Clearances of several solutes (urea, creatinine, phosphate, urates, beta(2)-microglobulin [beta(2)-M]) were measured during venovenous continuous renal replacement therapy (CRRT) at various ultrafiltration (Q(UF); 0 to 2 L/h) and dialysate flow rates (Q(D); 0 to 2.5 L/h). Preset Multiflow-60 and Multiflow-100 hollow-fiber dialysers (M-60 and M-100; Hospal-Gambro, St-Leonard, Canada) were compared (five patients for each type). First, we evaluated the impact of predilution on convective clearances: a progressive decrease in patient clearances, similar for both filters, was observed, reaching a maximum of 15%, 18%, and 19% for urea, urates, and creatinine, respectively, with predilution at a Q(UF) of 2 L/h. Second, we compared convective and diffusive clearances. Because effluent to plasma ratio (E/P) remained at 1 for small solutes (urea, creatinine, phosphate, urates) during convection, clearances were equal to the effluent rate for both dialyzers. However, we observed greater diffusive clearances for small molecules with M-100 than with M-60 at a Q(D) of 1.5 to 2.5 L/h, the difference being more significant as molecular weight increased. For beta(2)-M, diffusive clearance was very low and rapidly reached a plateau of 8 and 12 mL/min for M-60 and M-100, respectively, at a Q(D) greater than 1.5 L/h. Convective clearances for beta(2)-M increased nonlinearly up to 20 +/- 2 mL/min at a progressively greater Q(UF) (from 0.5 to 2 L/h) for both M-60 and M-100. This nonlinear increase was attributed to an increase of almost 40% in E/P for beta(2)-M from a Q(UF) of 0.5 to 2 L/h. Third, the interaction between convection and diffusion was assessed by measuring solute clearances at a fixed Q(UF) (1 and 2 L/h) and variable Q(D) (0.5 to 2.5 L/h). For small molecules, no significant interaction between convection and diffusion was noticed with M-100, whereas only a small interaction was noticed with M-60. However, for beta(2)-M, the addition of diffusion (Q(D), 0.5 to 2.5 L/h) did not result in any significant increase in total clearances over convective clearances for M-60 and M-100. This observation suggests that the diffusive clearances for beta(2)-M observed with M-60 and M-100 at a Q(UF) of 0 L/h and at various Q(D) probably occurs by convective fluxes across the membrane. These results demonstrate that convection is more efficient than diffusion in removing mixed-molecular-weight solutes during CRRT.