Our study has a number of limitations, including the heterogeneity in patient populations, in inclusion and exclusion criteria, and outcome measures. Nevertheless, to increase comparability, we limited our search to studies that used the criteria for ALI/ARDS published by the American-European Consensus Conference.11 Even though these have been criticized,40,91 they are widely used and provide some homogeneity by which to compare studies. The present definitions of ALI/ARDS are considered by some authors91as arbitrary, and others92have called for new definitions to be developed that include prognostic measures and reduce patient heterogeneity. Moreover, in a comparison of autopsy findings with clinical definitions, including the American-European Consensus Conference definition, ARDS appeared to be underrecognized by clinicians, and the specificity of the definitions highly variable.93 A second potential limitation was that the absolute mortality rate may be affected by the type of study, with randomized clinical trials often having strict exclusion and inclusion criteria resulting in the selecting of populations of patients that may not always reflect the outcome of general ICU patients. However, to avoid this kind of “selection” bias, we separately analyzed studies with and without exclusion criteria, and noted a similar trend for both groups. Moreover, even if absolute mortality rates may be affected by this form of selection bias, trends over time may remain valid. A third limitation was that survival was assessed at different time periods (eg, at ICU discharge, hospital discharge, or at fixed time periods of 28 days or 30 days). The hospital stay following discharge from the ICU contributes 3 to 15% of the mortality in the few studies,35,59 that have reported both; this could be related to factors occurring after the ICU stay, which may be independent of the ARDS diagnosis and management. Unfortunately, it was impossible to compare disease severity in the patients because of the high variability in the severity scores used in the studies, which included acute physiology and chronic health evaluation II or III, simplified acute physiology score II, and the lung injury score, among others.