The optimal target for glycemic control has not been established in diabetic dialysis patients.To address this question, the national database of a large dialysis organization (DaVita) was analyzed via time-dependent survival models with repeated measures.Of 82,933 patients undergoing maintenance hemodialysis (MHD) in DaVita outpatient clinics over 3 years (July 2001 through June 2004), 23,618 diabetic MHD patients had A1C measurements at least once. Unadjusted survival analyses indicated paradoxically lower death hazard ratios (HRs) with higher A1C values. However, after adjusting for potential confounders (demographics, dialysis vintage, dose, comorbidity, anemia, and surrogates of malnutrition and inflammation), higher A1C values were incrementally associated with higher death risks. Compared with A1C in the 5-6% range, the adjusted all-cause and cardiovascular death HRs for A1C > or = 10% were 1.41 (95% CI 1.25-1.60) and 1.73 (1.44-2.08), respectively (P < 0.001). The incremental increase in death risk for rising A1C values was monotonic and robust in nonanemic patients (hemoglobin > 11.0 g/dl). In subgroup analyses, the association between A1C > 6% and increased death risk was more prominent among younger patients, those who had undergone dialysis for > 2 years, and those with higher protein intake (> 1 g x kg(-1) x day(-1)), blood hemoglobin (> 11 g/dl), or serum ferritin values (> 500 ng/ml).In diabetic MHD patients, the apparently counterintuitive association between poor glycemic control and greater survival is explained by such confounders as malnutrition and anemia. All things equal, higher A1C is associated with increased death risk. Lower A1C levels not related to malnutrition or anemia appear to be associated with improved survival in MHD patients.