BACKGROUND: Elevated fasting plasma glucose (FBG) is linked to overall mortality in chronic hemodialysis (HD) patients. It is unknown if different levels of FBG are associated with a poor out-come in patients with diabetes mellitus (DM) and those without at the initiation of HD. The aims of this study are to investigate (1) if increased levels of FBG predict higher mortality in maintenance HD patients. (2) Is there a different influence of FBG levels on overall mortality between those with or without pre-existing DM before HD. METHODS: Patients who survived over the first 90 days on HD in the Taiwan Renal Registry Database from 1997 to 2005 were analyzed. All patients were followed until December 31, 2008. The association between FBG levels and overall mortality was analyzed using three adjusted Cox proportional-hazard regression models. DM patients were defined as those carrying a diagnosis of DM either as cause of primary renal disease or as a co-morbidity, at the initiation of HD. Non-DM patients were defined as those who did not have a pre-existing diagnosis of DM at the initiation of HD. RESULTS: A total of 46,596 HD patients including 26,107 non-DM and 20,489 DM patients were followed for 5.2 (3.4-7.8) years. The incidence density rate of overall mortality was 12.5 per 100 person-years in DM patients and 6.2 per 100 person-years in non-DM patients. In pre-existing DM patients, an FBG less than 80 mg/dL or more than 240 mg/dL was associated with an increased mortality risk (<80 mg/dL, P = 0.007 and >240 mg/dL, P = 0.03). In patients without pre-existing DM, an FBG less than 80 mg/dL or more than 110mg/dL was associated with an increased mortality risk (<80 mg/dL, P < 0.001 and >110 mg/dL, P < 0.001). CONCLUSIONS: A minimal increase in FBG is associated with an increased overall mortality in patients who did not have pre-existing DM before maintenance HD. In contrast, a much higher FBG threshold is linked to increased mortality in those with pre-existing DM.