Genetic Predictors of Circulating 25-Hydroxyvitamin D and Prognosis after Colorectal Cancer

Colorectal cancer belongs to the most common cancer types (third) and is the second leading cause of cancer-related death globally (1). Despite improved therapy regimens which have led to increased survival after diagnosis, survival time is still limited for advanced stages (2).

One of the factors with potential prognostic relevance is 25-hydroxyvitamin D [25(OH)D] levels because the vitamin D receptor is highly expressed in the colon (3, 4). The most stable and therefore most reliable indicator of circulating vitamin D is 25(OH)D; it is influenced by both dietary intake and skin synthesis by sun exposure (5). Low levels of 25(OH)D have been consistently found associated with reduced overall [HR = 0.68; 95% confidence interval (CI), 0.55–0.85] and colorectal cancer–specific survival (HR = 0.67; 95% CI, 0.57–0.78), as evidenced in a recent meta-analysis (6). Therefore, vitamin D status could be a potential modifiable factor for improving prognosis in patients with colorectal cancer.

It is unclear whether a genetic predisposition of higher/lower vitamin D levels is involved in mechanisms leading to better survival or whether low vitamin D levels are primarily an indicator for poor health (7). Studies assessing postdiagnostic vitamin D concentrations need to be interpreted with caution as lower vitamin D levels in patients with poorer health could also be due to behavior changes after diagnosis and treatment, for example, less sun exposure. The association between postdiagnostic vitamin D and survival after colorectal cancer in observational studies could thus reflect confounding or reverse causation.

Previous genome-wide association studies (GWAS) have identified six SNPs associated with circulating 25(OH)D levels at genome-wide significance (P value < 5 × 10⁻⁸; refs. 8, 9). The four SNPs with the strongest influence on vitamin D levels have been replicated by several studies (8–10). As genetic variants are randomly allocated during gamete formation independent of environmental factors, confounding should not play a role. Therefore, determining associations using vitamin D-related genetic variants could help minimize confounding bias and reverse causation.

We aimed to estimate the relationship of these six vitamin D-related genetic variants with overall and colorectal cancer–specific survival in studies collaborating in the International Survival Analysis in Colorectal Cancer Consortium (ISACC).

This analysis is based on 7,657 patients with colorectal cancer from 10 studies with available genotyping and follow-up data participating in the ISACC Consortium. The studies are Darmkrebs: Chancen der Verhuetung durch Screening (DACHS; refs. 11, 12), Diet, Activity and Lifestyle Study (DALS; ref. 13), Cancer Prevention Study II Nutrition cohort (CPSII; ref. 14), Health Professionals Follow-up Study (HPFS; ref. 15), Nurses' Health Study (NHS; refs. 16–18), Physicians Health Study (PHS; ref. 19); Prostate, Lung, Colorectal and Ovarian Cancer Screening Trial (PLCO; refs. 20, 21), Post-Menopausal Hormone- Seattle Colon Cancer Family Registry Study (PMH-CCFR), VITamins And Lifestyle Study (VITAL; ref. 22), and Woman's Health Initiative (WHI; ref. 23; Supplementary Table S1). Nine of the 10 studies are also included in the Genetics and Epidemiology of Colorectal Cancer Consortium (GECCO; ref. 24) and one study, CPS II, is part of the Colorectal Cancer Transdisciplinary consortium (CORECT; ref. 25). Details on the consortia and studies have been described previously (14, 24, 25) and study descriptions are provided in the supplementary information (Supplementary Table S1). Participant overlap between the studies has been excluded. All participants provided written, informed consent and studies were approved by their respective institutional review boards. Demographic and lifestyle factors were queried by in-person interviews or self-filled questionnaires. A multistep data harmonization procedure was carried out centrally for pooled analyses (26). Details of assessment of survival in the individual studies have been previously published (11, 13–15, 18, 22, 27–30). In short, vital status was obtained either using active follow-up with confirmation of death by review of death certificates or medical records or the studies used linkage to state death records or state cancer registries. Alive patients were censored at the date of last follow-up or data linkage.

All included studies provided genotype information. See previously published reports for details on genotyping, quality assurance, and imputation (24, 25, 31). Imputation was conducted using the imputation panel of the Haplotype Reference Consortium (32). Exclusion of SNPs was based on call rate (<98% in GECCO; <95% in CORECT), Hardy–Weinberg equilibrium in controls (P value < 1 × 10⁻⁴) or low minor allele frequency (≤1%). Patients were assigned values of 0, 1, or 2 for having 0 (wild-type homozygous), 1 (heterozygous), or 2 (homozygous for the risk allele) alleles associated with lower vitamin D levels for each SNP. For imputed SNPs, patients received continuous values between 0 and 2.

Six SNPs, which were found associated with 25(OH)D levels at genome-wide significance (P value <5 × 10⁻⁸) in a GWAS of European populations, were selected (refs. 8–10; Table 1). Two of these SNPs have been recently discovered by the SUNLIGHT Consortium (8), which also confirmed the other four previously identified SNPs (9). The six SNPs explain around 2.8% of the variance in circulating 25(OH)D levels (8).

Linkage disequilibrium between the individual SNPs was checked, and no strong correlation was detected (R² < 0.01). A weighted genetic risk score (GRS) consisting of the six SNPs was calculated for each person as the sum of the number of vitamin D decreasing alleles weighted by their effect on vitamin D levels as reported in Jiang and colleagues (8). As Jiang and colleagues used natural log-transformed vitamin D levels as outcome in their GWAS, units of the GRS are not easily interpretable. The risk score ranges between 0 and 0.432. Results are reported per one unit increase in risk score.

According to Milaneschi and colleagues (33), a one unit increase in this risk score is associated with a change of 5.29 nmol/L (95% CI, 4.18–6.39).

The association of the individual SNPs with overall survival and CRC-specific survival was assessed using Cox proportional hazard models. Analyses were adjusted for age, sex, and principal components (PC) of genetic ancestry to control for potential confounding due to population substructure (PCs were calculated using the EIGENSTRAT method; https://reich.hms.harvard.edu/software). Three PCs were used for analysis of GECCO studies, and 10 PCs were used for CPSII from the CORECT consortium. Analyses were conducted for GECCO and CORECT studies separately. Fixed-effects meta-analysis was employed to get summary results for GECCO and CORECT. Bonferroni correction was used to account for multiple testing of six single SNP tests (P < 0.05/6 = 0.0083). Additional stratified analyses were carried out by sex, cancer site (colon/rectum), stage (stage 1: local; stage 2, 3: regional; stage 4: distant) and BMI categories (self-reported prediagnostic BMI) in kg/m² (BMI 18.5–24.9 for normal weight, 25–30 for overweight, >30 for obese, excluding the 1% with BMI <18.5). Heterogeneity was assessed using likelihood ratio tests, which compare the models including/excluding interaction term.

To calculate time-to-event, date of diagnosis was considered as the starting point and follow-up time was censored at death or end of follow-up, whichever occurred first. For calculation of colorectal cancer–specific survival, censoring was done at the time of death for all patients who died from other causes than colorectal cancer. The reverse Kaplan–Meier method was employed for calculation of median follow-up time (34).

Genotyping data of the GECCO studies are available at the database of genotypes and phenotypes (dbGaP) for download at the accession number: phs001078.v1.p1.

Of 7,657 patients with colorectal cancer included for this analysis, 2,438 died from any cause and 1,648 died from colorectal cancer after a median follow-up time of 54.8 month (interquartile range: 27.7–73.6 months). Supplementary Table S2 shows selected characteristics of the study population and numbers per GECCO/CPSII. Among patients included, 54.6% were women and 45.4% were men.

The association of the vitamin D associated SNPs with overall and with colorectal cancer–specific survival were similar, see Table 2. None of the six SNPs were statistically significantly associated with overall survival. There was a significant association between SNP rs2282679 and colorectal cancer–specific survival (HR = 1.08; 95% CI, 1.00–1.16; Table 2), which did not remain significant when applying a Bonferroni corrected P-value of <0.0083. None of the other five SNPs showed significant associations with colorectal cancer–specific survival. Results per GECCO/CPSII studies are shown in Supplementary Table S3 for overall survival and colorectal cancer–specific survival.

The GRS representing genetically determined lower levels of vitamin D was not significantly associated with risk of death after colorectal cancer diagnosis (HR per one unit of GRS = 1.54; 95% CI, 0.86–2.78; Table 3). For colorectal cancer-specific survival, a similar nonsignificant association was found for lower genetically determined levels of vitamin D (HR per one unit of GRS = 1.76; 95% CI, 0.86–3.58).

Exploration of effect heterogeneity yielded differential associations for overall survival according to sex and BMI (Table 3). A higher GRS was significantly associated with increased overall mortality (HR = 3.26; 95% CI, 1.45–7.33) in women and not in men (HR = 0.68; 95% CI, 0.29–1.62). In addition, the association of higher GRS with increased overall mortality was only significant in normal weight patients (HR = 4.14; 95% CI, 1.50–11.43) but not obese patients (HR = 1.67; 95% CI, 0.53–5.26). In overweight patients, the SNP association was in the opposite direction (HR = 0.55; 95% CI, 0.21–1.45). No significant differential associations were found for colorectal cancer–specific survival, but associations in the subgroups were generally similar to that for overall survival in magnitude and direction (Table 3).

In this large study, we investigated the association between six genetic variants and a GRS associated with vitamin D levels and overall and colorectal cancer–specific survival. The single SNPs and GRS were not significantly associated with overall/colorectal cancer–specific survival. We found significant effect heterogeneity by sex and BMI in the association of GRS for vitamin D and overall survival.

Recent large observational studies reported inverse associations between serum 25(OH)D levels and survival after colorectal cancer (4, 35, 36). Except for one study that used prediagnostic 25(OH)D levels (35), all studies used postdiagnostic vitamin D levels which could already have been influenced by the disease or a poor health status after treatment (4, 36). We and others have thus employed vitamin D–related genetic variants as instrumental variables to evaluate the association with colorectal cancer mortality to minimize bias due to reverse causation and confounding.

A recent study that investigated only rs2282679 (GC gene) found a nonsignificant association with reduced overall survival in 489 patients with colorectal cancer, corroborating our findings. In addition, they reported a significant association of the vitamin D lowering allele with poorer disease-free survival (37), however, colorectal cancer–specific survival was not investigated. A further study that investigated three SNPs rs2282679 (GC), rs10741657 (CYP2R1), and rs12785878 (DHCR7) in association with time to recurrence in patients with stages II and III colon cancer found only rs2282679 to be significantly associated with decreased time to recurrence in the subgroup of patients who underwent only surgery (38).

Mendelian randomization (MR) studies on vitamin D levels have also been conducted. These MR studies were not conducted specifically in patients with colorectal cancer and colorectal cancer–specific death was not investigated as outcome in any of these studies. A former large MR study of 95 766 participants investigated the association of SNPs associated with lower vitamin D levels (using a GRS score of two SNPs in the DHCR7 gene and two SNPs in the CYP2R1 gene explaining around 1% in variation in vitamin D levels) with all-cause mortality and cancer-specific mortality and found that the genetic instrument was associated with increased all cause and cancer-specific mortality (39). In contrast to that, a study in women of 33,682 participants and 3,985 cancer cases found no significant association of a five SNP instrument and overall or cancer mortality (40). A recent study using UK Biobank data (438,870 participants and 6,998 cancer-specific deaths) did not find evidence for genetically determined low vitamin D levels (using five vitamin D–related variants, four of them overlapping with the variants used in our study) and increased cancer mortality (41). The variance in vitamin D levels explained by a genetic instrument using the previously known four SNPs, which were confirmed by Jiang and colleagues, was estimated to be approximately 5% (42). Jiang and colleagues estimated that the six SNPs used in our study explain around 2.8% of genetic variation; 7.5% of variation was explained by all common GWAS variants (8).

We found sex-differences in the association of the GRS with overall survival. Some observational studies have also observed differences by sex (43) or significant associations of vitamin D levels with cancer incidence (44) or mortality only in women (45) but not in men. In this study, the GRS conferring lower vitamin D levels was found associated with worse prognosis only in women. One reason for this could be stronger associations of vitamin D levels with diseases that are more common in women than in men, like breast cancer (46) and osteoporosis (47). But whether the vitamin D–related SNPs are also associated with breast cancer mortality or mortality after osteoporosis (e.g., as consequence of fractures) is unclear. A lookup in the BCAC consortium including 42,124 patients with 3,733 breast cancer–specific deaths showed that rs2282679 is not associated with breast cancer–specific mortality (http://bcac.ccge.medschl.cam.ac.uk/). A SNP in the vitamin D receptor gene, which was previously reported to be significantly associated with higher breast cancer–specific mortality in 498 patients with breast cancer (48), has not been found associated with vitamin D levels in GWAS. In addition, our finding of higher mortality with a GRS for lower 25(OH)D levels only in normal weight people is interesting and it is consistent with results of a recent RCT in the VITAL study of vitamin D supplementation and all cancer incidence (49) which found lower all-cancer incidence only in normal weight participants receiving vitamin D compared with placebo and not in overweight or obese participants. It is known that overweight and obese people have lower levels of vitamin D compared with normal weight people (7), which could be due to sequestration of vitamin D in adipose tissues or dilution of ingested vitamin D (7). In this case higher levels of 25(OH)D would be needed for protection in overweight/obese individuals. Also there is a difference in direction of point estimates when comparing the overweight and obese groups. Because the CIs are large, these differences could be due to chance because of power issues for the subgroup analyses and therefore needs to be investigated in further larger studies.

Regarding mortality after colorectal cancer diagnosis, RCTs of vitamin D supplementation are also being conducted. Two recent meta-analysis reported that vitamin D supplementation reduced all-cancer death by 16% (50) and 13% (51), colorectal cancer death was not specifically investigated. One small RCT has been published (52) on vitamin D supplementation in patients with colorectal cancer, which was conducted in Croatia and randomized 71 patients with metastatic colorectal cancer to either standard chemotherapy or standard chemotherapy plus 2000IU vitamin D. No difference in overall or progression-free survival between groups was observed (52). A further RCT of 139 patients with metastatic colorectal cancer reported in a conference abstract a longer progression-free survival in patients who received high vitamin D supplementation compared with low vitamin D supplementation (53). A more recent RCT found suggestive evidence that high dose compared with standard-dose vitamin D supplementation combined with chemotherapy could improve survival for patients with advanced or metastatic colorectal cancer (54). Supplementation for patients with colorectal cancer might have a different effect on mortality after colorectal cancer compared with genetically determined differences in vitamin D levels, and this RCT suggests that higher than physiologic levels may lead to improvement in survival.

Strengths of our study include the large sample size, which is the largest published sample size among colorectal cancer survivors to date for investigation of the association of vitamin D–related genetic variants with overall and colorectal cancer–specific survival. We used the latest published set of SNPs discovered in GWAS to be associated with vitamin D levels. The association of vitamin D–related SNPs with colorectal cancer–specific survival has been investigated for the first time. Moreover, all included studies carried out a comprehensive follow-up with long follow-up duration. Our study has also some limitations. As this study is based on populations of European descent, the generalizability to other populations is limited. The advantage is that we were able to minimize bias due to confounding by population stratification. We did not have serum measurements of vitamin D levels available for all included studies; therefore, we were not able to analyze the strength of the association of the SNPs with vitamin D levels in our study. Although our study has a large sample size, power to conduct a Mendelian randomization study is still limited due to the moderate association of the SNPs with decreasing vitamin D levels.

In conclusion, this is the first study to examine the most recent set of vitamin D-related SNPs discovered in GWAS with respect to overall and also colorectal cancer–specific mortality among patients with colorectal cancer. We did not find evidence for an association of genetically determined lower vitamin D levels and higher mortality. The potential effect heterogeneity by sex and BMI requires confirmation. Further larger studies are warranted to investigate the association of vitamin D levels and survival after colorectal cancer also in specific subgroups.

R.T. Chlebowski is a consultant for Pfizer, Immunomedics, AstraZeneca, Novartis, and Amgen and reports receiving speakers bureau honoraria from Novartis, AstraZeneca, and Genentech. No potential conflicts of interest were disclosed by the other authors.

Conception and design: S. Neumeyer, J.E. Manson, J.C. Figueiredo, J.L. Hopper, F.A. Macrae, U. Peters, J. Chang-Claude

Development of methodology: S. Neumeyer, J.C. Figueiredo, J.L. Hopper, P.A. Newcomb, J. Chang-Claude

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): S.I. Berndt, P.T. Campbell, R.T. Chlebowski, A.T. Chan, E.L. Giovannucci, S. Ogino, M. Song, M.L. McCullough, H. Maalmi, J.E. Manson, R.E. Schoen, M.L. Slattery, E. White, A.K. Win, J.C. Figueiredo, J.L. Hopper, F.A. Macrae, U. Peters, H. Brenner, M. Hoffmeister, P.A. Newcomb, J. Chang-Claude

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): B.L. Banbury, P.T. Campbell, R.T. Chlebowski, S. Ogino, M. Song, J.E. Manson, E. White, A.K. Win, J.C. Figueiredo, F.A. Macrae, U. Peters, H. Brenner, J. Chang-Claude

Writing, review, and/or revision of the manuscript: S. Neumeyer, B.L. Banbury, S.I. Berndt, P.T. Campbell, R.T. Chlebowski, A.T. Chan, E.L. Giovannucci, A.D. Joshi, S. Ogino, M. Song, M.L. McCullough, H. Maalmi, J.E. Manson, L.C. Sakoda, R.E. Schoen, M.L. Slattery, E. White, A.K. Win, J.C. Figueiredo, J.L. Hopper, U. Peters, H. Brenner, M. Hoffmeister, P.A. Newcomb, J. Chang-Claude

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): K. Butterbach, P.T. Campbell, R.T. Chlebowski, J.L. Hopper, P.A. Newcomb, J. Chang-Claude

Study supervision: J.C. Figueiredo, J.L. Hopper, M. Hoffmeister, J. Chang-Claude

CPS-II: The authors thank the CPS-II participants and Study Management Group for their invaluable contributions to this research. The authors also acknowledge the contribution to this study from central cancer registries supported through the Centers for Disease Control and Prevention National Program of Cancer Registries, and cancer registries supported by the NCI Surveillance Epidemiology and End Results program.

DACHS: The authors thank all participants and cooperating clinicians, and Ute Handte-Daub, Utz Benscheid, Muhabbet Celik, and Ursula Eilber for excellent technical assistance.

Harvard cohorts (HPFS, NHS, PHS): The study protocol was approved by the institutional review boards of the Brigham and Women's Hospital and Harvard T.H. Chan School of Public Health, and those of participating registries as required. The authors thank the participants and staff of the HPFS, NHS, and PHS for their valuable contributions as well as the following state cancer registries for their help: AL, AZ, AR, CA, CO, CT, DE, FL, GA, ID, IL, IN, IA, KY, LA, ME, MD, MA, MI, NE, NH, NJ, NY, NC, ND, OH, OK, OR, PA, RI, SC, TN, TX, VA, WA, WY. The authors assume full responsibility for analyses and interpretation of these data.

PLCO: The authors thank the PLCO Cancer Screening Trial screening center investigators and the staff from Information Management Services Inc. and Westat Inc. Most importantly, they thank the study participants for their contributions that made this study possible.

PMH-SCCFR: The authors thank the study participants and staff of the Hormones and Colon Cancer and Seattle Cancer Family Registry studies (CORE Studies).

WHI: The authors thank the WHI investigators and staff for their dedication, and the study participants for making the program possible. A full listing of WHI investigators can be found at: http://www.whi.org/researchers/Documents%20%20Write%20a%20Paper/WHI%20Investigator%20Long%20List.pdf.

Fred Hutch core grant: This research was funded in part through the NIH/NCI Cancer Center Support Grant P30 CA015704.

Genetics and Epidemiology of Colorectal Cancer Consortium (GECCO): NCI, NIH, U.S. Department of Health and Human Services (U01 CA137088, R01 CA059045, R01 CA176272).

CPSII: The American Cancer Society funds the creation, maintenance, and updating of the Cancer Prevention Study-II (CPS-II) cohort. This study was conducted with Institutional Review Board approval.

DACHS: This work was supported by the German Research Council (BR 1704/6-1, BR 1704/6- 3, BR 1704/6-4, CH 117/1-1, HO 5117/2-1, HE 5998/2-1, KL 2354/3-1, RO 2270/8-1, and BR 1704/17-1); the Interdisciplinary Research Program of the National Center for Tumor Diseases (NCT), Germany; and the German Federal Ministry of Education and Research (01KH0404, 01ER0814, 01ER0815, 01ER1505A, and 01ER1505B).

DALS: NIH (R01 CA48998 to M.L. Slattery).

Harvard cohorts (HPFS, NHS, PHS): HPFS is supported by the NIH (P01 CA055075, UM1 CA167552, U01 CA167552, R01 CA137178, R01 CA151993, and R35CA197735), NHS by the NIH (R01 CA137178, P01 CA087969, UM1 CA186107, R01 CA151993, and R35 CA197735), and PHS by the NIH (R01 CA042182).

PLCO: Intramural Research Program of the Division of Cancer Epidemiology and Genetics and supported by contracts from the Division of Cancer Prevention, NCI, NIH.

PMH-SCCFR: NIH (R01 CA076366 to P.A. Newcomb and U01 CA074794 to J. Potter).

VITAL: NIH (K05 CA154337).

WHI: The WHI program is funded by the National Heart, Lung, and Blood Institute, NIH, U.S. Department of Health and Human Services through contracts HHSN268201100046C, HHSN268201100001C, HHSN268201100002C, HHSN268201100003C, HHSN268201100004C, and HHSN271201100004C.

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