Background: An inverse association was observed between serum 25-hydroxyvitamin D [25(OH)D] levels and risk of female breast cancer in four recent observational studies. However, the exact 25(OH)D serum levels associated with a specific level of reduction in breast cancer incidence are still imprecise.
 Methods: Data were combined from all studies that provided data on risk of breast cancer by quantile of serum 25(OH)D, using Review Manager (Copenhagen: Nordic Cochrane Centre, 2007). The analysis was performed according to published international standards for meta-analyses of data from observational studies. The number of individuals and person-years at risk in each quantile (generally quartiles or quintiles) of each study was determined, and a record was created for each individual from each study, with a study number identifying the study, a serial number for the individual, the median 25(OH)D concentration in the individual’s quantile in their study, and a code indicating whether the individual was a case or noncase. Records from the cohort studies were combined into an inclusive dataset in which individuals were divided into quintiles of 25(OH)D. There was an approximately equal number of individuals in each quintile. Measures of risk (incidence rates or odds ratios) were calculated for each quintile, providing an estimate of the pooled dose-response relationship. The procedure was performed separately for cohort and case-control studies. Results for each type of study, and all studies combined, were plotted and tested for goodness of fit to a classical dose-response curve.
 Results: A sigmoid dose-response gradient was observed between serum 25(OH)D and risk of breast cancer, and was approximately consistent across all studies. A higher serum 25(OH)D level (100-120 nmol/L) was associated with a projected 50% lower risk of breast cancer, compared to the serum 25(OH)D level associated in previous observational studies with 50% lower risk of colon cancer (80 nmol/L).
 Conclusions: Serum 25(OH)D is recognized as the optimal measure of vitamin D status. This study provides the most precise estimate available to date of the dose-response curve of the inverse association of 25(OH)D with risk of breast cancer, as estimated by observational studies. A substantially higher serum 25(OH)D level was needed for a projected reduction by half in incidence of breast cancer than for a similar reduction in risk of colon cancer, based on observational studies. The dose-response curve will help project the association of vitamin D deficiency states with risk of breast cancer, and will provide useful information for planning future clinical trials. Additional cohort and other epidemiological studies also would be worthwhile to further characterize this association.

99th AACR Annual Meeting-- Apr 12-16, 2008; San Diego, CA