The relations of high density lipoprotein (HDL) subclasses to severity and rate of progression of coronary atherosclerosis were investigated in 60 men who had survived a myocardial infarction before the age of 45 years and who had subsequently undergone two coronary angiographies, with an intervening time interval of 4-7 years between angiographies. Five HDL subclasses with different particle sizes were determined by gradient gel electrophoresis, and the major serum lipoprotein classes were separated by preparative ultracentrifugation in connection with the second angiography. Highly significant inverse correlations were found between the plasma levels of the largest HDL particles, the HDL(2b) subclass, and both disease severity as observed on the second coronary angiogram (r = - 0.53, p < 0.001) and progression of coronary lesions between angiographies (r = - 0.38, p < 0.01). Grouping the patients according to the presence or absence of very low density lipoprotein (VLDL) triglyceride elevation revealed striking differences in the relations of HDL subspecies to coronary atherosclerosis between normotriglyceridemic and hypertriglyceridemic subjects. There were strong inverse correlations between the plasma HDL(2b) concentration and both severity of lesions (r = - 0.72, p < 0.001) and rate of lesion progression (r = - 0.58, p < 0.01) in the normotriglyceridemic patients, whereas this relation was absent in subjects with hypertriglyceridemia. Although it was retrospective in design, affected by an inherent selection bias due to unavoidable dropouts and exclusions, and limited to young, male, postinfarction patients, the study suggests that the inverse relation between HDL level and coronary heart disease or coronary atherosclerosis seen in previous epidemiological and angiographic studies is accounted for by the largest HDL particles, the HDL(2b) subclass. Furthermore, the hypertriglyceridemic state appears to markedly influence the relations between plasma concentrations of HDL subclasses and coronary atherosclerosis. Prospective studies of unselected patients are needed to corroborate these findings.