In their Reader’s Comment, Drs. Mullie and Autier propose the interesting hypothesis that the inverse relation between vitamin D and cardiovascular (CV) risk we reported might be explained by confounding of the relation by obesity, quantified by body mass index (BMI), that is, by an inverse relation between serum 25(OH) vitamin D and BMI.

We agree that obesity is a well-established univariate risk factor for CV disease and for CV and total mortality. However, much of the CV-related risk of obesity is carried by its related risk factors of hypertension, hyperlipidemia, and diabetes mellitus. Indeed, BMI has not consistently been found to be an independent, multivariate predictor of CV risk, and it does not appear in most standard CV risk models, such as Framingham and the National Cholesterol Education Program Adult Treatment Panel algorithms. Because BMI is not a standard risk factor in these models, and because our database did not uniformly provide BMI information in our study population, it was not included and adjusted for in our multivariate regression models in our published report.

To specifically address the authors’ concern, we have gone back to our database and reassessed the relation of vitamin D and CV risk among the large subset (>10,000 to 15,000) of subjects for whom concurrent BMIs were available for each end point. We then performed an additional multivariate adjustment for the hazard ratio (HR) of vitamin D and incident CV events and mortality with BMI added to the standard risk factors already included in the original models.

To summarize, these results showed that the addition of BMI to traditional risk factors did not change the reported adjusted HRs for vitamin D and incident CV diseases and mortality. Specifically, the multivariate HR for death of very low (≤15 ng/ml) versus normal (>30 ng/ml) vitamin D was 1.74 (p <0.0001) without and 1.80 (p <0.0001) with BMI added (n = 15,794). For the composite end point (death, incident myocardial infarction or coronary artery disease diagnosis, incident heart failure, incident cerebrovascular accident) the multivariate HR of very low versus normal vitamin D was 1.73 (p <0.0001) without and 1.69 (p <0.0001) with BMI added (n = 10,724). Similarly, BMI did not measurably change the reported adjusted HR for coronary artery disease or myocardial infarction, heart failure, cerebrovascular accident, atrial fibrillation, or peripheral vascular disease, the specific outcomes assessed, nor did the adjusted HRs for any of these individual or composite CV outcomes change for the comparisons of low (16 to 30 ng/ml) versus normal (>30 ng/ml) vitamin D status.

We interpret these findings to indicate that any confounding of BMI and vitamin D on CV risk already has been taken into account by our multivariate adjustments, which included accepted, independently contributing standard risk factors, and that the reported independent risk relations for vitamin D in our published article remain valid.