BACKGROUND: We previously showed that nutritional protein concentrations were predictive of outcome, whereas variables reflecting body composition and dialysis dose were not, in a 30-month prospective follow-up of 1,610 hemodialysis patients. Information on dialysis membrane and erythropoietin use had to be evaluated in an additional follow-up. METHODS: A subset of 650 patients from the initial cohort of 1,610 was analyzed for survival in a 2-year extension of follow-up. Detailed data were collected: demographics; cause of renal failure; time on dialysis therapy; type of membrane; erythropoietin treatment; body mass index (BMI); predialysis albumin, prealbumin, and bicarbonate levels; and outcome. Normalized protein catabolic rate (nPCR), dialysis adequacy, and lean body mass were computed from predialysis and postdialysis urea and creatinine values. RESULTS: Patient characteristics were age of 61 +/- 16 years, 58% men, BMI of 22.7 +/- 4.4 kg/m2 , time on dialysis therapy of 102 +/- 73 months, and 8.8% had diabetes. Dialysis parameters were duration of 247 +/- 31 minutes, Kt/V of 1.4 +/- 0.3, and nPCR of 1.2 +/- 0.3 g/kg/d. Albumin level was 3.73 +/- 0.53 g/dL (37.3 +/- 5.3 g/L), and prealbumin level was 31 +/- 8 mg/dL. The survival rate was 78.7% after 2 years. Survival was influenced by age, presence of diabetes, use of high-flux membrane, and serum albumin level, but not other variables, including Kt/V and prealbumin level. Two-year variations in values for urea, creatinine, and weight were predictive of survival in univariate, but not multivariate, analyses. CONCLUSION: In patients on dialysis therapy for a long period, better survival was observed when high-flux dialysis membranes were used.