In a clinic-based, cross-sectional study of 320 type 2 diabetic patients, we staged the level of diabetic nephropathy (normoalbuminuric, microalbuminuric and macroalbuminuric stage) and estimated GFR based on serum creatinine and cystatin C (CysC). Serum creatinine and CysC levels were 0.91+/-0.21 mg/dL and 0.87+/-0.26 mg/L, respectively. Correlation coefficients between CysC-GFR and each of the creatinine-based GFR measurements (MDRD-GFR, Cockcroft-Gault-GFR, and CLcr) were 0.589, 0.569, and 0.479 (p<0.001). Serum CysC was significantly lower in normoalbuminurics (0.83+/-0.22) than in microalbuminurics and macroalbuminurics (0.94+/-0.33 and 1.05+/-0.28; p=0.004 and p<0.001). Of the estimations of GFR, significant differences among the groups were found on CysC-GFR and CLcr. CysC-GFR (mL/min) was statistically lower in macroalbuminurics (79.5+/-30.5) than in normoalbuminurics (104.3+/-30.9, p=0.01). The logistic regression analyses showed that retinopathy, A1C, CysC, diabetic duration, and CysC-GFR were indicators to predict the development of microalbuminuria. Serum CysC seems to be more accurate serum marker than serum creatinine in evaluating a prognostic stage of type 2 diabetic nephropathy. Our study suggests that, in Korean type 2 diabetic patients, CysC-based GFR might be more valuable than creatinine-based GFR in the prediction of the microalbuminuric stage.