Bayesian analysis of the early cold stored platelet transfusion following severe injury randomized clinical trial

The Early Cold Stored Platelet Transfusion Following Severe Trauma Injury Trial evaluated the safety and efficacy of cold stored platelet (CSP) transfusion in injured patients at risk of hemorrhagic shock. It was a phase II, multicenter, randomized clinical trial conducted from June 2022 to September 2023 at five US level 1 trauma centers. A total of 200 adult patients at risk of hemorrhagic shock (102 CSP, 98 standard care (SC)) were included. Median age was 34±13 years, 85% were males, and injury severity was similar between groups. The original frequentist analysis did not demonstrate a statistically significant difference in 24-hour mortality between CSP (5.9%) transfusion and SC (10.2%, p=0.28).

To provide further insight into these findings, a Bayesian logistic regression outcome model with an uninformative prior was performed. To validate the findings further, a sensitivity analysis was performed in addition to analysis using a beta-binomial Bayesian model.

The Bayesian logistic regression model yielded an 89.1% posterior probability that CSP transfusion reduced 24-hour mortality (ß₁=-0.668, 95% highest posterior density interval: -1.7 to 0.48). This posterior distribution indicates it is 8.1 times more likely that CSP transfusion improves 24-hour mortality than not. Sensitivity analyses across a spectrum of priors consistently produced posterior probabilities >81%. The beta-binomial model demonstrated a similar 86.2% probability that CSP transfusion reduced 24-hour mortality.

These findings suggest a high likelihood of mortality benefit with early CSP transfusion in injured patients at risk of hemorrhagic shock and illustrate the added interpretive value of Bayesian analysis in trauma trials with low event rates.

Level II-therapeutic/care management.