Discrepancies between estimated and geometrically derived mechanical power in invasively ventilated patients – insights from a randomized clinical trial

Le résumé fourni par la source

BACKGROUND: Mechanical power (MP) quantifies the energy transferred from the ventilator to the respiratory system and is increasingly linked to outcomes in invasively ventilated critically ill patients. Limiting MP may support lung-protective ventilation by integrating key determinants of ventilator-induced lung injury. While geometric MP, derived from pressure-volume loops, is considered the reference standard, it is not readily available at the bedside. Therefore, simplified equations are commonly used, although their agreement remains uncertain. This study evaluates the agreement of previously proposed equations to estimate inspiratory MP compared with a geometric reference. RESULTS: 494 paired values of geometric and estimated MP were collected, in 38 passively ventilated patients during closed-loop ventilation and during conventional ventilation. The median geometric MP was 17.0 [12.3-21.5] J/min. Correlation between geometric and estimated MP ranged from strong to very strong. Agreement between geometric and estimated MP was good, though bias and limits of agreement varied across the different equations. ROC analysis showed excellent discrimination for most equations across all tested cutoffs, with discrimination gradually declining at higher MP levels. CONCLUSION: In passively ventilated critically ill patients, estimated MP showed strong to very strong correlations with the geometric reference. ROC analysis confirmed excellent discrimination, but Bland-Altman analysis revealed considerable variability, suggesting potential limitations in using these equations interchangeably in clinical or research settings. STUDY REGISTRATION: INTELLiPOWER was registered at clinicaltrials.gov (study identifier NCT04827927).

Ce résumé expose les affirmations des auteurs. BNTIC ne l’interprète pas comme une validation indépendante des résultats.