This retrospective cohort study of premature infants born at <37 weeks of gestation between January 2013 and August 2023 was conducted at the Jeonbuk National University Children’s Hospital. A total of 1,979 infants were initially screened. After excluding 271 participants due to incomplete data and congenital conditions associated with mortality, such as major congenital heart defects or severe congenital lung anomalies, 1,708 participants were included in the final analysis. Only patients with sufficient echocardiographic data to confirm the presence of PDA or PH were enrolled in the study. Patients were excluded if they had unsuitable echocardiographic images or were diagnosed with complex heart diseases, such as coarctation of the aorta, tetralogy of Fallot, pulmonary atresia with ventricular septal defect, or severe congenital anomalies significantly affecting mortality, including anencephaly and congenital diaphragmatic hernia. Patients born outside the hospital and those without immediate postnatal data were also excluded. Clinicodemographic data were obtained through a chart review of medical records, which included information on gestational age (GA), sex, birth anthropometry, and medical treatment. Medical treatment consisted primarily of sildenafil and bosentan. For patients with insufficient response, inhaled nitric oxide or iloprost was added as second-line therapies. In later years of the study period, adjunctive therapies such as milrinone and treprostinil were increasingly utilized in severe cases.

This study was conducted in accordance with the Declaration of Helsinki, 1975, as revised in 2008, and was approved by the Institutional Review Board of Jeonbuk National University Hospital (IRB number: 2023-09-027-003, approval date: November 3, 2023). Given the retrospective nature of this study, the need for informed consent was waived.

Echocardiographic assessments were performed using a Philips IE33 System with an S12-4 transducer, which enables robust precision in neonatal cardiac imaging. An experienced cardiologist examined the patients, and their medical records and echocardiographic views were thoroughly reviewed by the principal investigator (Jihye You).

The diagnostic protocol for PH was based on a comprehensive echocardiographic evaluation performed by an experienced pediatric cardiologist. Echocardiographic assessments were conducted between 72 hours and 14 days after birth, following the early transitional period, to allow for the physiological decline in pulmonary vascular resistance and to minimize the influence of transient postnatal PH [23]. The presence of rightto-left or bidirectional shunting through a PDA or patent foramen ovale was considered a primary indicator of elevated pulmonary pressures. In cases without a detectable shunt, PH was diagnosed if the right ventricular systolic pressure (RVSP), estimated by echocardiography using the modified Bernoulli equation,

RVSP=4×(TRV)²+RAP,

was >40 mmHg or if the RVSP/systemic systolic pressure ratio exceeded 0.524, [25]. A tricuspid regurgitation velocity (TRV) of ≥3.0 m/sec was considered suggestive of elevated pulmonary pressure, based on a right atrial pressure (RAP) of 5 mmHg under normal conditions. Additional structural and functional echocardiographic markers supporting a diagnosis of PH included right ventricular hypertrophy, right atrial enlargement, flattening or leftward bowing of the interventricular septum, and pulmonary artery dilation [23,26,27].

To prevent the inclusion of patients with transient physiological PDA, cases with PDA identified within the first 72 hours postpartum were excluded. For PDAs diagnosed after this period, significant left-to-right or bidirectional shunting was confirmed. Moderate-to-large PDA was characterized by shuntinduced clinical changes, such as cardiomegaly and significant pulmonary overcirculation. Cardiomegaly and pulmonary overcirculation were assessed based on a ductal diameter ≥1.5 mm, a left atrial-to-aortic root ratio ≥1.4, and evidence of diastolic flow reversal in the descending aorta, indicative of hemodynamic significance [28].

BPD was diagnosed in accordance with the National Institutes of Health consensus definition [29]. Infants requiring supplemental oxygen for at least 28 days postpartum were evaluated for BPD. Necrotizing enterocolitis (NEC) was diagnosed in advanced cases based on a combination of clinical signs, such as abdominal distension, bloody stools, and feeding intolerance, as well as radiographic findings, including pneumatosis intestinalis, portal venous gas, and pneumoperitoneum. Diagnosis of NEC was established using the modified Bell’s staging criteria [30].

Statistical analysis was performed using IBM SPSS Statistics for Windows version 29.0 (IBM Corp.). The significance level was set at P<0.05. Receiver operating characteristic (ROC) curve analysis was conducted to evaluate the effectiveness of GA and birth weight as predictors of mortality. This involved calculating the area under the ROC curve (AUC) to determine the sensitivity and specificity for forecasting mortality outcomes.

Quantitative data are presented as mean±standard deviation, whereas categorical data are expressed as frequencies (%). For intergroup comparisons, the unpaired t-test, Fisher’s exact test, and Pearson chi-square test were used, as appropriate. Logistic regression analysis was used to determine the odds ratios (ORs), with 95% confidence intervals (CIs), and identify the factors among GA, birth weight, PH, PDA, NEC, and BPD that significantly affected mortality in the cohort. ChatGPT provided by OpenAI in English was utilized for proofreading in the manuscript text.

Figure 1 shows the flow diagram of patient selection and classification. Table 1 describes the baseline characteristics of the entire cohort. The overall cohort had a mean GA and birthweight of 34.2±2.3 weeks and 2,159.9±574.1 g, respectively, with 53.0% males (906 patients) in the cohort (Table 1). Additionally, the median length of hospital stay for the entire cohort was 7 days (interquartile range [IQR], 5 to 16). In the cohort, 257 patients (15.0%) were diagnosed with PDA, of whom 110 (6.4%) were classified as having moderate-to-large PDA. The PDA group had a mean GA and birthweight of 32.4±3.3 weeks and 1,829.2±73.9 g, respectively, and males comprised 52.9% of this subgroup. At the time of discharge, PDA closure had been achieved in 59 participants. The median length of hospital stay for the PDA group was 18 days (IQR, 8 to 42).

PH was diagnosed in 46 patients (2.7%). The PH subgroup demonstrated a lower mean GA and birthweight of 31.4±3.6 weeks and 1,603.3±714.9 g, respectively (both P<0.001), and males comprised 47.8% of this subgroup. Furthermore, the median length of hospital stay for patients diagnosed with PH was 32 days (IQR, 10.25 to 97).

The assessment of mortality-related factors generated an ROC curve that revealed the optimal cutoff values of GA and birthweight as ≤33.5 weeks and ≤1,564 g, respectively. The sensitivity for both GA and birthweight was 78.6%, whereas the specificity was 86.4% for GA and 73.1% for birthweight. These thresholds were determined to provide the best balance between sensitivity and specificity for the predictive analysis (Supplementary Figure 1).

PH was associated with mortality and intubation rates of 21.7% and 93.5%, respectively. The occurrence of PDA was associated with mortality and intubation rates of 3.5% and 49.0%, respectively. The subgroup with moderate-to-large PDA had mortality and intubation rates of 8.2% and 70.9%, respectively. A total of 14 patients (0.8%) succumbed to illness, while 1,694 patients (99.2%) survived. Supplementary Table 1 provides a comparative analysis of baseline characteristics, including demographic and clinical factors, between mortality and survival groups.

Table 2 presents univariate and multivariate analyses of mortality-related factors. In the multivariate analysis, PH remained a significant predictor of mortality (OR, 36.10; 95% CI, 6.82 to 191.12). Notably, model 1 demonstrated a higher OR (OR, 47.7; 95% CI, 9.73 to 23.96) compared to model 2 (OR, 36.10; 95% CI, 6.82 to 191.12). This difference may be attributed to the inclusion of all PDA cases in model 1, whereas model 2 focused on moderate-to-large PDA cases, excluding those with less hemodynamically significant PDA, which may have diluted the observed association between PH and mortality.

We performed logistic regression analysis with PH as the outcome variable, adjusted for potential confounders in model 1 and model 2. Model 1 included all PDA cases, while model 2 focused on moderate-to-large PDA, providing insights into the impact of PDA size on PH risk. As PH was a significant outcome in the multivariate analysis, we further investigated its associated risk factors (Table 3). In model 1 multivariate analysis, significant predictors of PH included birthweight ≤1,564 g (OR, 2.86; 95% CI, 1.18 to 6.93; P=0.020), PDA (OR, 16.30; 95% CI, 7.05 to 37.69; P<0.001), and BPD (OR, 5.33; 95% CI, 1.69 to 16.84; P=0.004). In model 2, however, BPD did not remain statistically significant (OR, 2.79; 95% CI, 0.87 to 8.97; P=0.085).

Further investigation into the treatments for PH and their association with mortality revealed that, although mortality rates appeared to differ between treatment groups, most comparisons did not reach statistical significance (Table 4). The exception was iloprost, which demonstrated a significant association with increased mortality (OR, 5.30; 95% CI, 1.15 to 24.42; P=0.033). Specifically, neonates who received iloprost had significantly lower GA (28.5±2.1 weeks) and birth weight (1,200±300 g) than those who did not (30.5±2.0 weeks; Mann– Whitney U-test P<0.001 and 1,600±290 g; Mann–Whitney U-test P<0.001, respectively). Furthermore, the prevalence of PDA was higher in the iloprost group (25%) than in the non-iloprost group (10%), with a Fisher's exact test P-value of 0.011. After adjusting for GA and birth weight in a multivariate logistic regression analysis, the association between iloprost treatment and mortality was no longer statistically significant (adjusted OR, 2.31; 95% CI, –0.06 to 4.68; P=0.056), suggesting that the initially observed association may have been confounded by differences in baseline clinical characteristics.

In the general PDA cohort, the mortality rate was 64.3% in patients with PDA, while 14.6% of those without PDA survived (Supplementary Table 1). In the moderate-to-large PDA group, the mortality rates were 3.2% and 10.1% in the treated and untreated groups, respectively (Table 5).