Real-World Outcomes and Prognostic Factors of Polymyxin B Hemoperfusion in Severe Sepsis and Septic Shock: A Seven-Year Single-Center Cohort Study from Taiwan

Background: Severe sepsis and septic shock remain major contributors to ICU mortality. Polymyxin B hemoperfusion (PMX-HP) has been widely adopted as adjunctive therapy in Asian ICUs for endotoxemia, but its real-world effectiveness and prognostic factors remain uncertain, especially in high Gram-negative settings. Methods: This retrospective cohort study included 64 adult patients with severe sepsis or septic shock who received at least one session of PMX-HP in a 25-bed tertiary medical ICU in Taiwan between July 2013 and December 2019. Demographic, clinical, microbiological, and treatment data were extracted. The primary outcome was 28-day mortality. Prognostic factors were analyzed using logistic regression. Results: The mean age was 66.1 ± 12.3 years; 67.2% were male. Pneumonia (29.7%) and intra-abdominal infection (18.8%) were the most common sources of sepsis, with E. coli and K. pneumoniae as leading pathogens. Median APACHE II score at ICU admission was 26 (IQR 21–32), and 79.7% received two PMX-HP sessions. The 28-day mortality rate was 46.9%, with ICU and hospital mortality both 53.1%. Non-survivors were older, had higher APACHE II scores, and more frequent use of continuous renal replacement therapy (CRRT). Positive changes in vasoactive-inotropic score (VIS) after PMX-HP were also more common among non-survivors. Multivariate analysis identified advanced age, higher APACHE II score, and CRRT requirement as independent predictors of mortality. Conclusions: In this real-world Asian ICU cohort, PMX-HP was used mainly for severe cases with a high disease burden and Gram-negative predominance. Despite its frequent use, overall mortality remained high. Prognosis was primarily determined by underlying disease severity, organ dysfunction (especially renal failure), and persistent hemodynamic instability. In this high-severity cohort, mortality appeared to be primarily driven by baseline organ dysfunction and persistent hemodynamic instability; PMX-HP session number or sequencing showed no association with survival. Given the absence of a contemporaneous non-PMX-HP control group, mortality observations in this cohort cannot be causally attributed to PMX-HP and should be interpreted with caution as hypothesis-generating rather than definitive evidence of efficacy. Further multicenter studies are needed to clarify the optimal role of PMX-HP in modern sepsis management.