Bilirubin Proficiency Testing Using Specimens Containing Unconjugated Bilirubin and Human Serum: Results of a College of American Pathologists Study

Abstract

Specimens of the College of American Pathologists Neonatal Bilirubin and Chemistry surveys are inadequate for evaluating the performance of clinical laboratories in measuring serum bilirubin because they exhibit strong matrix interference. Recently published data indicate that at least 1 major clinical analyzer provided inaccurate bilirubin values for Neonatal Bilirubin Survey specimens. The composition of the specimens, bovine serum enriched with ditaurobilirubin, was responsible for the erroneous results. This article evaluates the performance of major clinical analyzers using a survey specimen free of matrix interference. A human serum-based specimen enriched solely with unconjugated bilirubin was included in the 2003 Neonatal Bilirubin and Chemistry surveys. Its bilirubin concentration (19.4 mg/dL [332 micromol/L]) was determined by the reference method for total bilirubin. The coefficients of variation for the 4 major clinical analyzers (Dimension, Hitachi, Synchron, and Vitros) ranged from 1.9% to 3.7%. When compared to the bilirubin value measured by the reference method, mean bilirubin values of the 4 major clinical analyzers and College of American Pathologists (CAP) All Data (which refers to the grand mean and overall coefficient of variation of all method principles, all instruments according to CAP terminology) ranged from -3.5% to 5.1%. Direct bilirubin results from most field methods showed good specificity overall. Human serum-based survey specimens, having their bilirubin concentrations determined by the reference method, should be included as frequently as feasible in the Neonatal Bilirubin Survey. Such specimens may be used by instrument manufacturers as standards for calibrating bilirubin methods and for assigning values to calibrators provided to instrument users. A substantial improvement in bilirubin measurements due to the reduction of systematic error is expected.