Introduction: Polymyxin B hemoperfusion (PMX-HP) is an adjuvant therapy for sepsis or septic shock that removes circulating endotoxin. However, PMX-HP has seldom achieved expectations in randomized trials targeting nonspecific overall sepsis patients. If used in an optimal population, PMX-HP may be beneficial. This study aimed to identify the optimal population for PMX-HP in patients with septic shock.
Methods: We used a prospective nationwide cohort targeting consecutive adult patients with severe sepsis (Sepsis-2) in 59 intensive care units in Japan. Associations between PMX-HP therapy and in-hospital mortality were assessed using multivariable Cox proportional hazard regression models. To identify best targets for PMX-HP, we developed a non-linear restricted cubic spline model including two-way interaction term (treatment × Acute Physiology and Chronic Health Evaluation [APACHE] II score/Sequential Organ Failure Assessment [SOFA] score) and three-way interaction term (treatment × age × each score).
Results: The final study cohort comprised 741 sepsis patients (92 received PMX-HP, 625 did not). Cox proportional hazards regression model adjusted for the covariates suggested no association between PMX-HP therapy and improved mortality overall. Effect modification of PMX-HP by APACHE II score was statistically significant (P for interaction = 0.189) but non-significant for SOFA score (P for interaction = 0.413). Three-way interaction analysis revealed suppressed risk hazard in the PMX-HP group versus control group only in septic shock patients with high age and in the most severe subset of both scores, whereas increased risk hazard was observed in those with high age but in the lower severity subset of both scores.
Conclusions: Our results suggested that although PMX-HP did not reduce in-hospital mortality among overall septic shock patients, it may benefit a limited population with high age and higher disease severity.