Predictive power of initial severity of pulmonary disease for subsequent development of bronchopulmonary dysplasia.

Using retrospectively acquired data from 138 mechanically ventilated premature infants, logistic regression was used to determine the relationships between the risk of bronchopulmonary dysplasia (BPD) and indices of initial respiratory disease severity [oxygen index (OI) and alveolar-arterial pO(2) difference (A-a DO(2))]. Indices were calculated from the first arterial blood gas analysis after initial surfactant administration. Infants were also classified as having mild [OI <4 cm H(2)O x mm Hg(-1), A-a DO(2) <150 mm Hg (20 kPa)] or severe [OI >or=10 cm H(2)O x mm Hg(-1), A-a DO(2) >or=300 mm Hg (40 kPa)] respiratory disease, and the ability of this classification to predict subsequent BPD risk was calculated. OI and A-a DO(2) were significantly higher in the BPD group. Logistic regression analysis showed that BPD risk increased linearly with both OI (9%/cm H(2)O x mm Hg(-1)) and A-a DO(2) [16%/50 mm Hg (16%/6.7 kPa)]. However, the predictive power (receiver-operator characteristic) of these models was modest. Unexpectedly, 29% of infants with mild initial disease developed BPD. These data suggest that, while BPD prediction in infants with severe disease is straightforward, the identification of those few infants with mild to moderate disease destined to develop BPD remains problematic.