Abstract P129: Confirmatory Factor Analysis of Multivariate Lipid Distributions Across Time and Race/Ethnic Groups: United States, 2003-2012

Abstract

Plasma concentrations of low density lipoprotein cholesterol (LDL-C), high density lipoprotein cholesterol (HDL-C), and triglycerides (TG) are important risk factors for metabolic and cardiovascular diseases and are widely-used as targets for therapeutic intervention. Consideration of the multivariate distribution of these lipid traits could be informative for studies involving pleiotropic genetic variants, variants associated with more than one lipid outcome, but featuring only univariate lipid outcome models. Confirmatory factor analysis (CFA) is one way to characterize the multivariate distribution of biologically plausible, related indicators such as lipids with a continuous latent factor. Although CFA has been used to characterize metabolic syndrome, a cardiovascular disease (CVD) risk factor involving many correlated indicators, dyslipidemia is a CVD risk factor that remains relatively unexplored in this literature. Thus, the primary aim of this study is to conduct CFA with LDL-C, HDL-C and TG as indicators representing a single continuous latent lipid factor and assess its suitability as an outcome measure in a United States representative sample by racial/ethnic groups stratified by age and gender over five 2-year time periods (2003-2004, 2005-2006, 2007-2008, 2009-2010, 2011-2012) from NHANES. First, we used principal component analyses (PCA) to visually examine clustering by race and gender in biplots. Second, we tested for scale differences in the lipid factor across a) calendar time and b) racial/ethnic groups (Mexican-American, non-Hispanic White, non-Hispanic Black, Other Hispanic and Other race/ethnicity). All analyses were stratified by medication use, gender and three age groups: 12-19 years, 20 to 49 years and 50-80 years, and adjusted for age and body mass index (BMI). We found significant scale differences across racial/ethnic groups. In particular, the center of the distribution differed across racial/ethnic groups for ages 12-19 years (p<0.0001), 20-49 years (p<0.0001) and 50-80 years (p<0.0001), suggesting joint distributions of TG, HDL-C, and LDL-C vary across racial/ethnic groups. In summary, one continuous latent factor representing all three lipids and their concomitant associations provides a means to characterize lipid values simultaneously and can serve as an outcome when studying exposures influencing multiple lipid values, pleiotropic genetic variants being just one example of many. Differences in the lipid latent factor across racial/ethnic groups emphasizes distinct multivariate distributions necessitating stratification. Limited analyses exist using this CFA framework for lipids, and future efforts considering the joint distribution of lipids may improve our understanding of the genetic architecture of dyslipidemia.