Comparison of two homogeneous HDL cholesterol methods in a large population study

OBJECTIVES: High-density lipoprotein cholesterol (HDL-C) is an independent risk factor for coronary heart disease. Data are lacking on the performance of homogeneous methods using a large number of samples. DESIGN AND METHODS: We compared the performance of two HDL-C direct assays, the Dimension RxL (the Dade method) and the COBAS INTEGRA (the Roche method), for population screening. Performance was assessed using 4214 sera obtained from the International Collaborative Study on Atherosclerosis and Stroke In Asia (InterASIA) participants. RESULTS: The method comparison results demonstrated that both methods were highly correlated (r = 0.972). Deming regression analysis showed a slope of 1.009 +/- 0.007, an intercept of 0.048 +/- 0.009 and a S(y/x) of 0.08. The means were 1.29 +/- 0.33 and 1.23 +/- 0.33 mmol/l for the Roche and Dade methods, respectively, and the range of observed values were 0.30-3.05 and 0.19-2.86 mmol/l, respectively. The 95% confidence interval for the mean of the method differences was -0.10 to 0.22 mmol/l. Percentages of low (> or = 1.55 mmol/l), normal (1.03-1.54 mmol/l), and high risk (< 1.03 mmol/l) results were 15.5, 55.4, 29.0 for the Dade and 19.3, 59.3, 21.4 for the Roche method. The percentage of concordantly classified subjects at each cut point was 77.1%, 84.4%, and 95.5%. The percentage of overall consistency subjects was 85.4%. Thirteen percent of subjects were discordantly classified into the higher-risk group while the 1.6% of subjects were discordantly classified into the lower-risk group. CONCLUSIONS: Both homogeneous HDL-C methods were correlated and agree well with one another. The percentage of concordantly classified subjects was high. Thus, either method is suitable for large population studies.