Background Apolipoprotein (Apo) levels are considered more reliable than plasma lipoprotein levels for predicting coronary artery disease (CAD). However, a unanimous Apo marker for CAD has not been identified. In the Chin-Shan Community Cardiovascular Cohort (CCCC), we sought to identify a common Apo marker for predicting CAD in the general population. Methods We examined the cross-sectional association between Apo markers and CAD in the CCCC from 1990 to 2001. Among 3,602 subjects, 90 had angiographically proven CAD (>50% stenosis in ≥1 vessel), and 200 did not have CAD. These subjects were divided into the following 4 groups for analysis: normolipidemic (total cholesterol [TC] <200 mg/dL, triglyceride [TG] <150 mg/dL), hypertriglyceridemic (TC <200 mg/dL, TG ≥150 mg/dL), hypercholesterolemic (TC ≥200 mg/dL, TG <150 mg/dL), and hyperlipidemic (TC ≥200 mg/dL, TG ≥150 mg/dL). Results Compatible with findings in other populations, our results showed that CAD patients in the CCCC had higher ApoB and lower high-density lipoprotein (HDL) cholesterol and ApoAI concentrations than non-CAD subjects, but the differences were not significant in all groups. Plasma concentrations of ApoE and lipoprotein (a) were not consistently correlated with CAD. In contrast, the ratio of HDL-ApoCIII to very-low-density lipoprotein (VLDL)-ApoCIII was the only universal determinant for CAD in the normolipidemic group (P=0.0018), the hypertriglyceridemic group (P=0.0001), the hypercholesterolemic group (P=0.0001), and the hyperlipidemic group (P=0.0001). Overall, a high HDL-ApoCIII/VLDL-ApoCIII ratio was observed in all CAD patients, including those with a normal lipid profile. In multivariate analyses, the HDL-ApoCIII/VLDL-ApoCIII ratio was the strongest predictor for CAD among all lipid factors investigated (odds ratio, 2.04; 95% confidence interval, 1.46–2.84; P<0.0001). Conclusions A high HDL-ApoCIII to VLDL-ApoCIII ratio is a better marker for predicting CAD than are the conventional lipid markers or ApoAI and ApoB. High HDL-ApoCIII and low VLDL-ApoCIII values in CAD, irrespective of lipid variations, suggest that ApoCIII is markedly transported from VLDL to HDL in this disease. Measurement of plasma ApoCIII may improve CAD prediction in the general population.