Meduri et al highlight the nature of the ongoing debate over the value of corticosteroids in acute lung injury/ARDS. The study by Confalonieri and colleagues1was not included in our review because the study population had severe pneumonia, not ARDS. The retrospective study by Annane et al2was a secondary analysis of a randomized controlled trial of corticosteroids in septic shock; benefit was noted in the subgroup of patients with sepsis-associated ARDS who failed to respond to a corticotropin stimulation test. The benefits of corticosteroids in this group likely derive from their beneficial effects in the overall population of nonresponders in this study rather than from an effect specific to ARDS; no statistical test of interaction between corticosteroid therapy and ARDS was reported in the article. In addition, while post hoc subgroup analysis may be useful for hypothesis generation, generalizing the findings to patient treatment can be perilous.3 The 2007 study by Meduri et al was not published at the time of our review; however, this trial4has significant limitations as well. For one, the majority of patients randomized to placebo who remained on mechanical ventilation at day 9 of the study were crossed over to open-label methylprednisolone, making outcomes analysis after that point (such as mortality and ventilator-free days) very difficult to interpret. In our review,5we focused on the largest and most rigorous trial on this issue: the prospective, randomized controlled trial performed by the ARDS Network, which demonstrated no mortality benefit to corticosteroids.6 We agree that the size of the subgroup of patients randomized after day 14 in this study is small, and that conclusions drawn from this subgroup, albeit a prespecified one, should be tempered by this consideration; however, most of the baseline imbalances cited by the letter were not statistically significant.