OBJECTIVE To assess, in diabetic nephropathy, the accuracy of a method that estimates glomerular function with age, body weight, and serum creatinine as parameters. METHODS Glomerular filtration rate (GFR) was measured 57 times in 20 subjects with insulin-dependent diabetes mellitus and nephropathy with a single injection of 51Cr-EDTA. At the same time, the estimated creatinine clearance (ml/min) was calculated with the Cockroft-Gault formula (140 - age [yr]) x body wt [kg] x K/serum creatinine [mumol/L]) K = 1.23 for men, 1.05 for women These values were then corrected for body surface area (1.73 m2). RESULTS For GFR measurements less than 100 ml.min-1.1.73 m-2 (n = 41), there was a strong positive correlation with the estimated creatinine clearance corrected for body surface area (r = 0.94, P less than 0.0001). The slope of this regression line did not differ significantly from identity or the y-intercept from zero. On average, the Cockroft-Gault formula (corrected for body surface area) underestimated the GFR by only 3.1 ml.min-1.1.73 m-2 (9.7 SD). CONCLUSIONS This formula, corrected for body surface area, gives accurate estimates of GFR when GFR less than 100 ml.min-1.1.73 m-2. This formula could be used with an acceptable degree of confidence when repeated isotope assessments of renal function in diabetic nephropathy are impracticable.