OBJECTIVES:The relationships among body mass index (BMI), diabetes, age, and all-cause mortality in hemodialysis patients remain unclear. We examined BMI-stratified relationships between diabetes, age, and the risk of long-term mortality in incident hemodialysis patients. METHODS:This is a cohort study. Data were obtained from the national dialysis registry in Japan 2007 that included 35,415 incident hemodialysis patients and 6,061 patients aged ≥20 years with BMI data. Patients were divided into 6 categories according to baseline BMI (low: <18.5, normal: 18.5-25, Obesity: ≥25) and the presence or absence of diabetes. The primary outcome was all-cause mortality during a 5-year follow-up. Hazard ratios were estimated using Cox's model for the relationships among diabetes, BMI categories, and all-cause mortality, and adjusted for potential confounders. Patients with a normal BMI and non-diabetic were the reference category. We also examined the effects of age on these relationships. RESULTS:A total of 6,061 patients, including 3,239 with diabetes, were enrolled. During the follow-up, 31.0% and 30.7% of all and diabetic patients, respectively, died. Cox's regression analysis showed that low BMI, but not obesity, was independently associated with an increased risk of all-cause death in patients with and without diabetes. When patients were divided into 2 groups-younger and older than 60 years-the risk of mortality in both groups was increased in low BMI with diabetes. CONCLUSIONS:Among Japanese incident hemodialysis patients, low BMI increases the risk of all-cause mortality. The markedly high mortality rate in diabetic patients with low BMI regardless of age warrants attention.