Pulmonary fibroproliferation (PFP) is directly or indirectly the leading cause of death in patients with late ARDS. We previously reported our experience using intravenous corticosteroids (IVC) in 8 patients with late ARDS and now have expanded our observation to a total of 25 patients with severe fibroproliferation (mean lung injury score [LIS] 3) and progressive respiratory failure (RF). Thirteen patients had open-lung biopsy before treatment. Patients were started on IVC treatment (IVCT) an average of 15 +/- 7.5 days into mechanical ventilation (MV). Significant physiologic improvement (SPI) to IVCT was defined as a reduction in LIS of greater than 1 point or an increase in PaO2:FIO2 ratio of greater than 100. We observed three patterns of response: rapid responders (RR) had an SPI by day 7 (n = 15); delayed responders (DR) had an SPI by day 14 (n = 6); nonresponders (NR) were without SPI by day 14 (n = 4). Overall the following significant mean changes were seen within 7 days of IVCT: LIS from 3 to 2 (p = 0.001), PaO2:FIO2 from 162 to 234 (p = 0.0004), PEEP from 11 to 6.8 cm H2O (p = 0.001), chest radiograph score from 3.8 to 3.0 (p = 0.009), and VE from 16 to 13.6 L/min (p = 0.01). Development of pneumonia was related to the pattern of response. Surveillance bronchoscopy was effective in identifying pneumonia in eight afebrile patients. Nineteen of 25 (76 percent) patients survived the ICU admission. Comparisons were made between survivors (S) and nonsurvivors (NS) and among the three groups of responders. At the time ARDS developed, no physiologic or demographic variable could discriminate between S and NS. At the time of IVCT, only liver failure was more frequent in nonsurvivors (p = 0.035). Histologic findings at open-lung biopsy and pattern of physiologic response clearly predicted outcome. The presence of preserved alveolar architecture (p = 0.045), myxoid type fibrosis (p = 0.045), coexistent intraluminal bronchiolar fibrosis (p = 0.0045), and lack of arteriolar subintimal fibroproliferation (p = 0.045) separated S from NS. ICU survival rate was 86 percent in responders and 25 percent in nonresponders (p = 0.03). Only one death resulted from refractory respiratory failure.