ARS | Publication request: Modeling the sssociation between 25[OH]D and all-cause mortality in a representative US population sample

Submitted to: Endocrine Society Meeting
Publication Type: Abstract Only
Publication Acceptance Date: February 10, 2010
Publication Date: February 11, 2010
Citation: Durazo-Arvizu, R.A., Sempos, C., Luke, A., Kramer, H., Yetley, E., Dawson-Hughes, B., Cao, G., Bailey, R.L., Dwyer, J., Piccciano, M.F., Rovner, A. 2010. Modeling the association between 25[OH]D and all-cause mortality in a representative US population sample. Endocrine Society Meeting. 31(3):987-988.

Technical Abstract: Vitamin D has been identified as a potential key risk factor for several chronic diseases and mortality. The association between all-cause mortality and circulating levels of 25-ydroxyvitamin D (25[OH]D) has been described as non-monotonic with excess mortality at both low and high levels (1). However, the shape of this association appears asymmetric. Although the non-linear relationship has been modeled with up-to-date statistical methods, to our knowledge no attempt has been made to quantitatively estimate the 25[OH]D level at which minimum mortality occurs, accounting for the observed asymmetry. We modeled the non-monotonic, asymmetric association between 25[OH]D and all-cause mortality (Fig 1) in 15,106 participants aged 20 and older(6,355 Whites) in the NHANES III follow-up study using logistic regression, adjusting for age, education level, GFR, season, BMI, sex, race/ethnicity, medication use (anticonvulsants, estrogens, loop diuretics and/or thiazide diurectics) and blood pressure level. Two modeling approaches were used, namely 1) restricted cubic splines, and 2) transformation of 25[OH]D to normality. Maximum likelihood methods were used to estimate the parameters in the model and standard calculus techniques to obtain closed form solutions of the 25[OH]D of minimum mortality. The delta method was used to estimate standard errors of the parameters, along with normal-theory based confidence intervals(CI). Table 1 presents results for the entire population and for Whites. Estimates of 25(OH)D levels of minimum mortality based on the cubic splines (82 nmol/L) are lower than those obtained from the normal transformation approach (91 to 99 nmol/L). Similarly, the cubic splines approach yields narrower 95% CIs. In conclusion, the range of optimal 25[OH]D values appears rather wide, with excess mortality observed below 50 nmol/L and above 120 nmol/L Interpretation for large levels of 25[OH]D deserves careful examination owing to the limited number of events.