A multivariate analysis of familial associations of lipoprotein levels in the Lipid Research Clinics Collaborative Family Study: I. Familial correlation and regression analyses.

Abstract

In view of the complex, intraindividual relationships among different lipoprotein levels (LDL-C, HDL-C, and VLDL-C), multivariate methods aimed at assessing joint familial associations and their possible determinants were performed in the white, random sample component of the Collaborative Lipid Research Clinics Family Study data (1,336 families with 5,097 subjects). After appropriate transformations and covariate adjustments of the data, several kinds of correlation and regression analyses were performed, taking into consideration variable family size and possible age and clinic differences. The association patterns across clinics and age strata were found to be homogeneous for the vast majority of comparisons. The results of multivariate analyses (especially the significant association of each lipoprotein among biological relatives), the persistence of parental levels as the best predictors for the same lipoprotein levels among the offspring, and the essentially unchanged partial correlation estimates as compared to ordinary correlations suggest strong influence of factors specific to each lipoprotein in the familial associations. But the highly significant intraindividual correlations and the nonnegligible cross-correlations among relatives also suggest the additional presence of common underlying factors for the familial associations, especially between HDL-C and VLDL-C and to a lesser extent between LDL-C and VLDL-C. The issues stemming from these analyses and the directions for further analyses are discussed.