Besides the age at start of dialysis and the presence of diabetes mellitus, serum albumin has been shown to be a significant predictor of survival in prevalent dialysis patients. However, this was not evaluated in incipient dialysis patients. The initial biochemical variables were retrieved for a large dialysis population (N = 1,982) who were started on chronic dialysis in Okinawa, Japan from 1971 to 1990. Biochemical data that included blood urea nitrogen, serum creatinine, serum electrolytes (sodium, potassium, calcium, and phosphate), total cholesterol, triglyceride, total protein, serum albumin, and hematocrit obtained just before the first dialysis session were available for 1,491 (75.2%) patients. Six hundred sixty-four (43.2%) patients died by the end of 1993. Cox proportional analysis adjusted for sex, age, year of start of dialysis, presence of diabetes mellitus, and the laboratory variables was performed to examine the significance of the initial biochemical data on survival. The risk ratio was 0.577 (P = 0.0025) in serum albumin, 1.291 (P = 0.0138) in serum potassium, and 0.966 (P = 0.0202) in serum sodium. The study results support the notion that nutritional status is important for survival in chronic dialysis patients. Causes of hypoalbuminemia, hyperkalemia, and hyponatremia should be evaluated carefully at initiation of dialysis.