346 - Prediction of Bronchopulmonary Dysplasia Associated Pulmonary Hypertension

Publication Number: 346.6243

Samuel Gentle, Yale School of Medicine, Guilford, CT, United States; Wesley Jackson, University of North Carolina at Chapel Hill School of Medicine, Chapel Hill, NC, United States; Matthew M. Laughon, The University of North Carolina at Chapel Hill, Chapel Hill, NC, United States; Waldemar Carlo, UAB School of Medicine, Birmingham, AL, United States; Namasivayam Ambalavanan, University of Alabama School of Medicine, Birmingham, AL, United States

Background: Previously developed models to predict infants’ risk for bronchopulmonary dysplasia (BPD) have not included estimates for BPD associated pulmonary hypertension (BPD-PH). The lack of models to predict risk of BPD-PH limits the ability for early identification and therapeutic strategies that may improve survival or reduce disease burden.

Objective: We hypothesized (1) that clinical covariates from the first postnatal day can accurately predict risk for BPD-PH and (2) that available BPD risk outcome estimation models, developed to predict the risk for BPD and BPD severity, can accurately predict risk for BPD-PH.

Design/Methods: This was a retrospective, single center cohort study of infants born at < 29 weeks’ gestation between 2017 and 2024. BPD-PH was defined as the presence of a tricuspid regurgitation jet >2.5 m/s, bidirectional shunting through either a septal defect or ductus arteriosus, or an eccentricity index >1.2. Given the collinearity between BPD severity and BPD-PH, the following predictive covariates from the NICHD BPD Risk Estimators were evaluated from postnatal day one: birth weight, gestational age, respiratory support, fraction of inspired oxygen, race, gender, and antenatal corticosteroid exposure. Models were developed using logistic regression and adjudicated using the area under the receiver operating characteristic curve (AUROC). The AUROC was then compared to predicted risk for moderate or severe BPD, severe BPD, moderate or severe BPD or death, grade 1 to 3 BPD, grade 3 BPD, and grade 1-3 BPD or death as estimated using the NICHD BPD Risk Estimators developed in 2011 and 2022.

Results: Among the 809 infants available for analysis, 454 infants did not develop BPD, 216 developed BPD without PH, and 139 infants developed BPD-PH. Infants that developed pulmonary hypertension were born at a lower birth weight (median 619 vs 870; p< 0.001) and lower gestational age (median 24 vs 27; p< 0.001;Table 1). Of the clinical covariates evaluated, only birth weight and gestational age were retained in the final internal model. The final model, as well as predicted risk for moderate or severe BPD or death and grade 1-3 BPD or death as predicted using the 2011 and 2022 NICHD BPD Risk Estimators respectively, accurately predicted the outcome of BPD-PH on postnatal day 1 with an AUROC > 0.80 (Figure 1).

Conclusion(s): Clinical covariates available on the first postnatal day can accurately predict the BPD-PH phenotype. Model informed cut-off values may be incorporated into the design of clinical trials aimed at specifically reducing risk for BPD-PH.

Table 1: Covariates Considered in Prediction Models
Prediction Table 1.pdf1median (interquartile range)

Figure 1: Areas Under the Receiver Operating Characteristics Curve for Prediction Models
Figure 1. Areas under the receiver operating characteristics curve for A, top performing models, B, 2011 NICHD BPD Risk Estimation models, and C, 2022 NICHD BPD Risk Estimation models. The top performing models included the unique model generated from the UAB data set including covariates gestational age and birthweight, the 2011 NICHD model for the composite of moderate or severe BPD or death, and the 2022 NICHD model for grade 1-3 BPD or death.