Objective:To assess the effect of plasmafiltration (PF) on biochemical markers of inflammation, cytokines, organ dysfunction, and 14-day mortality in human sepsis.Design:Multicenter, prospective, randomized, controlled clinical trial.Setting:Seven university-affiliated intensive care units.Patients:Thirty patients (22 adults, eight children) with new (<24 hrs) clinical evidence of infection and sepsis syndrome were enrolled. Fourteen of 30 (nine adults, five children) were randomized to PF.Interventions:All patients received protocol-driven supportive intensive care, and those randomized to PF received continuous plasma exchange for 34 hrs using a hollow-fiber plasma filter.Measurements and Main Results:Illness severity and risk of death were calculated with the Pediatric Risk of Mortality (children) and the Acute Physiology and Chronic Health Evaluation II (adults) scales. Plasma samples (0, 6, 24, and 48 hrs) were assayed for acute-phase proteins (albumin, globulin, C-reactive protein, α1-antitrypsin, haptoglobin), inflammatory mediators (complement fragment C3, thromboxane B2), and cytokines (interleukin-6, granulocyte colony-stimulating factor, leukemia inhibitory factor). Sieving coefficients were estimated from filtrate concentrations at 3 hrs. The two groups were matched for incidence of septic shock (13 of 14 vs. 11 of 16), refractory shock (three of 14 vs. six of 16), bacteremia (six of 14 vs. five of 16), severity of illness, and calculated risk of death (0.68 vs. 0.64). There was no difference in mortality. Eight of 14 PF patients (57%) and eight of 16 controls (50%) survived for 14 days (p= .73, Fisher's exact test). Multiple logistic regression revealed age (odds ratio, 16.4:1; 95% confidence interval, 2.12-∞) and shock (10.6:1; 1.32-∞) as significant predictors of death; plasmafiltration was associated with a nonsignificant reduction in the risk of death (odds ratio, 1.78:1; 95% confidence interval, 0.20-18.1). The mean (SD) number of organs failing in the first 7 days in the PF group was 2.57 (0.94) vs. 2.94 (0.85) in controls (p= .37, Mann-Whitney U test). Both groups had similarly elevated plasma concentrations of all inflammatory mediators except complement fragment C3at study entry. Leukemia inhibitory factor was detectable in four patients only. PF did not influence mean concentrations of interleukin-6, granulocyte colony-stimulating factor, thromboxane B2, total white cell count, neutrophil count, or platelet count, but it was associated with significant reductions of α1-antitrypsin, haptoglobin, C-reactive protein, and complement fragment C3in the first 6 hrs (p< .05). The sieving coefficients for all inflammatory mediators approached unity.Conclusions:PF caused a significant attenuation of the acute-phase response in sepsis. There was no significant difference in mortality, but there was a trend toward fewer organs failing in the PF group that suggests that this procedure might be beneficial.