We examined the analytical performance of eight compact systems for me
asuring total cholesterol: AccuMeter, Cobas Ready, Discovery f2, DT60,
L-D-X, Reflotron, QCA, and Vision, We determined average bias at two
decision levels, the mean absolute bias, and the percentage of results
differing from the comparison method results by >8.9% allowable total
error limit for multiple reagent lots. Average bias was <3% for all l
ots tested for AccuMeter, Discovery f2, and DT60, but >3% for one or m
ore lots or sample types tested with the other systems. Of results fro
m each reagent lot, >95% were within the 8.9% total error specificatio
ns with Discovery f2, DT60, and QCA, whereas the performance of L-D-X,
Vision, and Reflotron depended on reagent lot and (or) sample type. O
f all results from each lot tested with AccuMeter and Cobas Ready, >5%
exceeded the total allowable error limit. We determined imprecision f
or five systems: Cobas Ready, Discovery f2, and QCA had CVs <3%, where
as CVs for AccuMeter and L-D-X were >3% but <5%.