External quality-control survey of cholesterol analyses performed by 12 lipid research clinics.

We report accuracy and precision achieved in the automated analysis for cholesterol in a long-term multilaboratory study, presenting and evaluating the significance of data accumulated by 12 Lipid Research Clinics (LRC's) in the analysis of 18 unknown surveillance pools during three years. The average bias for all pools and for 13 autoAnalyzer II (Technicon Instruments Corp., Tarrytown, N.Y. 10591) instruments in the 12 clinics was -0.41% (range -1.2 to +0.3%), as compared to values established by reference methodology. The regression equation relating observed cholesterol values (y) to reference values (x) was: y = 0.35 + 0.977x. The bias varied from pool to pool (-2.3 to +5.3%), positive biases being observed for pools with cholesterol concentrations less than 1.4 g/liter, and negative biases for those pools with higher concentrations. Total standard deviations ranged between 25 and 75 mg/liter, and total CV's for most individual instruments were between 1 and 3%. Of the variability for a particular pool, less than 20% was due to differences among instruments, and within- and between-run variabilities were approximately equal. These trends were the same as those previously observed [Clin. Chem. 23, 1744 (1977)] in the analysis of bench control pools of known cholesterol concentration.