The collaborative lipid research clinics program family study: III. Transformations and covariate adjustments of lipid and lipoprotein levels

Green, P. P., K. K. Namboodiri (Dept of Biostatistics, U. of North Carolina, Chapei Hill, NC 27514), P. Hannan, J. Martin, A. R. G. Owen, G. A. Chase, E. B. Kapian, L. Williams and R. C. Eiston. The Coliaborative Lipid Research Clinics Program Famliy Study. III. Transformations and covariate adjustments of lipid and lipoprotein levels. Am J Epidemiol 1984; 119: 959-74.Several methods of transformation and covarlate adjustment have been applied to the Coliaborative Lipid Research Clinics Program Famliy Study data to facilitate analysis of lipld levels of examinees of different sex and age groups. After several exploratory analyses, choiesterol, high density lipoprotein cholesterol, and low density lipoprotein cholesterol were logarithmicaily transformed, and triglycerides were transformed by the power -0.25. Two types of covariate adjustment procedures, the regression and Z score methods, were used. A modified regression method was developed and was found to be preferabie to both simple cubic regression and the Z score method on theoretical and empirical grounds. Refinements in this method to correct for change in variance with age, and for the effect of socioeconomic status, seasonaliy, and anthropometric measures were made. Methodoiogical Issues connected with the transformation and adjustment procedures are discussed.