Context: The National Kidney Disease Education Program recommends calculating glomerular filtration rate from serum creatinine concentration. Accurate creatinine measurements are necessary for this calculation.
Objective: To evaluate the state of the art in measuring serum creatinine, as well as the ability of a proficiency testing program to measure bias for individual laboratories and method peer groups.
Design: A fresh-frozen, off-the-clot pooled serum specimen plus 4 conventional specimens were sent to participants in the College of American Pathologists Chemistry Survey for assay of creatinine. Creatinine concentrations were assigned by isotope dilution mass spectrometry reference measurement procedures.
Participants: Clinical laboratories with an acceptable result for all 5 survey specimens (n = 5624).
Results: The fresh frozen serum (FFS) specimen had a creatinine concentration of 0.902 mg/dL (79.7 micromol/L). Mean bias for 50 instrument-method peer groups varied from -0.06 to 0.31 mg/dL (-5.3 to 27.4 micromol/L), with 30 (60%) of 50 peer groups having significant bias (P < .001). The bias variability was related to instrument manufacturer (P < or = .001) rather than method type (P = .02) with 24 (63%) of 38 alkaline picric acid methods and with 6 (50%) of 12 enzymatic methods having significant biases. Two conventional specimens had creatinine concentrations of 0.795 and 2.205 mg/dL (70.3 and 194.9 micromol/L) and had apparent survey biases significantly different (P < .001) from that of the FFS specimen for 34 (68%) and 35 (70%) of 50 peer groups, respectively.
Conclusions: Thirty of 50 peer groups had significant bias for creatinine. Bias was primarily associated with instrument manufacturer, not with type of method used. Proficiency testing using a commutable specimen measured participant bias versus a reference measurement procedure and provided trueness surveillance of instrument-method peer groups.