Comparison of oxygenation targets in critically ill adults: a systematic review and network meta-analysis of randomized controlled trials

Le résumé fourni par la source

Acute respiratory failure affects an estimated 13-20 million people worldwide each year, and acute lung injury carries a mortality rate of approximately 35%-45%1 To prevent hypoxemic organ injury, oxygen is typically titrated according to pulse oximetry readings.2 However, excess oxygen can generate reactive oxygen species, alter hemodynamics, trigger inflammation,2 and exacerbate lung injury during mechanical ventilation.3 Therefore, conservative oxygen therapy, focused on avoiding hyperoxia while ensuring adequate tissue oxygenation, has been proposed.2 Guidelines from the British Thoracic Society and the French intensive care and emergency medicine societies recommend peripheral oxygen saturation (SpO2) 94%-98% in adults with acute hypoxemic respiratory failure.4,5 However, evidence remains inconclusive and the optimal target is still uncertain. Several randomized controlled trials (RCTs) have compared conservative versus liberal oxygenation in critically ill adults. Although one early trial reported lower ICU mortality with conservative oxygen,6 subsequent large RCTs in heterogeneous ICU populations have not demonstrated superiority of conservative strategies; rather, mortality point estimates have generally favored liberal oxygen or suggested modest harm with conservative oxygen.7-12 Uncertainty about the optimal target stems from the wide range of targets tested: Most studies compare just 2 strategies, and conventional pairwise meta-analyses pool diverse targets into broad “low” versus “high” categories, which prevents head-to-head comparisons of several distinct oxygen ranges. One network meta-analysis13 grouped 8 heterogeneous RCTs (n = 2532) into 4 oxygenation categories. Since then, new RCTs have more than doubled the evidence base, underscoring the need for an updated systematic review and meta-analysis.

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