Mercat 2008.

Methods	Randomized controlled trial. Time period of study: September 2002 through December 2005.
Participants	767 participants, aged > 18 (37 centres). Included: ARDS. Excluded: known pregnancy, participation in another trial within 30 days, increased intracranial pressure, sickle cell disease, severe chronic respiratory disease requiring oxygen therapy or home MV, actual body weight exceeding 1 kg/cm of height, severe burns, severe chronic liver disease, bone marrow transplant or chemotherapy‐induced neutropenia, pneumothorax, expected duration of MV ≤ 48 hours, decision to withhold life‐sustaining treatment.
Interventions	Control (382) and intervention (385): MV: ventilator mode (both groups): volume‐assist control, VT: 6 mL/kg PBW, plateau‐pressure limit ≤ 30 cm H2O, respiratory rate (breaths/min) ≤ 35 adjusted for a pH between 7.30 and 7.45, recruiting manoeuvres: allowed but not recommended. Target ranges for oxygenation: PaO2 between 55 and 80 mm Hg; SpO2 between 88% and 95%. PEEP: Control: Total PEEP between 5 and 9 cm H2O. Intervention: PEEP level to achieve plateau pressures between 28 and 30 cm H2O. Use of rescue therapies1 (both groups) when the oxygenation goal was not met despite FIO2 ≥ 0.8.
Outcomes	Primary: mortality at day 28. Secondary: mortality at day 60, mortality before hospital discharge censored on day 60, VFD, days‐free without organ failure and barotrauma between day 1 and day 28.
Notes	1. Rescue therapies: prone ventilation, inhaled NO, almitrine bismesylate. Discontinued during the 18th interim analysis because of absence of 10% absolute reduction in mortality between groups.
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Authors performed random allocation in permuted blocks stratified by centre.
Allocation concealment (selection bias)	Low risk	Assignment was performed by centralized‐interactive telephone system.
Blinding of participants and personnel (performance bias) All outcomes	Low risk	Incomplete blinding (blinding of participants but not personnel), but outcomes not influenced
Blinding of outcome assessment (detection bias) All outcomes	Low risk	The main analyses were conducted in a blind fashion.
Incomplete outcome data (attrition bias) All outcomes	Low risk	One participant (control) was excluded because family withdrew consent after randomization. One participant (intervention) was lost to follow‐up after discharge on day 29 and was included in the analysis on the basis of the intention‐to‐treat principle.
Selective reporting (reporting bias)	Low risk	Published reports included all expected outcomes.
Other bias	Low risk	Review author believed the study to be free of other sources of bias.