proteins in relation to prostate cancer risk in the National Cancer Institute Breast and Prostate Cancer Cohort Consortium. Am J Epidemiol. 2012;175(9):926–935. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 6. Tsilidis KK, Travis RC, Appleby PN, et al. . Insulin-like growth factor pathway genes and blood concentrations, dietary protein and risk of prostate cancer in the NCI Breast and Prostate Cancer Cohort Consortium (BPC3). Int J Cancer. 2013;133(2):495–504. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 7. Giovannucci E. The epidemiology of vitamin D and cancer incidence and mortality: a review (United States). Cancer Causes Control. 2005;16(2):83–95. [DOI] [PubMed] [Google Scholar]
  • 8. Willis MS, Wians FH. The role of nutrition in preventing prostate cancer: a review of the proposed mechanism of action of various dietary substances. Clin Chim Acta. 2003;330(1-2):57–83. [DOI] [PubMed] [Google Scholar]
  • 9. Lokeshwar BL, Schwartz GG, Selzer MG, et al. . Inhibition of prostate cancer metastasis in vivo: a comparison of 1,23-dihydroxyvitamin D (calcitriol) and EB1089. Cancer Epidemiol Biomarkers Prev. 1999;8(3):241–248. [PubMed] [Google Scholar]
  • 10. Stewart LV, Weigel NL. Vitamin D and prostate cancer. Exp Biol Med (Maywood). 2004;229(4):277–284. [DOI] [PubMed] [Google Scholar]
  • 11. Yin L, Raum E, Haug U, et al. . Meta-analysis of longitudinal studies: serum vitamin D and prostate cancer risk. Cancer Epidemiol. 2009;33(6):435–445. [DOI] [PubMed] [Google Scholar]
  • 12. Gilbert R, Martin RM, Beynon R, et al. . Associations of circulating and dietary vitamin D with prostate cancer risk: a systematic review and dose-response meta-analysis. Cancer Causes Control. 2011;22(3):319–340. [DOI] [PubMed] [Google Scholar]
  • 13. Xu Y, Shao X, Yao Y, et al. . Positive association between circulating 25-hydroxyvitamin D levels and prostate cancer risk: new findings from an updated meta-analysis. J Cancer Res Clin Oncol. 2014;140(9):1465–1477. [DOI] [PubMed] [Google Scholar]
  • 14. Hunter DJ, Riboli E, Haiman CA, et al. . A candidate gene approach to searching for low-penetrance breast and prostate cancer genes. Nat Rev Cancer. 2005;5(12):977–985. [DOI] [PubMed] [Google Scholar]
  • 15. Shui IM, Lindström S, Kibel AS, et al. . Prostate cancer (PCa) risk variants and risk of fatal PCa in the National Cancer Institute Breast and Prostate Cancer Cohort Consortium. Eur Urol. 2014;65(6):1069–1075. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 16. Albanes D, Mondul AM, Yu K, et al. . Serum 25-hydroxy vitamin D and prostate cancer risk in a large nested case-control study. Cancer Epidemiol Biomarkers Prev. 2011;20(9):1850–1860. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 17. Travis RC, Crowe FL, Allen NE, et al. . Serum vitamin D and risk of prostate cancer in a case-control analysis nested within the European Prospective Investigation into Cancer and Nutrition (EPIC). Am J Epidemiol. 2009;169(10):1223–1232. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 18. Shui IM, Mucci LA, Kraft P, et al. . Vitamin D-related genetic variation, plasma vitamin D, and risk of lethal prostate cancer: a prospective nested case-control study. J Natl Cancer Inst. 2012;104(9):690–699. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 19. Li H, Stampfer MJ, Hollis JB, et al. . A prospective study of plasma vitamin D metabolites, vitamin D receptor polymorphisms, and prostate cancer. PLoS Med. 2007;4(3):e103. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 20. Ahn J, Albanes D, Berndt SI, et al. . Vitamin D-related genes, serum vitamin D concentrations and prostate cancer risk. Carcinogenesis. 2009;30(5):769–776. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 21. Rosner B, Cook N, Portman R, et al. . Determination of blood pressure percentiles in normal-weight children: some methodological issues. Am J Epidemiol. 2008;167(6):653–666. [DOI] [PubMed] [Google Scholar]
  • 22. Benjamini Y, Hochberg Y. Controlling the false discovery rate: a practical and powerful approach to multiple testing. J R Stat Soc Ser B Stat Methodol. 1995;57(1):289–300. [Google Scholar]
  • 23. Gatto NM, Campbell UB, Rundle AG, et al. . Further development of the case-only design for assessing gene-environment interaction: evaluation of and adjustment for bias. Int J Epidemiol. 2004;33(5):1014–1024. [DOI] [PubMed] [Google Scholar]
  • 24. Lee WC, Wang LY, Cheng KF. An easy-to-implement approach for analyzing case-control and case-only studies assuming gene-environment independence and Hardy-Weinberg equilibrium. Stat Med. 2010;29(24):2557–2567. [DOI] [PubMed] [Google Scholar]
  • 25. Hosmer DW, Lemeshow S. Confidence interval estimation of interaction. Epidemiology. 1992;3(5):452–456. [DOI] [PubMed] [Google Scholar]
  • 26. Assmann SF, Hosmer DW, Lemeshow S, et al. . Confidence intervals for measures of interaction. Epidemiology. 1996;7(3):286–290. [DOI] [PubMed] [Google Scholar]
  • 27. Knol MJ, van der Tweel I, Grobbee DE, et al. . Estimating interaction on an additive scale between continuous determinants in a logistic regression model. Int J Epidemiol. 2007;36(5):1111–1118. [DOI] [PubMed] [Google Scholar]
  • 28. Efron B, Tibshirani RJ. An Introduction to the Bootstrap. Boca Raton, FL: Chapman & Hall/CRC; 1993. [Google Scholar]
  • 29. Lin X, Cai T, Wu MC, et al. . Kernel machine SNP-set analysis for censored survival outcomes in genome-wide association studies. Genet Epidemiol. 2011;35(7):620–631. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 30. Liu D, Ghosh D, Lin X. Estimation and testing for the effect of a genetic pathway on a disease outcome using logistic kernel machine regression via logistic mixed models. BMC Bioinformatics. 2008;9:292. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 31. Wu MC, Kraft P, Epstein MP, et al. . Powerful SNP-set analysis for case-control genome-wide association studies. Am J Hum Genet. 2010;86(6):929–942. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 32. Huppi K, Pitt JJ, Wahlberg BM, et al. . The 8q24 gene desert: an oasis of non-coding transcriptional activity. Front Genet. 2012;3:69. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 33. Li Q, Liu X, Hua RX, et al. . Association of three 8q24 polymorphisms with prostate cancer susceptibility: evidence from a meta-analysis with 50,854 subjects. Sci Rep. 2015;5:12069. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 34. Ahmadiyeh N, Pomerantz MM, Grisanzio C, et al. . 8q24 prostate, breast, and colon cancer risk loci show tissue-specific long-range interaction with MYC. Proc Natl Acad Sci USA. 2010;107(21):9742–9746. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 35. Grisanzio C, Freedman ML. Chromosome 8q24-associated cancers and MYC. Genes Cancer. 2010;1(6):555–559. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 36. Sotelo J, Esposito D, Duhagon MA, et al. . Long-range enhancers on 8q24 regulate c-Myc. Proc Natl Acad Sci USA. 2010;107(7):3001–3005. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 37. Salehi-Tabar R, Nguyen-Yamamoto L, Tavera-Mendoza LE, et al. . Vitamin D receptor as a master regulator of the c-MYC/MXD1 network. Proc Natl Acad Sci USA. 2012;109(46):18827–18832. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 38. Toropainen S, Väisänen S, Heikkinen S, et al. . The down-regulation of the human MYC gene by the nuclear hormone 1alpha,25-dihydroxyvitamin D3 is associated with cycling of corepressors and histone deacetylases. J Mol Biol. 2010;400(3):284–294. [DOI] [PubMed] [Google Scholar]
  • 39. Wang WL, Welsh J, Tenniswood M. 1,25-Dihydroxyvitamin D3 modulates lipid metabolism in prostate cancer cells through miRNA mediated regulation of PPARA. J Steroid Biochem Mol Biol. 2013;136:247–251. [DOI] [PubMed] [Google Scholar]
  • 40. Washington MN, Kim JS, Weigel NL. 1α,25-dihydroxyvitamin D3 inhibits C4-2 prostate cancer cell growth via a retinoblastoma protein (Rb)-independent G1 arrest. Prostate. 2011;71(1):98–110. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 41. Gilbert R, Metcalfe C, Fraser WD, et al. . Associations of circulating 25-hydroxyvitamin D with prostate cancer diagnosis, stage and grade. Int J Cancer. 2012;131(5):1187–1196. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 42. Mondul AM, Shui IM, Yu K, et al. . Genetic variation in the vitamin D pathway in relation to risk of prostate cancer—results from the Breast and Prostate Cancer Cohort Consortium. Cancer Epidemiol Biomarkers Prevention. 2013;22(4):688–696. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 43. Shui IM, Mondul AM, Lindström S, et al. . Circulating vitamin D, vitamin D-related genetic variation, and risk of fatal prostate cancer in the National Cancer Institute Breast and Prostate Cancer Cohort Consortium. Cancer. 2015;121(12):1949–1956. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 44. Kristal AR, Till C, Song X, et al. . Plasma vitamin D and prostate cancer risk: results from the Selenium and Vitamin E Cancer Prevention Trial. Cancer Epidemiol Biomarkers Prev. 2014;23(8):1494–1504. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 45. Jahn JL, Giovannucci EL, Stampfer MJ. The high prevalence of undiagnosed prostate cancer at autopsy: implications for epidemiology and treatment of prostate cancer in the prostate-specific antigen era. Int J Cancer. 2015;137(12):2795–2802. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 46. Boffetta P, Winn DM, Ioannidis JP, et al. . Recommendations and proposed guidelines for assessing the cumulative evidence on joint effects of genes and environments on cancer occurrence in humans. Int J Epidemiol. 2012;41(3):686–704. [DOI] [PMC free article] [PubMed] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

Web Material
Click here for additional data file. (524.4KB, pdf)

Articles from American Journal of Epidemiology are provided here courtesy of Oxford University Press

RESOURCES