Association of liberal versus restrictive oxygen saturation targets with mortality in mechanically ventilated stroke patients: a multicenter retrospective propensity-matched analysis.
Ghafarian A, Saad J, et al. • Frontiers in neurology • 2026
Restrictive oxygen saturation targets (SpO2 92-96%) during the first 48 h of mechanical ventilation were associated with higher short-term mortality than liberal targets (SpO2 ≥97%) in stroke patients, driven primarily by the AIS group, and these findings warrant prospective evaluation in stroke-focused randomized trials.
Key Findings
Results
Restrictive oxygen saturation targets in the first 48 h of mechanical ventilation were associated with higher all-cause mortality at 1 and 3 months in the matched combined stroke cohort.
Multicenter retrospective cohort from the TriNetX database (2006-2026) of 17,207 mechanically ventilated stroke patients: 2,366 restrictive (SpO2 92-96%) and 14,841 liberal (SpO2 ≥97%) before matching.
After propensity score matching, 2,350 patients remained in each group.
1-month mortality: HR 1.13, 95% CI 1.03-1.23; p = 0.008.
3-month mortality: HR 1.11, 95% CI 1.02-1.20; p = 0.014.
Results
In the acute ischemic stroke (AIS) subgroup, restrictive oxygenation was associated with a larger increase in mortality at 1 and 3 months.
1-month mortality: HR 1.26, 95% CI 1.11-1.43.
3-month mortality: HR 1.22, 95% CI 1.08-1.37.
Propensity matching was performed separately within the prespecified AIS subgroup.
Results
No statistically significant mortality difference between restrictive and liberal oxygen targets was observed in intracerebral hemorrhage (ICH).
ICH was a prespecified subgroup with its own propensity score matching.
The overall mortality signal was described as driven primarily by the AIS group.
Results
Restrictive oxygenation was associated with a lower 30-day tracheostomy rate, with no difference in ventilator-associated pneumonia (VAP).
30-day tracheostomy: 8.5% vs. 13.1%; RR 0.653, p < 0.001.
No difference in VAP between groups.
Conclusions
The authors conclude that early oxygenation strategy may have subtype-specific implications in stroke and needs prospective testing.
Primary outcome was all-cause mortality at 30 and 90 days; secondary outcomes were 30-day tracheostomy and VAP.
Analyses used Kaplan-Meier methods and Cox proportional hazards models with significance set at p = 0.05.
The design is retrospective and observational, so the findings show association rather than causation.
What This Means
This study looked at electronic health records from many hospitals to ask whether the oxygen level doctors aim for in stroke patients on breathing machines affects survival. Patients whose blood oxygen saturation stayed in a restricted range (92-96%) during the first 48 hours were compared with those kept at a higher, more liberal level (97% or above). After statistically matching about 2,350 patients in each group for age, other illnesses, and stroke severity, the restricted-oxygen group had somewhat higher death rates at 1 and 3 months.
The difference was clearest in patients with ischemic stroke (caused by a blocked blood vessel), where the increase in risk was larger. In patients with bleeding strokes (intracerebral hemorrhage), no clear difference was seen. The restricted-oxygen group also had fewer tracheostomies within 30 days (8.5% vs. 13.1%), while rates of ventilator-associated pneumonia were similar.
This research suggests that lower oxygen targets, which are increasingly used in general intensive care, may not be harmless in ventilated stroke patients, especially ischemic stroke. However, because this was a retrospective analysis of existing records, it cannot prove that the oxygen target caused the difference in survival. The authors call for randomized trials focused on stroke patients to find out the best oxygen targets.
Ghafarian A, Saad J, Wajahath M, Ghafarian M, Saasouh W, Saleh K, et al.. (2026). Association of liberal versus restrictive oxygen saturation targets with mortality in mechanically ventilated stroke patients: a multicenter retrospective propensity-matched analysis.. Frontiers in neurology. https://doi.org/10.3389/fneur.2026.1840183