OBJECTIVE: The role of protective ventilation in acute lung injury (ALI) and acute respiratory distress syndrome (ARDS) is controversial. Evidence was sought from published randomised trials for a consistent treatment effect of protective ventilation and any covariate modification. DESIGN: Meta-analysis of protective ventilation trials in ALI/ARDS and meta-regression of covariates on treatment effect (log odds ratio), with respect to 28-day mortality. Heterogeneity impact on the meta-analysis was assessed by the H statistic (substantial impact, >1.5) and graphical analysis. Five trials with a total of 1,202 patients were considered. MEASUREMENTS AND RESULTS: Average 28-day mortality was 0.40 in the treatment group (protective ventilation, n=605) vs. 0.46 in the control group (control ventilation, n=597). The treatment effect (odds ratio) was: fixed-effects, 0.71 (95% CI 0.56-0.91, p=0.006; heterogeneity, p=0.06) and random effects: 0.80 (95% CI 0.49-1.31, p=0.37). Heterogeneity impact (H statistic=1.50) was adjudged as modest. The treatment effect was significant and (a) favoured protective ventilation for a tidal volume less than 7.7 ml/kg predicted (treatment group) and a mean plateau pressure of 30 cmH(2)O or higher (control group) but was not influenced by plateau pressure 21-30 cmH(2)O (treatment group) and (b) depended upon plateau pressure difference greater than 5-7 cmH(2)O between protective ventilation and standard ventilation. CONCLUSIONS: Overall treatment effect estimate favoured protective ventilation but did not achieve statistical significance. Protective ventilation depended upon threshold levels of tidal volume, plateau pressure, and plateau pressure difference.