Association of liberal versus restrictive oxygen saturation targets with mortality in mechanically ventilated stroke patients: a multicenter retrospective propensity-matched analysis

Background. Optimal oxygen saturation targets in mechanically ventilated stroke patients remain uncertain. While restrictive oxygen strategies are increasingly adopted in general critical care, their impact in acute ischemic stroke (AIS) and intracerebral hemorrhage (ICH) are unclear. We evaluated the association between restrictive versus liberal oxygen saturation targets during the first 48 h of mechanical ventilation and short-term mortality in stroke patients.

Methods. We conducted a multicenter retrospective cohort study using the TriNetX database (2006–2026). Adult patients (≥18 years) with a stroke diagnosis who required invasive mechanical ventilation and had documented peripheral oxygen saturation (SpO₂) values within the first 48 h of ventilation were included. Patients were stratified according to early oxygen saturation targets: restrictive (SpO₂ 92–96%) versus liberal (SpO₂ ≥ 97%). Propensity score matching (PSM) by demographics, comorbidities, and stroke severity was performed separately within the combined stroke cohort and prespecified AIS and ICH subgroups to balance demographic and clinical covariates. The primary outcome was all-cause mortality at 30 and 90 days. Secondary outcomes included 30-day tracheostomy and ventilator-associated pneumonia. Time-to-event analyses were performed using Kaplan–Meier methods and Cox proportional hazards models. p-value was set at 0.05.

Results. Among 17,207 mechanically ventilated stroke patients, 2,366 received restrictive and 14,841 liberal oxygen saturation targets prior to matching. After PSM, 2,350 patients remained in each group. Restrictive oxygenation during the first 48 h was associated with higher mortality at 1 month (HR 1.13, 95% CI 1.03–1.23; p = 0.008) and 3 months (HR 1.11, 95% CI 1.02–1.20; p = 0.014). In AIS, restrictive oxygenation was similarly associated with increased mortality at 1 month (HR 1.26, 95% CI 1.11–1.43) and 3 months (HR 1.22, 95% CI 1.08–1.37). No statistically significant mortality difference was observed in ICH. Restrictive oxygenation was associated with lower 30-day tracheostomy rates (8.5% vs. 13.1%; RR 0.653, p < 0.001), with no difference in VAP.

Conclusion. Restrictive oxygen saturation targets during the first 48 h of mechanical ventilation may be associated with higher short-term mortality in stroke, driven primarily by the AIS group. These findings suggest that early oxygenation strategy may have subtype-specific implications and warrant prospective evaluation in stroke focused randomized trials.