The type of dialysis membrane used for routine therapy has been recently shown to correlate with the survival of chronic hemodialysis patients. We examined whether this effect of dialysis membrane could be explained by differences in dialyzer removal of middle molecules using data from the 1991 Case Mix Adequacy Study of the United States Renal Data System. The sample analyzed included patients who had been treated by hemodialysis for 1 year or more, who were dialyzed with the 19 most commonly used dialyzers in 1991, and for whom delivered urea Kt/V could be calculated from predialysis and postdialysis blood urea nitrogen concentrations. Vitamin B12 (1,355 daltons) was used as a marker for middle molecules, and the clearance of vitamin B12 was estimated based on in vitro data. After adjustments for case mix, comorbidities, and urea Kt/V, the relative risk of mortality for a 10% higher calculated total cleared volume of vitamin B12 was 0.953 (P < 0.0001 v 1.000). Similar results were obtained when middle molecule removal was adjusted for body size. We conclude that both small and middle molecule removal indices appear to be independently associated with the risk of mortality in chronic hemodialysis patients. Differences in mortality when using different types of dialysis membrane may be explained by differences in middle molecule removal.