Influence of analytical bias and imprecision on the clinical performance of reference intervals

This simulation study explores the influence of bias and analytical imprecision combined with biological variation on misclassification rates (expressed as clinical specificity and hypothetical clinical sensitivity) of reference intervals. The simulated data used in this study were drawn from three inter-related parameters: (1) the mean (physiological set point) of a patient from the overall population distribution, (2) the 'true' biological variation of a sample drawn from the within-subject distribution of the patient, and (3) the observed laboratory measurement drawn inclusive of bias and analytical imprecision. Clinical specificity is the proportion of non-pathological patients who are correctly classified as non-pathological. Hypothetical sensitivity is the proportion of pathological patients who are correctly classified as pathological. The hypothetical sensitivity of reference intervals is influenced more by the distance between non-pathological and pathological result distributions. In the presence of increased imprecision and bias (away from the reference limit), the hypothetical sensitivity increases since the pathological distribution is moved further away from the reference limits. The presence of introduced bias in a direction towards one of the reference limits has the same effect as reducing the distance between the non-pathological and pathological distributions, which reduces hypothetical sensitivity. The data generated from this study illustrate the interrelationships between biological variation, analytical imprecision, clinical performance and characteristics between pathological and non-pathological populations. These parameters display a linear relationship under certain conditions and may provide a pragmatic solution to estimate the impacts on clinical specificity and hypothetical sensitivity.