An Approach for Evaluating Potential Screening Thresholds Using Biomarker Population Distribution and Analytical Imprecision.

Abstract

A common approach in laboratory medicine is to use a simple but sensitive test to screen samples to identify those that require additional investigation with a more complex and informative method. Selection of screening thresholds can be guided by biomarker distribution in the tested population and the analytical imprecision of the method. A simulation using joint probabilities derived from the population distribution for galactose-1-phosphate uridylyltransferase (GALT) activity and the analytical imprecision for the GALT assay was used to estimate the number of samples that would require repeat analysis and the number of samples with possibly false-negative screening determinations due to analytical imprecision. In the case of GALT activity, screening a conservative initial threshold 6 standard deviations from the confirmation threshold can essentially eliminate the chance of a false-negative screening determination due to analytical imprecision. The trade-off is a greater number of samples requiring follow-up testing (n = 222, equivalent to 0.15% of samples annually). Selection of thresholds in a screening algorithm is informed by estimates of the number of samples that would require repeat testing and the number that could be false negative due to analytical imprecision.