The timely institution of therapy as well as the appropriate design of clinical trials of new therapies depend on an accurate understanding of the time course for evolution of acute hypoxemic respiratory failure. Methods: We analyzed the data collected at 12 hour intervals from 331ECMO-eligible patients in 32 centers (Crit Care Med 24:323-9, 1996) as part of the PCCSG study of Acute Hypoxemic Respiratory Failure (Chest 108:789-97, 1995). Data collection began when patients received at least 50% FiO2 and end expiratory pressure of 6 cm H2O or greater. The time courses to the primary positive outcome of interest (successful tracheal extubation) and primary negative outcomes (death in the PICU or the need for extracorporeal membrane oxygenation) were determined. Measures of oxygenation and ventilation as well as ventilator pressures were compared as predictors of outcomes. Results: Thirty six per cent of patients died and 11% received ECMO. The worst value for oxygenation index (OI)was the strongest predictor of death(P<.0001). An OI of 40 or above predicted a 48% probability of death and was selected as a third negative outcome (24% of patients). 42% of patients reached one or more negative outcome - 50% of these outcomes occurred within 24 hours of enrollment, 75% within 96 hours. Positive outcomes also occurred rapidly after enrollment: 25% of patients were extubated within 72 hours, 50% within 7 days. Conclusions: Acute respiratory failure in children improves or evolves to critical stages rapidly in the majority of children. Therapeutic trials must enroll and randomize patients within a narrow time window. Critical phases in the pathophysiology of lung injury probably occur early in the course when ventilator settings are low.