Renal Function in Older Hospital Patients is More Accurately Estimated Using the Cockcroft-Gault Formula Than the Modification Diet in Renal Disease Formula

Objectives: To compare the accuracy of the two most popular creatinine clearance (CrCl) estimation formulae (Cockcroft-Gault (CG) and Modification Diet in Renal Disease (MDRD)) in older hospitalized patients.

Design: Prospective, cross-sectional, observational study.

Setting: Two hospital geriatric wards.

Participants: Consecutive patients aged 70 and older with an indwelling urinary catheter for the purpose of care.

Measurements: CrCl was determined according to three methods: measured CrCl from plasma and urine creatinine and 24-hour urine volume, CG (CG-CrCl), and MDRD (MDRD-CrCl). Results were expressed as median and interquartile range (IQR). Moderate and severe renal impairment were defined as a CrCl between 30.0 and 59.9 mL/min and less than 30.0 mL/min, respectively.

Results: One hundred twenty-one patients were included (46% male). Mean age was 86.1+/-6.7 (range 72-100). Median measured CrCl was 43.8 mL/min (IQR 33.6-61.1 mL/min), CG-CrCl was 40.9 mL/min (IQR 31.0-52.6 mL/min), and MDRD-CrCl was 61.3 mL/min (IQR 49.4-77.0 mL/min). The biases of CG-CrCl and MDRD were -3.5+/-22.5 and 20.1+/-28.2, respectively (P<.001). Misclassification of renal impairment (absent/moderate/severe) occurred in 33% of patients according to CG-CrCl, and concordance was mild to moderate (kappa=0.50). Misclassification occurred in 50% of patients according to MDRD-CrCl, and concordance was poor (kappa=0.33). Bias was significantly related to bed confinement for both formulae and to plasma creatinine for MDRD.

Conclusion: In elderly hospitalized patients, CG slightly underestimates CrCl, and MDRD strongly overestimates it. CG gave a better prediction of measured CrCl than MDRD.