A conceptual model for establishing tolerance limits for analytic bias and imprecision based on variations in population test distributions

A conceptual model is proposed for defining analytic bias limits utilizing the variations found in cumulative test value distributions. The model is based on the propositions that changes in analytic bias are more important than analytic imprecision in medical diagnoses and that analytic bias alters clinical specificity more than clinical sensitivity. The rationale for these propositions are presented along with a step-by-step procedure for estimating bias tolerance limits. These concepts are illustrated with an example using prostate-specific antigen. A second protocol is provided to define analytic imprecision tolerance limits, based on the quality control performance characteristics required to maintain the bias tolerance limits. This model can be applied to most chemistry, immunoassay, and hematologic quantitative assays. The relationship of this procedure to the published procedures using biologic variation for defining analytic tolerance limits is discussed.