Background Factors predicting the occurrence of premature coronary artery disease (CAD) may not be quantitatively the same as those predicting CAD severity, particularly in women, in whom there have been few studies.50%) luminal obstructions and from a coronary disease severity score. We related severity to quantitative and categorical atherogenic variables and assessed severity of angina (no angina, stable angina, or unstable angina) at the time of study in the same way. There were eight variables independently predictive of severity: in descending order of relative importance, male gender, diabetes, smoking dose, ratio of total cholesterol to high-density lipoprotein cholesterol (TC/HDL-C), lipoprotein(a) (Lp(a)), age, positive family history, and hypertension. These correctly classified 43.3% of patients into no-, one-, two-, and three-vessel disease categories and accounted for 25.8% of variance of severity. Among 246 patients not taking lipid-lowering or beta -blocking drugs, these variables (in slightly different order) correctly classified 49.2% of patients and accounted for 36% of the variance. Among men (n=427), seven significant variables correctly classified 39.3% of patients compared with 54.5% in women (n=167). For those not taking the above drugs, these proportions were 49.4% and 65.4%, respectively. Among the quantitative variables, total smoking dose was the most predictive independent variable irrespective of current or ex-smoking habit and was more predictive in women than in men; of the lipid variables, high TC/HDL-C (or low HDL-C) and high Lp(a) were consistently highly predictive for all patients and in the subgroup analyses. Patients with unstable angina had higher coronary severity scores and Lp(a) levels and were more likely to have diabetes, hypertension, or a positive family history.Conclusions We conclude that the quantitative variables most relevant to severity of premature CAD and to its prevention in Australian men and women are total amount of lifetime smoking, TC/HDL-C (or HDL-C), and Lp(a) and that patients with unstable versus stable angina usually have more severe disease and higher Lp(a). (Circulation. 1994;89: 1974-1981.)