AI-generated research brief — verify at source
Polymyxin B Haemoadsorption Shows High Probability of Mortality Benefit in Endotoxic Septic Shock
In a phase 3 Bayesian randomised trial, polymyxin B haemoadsorption was associated with lower 28-day mortality compared with standard care alone in patients with endotoxic septic shock (39% vs 45%; posterior probability of benefit 95.3%, APACHE-II adjusted OR 0.67, 95% credible interval 0.39–1.08). At 90 days, the evidence for benefit strengthened further, with a posterior probability of 99.4% and an adjusted odds ratio suggesting roughly half the odds of death (OR 0.54, 95% CrI 0.32–0.87).
What Was Studied
The trial investigated whether removing circulating endotoxin from the bloodstream using polymyxin B haemoadsorption reduces 28-day all-cause mortality in a biologically defined subgroup of septic shock patients with high endotoxin activity and multiorgan failure. This question is important because endotoxin-driven organ dysfunction represents one of the more lethal and mechanistically distinct presentations of sepsis, yet no intervention targeting endotoxin directly has previously achieved phase 3 efficacy evidence.
How It Was Studied
The Tigris trial was an open-label, randomised, controlled phase 3 study conducted across 19 hospitals in the United States, enrolling adults aged 18 years or older who had septic shock requiring vasopressor support, evidence of multiple organ dysfunction, and endotoxin activity measurements between 0.60 and 0.89 units — a range indicating high endotoxin burden. A total of 157 patients were enrolled between September 2019 and April 2025, randomly assigned in a 2:1 ratio to polymyxin B haemoadsorption plus standard of care (n=106) or standard of care alone (n=51). The intervention consisted of two haemodialysis sessions at a blood flow rate of 80–120 mL/min lasting 90–120 minutes each, administered 22 hours apart. The trial used a Bayesian analytical framework, incorporating a prior derived from a pre-specified subgroup of the earlier EUPHRATES trial, with 90-day mortality as the key secondary outcome.
What Was Observed
- 28-day mortality was 39% in the polymyxin B group versus 45% in the control group. The Bayesian analysis yielded a posterior probability of benefit of 95.3%, with an APACHE-II adjusted odds ratio suggesting approximately one-third lower odds of death, though the credible interval remains wide and includes the null (OR 0.67, 95% CrI 0.39–1.08).
- 90-day mortality showed a stronger signal, with a posterior probability of benefit of 99.4% and an adjusted odds ratio indicating roughly 46% lower odds of death in the polymyxin B group compared with controls (OR 0.54, 95% CrI 0.32–0.87), with a credible interval that no longer crosses 1.0.
- Serious adverse events occurred in 30% of evaluated polymyxin B patients versus 22% of controls, a difference of 8 percentage points that was not statistically distinguishable from chance (95% CI −22 to 6). Only two serious adverse events in the treatment group were judged treatment-related.
Why This Matters
This trial provides the first phase 3 Bayesian evidence that endotoxin-targeted haemoadsorption may reduce mortality in a carefully phenotyped subgroup of septic shock, lending support to the concept that endotoxin activity can serve as a clinically actionable patient-selection biomarker. The findings advance the case for precision medicine approaches in critical care, where broad-spectrum interventions have repeatedly failed while biologically targeted strategies remain under-explored. The sustained and strengthened benefit observed at 90 days, beyond the primary 28-day window, adds further scientific weight to the direction of effect.
How to Read This Result
While the Bayesian posterior probabilities are encouraging, this is a moderately sized, open-label trial whose primary endpoint credible interval crosses 1.0, and whose prior was derived from a subgroup of an earlier trial, making the magnitude of benefit sensitive to that prior specification and warranting cautious interpretation pending independent replication.
Limitations
Several constraints temper the conclusions. The open-label design precluded blinding of personnel to treatment allocation, introducing potential performance bias. The sample size of 157 patients is relatively modest for a phase 3 mortality trial, limiting statistical precision. The Bayesian prior was constructed from a subgroup of the EUPHRATES trial, meaning results are sensitive to that prior choice and may not be fully reproducible under a neutral prior. Enrolment was restricted to US hospitals, which may limit generalisability to healthcare systems with different patient populations or standard-of-care practices. Finally, the 28-day primary outcome credible interval includes 1.0, meaning the conventional threshold for excluding a null effect was not met on the primary endpoint.