In hemodialysis (HD) patients the glycated hemoglobin (Hb(A1c)) level may underestimate glycemic control. The aim of this study is to estimate accurate glycemic control in type 2 diabetic patients on HD. Type 2 diabetes patients (N = 87) who had been receiving maintenance HD for at least one year were enrolled. Hb(A1c) and the percentage of glycated albumin relative to total the serum albumin (%GA) were measured in blood samples and the factors that affected the %GA/Hb(A1c) ratio were examined. There were significant and positive correlations between the plasma glucose and either the Hb(A1c) levels (r = 0.539, P < 0.01) or the %GA level (r = 0.520, P < 0.01). No relationship between the serum albumin levels and %GA levels was observed. A weekly dose of erythropoietin (EPO) was positively correlated with the ratio of %GA/Hb(A1c) and hematocrit (Ht) correlated negatively. There was no significant correlation between the %GA/Hb(A1c) level and the EPO dose in patients with Ht > or = 30%, although a significant correlation was found between those parameters in the Ht < 30% group. The mean of the %GA/Hb(A1c) ratios in patients with Ht > or = 30%, with Ht < 30% and treated with EPO < 100 IU/kg/week, and with Ht < 30% and treated with EPO > or = 100 IU/kg/week were 3.41, 3.56 and 4.13, respectively. In HD patients, accurate glycemic control may be estimated as: Hb(A1c) x 1.14 if Ht > or = 30%; Hb(A1c) x 1.19 if Ht < 30% and treated with low dosages of EPO; and Hb(A1c) x 1.38 if Ht < 30% and treated with high dosages of EPO.