Circulating vitamin D and breast cancer risk: an international pooling project of 17 cohorts

Abstract

Laboratory and animal research support a protective role for vitamin D in breast carcinogenesis, but epidemiologic studies have been inconclusive. To examine comprehensively the relationship of circulating 25-hydroxyvitamin D [25(OH)D] to subsequent breast cancer incidence, we harmonized and pooled participant-level data from 10 U.S. and 7 European prospective cohorts. Included were 10,484 invasive breast cancer cases and 12,953 matched controls. Median age (interdecile range) was 57 (42-68) years at blood collection and 63 (49-75) years at breast cancer diagnosis. Prediagnostic circulating 25(OH)D was either newly measured using a widely accepted immunoassay and laboratory or, if previously measured by the cohort, calibrated to this assay to permit using a common metric. Study-specific relative risks (RRs) for season-standardized 25(OH)D concentrations were estimated by conditional logistic regression and combined by random-effects models. Circulating 25(OH)D increased from a median of 22.6 nmol/L in consortium-wide decile 1 to 93.2 nmol/L in decile 10. Breast cancer risk in each decile was not statistically significantly different from risk in decile 5 in models adjusted for breast cancer risk factors, and no trend was apparent (P-trend = 0.64). Compared to women with sufficient 25(OH)D based on Institute of Medicine guidelines (50- < 62.5 nmol/L), RRs were not statistically significantly different at either low concentrations (< 20 nmol/L, 3% of controls) or high concentrations (100- < 125 nmol/L, 3% of controls; ≥ 125 nmol/L, 0.7% of controls). RR per 25 nmol/L increase in 25(OH)D was 0.99 [95% confidence intervaI (CI) 0.95-1.03]. Associations remained null across subgroups, including those defined by body mass index, physical activity, latitude, and season of blood collection. Although none of the associations by tumor characteristics reached statistical significance, suggestive inverse associations were seen for distant and triple negative tumors. Circulating 25(OH)D, comparably measured in 17 international cohorts and season-standardized, was not related to subsequent incidence of invasive breast cancer over a broad range in vitamin D status.

Keywords: 25-Hydroxyvitamin D; Biomarker; Breast cancer; Calibration; Pooled analysis; Prospective cohort study; Vitamin D.

Fig. 1

Median circulating 25(OH)D concentrations among controls for the 12 calibrated studies, before and after calibration, and for the five newly measured studies. For the calibrated studies, median 25(OH)D before calibration is indicated by red circles and after calibration by blue circles. For the newly measured studies, median 25(OH)D is indicated by blue squares. All 25(OH)D concentrations are season-standardized. The relative change after calibration is +10% for CPS-II, −47% for E3N, −18% for EPIC, +16% for JANUS, −31% for MDCS, −41% for MEC, +14% for NHS, +9% for NHSII, −11% for NSHDS, −14% for NYU WHS, −9% for PLCO, and +11% for WHI. Horizontal green lines indicate IOM-based clinical guidance for circulating 25(OH)D: deficiency at <30 nmol/L, insufficiency at 30-<50 nmol/L, sufficiency at 50-<75 mmol/L, beyond sufficiency at ≥75 nmol/L. The full names of the 17 cohorts are given in Table 1. 25(OH)D: 25-hydroxyvitamin D; IOM: Institute of Medicine.

Fig 2

RRs of breast cancer by 25(OH)D categories based on IOM clinical guidance for vitamin D. RRs (solid circles) of breast cancer and 95% CIs (vertical lines) are plotted on a log scale for 11 categories of circulating 25(OH)D. These 25(OH)D categories expand those established by the IOM. The categories <20 (3.1% of controls) and 20-<30 (10.4%) nmol/L are considered deficient; 30-<40 (15.7%) and 40-<50 (18.5%) nmol/L are considered insufficient; 50-<62.5 (21.2%) (the referent) and 62.5-<75 (14.9%) nmol/L are considered sufficient; and 75-<87.5 (9.0%), 87.5-<100 (3.9%), 100-<112.5 (1.8%), 112.5-<125 (0.9%), and ≥125 nmol/L (0.7%) are considered beyond sufficient. The RR and 95% CI for each category are plotted at the median concentration of 25(OH)D among controls in that category. The model used, model 3, conditioned on study-specific matching factors, which included age at blood collection and usually date of blood collection, and adjusted for BMI (using a variable combining BMI and menopausal status at blood collection), physical activity, race, family history of breast cancer, age at menarche, parity and age at first birth, oral contraceptive use, menopausal hormone therapy, and alcohol intake. The P-trend across categories was calculated with a Wald test that assigned the median 25(OH)D in controls to each category. 25(OH)D: 25-hydroxyvitamin D; BMI: body mass index; CI: confidence interval; IOM: Institute of Medicine; ref: referent; RR: relative risk.

Fig 3

Cohort-specific and pooled RRs of breast cancer per 25 nmol/L increase in circulating 25(OH)D. Cohort-specific RRs of breast cancer per 25 nmol/L increase in 25(OH)D are plotted on a log scale and indicated by solid squares with the size proportional to the cohort’s contribution to the pooled RR. Horizontal lines indicate 95% CIs. The pooled RR is indicated by a solid diamond. A RR of 1.0, marked by the vertical line, indicates no association. The model used, model 3, conditioned on study-specific matching factors, which included age at blood collection and usually date of blood collection, and adjusted for BMI (using a variable combining BMI and menopausal status at blood collection), physical activity, race, family history of breast cancer, age at menarche, parity and age at first birth, oral contraceptive use, menopausal hormone therapy, and alcohol intake. The P-heterogeneity in the results across cohorts was calculated using the Q statistic. The full names of the 17 cohorts are given in Table 1. 25(OH)D: 25-hydroxyvitamin D; BMI: body mass index; CI: confidence interval; ref: referent; RR: relative risk.

Fig. 4

RRs of breast cancer per 25 nmol/L increase in circulating 25(OH)D in population subgroups and by tumor characteristics. RRs (solid squares) of breast cancer and 95% CIs (horizontal lines) per 25 nmol/L increase in 25(OH)D are plotted on a log scale. A RR of 1.0, marked by the vertical line, indicates no association. Differences in RR by participant characteristics, such as breast cancer risk factors, latitude, and season of blood collection, were examined using a two-stage approach and unconditional logistic regression. Differences in RR by study characteristics, such as geographic region and calibration (newly measured vs. calibrated 25(OH)D data), were examined using a two-stage approach and conditional logistic regression. Differences in associations by participant and study characteristics were tested for statistical significance using meta-regression. Heterogeneity by time elapsed from blood collection to diagnosis, age at diagnosis, and tumor characteristics were assessed in an aggregated dataset with conditional analyses, and tested for statistical significance with a contrast test. Conditional models were conditioned on study-specific matching factors, which included age at blood collection and usually date of blood collection; while unconditional models included as covariates age at blood collection, elapsed time from blood collection to assay, and study center (EPIC and MEC only). In addition, all models were further adjusted for BMI (using a variable combining BMI and menopausal status at blood collection), physical activity, race, family history of breast cancer, age at menarche, parity and age at first birth, oral contraceptive use, menopausal hormone therapy, and alcohol intake. Analyses by use of menopausal hormone therapy were restricted to postmenopausal and perimenopausal women. For analyses stratified by season of blood collection, winter was defined as weeks 2–14; summer as weeks 28–40; and circulating 25(OH)D was not season-standardized. In the analyses by molecular subtype, ER+ and/or PR+/HER2- tumors are considered luminal A-like, ER+ and/or PR+/HER2+ tumors are considered luminal B-like, ER-/PR-/HER2+ tumors are considered HER2-enriched, and ER-/PR-/HER2- tumors are considered triple negative. Additional information on how we defined and harmonized breast tumor characteristics is presented in Supplementary Methods. 25(OH)D: 25-hydroxyvitamin D; BMI: body mass index; CI: confidence interval; ER: estrogen receptor; HER2: human epidermal growth factor receptor 2; PR: progesterone receptor; ref: referent; RR: relative risk; +; positive; −: negative.

Fig. 4

RRs of breast cancer per 25 nmol/L increase in circulating 25(OH)D in population subgroups and by tumor characteristics. RRs (solid squares) of breast cancer and 95% CIs (horizontal lines) per 25 nmol/L increase in 25(OH)D are plotted on a log scale. A RR of 1.0, marked by the vertical line, indicates no association. Differences in RR by participant characteristics, such as breast cancer risk factors, latitude, and season of blood collection, were examined using a two-stage approach and unconditional logistic regression. Differences in RR by study characteristics, such as geographic region and calibration (newly measured vs. calibrated 25(OH)D data), were examined using a two-stage approach and conditional logistic regression. Differences in associations by participant and study characteristics were tested for statistical significance using meta-regression. Heterogeneity by time elapsed from blood collection to diagnosis, age at diagnosis, and tumor characteristics were assessed in an aggregated dataset with conditional analyses, and tested for statistical significance with a contrast test. Conditional models were conditioned on study-specific matching factors, which included age at blood collection and usually date of blood collection; while unconditional models included as covariates age at blood collection, elapsed time from blood collection to assay, and study center (EPIC and MEC only). In addition, all models were further adjusted for BMI (using a variable combining BMI and menopausal status at blood collection), physical activity, race, family history of breast cancer, age at menarche, parity and age at first birth, oral contraceptive use, menopausal hormone therapy, and alcohol intake. Analyses by use of menopausal hormone therapy were restricted to postmenopausal and perimenopausal women. For analyses stratified by season of blood collection, winter was defined as weeks 2–14; summer as weeks 28–40; and circulating 25(OH)D was not season-standardized. In the analyses by molecular subtype, ER+ and/or PR+/HER2- tumors are considered luminal A-like, ER+ and/or PR+/HER2+ tumors are considered luminal B-like, ER-/PR-/HER2+ tumors are considered HER2-enriched, and ER-/PR-/HER2- tumors are considered triple negative. Additional information on how we defined and harmonized breast tumor characteristics is presented in Supplementary Methods. 25(OH)D: 25-hydroxyvitamin D; BMI: body mass index; CI: confidence interval; ER: estrogen receptor; HER2: human epidermal growth factor receptor 2; PR: progesterone receptor; ref: referent; RR: relative risk; +; positive; −: negative.

Fig 5

RRs of breast cancer by season-specific deciles of circulating 25(OH)D. RRs (solid circles) of breast cancer and 95% CIs (vertical lines) are plotted on a log scale by winter-specific and summer-specific, consortium-wide deciles of circulating 25(OH)D. Results for the participants who donated blood in the winter (defined as weeks 2–14; 2,741 cases) are shown in blue; results for those who donated in the summer (defined as weeks 28–40; 2,491 cases) are shown in orange. Participants donating blood in the spring and fall were excluded from these analyses. 25(OH)D concentrations were not season-standardized. Decile cutpoints, based on the 25(OH)D distributions in controls, were for winter <22, 22-<28, 28-<34, 34-<39, 39-<45, 45-<50, 50-<56, 56-<63, 63-<75, and ≥75 nmol/L and for summer <35, 35-<43, 43-<50, 50-<55, 55-<61, 61-<67, 67-<72, 72-<81, 81-<94, and ≥94 nmol/L. The RR and 95% CI for each decile was plotted at the median 25(OH)D concentration among the controls in that decile. Unconditional logistic regression models in an aggregated dataset were adjusted for study, age at blood collection, elapsed time from blood collection to assay, study center (EPIC and MEC only), BMI (using a variable combining BMI and menopausal status at blood collection), physical activity, race, family history of breast cancer, age at menarche, parity and age at first birth, oral contraceptive use, menopausal hormone therapy, and alcohol intake. The P-trend across categories was based on a Wald test that assigned the median 25(OH)D in controls to each category. 25(OH)D: 25-hydroxyvitamin D; BMI: body mass index; CI: confidence interval; N: number; ref: referent.