BACKGROUND: The American-European Consensus Conference definition for the acute respiratory distress syndrome (ARDS) has never been validated. OBJECTIVE: To compare clinical criteria for ARDS with autopsy findings. DESIGN: Independent comparison of autopsy findings with clinical characteristics retrospectively abstracted from medical records. SETTING: Tertiary medical-surgical intensive care unit. PARTICIPANTS: 382 patients who underwent clinical autopsy. MEASUREMENTS: Sensitivity, specificity, and likelihood ratios for clinical criteria were calculated in 3 cohorts by using diffuse alveolar damage at autopsy as the reference standard. The 3 cohorts were 1) all patients, 2) patients with any risk factor for ARDS, and 3) patients who were separated according to their pulmonary or extrapulmonary risk factors. RESULTS: 127 patients (33%) met the clinical criteria, and 112 (29%) had diffuse alveolar damage. In all patients, the sensitivity of the clinical definition was 75% (95% CI, 66% to 82%) and the specificity was 84% (CI, 79% to 88%). In 284 patients with risk factors, the sensitivity was 76% (CI, 67% to 83%) and the specificity was 75% (CI, 68% to 81%). Compared with patients with pulmonary risk factors, patients with extrapulmonary risk factors had significantly higher sensitivity (61% vs. 85%; P = 0.009) and the specificity did not statistically significantly differ (69% vs. 78%; P > 0.2). LIMITATIONS: Only patients who died and underwent autopsy could be included in this study, so these results may not apply to less severe cases of ARDS. CONCLUSIONS: In a series of autopsy patients, the accuracy of the American-European Consensus Conference definition of ARDS was only moderate. The definition was more accurate for patients with extrapulmonary risk factors than for patients with pulmonary risk factors.