The effect of high-flux hemodialysis membranes on patient survival has not been unequivocally determined. In this prospective, randomized clinical trial, we enrolled 738 incident hemodialysis patients, stratified them by serum albumin < or = 4 and >4 g/dl, and assigned them to either low-flux or high-flux membranes. We followed patients for 3 to 7.5 yr. Kaplan-Meier survival analysis showed no significant difference between high-flux and low-flux membranes, and a Cox proportional hazards model concurred. Patients with serum albumin < or = 4 g/dl had significantly higher survival rates in the high-flux group compared with the low-flux group (P = 0.032). In addition, a secondary analysis revealed that high-flux membranes may significantly improve survival of patients with diabetes. Among those with serum albumin < or = 4 g/dl, slightly different effects among patients with and without diabetes suggested a potential interaction between diabetes status and low serum albumin in the reduction of risk conferred by high-flux membranes. In summary, we did not detect a significant survival benefit with either high-flux or low-flux membranes in the population overall, but the use of high-flux membranes conferred a significant survival benefit among patients with serum albumin < or = 4 g/dl. The apparent survival benefit among patients who have diabetes and are treated with high-flux membranes requires confirmation given the post hoc nature of our analysis.