Abstract
BACKGROUND: We aimed to assess the prognostic performance of different indexes of oxygenation, respiratory mechanics and ventilation intensity in predicting 90-day mortality, and to estimate their independent associations, in a "real-world" observational cohort of acute respiratory distress syndrome (ARDS) patients on intensive care unit (ICU) mortality. METHODS: This is a secondary analysis of LUNG SAFE (Large observational study to UNderstand the Global impact of Severe Acute respiratory FailurE), an international prospective cohort study of patients with severe respiratory failure involving 459 ICUs from 50 countries. We evaluated the prognostic performance of oxygenation (arterial oxygen tension (P (aO(2)) )/inspiratory oxygen fraction (F (IO(2)) )), respiratory mechanics (normalised elastance) and ventilation intensity (plateau pressure (P (plat)), driving pressure (DP), 4DP+respiratory rate (RR) and mechanical power (MP)) measured on day 1 of controlled mechanical ventilation in ARDS patients, with respect to ICU mortality within 90 days of admission. For each parameter, associations with mortality were assessed using logistic regression models, estimating effect sizes (odds ratios with 95% confidence interval), model discrimination (area under the receiver operating characteristic curve), calibration and overall predictive accuracy. RESULTS: Among 2813 early ARDS patients, 516 (18.3%) met the inclusion criteria: mean±sd age 60±16 years, 61% male. Normalised elastance, P (plat), DP and 4DP+RR were significantly associated with mortality, with adjusted ORs ranging from 1.02 (95% CI 1.01-1.03) for 4DP+RR to 1.48 (95% CI 1.15-1.95) for normalised elastance. These parameters showed higher predictive accuracy for mortality compared with P (aO(2)) /F (IO(2)) and MP. MP showed a U-shaped relationship with mortality but was not significantly associated with it. Its predictive accuracy decreased after accounting for positive end-expiratory pressure (PEEP) and dynamic resistance, with PEEP also demonstrating a U-shaped association with mortality. CONCLUSIONS: Normalised elastance, DP and 4DP+RR, measured at day 1 of ARDS, were the best predictors of ICU mortality, and outperformed oxygenation and MP. DP showed the best balance between predictive accuracy and clinical simplicity. These results reinforce the importance of focusing on DP and 4DP+RR as key metrics to guide lung-protective strategies and ARDS severity classification.