In eight unique challenges mailed by The College of American Pathologi
sts Comprehensive Chemistry Survey to participating laboratories withi
n 3 years, results for direct-reacting bilirubin (DBIL) were highly va
riable among the 12 largest peer groups, and most of the results diffe
red greatly from the values obtained by a preferred method. Peer-group
mean values for total bilirubin (TBIL) were in much better agreement
with each other and with those obtained by the Reference Method for TB
IL. From a review of the information on the assay of DBIL provided to
us by the manufacturers, we conclude that among the major causes of th
e large variability and bias in DBIL assays are problems with calibrat
ion, lack of a serum blank measurement, inadequate concentrations of H
CI in the reaction mixture, inappropriate use of bichromatic correctio
n methods, and possibly the use of wetting agents or surfactants in th
e reagent. Within-group SDs were small and generally acceptable. The a
mong-peer-group variability in DBIL values is attributable to bias, no
t imprecision. We recommend several simple changes that could improve
the accuracy of DBIL determinations in clinical laboratories.