Standardization of serum creatinine and estimated GFR in the Kidney Early Evaluation Program (KEEP)

Creatinine calibration by clinical laboratories is important because variability among assays adversely affects the accuracy of glomerular filtration rate (GFR) estimation. We describe the calibration of creatinine assays used in the National Kidney Foundation Kidney Early Evaluation Program (KEEP).

Creatinine values were requested for 200 samples at each of the 2 KEEP laboratories, Satellite Laboratory Services, LLC (2000 to 2005) and Clinical Laboratory Services (CLS; 2005 to present), for comparison with samples at the Cleveland Clinic Research Laboratory (CCRL). Linear regression and Deming regression were used to obtain slopes adjusted for measurement error and regression to the mean.

After exclusion of outliers, mean creatinine level in 184 samples was 0.94 mg/dL at Satellite compared with 0.89 mg/dL at CCRL. Deming regression showed a slope of 1.003 (95% confidence interval (CI), 0.99 to 1.02; P < 0.001) and intercept of -0.04 (95% CI, -0.59 to -0.02; P = 0.003) with R(2) = 0.9853. Final calibration consists of intercept alone because of a small slope. After exclusion of outliers, mean creatinine level in 199 samples was 1.06 mg/dL at CLS compared with 0.96 mg/dL at CCRL. Deming regression showed a slope of 1.08 (95% CI, 1.07 to 1.09; P < 0.001) and intercept of -0.18 (95% CI, -0.19 to -0.17; P < 0.001) with R(2) = 0.9939. GFR estimates were minimally affected by the Satellite calibration. At a serum creatinine value of 1 mg/dL, the change in estimated GFR was 1 mL/min/1.73 m(2) after calibration. Conversely, higher range GFR estimates were affected by calibration of the CLS creatinine assay. At a serum creatinine value of 1 mg/dL, the GFR estimate was 6 mL/min/1.73 m(2) higher after calibration.

Calibration of KEEP creatinine measurements had a greater impact on the current laboratory than on the laboratory previously used. The calibration process has worked to decrease overestimation of eGFR at the high range and decrease misclassification bias.