We compared cystatin C, creatinine, and the Cockroft formula for assessment of early renal failure, defined as a (51)Cr-EDTA clearance < 80 mL/min, in 89 diabetic patients with various degrees of renal impairment (glomerular filtration rate [GFR], 11.4 to 196.5 mL/min). The relationships between cystatin C, creatinine, and (51)Cr-EDTA clearance were linearized by plotting the reciprocals of the values, and correlation coefficients were determined. Sensitivity and specificity for the diagnosis of early renal failure were calculated from receiver operating characteristic (ROC) curves. Over the whole population, cystatin C was as well correlated with GFR (r =.74) as was creatinine (r =.67) or the Cockroft formula (r =.88). Moreover, its diagnostic accuracy was comparable to that of the 2 other parameters. Its sensitivity (86.8%) was better than that of creatinine (77.4%) for screening GFR < 80 mL/min, although the Cockroft formula was more sensitive (96.2%). The study of albuminuric diabetics (n = 63) led to similar conclusions, except for a poor sensitivity of cystatin C. In the 36 patients whose plasma creatinine was < 1 mg/dL, 10 (27.7%) had GFR < 80 mL/min. The correlation of creatinine with GFR, its diagnostic accuracy, and sensitivity were significantly lower than those of cystatin C. In this population of patients with normal creatinine levels, the correlation coefficient of cystatin C, its sensitivity, and its diagnostic accuracy were comparable to those of the Cockroft formula. A moderate reduction in GFR may be present in diabetic patients with low creatinine levels. Although Cockroft formula remains the most reliable and the less expensive tool for the evaluation of renal function, cystatin C is a more reliable criterion for screening and assessment than creatinine and represents a useful alternative to the Cockcroft-Gault formula.