We examined the predictive value of various clinical variables in assessing survival in chronic hemodialysis patients (N = 1,243, 524 females, 719 males) who were under treatment with hemodialysis as of January 1991 in Okinawa, Japan and who were followed up until April 1992. Basal clinical data such as sex, starting date of dialysis, primary renal disease, blood pressure, blood chemistry values, and dialysis prescription data obtained just prior to dialysis were registered at the start of the study. As of the end of April 1992, 104 had died, 16 were transplant recipients, and five had been transferred. Those who died had significantly lower levels of total protein, serum albumin, total cholesterol, triglyceride, BUN, serum creatinine, body weight, body height, diastolic blood pressure, and duration of hemodialysis than those who survived. Older patients and those with diabetes mellitus had a poorer prognosis. A forward stepwise logistic procedure by SAS was used to determine the predictive value of the above clinical variables. With the addition of laboratory variables, the predictive value of diabetes was lost, as the diabetic patients had low serum levels of albumin and creatinine. The standardized coefficient was -0.380 (P = 0.0001) at age of entry, 0.316 (P = 0.0001) for serum albumin, 0.280 (P = 0.0001) for serum creatinine, 0.138 (P = 0.043) for body mass index (BMI), and -0.139 (P = 0.016) for male sex. The prescribed dialysis dose (M2 hr per week) was significantly correlated with serum creatinine (r = 0.48, P = 0.0001), serum albumin (r = 0.135, P = 0.0001) and BMI (r = 0.275, P = 0.0001).(ABSTRACT TRUNCATED AT 250 WORDS)