Classification and effectiveness of different oxygenation goals in mechanically ventilated critically ill patients: network meta-analysis of randomised controlled trials.

Background The optimal oxygenation in mechanically ventilated critically ill patients remains unclear. Methods Randomised controlled trials investigating different oxygenation goal-directed mechanical ventilation in critically ill adult patients were eligible. The trinary classification classified oxygenation goals into conservative (PaO2=55–90 mmHg), moderate (PaO2=90–150 mmHg), and liberal (PaO2>150 mmHg). The quadruple classification further divided the conservative goal from the trinary classification into far-conservative (PaO2=55–70 mmHg) and conservative (PaO2=70–90 mmHg) goals. The primary outcome was 30-day mortality. The secondary outcomes included intensive care unit, hospital, and 90-day mortalities. The effectiveness was estimated by the relative risk (RR) and 95% credible interval (CrI) using network meta-analysis and visualised using SUCRA scores and survival curves. Results We identified eight eligible studies involving 2532 patients. There were no differences between conservative and moderate goals (RR, 1.08; 95% CrI, 0.85 to 1.36; moderate quality), between moderate and liberal goals (RR, 0.83; 95% CrI, 0.61 to 1.10, low quality), and between conservative and liberal goals (RR, 0.89; 95% CrI, 0.61 to 1.30, low quality) based on the trinary classification. No differences in secondary outcomes among the different goals. The results were consistent between the trinary and quadruple classifications. The SUCRA scores and survival curves suggested that the moderate goal in the trinary and quadruple classifications and the conservative goal in the quadruple classification may be superior to the liberal and far-conservative goals. Conclusions Different oxygenation goals do not lead to different mortalities in mechanically ventilated critically ill patients. The potential superiority of maintaining PaO2 in the range of 70–150 mmHg remains to be validated.