Patent ductus arteriosus (PDA) is a common problem in premature infants and is associated with mortality and neonatal morbidity; yet the approach to treatment selection (who, when, how) is controversial, with polarized opinions amongst neonatologists.Recent trials of early ibuprofen treatment have shown no benefit in reducing the composite outcome of death and/or bronchopulmonary dysplasia. 1,2While patient selection is reflective of reallife practice, with the eligible patient population selection based on PDA diameter >1.5 mm and predominant left-right transductal shunt, treatment was not uniformly effective.The issue of whether the enrolled population is reflective of patients with moderatehigh volume shunt is also questionable.The importance of patient selection was demonstrated at a single center where there was no increased risk of untoward outcomes when low-volume PDA shunts, several of which were greater than a diameter of 1.5 mm, for a minimum of 2 weeks, were left untreated. 3In addition, when hemodynamic significance is clearly defined, improvements in death, severe intraventricular hemorrhage, and grade 3 bronchopulmonary dysplasia were seen with targeted PDA therapy. 4The controversy regarding definition, approach to screening, and management strategies lends itself well to the potential of artificial intelligence and machine learning models in PDA identification and prediction.In this issue of Pediatric Research, Kung et al. 5 describe a machine learning-based tool to facilitate recognition of premature infants at increased risk of hemodynamically significant PDA (hsPDA).The final model included 6 variables: birth weight, mean fraction of inspired oxygen (FiO 2 ) in the first 3 days, mean blood pressure (BP) in first 3 days, percent of time with saturations >96% in the first 3 days, amount of surfactant received in the first 3 days, and mode of delivery.The model achieved a sensitivity of 74.8% and a specificity of 53.4% with an area under the ROC curve of 0.685.The building of this model was based on the assumption that receipt of therapy was pathognomonic of a hsPDA.This approach places a high degree of importance on the validity of the echocardiography definition of hemodynamic significance.In their routine clinical practice, adjudication of hemodynamic significance is based on the presence of a left-to-right PDA shunt, and identification of one of the following echocardiography parameters: end-diastolic flow in the left pulmonary artery >0.2 m/ s, diastolic flow reversal in the abdominal aorta/celiac trunk, or left ventricular load (visual estimate of ejection fraction/left atrium to aorta ratio > 1.4/left ventricle end diastolic diameter >15 mm/kg).
Rios et al. (Fri,) studied this question.