Total Error Evaluation of Roche Direct HDL-Cholesterol Reagent and Calibrator across 31 Lot Combinations: A 2-Year Experience

Publisher Website
Google Scholar
Abstract
Professionally set quality specifications are needed as major considerations in development of new reagents (1). Two years ago, the Dutch Lipid Reference Laboratory, a permanent international member of the CDC Cholesterol Reference Method Laboratory Network (2)(3), in collaboration with Roche Diagnostics Nederland B.V. (formerly Boehringer Almere, The Netherlands), began evaluating lot-to-lot differences of the new direct HDL-cholesterol (HDL-C) reagent from Kyowa Medex (cat. no. 1731157) and of the HDL-C/LDL-cholesterol cfas calibrator (cat. no. 1778501) before distribution on the Dutch market. To this end, a scaled down split-sample comparison protocol was used, essentially according to the design described in the HDL Cholesterol Method Evaluation Protocol for Manufacturers (4)(5). The accuracy platform was the CDC Designated Comparison Method (DCM) for HDL-C (2)(4); the test method was run on an Hitachi 911 analyzer (Boehringer Mannheim/Roche). A split-sample comparison was done with six specimens covering the HDL-C concentration range. All specimens were from individual donors. The matrix types investigated with the test method were serum and heparin plasma, whereas serum was used in combination with the DCM. Fresh specimens, intermittently stored at 4 °C, were analyzed with the test method, whereas frozen split sera were analyzed with the HDL-C DCM (5)(6). With the test method, all specimens were run in duplicate during 3 consecutive days for a total of 36 measurements (6 × 3 × 2 measurements); with the DCM, duplicate analyses were performed in one analytical run for a total of 12 measurements (6 × 2 measurements). Both test and reference data were produced in the Lipid Reference Laboratory in Rotterdam. Analytical performance was checked against the National Cholesterol Education Program guidelines (4). New lot combinations were acceptable when the total error criterion of 13% was met. Total error can be considered as an error budget that can be divided between imprecision and bias (4)(7).