Impact of creatinine calibration on performance of GFR estimating equations in a pooled individual patient database

Background: Variation in performance of glomerular filtration rate (GFR) estimating equations is related to variation in calibration of the creatinine assay across clinical laboratories. Study Design: Cross-sectional analysis. Setting & Participants: 6 research studies and 4 clinical populations including 5,504 participants who had GFR measured using urinary clearance of iothalamate. Measurements: Standardized serum creatinine values obtained by means of calibration to the Cleveland Clinic Research Laboratory using frozen specimens, a calibration panel, and/or survey results from the College of American Pathologists. Predictor: Noncalibrated serum creatinine assayed in research and clinical laboratories compared with standardized serum creatinine. Outcome: Difference between measured GFR versus GFR estimated from the Modification of Diet in Renal Disease (MDRD) Study and Cockcroft-Gault equations. Results: For a noncalibrated serum creatinine value of 1 mg/dL (88.4 mu mol/L), standardized serum creatinine value was 0.07 mg/dL (6.2 mu mol/L) less than noncalibrated values. In the pooled data set, for the MDRD Study equation, calibration improved median percentage of difference between measured and estimated GFR from 9.0% (interquartile range [IQR], 28%) to 5.8% (IQR, 28%) and improved the percentage of estimates within 30% of measured GFR (P-30) from 80% to 83%. The effect of calibration was greater at higher levels of GFR and varied across studies. For the Cockcroft-Gault equation, calibration worsened the median percentage of difference from -2.0% (IQR, 38%) to -11.4% (IQR, 39%), and the P-30 from 74% to 69%. Limitations: College of American Pathologist samples were used for calibration of clinical populations; calibration factors do not account for drift over time in the serum creatinine assay; calibration cannot account for variation in assay performance among individuals. Conclusion: Calibration improves the performance of the MDRD Study equation. After calibration, larger errors remain for GFR estimates greater than 60 mL/min/1.73 m(2) ( > 1 mL/s/1.73 m(2)).