End-tidal carbon dioxide and arterial to end-tidal carbon dioxide gradient are associated with mortality in patients with neurological injuries

Pre-hospital end-tidal carbon dioxide (EtCO2) monitoring and arterial to end-tidal carbon dioxide gradient (Pa-EtCO2) have been associated with mortality in patients with traumatic brain injury. Our study aimed to analyze the association between alveolar EtCO2 or Pa-EtCO2 and mortality in patients admitted in intensive care unit (ICU) with neurological injuries. In our retrospective analysis from using large de-identified ICU databases (MIMIC-III and -IV and eICU databases), we included 2872 ICU patients with neurological injuries, identified according to the International Classification of Diseases (ICD-9 and -10), who underwent EtCO2 monitoring. We performed logistic regression and extended Cox regression to assess the association between mortality and candidate covariates, including EtCO2 and Pa-EtCO2. In-hospital mortality was 26% (n = 747). In univariate analysis, both the Pa-EtCO2 gradient and EtCO2 levels during the first 24 h were significantly associated with mortality (for a 1 mmHg increase: OR = 1.03 [CI95 1.016-1.035] and OR = 0.94 [CI95 0.923-0.953]; p < 0.001). The association remained significant in multivariate analysis. The time-varying evolution of EtCO2 was independently associated with mortality (for a 1 mmHg increase: HR = 0.976 [CI95 0.966-0.985]; p < 0.001). The time-varying Pa-EtCO2 gradient was associated with mortality only in univariate analysis. In neurocritical patients, lower EtCO2 levels at admission and throughout the ICU stay were independently associated with mortality and should be avoided.