Combined hyperlipoproteinemia (CHL) can be difficult to treat because of the heterogeneous nature of the lipoprotein abnormalities. We compared the relative efficacies of simvastatin and gemfibrozil and sought predictors of responsiveness in terms of the baseline lipids and other potential metabolic determinants (plasma insulin, Lp(a) and apo E genotype). Sixty-six subjects entered a cross-over, randomized trial involving 12 weeks on each drug. Efficacy was assessed after 6 and 12 weeks on each treatment. Simvastatin lowered total cholesterol 24%, triglycerides 12%, LDL cholesterol 33%, raised HDL cholesterol 13% and substantially reduced the cholesterol:triglyceride ratio in VLDL and IDL. Gemfibrozil lowered total cholesterol 5%, triglycerides 44%, raised HDL 26% and reduced VLDL and IDL lipids more than simvastatin did. LDL size increased with both treatments and HDL size increased with simvastatin. Responsiveness (25% fall in cholesterol or 40% fall in triglycerides) was shown by 31/61 subjects when taking simvastatin (cholesterol-lowering) and by 44/60 taking gemfibrozil (triglyceride-lowering). Responsiveness was greatest in those with apo E2 genotype with both drugs (P < 0.05). Unexpectedly, responders to simvastatin tended to have lower baseline total cholesterol but higher triglyceride levels than those whose cholesterol or triglyceride was lowered by gemfibrozil. Nevertheless, more hypercholesterolemic subjects responded to simvastatin and more hypertriglyceridemic subjects to gemfibrozil. Lp(a) (P = 0.04) and plasma insulin concentrations (P = 0.03) were negative predictors of percentage triglyceride-lowering with gemfibrozil. The difference between the two drugs in triglyceride-lowering lessened with rising insulin and falling HDL cholesterol. Thus, the responsiveness to the two major classes of lipid lowering drugs can be partly predicted from baseline lipids and related metabolic parameters.