Analytic goals, concentration relationships, and the state of the art for clinical laboratory precision

We analyzed the College of American Pathologists' Quality Assurance Service 1990 database of 1500 clinical laboratories to evaluate concentration relationships and the state of the art for the precision of 37 analytes. We compared these data to previously reported Quality Assurance Service precision performance data from 1974 through 1980. The general trend of decreasing average long-term, within-method, within-laboratory coefficients of variation continues. However, among 24 routine chemical and therapeutic drug monitoring tests observed for 11 to 16 years, 11 showed no evidence of continuous improvement in precision. The 1990 average coefficients of variation usually exceeded analytic goals modeled by clinical responses when these models assumed that variability in test results arises solely from the analytic performance of the method. In contrast, they usually did not meet goals that take into account the presence of biologic factors among the sources of test variation. Notable exceptions are the average coefficients of variation for cholesterol and triglycerides, which markedly decreased during the National Cholesterol Education Program initiatives of the 1980s. Continuous improvement of precision in disciplines dominated by relatively mature automated methods may not be the norm unless structure and direction are provided by appropriate and accepted analytic goals.