1995 Institute: Frontiers in Laboratory Practice Research

A major clinical use of laboratory tests is for classifying patients into diagnostic and treatment categories. Multiple factors influence the decision limits used for these decisions; however, the analytic performance of the assays seldom is explicitly considered in these decisions, even though both imprecision and bias may significantly alter the decisions. Of these two factors, changes in analytic bias have the most dominant effect. Assay imprecision adds only indirectly to the overlap of the distributions of test values by increasing the variance. The analytic contribution to the total variance is indirect because it is combined with biologic variations, which serve to buffer any changes. Analytic bias directly affects classification decisions by shifting the distribution of test values. This effect is greatest for values near the decision levels where the bias may alter classification. Thresholds for medical decisions generally are determined using data collected when the assays are initially calibrated; if the assays shift or are recalibrated to a different level, the number of patients exceeding the decision thresholds are directly increased or decreased. Unlike precision problems, repeat testing does not help to minimize these misclassifications. For example, decisions to pursue hyperparathyroidism often are triggered by calcium levels exceeding a defined threshold. Small analytic shifts upward markedly increase the number of patients investigated. Analyzing variation in the percentages of patients exceeding selected thresholds during times when the assays are analytically unbiased can provide reference standards for establishing tolerance limits for analytic bias. All levels of analytic bias will directly affect the medical decisions, but assays in which the bias is held within these tolerance limits should not substantially alter the number of patients misclassified. Maintaining analytic bias within these tolerance limits then becomes the primary analytic performance goal. Secondary goals for assay precision can be defined in terms of the quality control systems required to maintain the bias goals. Introduction The two major analytic quality control "systematic error" or bias for inaccuracy. parameters are imprecision and inaccuracy. For many clinical chemistry analytes, 1 The EFCC expert panel defines imprecision reference methods and reference standards as the "standard deviation or coefficient of do not exist. Therefore, it is not possible to variation of the results in a set of replicate define "true values." In this discussion the measurements." Inaccuracy is defined as term “bias” is used to describe the systematic 2 the "numerical difference between the mean error representing the mean analytic of a set of replicate measurements and the difference between the current measurement true value." Westgard has used the terms system and the system that was used to "random error" for imprecision and