Comparison between the three most popular formulae to estimate renal function, in subjects 75 years of age or older

The aim of the study was to compare the accuracy of three formulae that estimate creatinine clearance (CL(CR)), in elderly hospitalized patients: the Cockcroft-Gault (CG) formula and the Modification of Diet in Renal Disease formulae with 4 and 6 variables (MDRD4 and MDRD6).

A prospective, cross-sectional, observational study was conducted in four hospital geriatric wards. Consecutive patients admitted to the wards who were aged ≥75 years and had an indwelling urinary catheter for the purpose of care were eligible for enrolment. CL(CR) was determined via four methods: measurement of CL(CR) from plasma and urine creatinine plus 24-h urine volume; the CG formula; and the MDRD4 and MDRD6 formulae. Moderate and severe renal impairments were defined as a CL(CR) of 30.0-59.9 and <30.0 mL/min, respectively.

A total of 157 patients were included. Their mean age (±SD) was 86.5 ± 6.1 (range 75-105) years and 46.5 % were male. The median values and interquartile ranges (IQRs) (in mL/min) were 44.0 (IQR 32.1-64.5) for measured CL(CR), 42.1 (IQR 31.3-56.3) for CG-estimated CL(CR), 64.3 (IQR 49.8-81.7) for MDRD4-estimated CL(CR) and 49.3 (IQR 37.4-63.4) for MDRD6-estimated CL(CR) (respectively, p < 0.05, p < 0.001 and p = 0.44 compared with measured CL(CR)). Biases (±SD) for CG, MDRD4 and MDRD6 CL(CR) estimates were -3.6 (±22.2), 19.3 (±26.4) and 2.4 (±22.5) mL/min, respectively. When estimated CL(CR) values were assessed against the measured value, it was found that misclassification of renal impairment (absent/moderate/severe) occurred in 41 % of patients when using the CG, in 40 % when using the MDRD6, and in 45 % when using the MDRD4. The 30 % accuracies of the three formulae were 63 % for CG, 37 % for MDRD4 and 59 % for MDRD6.

In elderly hospitalized patients, CG and MDRD6 gave better predictions for measured CL(CR) than MDRD4, with no significant difference between them.