Improving the interlaboratory variation for creatinine serum assay

PURPOSE: To assess inter-assay variation and accuracy of blood creatinine measurements as well as the effect of the standardization of the calibration procedures on inter-assay variation. METHODS: Inter-assay variation and accuracy were assessed using 30 frozen human sera and 3 certified reference materials, which were analysed by 17 creatinine assays (colorimetric: 12, enzymatic: 4, HPLC: 1). Usual calibration procedure was compared with two common calibration procedures using either a reference material (404.1 micromol/L), or secondary sera calibrators (69, 115 et 180 micromol/L). RESULTS: Most of the commercially available methods display inaccuracy, > 10% for creatininemia < 150 micromol/L in most cases. For this concentration range, the mean creatininemia was statistically significantly different as a function of the assay used (p < 0.001). Enzymatic assays produced lower results than colorimetric ones for low creatinine levels but higher results for high creatinine levels. Assays being calibrated according to the manufacturer's recommendations, the median dispersion factor was 14% for the 20 samples between 45 and 150 micromol/L, and 8% for the 10 samples between 250 and 350 micromol/L. The calibration procedure modified inter-assay variation significantly (p < 0.001) but we gained little advantage from both common calibration procedures. A significant decrease of inter-assay variation occurred within each technical group (colorimetric or enzymatic) when a common calibration was performed using calibrators which concentration(s) was(were) close to the concentrations to be measured. CONCLUSIONS: Inter-assay variation is too high to allow prediction of glomerular filtration rate (GFR) or creatinine clearance from serum creatinine level. Our results highlight the interest of a calibration procedure using several concentrations with at least one between 90 and 150 micromol/L. The marketing of such a calibrator should be considered in order to decrease inter-assay variation in the range of creatinine levels which defines a mild chronic renal failure. Such an approach will certainly reduce inter-assay variation only within each technical group but could allow to include technical group as a co-variable in the algorithms developed for predicting GFR or creatinine clearance. A global transferability will certainly need the correlation of all types of creatinine assays versus a definitive method, whom definition remains uncertain.