A Collaborative Analysis of Individual Participant Data from 19 Prospective Studies Assesses Circulating Vitamin D and Prostate Cancer Risk

It has been hypothesized that vitamin D deficiency may increase prostate cancer risk (1, 2). A meta-analysis of 6 prospective studies published up to 2010 reported that circulating vitamin D concentrations were not related to prostate cancer risk (3); however, it was insufficiently powered to provide robust estimates of risk, especially for important disease subgroups. Although the active hormonal form of vitamin D is 1,25-dihydroxyvitamin D [1,25(OH)₂D], which is mainly formed by hydroxylation of 25-hydroxyvitamin D [25(OH)D] in the kidney under the control of parathyroid hormone, circulating 25(OH)D concentration is regarded as the most informative indicator of vitamin D status.

The Endogenous Hormones, Nutritional Biomarkers and Prostate Cancer Collaborative Group was established to conduct collaborative reanalysis of individual data from prospective studies on the relationships of circulating hormone concentrations and nutritional biomarkers with prostate cancer risk (4, 5). With pooled individual participant data on prediagnostic circulating 25(OH)D and 1,25(OH)₂D concentrations from 19 prospective studies (with up to 13,462 men with incident prostate cancer), this analysis aimed to provide precise estimates of the association of circulating vitamin D with prostate cancer risk and to investigate whether these associations differed by tumor characteristics or time from blood collection to diagnosis. We also examined the cross-sectional relationships between lifestyle factors and vitamin D concentrations.

Published and unpublished studies were eligible for the current analysis if they had data on prediagnostic circulating concentrations of 25(OH)D or 1,25(OH)₂D and incident prostate cancers. Studies were identified using literature search methods from computerized bibliographic systems and by discussion with collaborators, as described previously (4, 5). Data were available for 19 prospective studies by dataset closure in May 2018.

Individual participant data were requested on circulating 25(OH)D and 1,25(OH)₂D, date, age, and fasting status at sample collection, marital status, ethnicity, educational attainment, family history of prostate cancer, height, weight, waist and hip circumference, smoking status, alcohol intake, and vital status. Each study also provided data on prostate cancer stage and grade and death, if available, and the data were harmonized in a central database. Further details on data collection and processing are provided in the Supplementary Methods.

The characteristics of the included studies are shown in Supplementary Table S1, and details of the assay methods are shown in Supplementary Table S2. Most of the studies were case–control studies nested within prospective cohort studies. Data on the control participants from The Prostate Testing for Cancer and Treatment (ProtecT) trial are included in cross-sectional analyses of vitamin D concentrations in relation to participant characteristics, but because cases were diagnosed at the start of the study rather than during follow-up, these data were not included in the main risk analyses. Written-informed consent was obtained from study participants at entry into each cohort or was implied by participants' return of the enrolment questionnaire. The study protocols were approved by Institutional Review Boards of each study center.

Prostate cancer was defined as being “early” stage if it was tumor–node–metastasis (TNM) stage T1 with no reported lymph node involvement or metastases, or stage I; “other localized” stage if it was TNM stage T2 with no reported lymph node involvement or metastases, stage II, or the equivalent; “advanced” stage if it was TNM stage T3 or T4 and/or N1+ and/or M1, stage III–IV, or the equivalent; or stage unknown. Aggressive disease was categorized as “no” for TNM stage T0, T1, T2, or T3 with no reported lymph node involvement and no metastases or equivalent, “yes” for TNM stage T4 and/or N1+ and/or M1 and/or stage IV disease and/or death from prostate cancer, or “unknown.” Histologic grade was defined as “low-intermediate” if the Gleason sum was < 8 or equivalent, “high” grade if the Gleason sum was ≥ 8 or equivalent, or grade “unknown.” Fatal cases were men who died of prostate cancer during follow-up.

25(OH)D and 1,25(OH)₂D concentrations were log-transformed to approximate a normal distribution for parametric analyses. To allow for the influence of month of blood draw on circulating concentrations, a regression model of log-transformed vitamin D concentration by month of blood collection was fitted for each study. All results are presented by season-standardized vitamin D, unless otherwise specified.

The main method of analysis was logistic regression conditioned on the matching variables within each study. Men were categorized into fifths of the distribution of 25(OH)D and 1,25(OH)₂D, with cut-points defined by the study-specific quintiles of the distribution within control participants, to allow for any systematic differences between the studies in assay methods and blood sample types (6). Linear trends were calculated by replacing the categorical variable representing the fifths of each analyte with a continuous variable that was scored as 0, 0.25, 0.5, 0.75, and 1; a unit increase in this variable can be taken to represent an 80 percentile increase in the study-specific concentration of vitamin D. To examine the effects of potential confounders (other than the matching criteria, which were taken into account in the study design and matched analyses), conditional logistic regression analyses included the following covariates: age at blood collection, body mass index (BMI), height, marital status, educational status, and cigarette smoking, all of which were associated with prostate cancer risk in these analyses.

In a sensitivity analysis, conditional logistic regression models were also fitted using quintile cut-points defined by the overall distribution among the control participants in all studies combined. The analyses were also repeated using predefined categories for concentrations of 25(OH)D of <30, 30–<50, 50–<75, and ≥75 nmol/L, in order to investigate risks associated with very low (deficiency), low (insufficiency), moderate (sufficiency), and high circulating concentrations of vitamin D based on the Institute of Medicine recommendations (7).

For each analyte, heterogeneity in linear trends between studies was assessed by comparing the χ² values for models with and without a (study) x (linear trend) interaction term. Tests for heterogeneity for the case-defined factors were obtained by fitting separate models for each subgroup and assuming independence of the ORs using a method analogous to a meta-analysis, in which controls in each matched set were assigned to the category of their matched case. Tests for heterogeneity for noncase defined factors were assessed with χ² tests of interaction between subgroups and the binary variable.

In order to assess potential effect modification with different biomarkers, a χ² test of interaction was used to determine whether risks by study-specific thirds of 25(OH)D varied according to study-specific thirds of 1,25(OH)₂D (and vice versa), and according to study-specific thirds of circulating concentrations of insulin-like growth factor-I (IGF)-I, IGF binding protein-3 (IGFBP3), testosterone, free testosterone, sex hormone-binding globulin (SHBG), and prostate-specific antigen (PSA), where these data were available.

The cross-sectional associations of 25(OH)D and 1,25(OH)₂D with participant characteristics (among controls only) were examined using analyses of variance to calculate geometric mean concentrations and 95% confidence intervals (CI), adjusting for study and age at blood collection, as appropriate.

All tests of statistical significance were two-sided, and statistical significance was set at the 5% level. All statistical tests were carried out with Stata Statistical Software, Release 14 (StataCorp, LP). Full details of the statistical analyses are provided in the Supplementary Methods.

Details of the 19 participating studies are shown in Table 1. Data on 25(OH)D concentrations were available for 13,462 men who subsequently developed prostate cancer and 20,261 control participants, and for 1,25(OH)₂D concentrations for 1,885 case and 2,114 control participants. Mean age at blood collection across the studies ranged from 46.5 (SD = 4.2) to 76.3 (3.6) years. Blood collection preceded prostate cancer diagnosis by an average of 8.5 years (SD = 6.0 years), although there was a wide variation among the studies (Table 2). On average, cases were 67.5 years old (SD = 7.3 years) at diagnosis, and most (87.1%) were diagnosed after 1994. The majority of cases with information on stage and grade of disease had localized (early or other localized) disease (ranging from 47.8% to 99.0% of case patients across studies) and low-intermediate grade tumors (ranging from 75.8% to 100% of case patients). Concentrations of 25(OH)D and 1,25(OH)₂D varied significantly by month among both the cases and controls (Supplementary Fig. S1).

25(OH)D concentration was linearly positively associated with risk for total prostate cancer (Fig. 1); the multivariate-OR for prostate cancer for men in the highest compared with the lowest study-specific fifth was 1.22 (95% CI, 1.13–1.31; P trend < 0.001). The association was similar when only the matching factors were taken into account (Supplementary Fig. S2), and there was no evidence of heterogeneity between the contributing studies (Fig. 2A). When 25(OH)D was categorized into study-specific tenths, the OR for the highest versus the lowest tenth was 1.34 (1.20 to 1.49; P trend < 0.001; Supplementary Table S3).

There was no evidence of an association between 1,25(OH)₂D concentration and risk for total prostate cancer (see Figs. 1 and 2B). The association was similar when only the matching factors were taken into account (Supplementary Fig. S2).

In sensitivity analyses that used overall quintile cut-points of 25(OH)D across all studies combined (rather than study-specific cut-points), the ORs for total prostate cancer were materially unchanged (Supplementary Fig. S3). When the analyses were repeated using predefined cut-points for 25(OH)D, multivariable-adjusted ORs for total prostate cancer were 0.84 (0.76–0.93), 0.89 (0.84–0.95), and 1.07 (1.00–1.13), respectively, for men with 25(OH)D < 30 (at risk for deficiency), 30–49, and ≥75 nmol/L compared with those with concentrations of 50 to 74 nmol/L (Supplementary Table S4).

Although there was no evidence of heterogeneity in the association of 25(OH)D with risk by stage of disease, there were differences by disease aggressiveness (P_{heterogeneity} = 0.014): the OR for an 80-percentile increase in 25(OH)D was 1.24, 1.13–1.36 for nonaggressive disease (T1–T3/N0/M0) and 0.95, 0.78–1.15 for aggressive disease (T4, N1, M1, and/or fatal prostate cancer). Similar differences were also seen between low-intermediate and high-grade disease, although these differences were not statistically significant (Fig. 3). There was no association between circulating 25(OH)D concentrations and fatal prostate cancer (Fig. 3). Supplementary Fig. S4 shows results from categorical analyses of the associations of study-specific fifths of 25(OH)D with risk for advanced stage, aggressive disease, and high-grade prostate cancer.

There was no evidence of heterogeneity in risk of total prostate cancer associated with 25(OH)D according to time to diagnosis or other participant characteristics (Fig. 3), including season of blood draw (Fig. 4A) or by circulating concentrations of 1,25(OH)₂D, IGF-I, IGFBP-3, testosterone, free testosterone, SHBG, or PSA (Supplementary Table S5A–S5G).

For 1,25(OH)₂D, there was no evidence of heterogeneity by season of blood draw (Fig. 4B), time to diagnosis, or other tumor characteristics (Supplementary Fig. S5). There was some evidence of heterogeneity by family history of prostate cancer, with a positive association for men with a positive family history of the disease (P_{heterogeneity} = 0.03; multivariable-adjusted OR for an 80 percentile increase = 2.26; 95% CI, 1.19–4.32; Supplementary Fig. S4), although this was based on small numbers. There was no evidence of heterogeneity by season of blood draw (Fig. 4).

Concentrations of 25(OH)D and 1,25(OH)₂D were significantly but not strongly correlated with each other (r = 0.13, P < 0.001). In the subset of control participants with data available on other analytes, circulating 25(OH)D concentration was weakly correlated with sex hormones and other analytes (Supplementary Table S6), but neither 25(OH)D nor 1,25(OH)₂D concentration was correlated with PSA (r = 0.01 for both). After adjustment for age, 25(OH)D concentration was lower in men who were obese, current smokers, poorly educated, unmarried, and nondrinkers (Fig. 5). 1,25(OH)₂D displayed generally similar associations (Supplementary Fig. S6).

This collaborative analysis of individual participant data does not support the hypothesis that vitamin D deficiency and/or insufficiency increases the risk of prostate cancer. Higher 25(OH)D levels were associated with an increased risk of nonaggressive disease, with no association for aggressive disease. We also found no evidence that circulating concentration of 1,25(OH)₂D was related to risk for prostate cancer, overall or by tumor characteristics.

This collaborative analysis includes information from the vast majority (>90%) for 25(OH)D and 85% for 1,25(OH)₂D of the published prospective data. Of the 24 studies with published data on 25(OH)D, 7 did not contribute data to this collaboration, all of which had fewer than 200 incident cases and reported inconsistent findings (8–13). Combining the results of the current analyses with those of six of the seven additional studies (for whom data could be extracted to perform a meta-analysis) did not change the overall finding (summary relative risk of highest compared with the lowest fifth of 25(OH)D = 1.21; 95% CI, 1.13–1.30), suggesting that inclusion of participant-level data from these studies would not have materially altered the results. Two studies with published data on 1,25(OH)₂D did not contribute data, one of which reported an inverse association (based on 181 cases, RR not given for 1,25(OH)₂D alone; refs. 10, 14) and another that found no association (based on 136 cases; ref. 9). Including these two studies would not have materially changed our results. Thus, we believe that the findings from the current study provide a reliable summary of the totality of the evidence on the association between circulating vitamin D concentrations and prostate cancer risk.

Our findings do not appear to support the evidence from experimental research using cell lines and animal models that vitamin D compounds may promote cell differentiation, inhibit prostate cancer cell growth and invasion, and stimulate apoptosis (15, 16). Although there are no published data from adequately powered randomized controlled trials for the effects of vitamin D supplementation on prostate cancer incidence, two large recent studies have exploited GWAS-identified variation in genes related to vitamin-D synthesis, metabolism, and binding to study the possible relationship with prostate cancer risk. A Mendelian randomization analysis of data from up to 69,837 prostate cancer cases in the PRACTICAL and GAME-ON consortia found no evidence for an association with risk for either total (OR in PRACTICAL per genetically-determined 25 nmol/L increase in 25(OH)D concentration = 0.95, 95% CI, 0.80–1.13; P = 0.55) or aggressive prostate cancer (OR in GAME-ON = 1.14, 0.85–1.54; P = 0.38; ref. 17).

It is possible that our finding of a positive association between overall and nonaggressive prostate cancer risk and circulating 25(OH)D concentration may be explained by detection bias, in that health-conscious men who may be more likely to have a higher sun exposure, a higher dietary intake of vitamin D and/or vitamin D supplementation, are more likely to have a PSA test or to seek medical attention with early symptoms. The observation that vitamin D deficiency was associated with a reduced risk of prostate cancer and higher levels with an increased risk (particularly for nonaggressive disease) supports this hypothesis. Nonetheless, a positive association between 25(OH)D and prostate cancer risk was reported in both the Prostate, Lung, Colorectal, and Ovarian Cancer Screening Trial (PLCO) and Prostate Cancer Prevention Trial (PCPT) studies, in which almost all men had either regular PSA testing (as data were provided solely from the screening arm in PLCO and PCPT) or had an end-of-study biopsy (PCPT), suggesting that factors other than detection bias may be involved.

It is difficult to draw conclusions from the current pooled analyses of 1,25(OH)₂D as only a small number of prospective studies have measured this analyte. Although circulating 1,25(OH)₂D concentrations are considered to be tightly regulated within a narrow range (18), we found some evidence of seasonal variation in 1,25(OH)₂D concentrations, similar to that of 25(OH)D, and also differences in concentrations according to age, adiposity, cigarette smoking status, and alcohol consumption. It is difficult to determine the extent to which these associations are due to cross-reactivity of the 1,25(OH)₂D assay with 25(OH)D (or other molecules), although the correlation between 25(OH)D and 1,25(OH)₂D was weak (r = 0.13) and there was no evidence for an association between 1,25(OH)₂D and prostate cancer risk.

A number of previous studies have evaluated the joint association of 25(OH)D and 1,25(OH)₂D with prostate cancer risk (9, 19–21), but their sample sizes were small. We found no evidence that the association of prostate cancer risk with 25(OH)D is modified by circulating concentrations of 1,25(OH)₂D, although even in this collaborative pooled dataset, there are still relatively few cases (n = 1,885) with data on both vitamin D analytes. It has also been hypothesized that vitamin D may influence tumor growth by modulating the action of growth factors, such as IGF-I, that normally stimulate proliferation (16), for example by stimulating the release of IGFBP-3 (22). We observed weak correlations of circulating 25(OH)D or 1,25(OH)₂D concentrations with IGF-I, IGFBP-1, IGFBP-2, and IGFBP-3 concentrations and with levels of other blood biomarkers (e.g., free testosterone or PSA), and there was no evidence of modification of the association of 25(OH)D with risk according to these biomarkers.

This study has some limitations. The calculated relative risks were based on single measurements of vitamin D, which may not accurately reflect long-term circulating concentration. Several studies have found moderate correlations between two measures of 25(OH)D, even when the samples were not taken at the same time of the year, with correlations between 0.42 and 0.70 in blood taken between 3 and 14 years apart (reviewed in ref. 23). These findings suggest that a single measure of circulating 25(OH)D is an informative measure of vitamin D status, at least over the medium term. The published prospective data on vitamin D and risk for aggressive prostate cancer subtypes are still relatively limited. Thus, even in this pooled analysis, the total number of cases with aggressive disease and data on 25(OH)D is relatively small (n = 1,446), therefore the results by tumor subtype should be interpreted with some caution. Moreover, we do not have detailed data on other sun exposure measures, such as solar radiation levels in each study location, which would also vary within each individual study depending on where each participant lives. Finally, more than 95% of participants included in this pooled analysis were of White ethnicity, and results may therefore not be generalizable to non-White populations.

In summary, this collaborative analysis of the worldwide data on circulating vitamin D and prostate cancer risk suggests that a high vitamin D concentration is not associated with a lower risk of prostate cancer. Rather, the findings suggest that men with elevated circulating concentrations of 25(OH)D are more likely to be diagnosed with nonaggressive prostate cancer, though this may be due to detection bias. There was no evidence for an association with aggressive disease.

E.A. Platz is Senior Editor at the AACR. R. Blomhoff has an ownership interest (including stock, patents, etc.) in AS Vitas. F.C. Hamdy received honoraria from the speakers' bureau of Steba Biotech and Takeda. No potential conflicts of interest were disclosed by the other authors.

The funders of the study had no role in the study design, data collection, data analysis, data interpretation, or writing of the report. The authors in the writing team had full access to all data in the study. The corresponding author had the final responsibility for the decision to submit for publication. The Endogenous Hormones, Nutritional Biomarkers and Prostate Cancer Collaborative Group (EHNBPCCG) data are not suitable for sharing, but are suitable for research involving further analyses in collaboration with the Cancer Epidemiology Unit, Nuffield Department of Population Health, University of Oxford.

Conception and design: R.C. Travis, A. Perez-Cornago, D. Albanes, K.J. Helzlsouer, P.H. Peeters, B. Bueno-de-Mesquita, M.-J. Sánchez, B.B. Yeap, L. Flicker, E.L. Giovannucci, D.R. English, T.J. Key, N.E. Allen

Development of methodology: R.C. Travis, A. Perez-Cornago, O.P. Almeida, T.J. Key

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): A. Perez-Cornago, D. Albanes, C.E. Joshu, P.L. Lutsey, K.J. Helzlsouer, K. Visvanathan, D. Palli, P.H. Peeters, B. Bueno-de-Mesquita, A. Trichopoulou, M.-J. Sánchez, H. Brenner, B. Schöttker, L. Perna, B. Holleczek, P. Knekt, H. Rissanen, B.B. Yeap, L. Flicker, O.P. Almeida, Y.Y.E. Wong, J.M. Chan, E.L. Giovannucci, M.J. Stampfer, R.E. Gislefoss, T. Bjørge, H.E. Meyer, R. Blomhoff, S Tsugane, N.Sawada, D.R. English, A.K. Heath, E.J. Williamson, J. Manjer, J. Malm, M. Almquist, L.Le Marchand, C.A. Haiman, L.R. Wilkens, J.M. Schenk, C.M. Tangen, A.Black, M.B. Cook, W.-Y. Huang, R.G. Ziegler, R.M. Martin, F.C. Hamdy, J.L. Donovan, D.E. Neal, M. Touvier, S. Hercberg, P. Galan, M. Deschasaux, T.J. Key

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): R.C. Travis, A. Perez-Cornago, P.N. Appleby, P.H. Peeters, M.J. Gunter, K.K. Tsilidis, H. Brenner, Y.Y.E. Wong, M.J. Stampfer, T. Bjørge, H.E. Meyer, R. Blomhoff, W.-Y. Huang, R.M. Martin, M. Touvier, N.E. Allen

Writing, review, and/or revision of the manuscript: R.C. Travis, A. Perez-Cornago, P.N. Appleby, D. Albanes, C.E. Joshu, P.L. Lutsey, A.M. Mondul, E.A. Platz, S.J. Weinstein, T.M. Layne, K.J. Helzlsouer, K. Visvanathan, D. Palli, P.H. Peeters, B. Bueno-de-Mesquita, A. Trichopoulou, M.J. Gunter, K.K. Tsilidis, M.-J. Sánchez, A. Olsen, H. Brenner, B. Schöttker, L. Perna, B. Holleczek, P. Knekt, H. Rissanen, B.B. Yeap, L. Flicker, O.P. Almeida, J.M. Chan, E.L. Giovannucci, G. Ursin, R.E. Gislefoss, T. Bjørge, H.E. Meyer, R. Blomhoff, S Tsugane, D.R. English, D.W. Eyles, A.K. Heath, J. Manjer, M. Almquist, L.Le Marchand, J.M. Schenk, C.M. Tangen, A.Black, M.B. Cook, W.-Y. Huang, R.G. Ziegler, R.M. Martin, F.C. Hamdy, J.L. Donovan, D.E. Neal, M. Touvier, S. Hercberg, P. Galan, M. Deschasaux, T.J. Key, N.E. Allen

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): A. Perez-Cornago, P.N. Appleby, S.J. Weinstein, B. Holleczek, H. Rissanen, M.J. Stampfer, T. Bjørge, A.K. Heath, E.J. Williamson, J. Manjer, M.B. Cook

Study supervision: R.C. Travis, A. Perez-Cornago, B. Bueno-de-Mesquita, O.P. Almeida, T.J. Key

We thank the men who participated in the collaborating studies, the research staff, the collaborating laboratories, and the funding agencies in each of the studies. The authors are solely responsible for the study design, data collection, analysis and interpretation of the data, writing of the article, and decision to submit the article for publication.

This work was supported by Cancer Research UK grant C8221/A19170 and involved centralized pooling, checking, and data analysis. Details of funding for the original studies are in the relevant publications (and see Supplementary Materials).

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.