Matrix effects and accuracy assessment. Identifying matrix-sensitive methods from real-time proficiency testing data.

Abstract

Retrospective proficiency testing data may be used to screen for matrix effect, which is suspected when significant methodologic bias is detected. Prioritization is centered on those situations where relatively high unacceptability rates are associated with high bias. Before attributing high unacceptability rates to methodologic bias, factors such as peer grouping strategies, number of participants in peer groups, analyte concentration, evaluation limits, and interparticipant precision, all of which affect scoring, need to be considered. When bias-associated high unacceptability rates are detected, methods are candidates for further evaluations to identify the causes of bias, to determine if matrix effect is present. Using outputs from the College of American Pathologists Survey Management Enhancement Program, we have presented data from the 1991 Comprehensive Chemistry Survey to demonstrate methodology to study this problem, and have selected examples of method performance in which tabular and graphic information contribute to its study. For the analytes studied, unacceptability rates tend to be greatest in association with high methodologic bias and/or between-participant imprecision relative to external evaluation limits, samples at the upper extreme of analyte concentration, low numbers of participants in peer groups, grading by comparative method, and concentration-independent fixed evaluation limits. Specific examples of significant bias that is possibly matrix induced are presented for uric acid, glucose, creatinine, and total protein.