In the Helsinki Heart Study 2,590 subjects (63.5% of total) had a type IIa hyperlipoproteinemia at screening. Baseline low-density lipoprotein (LDL) cholesterol (mean 193 mg/dl; 5 mmol/liter) and high-density lipoprotein (HDL) cholesterol (mean 50.2 mg/dl; 1.3 mmol/liter) showed no statistical correlation (r = 0.046). Both the placebo (1,293 patients) and gemfibrozil groups (1,297 patients) were divided into tertiles by baseline HDL and LDL cholesterol to determine the relative predictive risk of developing coronary artery disease. In a population with elevated LDL cholesterol, it is significant that the lipoprotein fraction with the greatest predictive value was HDL cholesterol. The severity of LDL cholesterol elevation did not provide any differential predictive value for coronary artery disease.