Objectives: To examine the accuracy of different criteria for the diagnosis of bronchopulmonary dysplasia (BPD), based on the final age at which oxygen therapy was stopped, in predicting pulmonary and neurologic outcomes at 18-month corrected age. Study design: Data were collected prospectively on infants with birth weights between 500 and 999 g enrolled in the Trial of Indomethacin Prophylaxis in Preterms (TIPP) who survived to discharge home. Differing postnatal ages and postmenstrual ages at which supplemental oxygen therapy was no longer required formed the criteria for defining BPD. Diagnostic accuracy of each criterion for defining BPD was calculated for both poor pulmonary and poor neurosensory outcomes. Results: The prevalence of poor pulmonary outcome was 54% and of poor neurosensory outcome was 34% in the 956 infants who were eligible for this analysis. Accuracy of different definitions of BPD was limited but greatest when using supplemental oxygen requirement at 36 weeks' postmenstrual age to predict long-term pulmonary outcome (63%) and 40 weeks to predict long-term neurosensory outcome (68%). Conclusions: Poor pulmonary outcome and poor neurosensory outcome are common late adverse outcomes in this population. BPD as defined by duration of oxygen therapy is a less accurate surrogate currently than in previous eras. (J Pediatr 2002;140:555-60)