Oxygen saturation and mortality for acute pulmonary embolism: multinational retrospective cohort study

Abstract

Background: Among patients with acute pulmonary embolism (PE), we aimed to evaluate for an association between oxygen saturation (SatO(2)) and 30-day post-PE mortality. Methods: Between January, 2001 and April, 2025, we conducted a retrospective cohort study of 38,739 non-hypotensive patients with acute PE enrolled from 18 countries in the Registro Informatizado de la Enfermedad Tromboemb & oacute;lica (RIETE) registry. Outcomes included 30-day all-cause and PE-specific mortality. We used hierarchical logistic regression to assess for an association between baseline SatO(2) and outcomes. We aimed to identify optimal SatO(2) cut-off points that maximized sensitivity and specificity in relation to mortality from receiver operating characteristic (ROC) curves. RIETE is registered with ClinicalTrials.gov, NCT02832245. Findings Baseline SatO(2) was significantly associated with mortality, where lower SatO(2) had worse outcomes. Using a reference SatO(2) of >96%, patients in the lower SatO(2) strata had higher rates of all-cause death (odds ratio [OR] 0.96 for SatO(2) 94-96%; 1.23 for SatO(2) 90-93%; and 1.81 for SatO(2) <90%) and PE-related mortality (OR 1.68 for SatO(2) 94-96%; 2.04 for SatO(2) 90-93%; and 3.95 for SatO(2) <90%). The optimal cut-off point to identify patients at low-risk for short-term mortality was SatO(2) of 96% (sensitivity of 86.1% and negative predictive value of 96.7%); while SatO(2) of 90% (specificity of 77.2% and positive predictive value of 7.9%) identified patients at increased risk for mortality. Interpretation Among non-hypotensive patients diagnosed with PE, we found an inverse association between baseline SatO(2) and odds of all-cause and PE-associated mortality over the subsequent 30 days. A SatO(2) cutoff value of 96% for the highest SatO(2) group while breathing room air might help better identify patients with acute PE at low-risk for mortality. Future studies should assess the contribution of SatO2 in addition to established prognosticators to predicting the risk of death.