Impact of Various Body Weights and Serum Creatinine Concentrations on the Bias and Accuracy of the Cockcroft‐Gault Equation

Abstract

Study Objective To evaluate the impact of various body weights and serum creatinine (S_cr) concentrations on the bias and accuracy of the Cockcroft‐Gault creatinine clearance (C‐G Cl_cr) equation compared with measured 24‐hour Cl_cr. Design Retrospective analysis. Setting Tertiary care hospital. Patients A total of 3678 patients with stable renal function and who underwent a 24‐hour urine collection between July 1, 1996, and June 30, 2010. Measurements and Main Results For each patient, C‐G Cl_cr was calculated and compared with a measured 24‐hour Cl_cr. Body weight adjustments to the calculation were performed based on the following weight classifications: underweight, normal weight, overweight, obese, and morbidly obese. In addition, C‐G Cl_cr was calculated by using rounded S_cr values based on two S_cr thresholds—0.8 mg/dl and 1 mg/dl—for patients with measured S_cr values below those thresholds. Those patients were then evaluated after stratification into two age groups: all ages and a subgroup of patients aged 65 years or older. The S_cr‐rounded C‐G Cl_cr values were compared with the C‐G Cl_cr values using actual S_cr values. Mean differences were calculated, and accuracy was evaluated. Use of actual body weight in the calculations for underweight patients resulted in an unbiased Cl_cr of −0.22 ml/minute (p=0.898). Use of ideal body weight in the calculations of patients of normal weight returned an unbiased Cl_cr of −1.3 ml/minute (p=0.544). An unbiased C‐G Cl_cr could not be calculated for other weight categories. In those patients, adjusted body weight using a factor of 0.4 (ABW_0.4) was the least biased and most accurate. In patients aged 65 years or older with an S_cr less than 0.8 mg/dl and less than 1 mg/dl, actual S_cr was unbiased (−3 ml/min [p=1] and −9 ml/min [p=0.279], respectively) and more accurate than rounded S_cr. In patients of all ages with an S_cr less than 0.8 mg/dl and less than 1 mg/dl, actual S_cr proved less biased (−4.5 ml/min [p=0.038] and −5.5 ml/min [pc_r. Conclusion An unbiased C‐G Cl_cr can be calculated using actual body weight in underweight patients and ideal body weight in patients of normal weight. Using ABW_0.4 for overweight, obese, and morbidly obese patients appears to be the least biased and most accurate method for calculating their C‐G Cl_cr. Rounding S_cr in patients with low S_cr did not improve accuracy or bias of the Cl_cr calculations.