Total error assessment of five methods for cholesterol screening.

We report the accuracy, imprecision, total analytical errors, and patient misclassification errors for cholesterol measured from capillary whole blood, venous whole blood, and venous plasma samples by five devices used in public cholesterol screening environments: Reflotron, Vision, Ektachem DT-60, QuickRead, and Liposcan. None of the methods met the National Cholesterol Education Program (NCEP) performance recommendations of 3% CV with 3% bias. The Vision and Reflotron methods used with venous samples gave individual results with total errors consistent with a combined CV and bias in the 4-5% range; capillary blood samples had total errors > 5% (combined CV and bias criteria). The DT-60 performance was near the 5% total error criterion for capillary samples and was > 5% for venous samples. Misclassification of individuals into desirable or referral groups for venous samples was as great as 5.1% for the DT-60, 5.7% for the Vision, and 7.1% for the Reflotron. Misclassifications for capillary blood samples were as great as 6.7%, 18.3%, and 14.1% for DT-60, Vision, and Reflotron, respectively. The QuickRead and Liposcan results were substantially poorer than those obtained by the other methods.