Predictive performance of bedside physiological parameters for the failure of prophylactic non-invasive ventilation after extubation

Abstract Background Extubation failure is associated with increased mortality. Whether bedside physiological variables can identify patients at risk of reintubation during prophylactic NIV after extubation is unknown. Methods In this post hoc analysis of a multicenter physiological study, 60 hypoxemic adults at high risk of extubation failure underwent two 1-hour NIV sessions after planned extubation, with oronasal and full-face masks in randomized crossover order. We assessed the magnitude of airway occlusion pressure (Pocc), the esophageal pressure swing (ΔPes), SpO₂/FiO₂, respiratory rate, tidal volume, dynamic lung elastance (EL,dyn), ROX index [(SpO₂/FiO₂)/respiratory rate] and the pressure-oxygenation index (POX, SpO₂/FiO₂ divided by Pocc). Prophylactic NIV failure was reintubation for respiratory failure within 7 days. Discrimination was assessed by ROC analysis. Results Fourteen patients (23%) experienced NIV failure. Pocc yielded areas under the curve (AUCs) of 0.83 (95% CI, 0.71–0.95) with the oronasal mask and 0.83 (95% CI, 0.72–0.95) with the full-face mask, showing greater discriminative performance than tidal volume (AUCs, 0.55 and 0.53), respiratory rate (0.54 and 0.38), SpO₂/FiO₂ (0.51 and 0.50), and the ROX index (0.54 and 0.38; all p < 0.05). By contrast, the discriminative performance of Pocc did not differ from that of EL,dyn (AUCs, 0.72 and 0.73; p = 0.08 and 0.11), POX (AUCs, 0.82 and 0.81; p = 0.68 and 0.29), or |ΔPes| (AUCs, 0.83 and 0.82; p = 1.00 and 0.64). Conclusions During prophylactic NIV, Pocc discriminated NIV failure better than conventional bedside variables, with no detectable difference from directly measured inspiratory effort. External validation is required.