A randomised controlled trial of plasma filtration in severe paediatric sepsis.

OBJECTIVE: To determine whether plasma filtration improves 28-day survival in infants and children with severe sepsis.

DESIGN: A multicentre randomised controlled trial.

SETTING: Paediatric intensive care units in teaching hospitals.

PATIENTS: Forty-eight infants and children with severe sepsis.

INTERVENTIONS: Patients were randomly assigned to receive plasma filtration (n = 25) or standard therapy (n = 23) for the treatment of septic shock. The primary outcome measure was 28-day survival. Secondary outcome measures included the number of failed organ systems on Day 7, a requirement for extracorporeal membrane oxygenation (ECMO), and the modified Glasgow outcome score (MGOS) at 6 months (where 1 is normal and 6 is dead).

RESULTS: The trial was stopped early due to poor recruitment. Patients in the plasma filtration group had higher initial disease severity as measured by serum lactate level, inotrope score and MGOS. Ten (40%) children died in the plasma filtration group and 4 (17%) died in the control group. With intention-to-treat analysis and adjustment for lactate level, ventilation index, inotrope score and MGOS at admission using logistic regression, the odds ratio for death with plasma filtration was 1.20 (95% CI, 0.23-6.20; P = 0.82). The median number of failing organ systems at 7 days was 2 (interquartile range [IQR], 1-4) in the plasma filtration group versus 2 (IQR, 1-3) in the control group. Two children in the plasma filtration group required ECMO for 2.5 and 123 hours, and one child in the control group required ECMO for 45 hours. The median MGOS at 6 months was 4 (IQR, 2-6) in the plasma filtration group and 2 (IQR, 1-4) in the control group.

CONCLUSIONS: Our study did not recruit enough patients to test the hypothesis that addition of plasma filtration to our standard care protocol reduces 28-day mortality in children with severe sepsis. However, mortality in the treatment and control groups was not significantly different after adjustment for severity of illness at the time of randomisation.