Analysis* | No. of trials | No. of deaths, n/N | Risk ratio (95% CI) | I2 value, % |
---|---|---|---|---|
Primary | ||||
Trials mandating protective ventilation† | 6 | 363/1016 | 0.74 (0.59–0.95) | 29 |
Sensitivity | ||||
Included all trials‡ | 10 | 797/1869 | 0.86 (0.73–1.00) | 42 |
Assumed patients lost to follow-up lived | 6 | 363/1020 | 0.74 (0.59–0.95) | 28 |
Assumed patients lost to follow-up died | 6 | 366/1020 | 0.74 (0.59–0.94) | 26 |
Excluded trial in which allocation was not concealed35 | 5 | 352/994 | 0.73 (0.55–0.98) | 43 |
Excluded trial with pediatric population37 | 5 | 355/914 | 0.73 (0.56–0.96) | 42 |
Included trial that used moderate tidal volume (< 10 mL/kg)36 | 7 | 438/1152 | 0.77 (0.65–0.91) | 16 |
Fixed-effects model | 6 | 363/1016 | 0.74 (0.63–0.87) | 29 |
Note: CI = confidence interval.
↵* Random-effects models were used for all analyses except in the final sensitivity analysis.
↵† Tidal volume < 8 mL/kg of predicted or actual body weight.
↵‡ For the 2 trials that enrolled some patients without ARDS,16,39 we included only patients whose condition met the authors’ definition of ARDS; when the analysis was redone to include all patients in these trials, the risk ratio changed minimally (0.87, 95% CI 0.74–1.02; I2 = 48%).