Circulating vitamin D concentrations and breast cancer risk: A pooled analysis of 17 cohorts

Talk Code: 
EP3D.8

The problem

In 2010 a US Institute of Medicine (IOM) committee reported that the evidence on vitamin D and risk of breast cancer was both inconsistent and inconclusive, and could not be used as a basis for developing guidance on vitamin D intake. Determining the concentrations at which circulating 25-hydroxyvitamin D [25(OH)D], the accepted measure of vitamin D status, influences cancer risk could provide critical information to inform cancer prevention strategies.

The approach

We analyzed individual 25(OH) data from 17 cohorts (with >11,000 breast cancer cases), which includes practically all the prospective data available worldwide. Included were cohorts for which we measured 25(OH)D levels using a direct, competitive chemiluminescence immunoassay that captures both 25(OH)D2 and 25(OH)D at Heartlands Assays (Iowa, USA). For the 12 cohorts that had previously measured 25[OH]D levels, a random sample of 2-3 controls selected from each decile of the control distribution (for a total of 29 samples) was re-assayed at the central laboratory. Calibrated 25(OH)D values were then calculated using robust linear regression models. This approach led to 25(OH)D concentrations on the same scale across studies (i.e. as if the same assay/lab had been used for all cohorts) to enable analyses of 25(OH)D concentrations using common absolute cut points across studies. Standardization for season of blood draw was performed using a periodic sine/cosine function for both measured 25(OH)D (5 cohorts) and calibrated 25(OH)D (12 cohorts) concentrations.

Findings

Our preliminary analyses included 11,725 cases of incident, invasive breast cancer, and 14,253 matched controls. Median calibrated, season-standardized 25(OH)D levels among the controls varied from 33 to 70 nmol/L across studies. The pooled odds ratio for breast cancer, comparing the highest versus lowest study-specific quintile of season-standardized 25(OH)D level, was 0.96 (95% confidence interval 0.87-1.06) with no statistically significant trend across quintiles (p-trend = 0.25) after adjusting for body mass index, physical activity, menopausal status and hormone therapy use at blood collection, parity/age at first birth, family history of breast cancer, and alcohol use. Results were not statistically significantly different in analyses stratified by age of diagnosis (<50, 50-60, 60+ y), or by estrogen receptor status. Further analyses are ongoing that take advantage of the absolute 25(OH)D concentrations. We are examining the dose-response relationship, with particular attention to low and high 25(OH)D concentrations, and the IOM guidelines for the vitamin D levels required for bone health. Preliminary analyses suggest that breast cancer risk is unrelated to 25(OH)D, even at concentrations ?100 nmol/L.

Consequences

Preliminary results from the largest pooling project of prospective studies to date suggest that vitamin D status over the range of absolute concentrations currently observed around the world is not associated with breast cancer risk. These findings will inform both clinical practice and public health guidance.

Credits

  • Toqir Mukhtar, Harvard T.H. Chan School of Public Health, USA, Massachusetts, USA
  • Kala Visvanathan, University of Michigan School of Public Health, Ann Arbor, USA
  • Alison Mondul, NYU School of Medicine, New York, USA
  • Anne Zeleniuch-Jacquotte, National Cancer Institute, Maryland, USA
  • Shiaw-Shyuan Yaun, Johns Hopkins Bloomberg School of Public Health, Maryland, USA
  • Stephanie Smith-Warner, Johns Hopkins Bloomberg School of Public Health, Maryland, USA
  • Regina Ziegler