Aim: Low-density lipoprotein cholesterol (LDL-C), the ratio of LDL-C to high-density lipoprotein cholesterol (HDL-C; LDL-C/HDL-C), and non-HDL-C were evaluated to determine their ability to predict cardiovascular disease (CVD) risk with pravastatin treatment. Methods: We conducted a large-scale randomized primary prevention trial in Japan (MEGA Study), in which we randomly allocated 7832 mild hypercholesterolemic patients to diet alone (n 3966) and diet plus pravastatin groups (n 3866) and followed them for an average of 5 years. We compared baseline levels and the CVD incidence in the diet alone group, and time-dependent receiver operating characteristic curves in the overall population. To determine the best parameter for predicting the efficacy of pravastatin, the diet plus pravastatin group was divided into tertiles to compare lipid parameters and CVD incidence versus the diet alone group. Results: Significantly graded correlations were found between CVD and LDL-C/HDL-C and non-HDL-C. Significantly more CVD events were associated with LDL-C/HDL-C 186 mg/dL and LDL-C/HDL-C 2.9. Furthermore, LDL-C/HDL-C or non-HDL-C was more predictive than LDL-C. By measuring LDL-C/HDL-C or non-HDL-C, we allocated 32% of the diet plus pravastatin group into a different risk category. The lowest significant incidence of CVD was found in patients with LDL-C 119.8-133.4 mg/dL, LDL-C/HDL-C 1.9, and non-HDL-C 145.2-160.8 mg/dL. Conclusion: Non-HDL-C and LDL-C/HDL-C have a greater ability to predict CVD risk in mild-tomoderate hypercholesterolemic Japanese individuals than LDL-C, and are more useful to evaluate the effect of pravastatin; however, these parameters should be interpreted independently when assessing CVD risk.