Are prediction equations for glomerular filtration rate useful for the long-term monitoring of type 2 diabetic patients?

Background. The aim of this study was to compare the accuracy of prediction equations [modification of diet in renal disease (MDRD), simplified MDRD, Cockcroft-Gault (CG), reciprocal of creatinine and creatinine clearance] in a cohort of patients with type 2 diabetes. Methods. A total of 525 glomerular filtration rates (GFRs) using I-125-iothalamate were carried out over 10 years in 87 type 2 diabetic patients. Accuracy was evaluated at three levels of renal function according to the baseline values obtained with the isotopic method: hyperfiltration (GFR: > 140 ml/min/1.73 m(2); 140 isotopic determinations in 27 patients), normal renal function (GFR: 140-90 ml/min/1.73 m(2); 294 isotopic determinations in 47 patients) and chronic kidney disease (CKD) stages 2-3 (GFR: 30-89 ml/min/1.73 m(2); 87 isotopic determinations in 13 patients). The annual slope for GFR (change in GFR expressed as ml/min/year) was considered to ascertain the variability in the equations compared with the isotopic method during follow-up. Student's t-test was used to determine the existence of significant differences between prediction equations and the isotopic method (P < 0.05 with Bonferroni adjusted for five contrast tests). Results. In the subgroup of patients with hyperfiltration, a GFR slope calculated with I-125-iothalamate -4.8 +/- 4.7 ml/min/year was obtained. GFR slope in patients with normal renal function was -3.0 +/- 2.3 ml/min/year. In both situations, all equations presented a significant underestimation compared with the isotopic GFR (P < 0.01; P < 0.05). In the subgroup of CKD stages 2-3, the slope for GFR with I-125-iothalamate was -1.4 +/- 1.8 ml/min/year. The best prediction equation compared with the isotopic method proved to be MDRD with a slope for GFR of -1.4 +/- 1.3 ml/min/year (P: NS) compared with the CG formula -1.0 +/- 0.9 ml/min/year (P: NS). Creatinine clearance presented the greatest variability in estimation (P < 0.001). Conclusions. In the normal renal function and hyperfiltration groups, none of the prediction equations demonstrated acceptable accuracy owing to excessive underestimation of renal function. In CKD stages 2-3, with mean serum creatinine >= 133 mu mol/l (1.5 mg/dl), the MDRD equation can be used to estimate GFR during the monitoring and follow-up of patients with type 2 diabetes receiving insulin, anti-diabetic drugs or both.