Analytical performance and comparability of the determination of cholesterol by 12 Lipid-Research Clinics.

Abstract

Twelve Lipid-Research Clinic laboratories performed automated cholesterol analyses on four control-serum pools of known cholesterol concentration, using the Liebermann-Burchard reaction. The analyses were done during a two-year period, with the same standards, methodology, and quality-control procedures. Estimates of analytical bias, variability, and short- and long-term trends for each instrument and for the entire group of LRC instruments are presented. High accuracy, precision, and interlaboratory comparability were achieved through the rigorous standardization and control of the entire analytical procedure. The significance of these results for long-term collaborative studies is discussed. Individual laboratory biases averaged from 0.5 to 2.0% below Abell-Kendall reference values. Between-run variability was about equal to within-run variability and inter-laboratory variation was substantially less than intra-laboratory variation. The total standard deviation for all instruments was about 0.04 g/liter. Only 8-15% of this variation was due to differences between instruments. The between-instrument standard deviation ranged from 0.011 to 0.015 g/liter; the between-run, within-instrument standard deviation ranged from 0.023 to 0.030 g/liter; and within-run standard deviation ranged from 0.023 to 0.028 g/liter. The significance of the achieved results for long-term collaborative studies is discussed.