Predicting successful extubation in very preterm infants: A case–control study on key respiratory and clinical predictors

This is the authors' abstract. We don't add key points for this paper.

Background: Respiratory failure requiring mechanical ventilation (MV) is a significant challenge for preterm infants, particularly those born before 29 weeks of gestation. Prolonged MV is associated with adverse respiratory and neurodevelopmental outcomes. Extubation failure (EF) remains common, with ˃67% of preterm infants failing initial attempts. This study aimed to identify predictors of EF and develop a practical extubation readiness model. Methods: This case–control study included preterm infants born before 33 weeks of gestation. Infants were categorized into EF ( n = 27) and extubation success ( n = 81) groups, matched for gestational age (GA), birth weight (BW), sex, and admission month. Pre- and postextubation variables, including ventilator settings, blood gas measurements, and comorbidities were analyzed. Logistic regression and Kaplan–Meier analyses identified predictors and characterized EF timing. Results: Lower preextubation fraction of inspired oxygen (FiO 2 ) (0.27 vs. 0.32, P < 0.05) and mean airway pressure (MAP) (8.05 vs. 8.44 cmH 2 O, P = 0.052) were associated with successful extubation. EF was associated with higher mortality (26% vs. 7%, P = 0.01) and severe broncho-pulmonary dysplasia (BPD) (22% vs. 4%, P < 0.002). A three-variable model (surfactant use, FiO 2 , and MAP) achieved an area under the receiver operating characteristic curve of 0.77. Most reintubations (59%) occurred within 24 h, with higher BW and GA associated with delayed reintubation. Conclusion: Lower FiO 2 and MAP before extubation were predictive of successful extubation, whereas EF was associated with higher mortality and severe BPD. This three-variable model offers practical utility in assessing extubation readiness. Further research should standardize the extubation criteria and explore the integration of advanced technologies to support decision-making.