Effect of a Lower Vs. Higher Positive End-Expiratory Pressure Strategy on Clinically Relevant Outcomes in ICU Patients Without Acute Respiratory Distress Syndrome: Bayesian Re-analysis of the REstricted Vs. Liberal Positive End-Expiratory Pressure in

OBJECTIVE: The "REstricted vs. Liberal positive end-expiratory pressure in patients without Acute Respiratory Distress Syndrome (ARDS)" (RELAx) trial compared lower vs. higher positive end-expiratory pressures (PEEP) in invasively ventilated critically ill patients without ARDS, concluding non-inferiority of lower PEEP in frequentist analysis. This study aimed to perform a Bayesian re-analysis of RELAx to assess probabilities of clinically meaningful differences between lower and higher PEEP. DESIGN: A post hoc Bayesian analysis of RELAx data. SETTING: RELAx was a non-inferiority multicenter trial conducted between October 2017 and March 2019 in eight ICUs. PATIENTS: The trial included 980 ICU patients expected to require invasive mechanical ventilation greater than or equal to 24 hours for reasons other than ARDS. All patients included in the original analysis entered the Bayesian re-analysis. INTERVENTIONS: Participants were randomized to the lowest possible PEEP between 0 and 5 cm H 2 O or to a higher PEEP of 8 cm H 2 O. MEASUREMENTS AND MAIN RESULTS: The primary outcome was ventilator-free days at day 28 (VFD-28). Secondary outcomes were 28-day mortality and ventilation duration. The odds ratio for higher VFD-28 with lower PEEP was 1.08 (95% credible intervals, 0.87-1.35), with consistent estimates across priors and with a probability of superiority ranging from 75% to 78%. For 28-day mortality and duration of ventilation, the probability of benefit of the lower-PEEP strategy ranged from 72% to 89%, and from 11% to 28%, respectively. In patients admitted for other reasons than cardiac arrest or intubated for other reasons than respiratory failure, probabilities of benefit with lower PEEP exceeded 90%. CONCLUSIONS: Although the probability of benefit in the overall population was modest, the analysis suggested a higher probability of benefit in selected subgroups, particularly patients admitted for other reasons than cardiac arrest or intubated for other reasons than respiratory failure. These findings highlight potential heterogeneity of treatment effect and support further investigations.