Figure 5.

Impact of age and BMI on baseline HDL. A repeated measures model was fit separately for males and females, with baseline HDL as the response, and age and BMI as predictors. The trends for age and BMI were modeled with restricted cubic splines (Stone CJ, Koo CY 1985). This approach provides a very flexible method for fitting trends without relying on any specific model formulation, with three knots located at approximately the 5th, 50th, and 95th percentiles (Durrleman 1989). 5A: HDL is plotted against age, stratified by gender, and adjusted for BMI 25 kg/m². (BMI 25 represents the clinical threshold between normal weight and overweight.) 5B: HDL is plotted against BMI, stratified by gender, and adjusted for age 46. (Age 46 represents the median age in the PMRP population.) If BMI was not available on the exact date of a given HDL data point, an estimate was imputed from BMI measures calculated before and/or after the HDL value, or based upon gender when no valid BMIs were available for an individual study subject. In both panels (5A and 5B), female subjects are represented by red dots; male subjects by blue dots.