(SLC-029) Predicting the Probability of Spontaneous Patent Ductus Arteriosus Closure to Inform Clinical Treatment Decisions for Preterm Infants

Background: All forms of deliberate patent ductus arteriosus (PDA) closure potentially have adverse effects. Detriments of both pharmacological (NSAIDs, acetaminophen) and invasive treatment (catheter-closure) may outweigh benefits, especially for infants destined for early, spontaneous closure. Identifying highly predictive characteristics of spontaneous closure may reduce unnecessary treatment.

Objectives: To estimate the risk of spontaneous PDA closure across a robust set of clinical and echocardiographic covariates, and to predict ductal closure timing in extremely preterm infants to inform treatment decisions.

Methods: We prospectively enrolled a cohort of < 30-week gestational age (GA) preterm infants without life-threatening congenital anomalies in Columbus, Ohio, USA from April 2019-September 2023. An open PDA was required on initial screening echocardiography (echo) at age < 72 hours. Clinical, echo, and biomarker variables were collected on the same day weekly for each participant during the first 4-weeks postnatal. Serial echos continued every 2-weeks until documented PDA-closure or discharge. Fine-Gray subdistribution hazard models were employed to estimate PDA closure probability over time, while accommodating competing risks of death, PDA-closure drug treatment, or catheter-closure prior to spontaneous closure. LASSO penalization was used in building multivariable prediction models.

Our primary outcome was time of documented spontaneous PDA closure through 36-weeks postmenstrual age. Maternal and infant demographics, clinical measures including illness severity markers, weekly biomarker measurements, and weekly echo-derived predictors collected the first 4 postnatal weeks were candidate predictors.

Results: We enrolled 363 eligible preterm infants, of whom 274 had spontaneous PDA closure (75%). Median time to closure from study enrollment was 15 days (IQR: 42 days). Twenty-nine infants were observed to have treatment for their PDA prior to spontaneous closure, and 20 infants died before closure or treatment. GA at delivery, size of the PDA in the first days of life, and 5-minute Apgar score are among the most influential covariates in predicting spontaneous closure. One echo measure remained in the final model, left ventricular posterior wall (LVPW) thickness. We
observed reasonable discrimination over time (AUC ROC: 0.74 (95%CI: 0.65, 0.80), as well as good calibration (slope: 0.92 [0.48, 1.61]) following internal validation.

Conclusions: When included in a multivariable model, LVPW thickness and traditional measures, including PDA size, maternal age, GA at delivery and urine output in the first 12 hours, remained strong predictors of spontaneous PDA closure.