The Cockcroft-Gault (CG) formula and the modification of diet in renal disease (MDRD) equation are commonly used to estimate glomerular filtration rate (GFR), but their validity at extreme body weight is questionable. This may be significant for diabetic patients. In 122 diabetic patients with renal damage, we compared both estimates to isotopically determined GFR by correlation studies and a Bland and Altman procedure before and after categorizing the patients according to body mass index (BMI). Over the whole population, the CG overestimated GFR (CG, 51.4 +/- 23.1 mL/[min . 1.73 m2]; isotopic GFR, 44.6 +/- 21.1 mL/[min . 1.73 m2], P < .0001). The MDRD (45.2 +/- 17.9; NS vs isotopic GFR) did not overestimate GFR, but it underestimated high GFR as revealed by the Bland and Altman procedure (r = -0.26, P < .005). The CG underestimated GFR in patients with normal BMI (-14%, P < .01) and overestimated it in overweight (15%, P < .005) and obese patients (55%, P < .0001); the result and the error of the estimation were correlated with BMI. This bias did not affect the MDRD. The use of ideal instead of measured body weight improved the CG prediction, but underestimated GFR. As the BMI of the 87 type 2 diabetic subjects was higher, the CG overestimated their mean GFR by 18% (P < .001), whereas the MDRD did not. There were 25% fewer patients with delayed referral using the MDRD than with the CG. Because the estimate of GFR by the CG is proportional to body weight, it is not suited for obese diabetic patients. Although it is less easy to calculate, the MDRD is not affected by weight, and its use would avoid delay in referral to nephrologists.