Question Is active treatment of hemodynamically significant patent ductus arteriosus (PDA) in preterm infants better than an expectant management approach?
Findings This meta-analysis included 10 randomized clinical trials involving 2035 preterm infants born before 33 weeks of gestation. Compared with expectant management, active closure of hemodynamically significant PDA during the first 2 weeks of life was associated with potential harm, including significantly higher rates of mortality (16% vs 12%) and a significantly worse composite outcome of death at 36 weeks' postmenstrual age or at discharge or moderate to severe bronchopulmonary dysplasia (56% vs 51%).
Meaning These findings suggest that an expectant management approach to PDA in preterm infants may be associated with a better morbidity and mortality profile.
For decades, the conventional management approach to hemodynamically significant patent ductus arteriosus (PDA) in preterm infants has been to attempt closure with prostaglandin (PG) inhibitors, failing which surgical ligation is usually carried out. The rationale for PDA closure is to avoid cardiac failure and pulmonary overperfusion from a left-to-right shunt, which may minimize various PDA-associated morbidities, such as intraventricular hemorrhage (IVH), necrotizing enterocolitis (NEC), and bronchopulmonary dysplasia (BPD), for which PDA is thought to be an important causative factor.
Since the 2000s, however, there has been a shift toward less aggressive treatment for PDA because despite some success with pharmacologic closure, no other benefit could be conclusively demonstrated. The pharmacologic agents used also have some serious adverse effects. In addition, rates of spontaneous PDA closure in infants with a birth weight of 1000 g or more were very high in previous studies. Surgical ligation of PDA is accompanied by substantial risks and does not guarantee a good outcome. Nevertheless, in the majority of neonatal intensive care units, infants with a birth weight of less than 1000 g are still being routinely treated for hemodynamically significant PDA, despite reports of noninferiority of an expectant management approach. Sung et al compared outcomes between 2 historical cohorts of infants with a gestational age (GA) of 23 to 26 weeks. During the earlier period, the protocol was mandatory PDA closure, with 82% of infants undergoing PDA ligation at a mean (SD) age of 12 (7) days. In the later period, no infant received indomethacin or underwent PDA ligation. Rates of BPD were notably higher in the mandatory PDA closure cohort.
Randomized clinical trials (RCTs) comparing active treatment vs expectant management of PDA are challenging to conduct because there is often insufficient clinician equipoise, resulting in difficulty in patient recruitment. Sosenko et al and Kluckow et al were the first to conduct trials using an echocardiography-guided targeted approach to the treatment of the PDA. Clyman et al conducted the largest RCT of pharmacologic treatment vs expectant management of PDA for preterm infants. Their results did not show better outcomes for the active treatment group. Instead, active treatment demonstrated harm in a certain subgroup of infants. Two similar large trials were completed in 2023 and 2024. The results of these large trials could potentially affect the way we manage PDA in preterm infants. We therefore performed a meta-analysis of RCTs comparing active treatment vs expectant management of PDA.
This meta-analysis was conducted using data from published studies where ethical approval and informed consent were obtained by the original investigators; therefore, no separate ethical approval was required. The study followed the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) reporting guideline. The research protocol is registered in the International Prospective Register of Systematic Reviews (PROSPERO; CRD42024541379).
A literature search was conducted by 2 independent reviewers (S.B. and Y.L.W.) using the PubMed (MEDLINE), Embase, and Cochrane Library databases, including both MeSH (Medical Subject Headings) terms and related keywords. The detailed search strategy is available in eTable 1 in Supplement 1. The search was limited to studies that were published in English between January 1, 2010, and July 31, 2024, and included human participants only. We chose to limit the search to studies published from 2010 onward because antenatal steroid use for preterm delivery was then more prevalent and consistent. The reference lists of included articles were screened for additional articles.
RCTs were included if they studied preterm infants born before 33 weeks of gestation with hemodynamically significant PDA (diagnosed by clinical or echocardiographic criteria) and had 2 groups comparing active treatment of PDA (pharmacologic, surgical ligation, percutaneous device closure) vs expectant management. Studies that administered treatment for the purpose of prophylaxis against PDA (ie, PDA was not determined to be present and hemodynamically significant before administration of the intervention) were excluded.
Primary outcomes were as follows: (1) composite of death at 36 weeks' postmenstrual age (hereinafter, 36 weeks) or at discharge (whichever occurred later) or moderate to severe BPD, (2) composite of death at 36 weeks or moderate to severe BPD, (3) death at 36 weeks, (4) death at 36 weeks or at discharge (whichever occurred later), and (5) moderate to severe BPD. Secondary outcomes included death before hospital discharge, death at 28 days, cause of death, IVH, periventricular leukomalacia (PVL), retinopathy of prematurity (stage 3) or requiring treatment, pulmonary hemorrhage, cardiovascular support (hypotension, inotropic support, or both), pulmonary hypertension, NEC (Bell stage ≥2), gastrointestinal perforation, gastrointestinal bleeding, time to full enteral feeding, diuretic use, sepsis, kidney failure, postnatal steroid use, surgical ligation, duration of respiratory support (oxygen vs invasive vs noninvasive), length of hospital stay, PDA status at discharge, weight gain, and discharge home with respiratory or oxygen support.
Two authors (S.B. and Y.L.W.) independently conducted the search in the chosen databases using the predetermined search syntax, reviewed the abstracts, and reviewed the full text of the included articles. Any discrepancies were mutually resolved by consensus or in consultation with a third author (J.L.). Data collection was performed by 2 authors (S.B. and Y.L.W.) independently and verified by a third author (J.L.). Any discrepancies were mutually resolved. The Cochrane Risk of Bias tool was used to assess study quality.
A random-effects model was chosen to estimate the pooled effects, including both relative risk (RR) and risk difference (RD). Heterogeneity among different studies was evaluated with the Cochran Q test and I statistics. Small-study effects were examined with the regression-based Egger test and the Begg test. Subgroup analyses were conducted for articles reporting on infants born before 29 weeks of gestation for the reported primary outcomes. Forest plots were generated to visualize the treatment effect of the individual studies and the pooled effect. P ≤ .05 (2-tailed) for RR was considered statistically significant. Stata, version 18 (StataCorp LLC), was used to perform the meta-analysis.
The search was conducted per the predefined search strategy. The full search results are reported in Figure 1. Of the 524 abstracts screened, 490 were excluded for the following reasons: (1) did not report predefined primary outcomes, (2) were not about treatment of PDA, (3) were not RCTs, (4) did not have a control group, or (5) were duplicates. A total of 34 articles underwent full-text review; 24 articles were excluded, 19 of which were published before 2010 but were excluded for reasons other than year of publication (eTable 2 in Supplement 1). Ten articles were included in this meta-analysis, all of which evaluated treatment of PDA in the first 2 weeks of life. A total of 6 studies used ibuprofen in their active treatment group, 1 study used indomethacin, and 3 studies used more than 1 medication (ibuprofen, indomethacin, or paracetamol). Details on the design of each study are provided in eTable 3 in Supplement 1. Secondary outcomes and BPD definitions are included in eTable 4 in Supplement 1. Differences between the PROSPERO protocol and our final meta-analysis are summarized in eTable 5 in Supplement 1.
A total of 2035 infants were enrolled in the included trials: 1018 in the active treatment group (510 female [50.1%] and 508 male [49.9%]) and 1017 in the expectant management group (466 female [45.8%] and 551 male [54.2%]). eTable 6 in Supplement 1 describes the baseline characteristics of the study population. Infants in the active treatment group vs the expectant management group had a mean (SD) GA of 26.2 (1.7) weeks vs 26.3 (1.7) weeks and a mean (SD) birth weight of 874.7 (222.1) g vs 897.7 (216.5) g, respectively. Other baseline characteristics were similar between groups. Open-label medical treatment was administered to 164 patients (16.1%) in the active treatment group compared with 297 (29.2%) in the expectant management group. PDA treatment was started during the first 2 weeks of life in all 10 trials and within the first 72 hours of life (HOL) in 7 trials.
Risk of bias was assessed for all included studies, as illustrated in eFigure 1 in Supplement 1. Three studies were deemed to have low risk of bias. The other 7 studies were overall deemed as having some concerns for bias, mainly arising due to deviations from the intended intervention (eg, treatment of PDA in the expectant management group).
The results of the primary outcomes are presented as RRs and RDs in Table 1. Forest plots for outcomes not described in the main text are presented in eFigure 2 in Supplement 1.
Eight studies (n = 1833) reported a significantly higher incidence of the composite outcome of death at 36 weeks or at discharge (whichever occurred later) or moderate to severe BPD in the active treatment group vs the expectant management group (516 of 918 [56.2%] vs 465 of 915 [50.8%]; RD, 5.4% [95% CI, 1.1%- 9.7%]; RR, 1.10 [95% CI, 1.01-1.19]; P = .02) (Table 1 and Figure 2A). Subgroup analysis for the composite outcome for infants with a GA less than 29 weeks yielded similar results for the active treatment group vs the expectant management group (474 of 760 [62.4%] vs 431 of 760 [56.7%]; RD, 5.0% [95% CI, -1.1% to 11.1%]; RR, 1.09 [95% CI, 1.01-1.18]; P = .03) (Table 1 and eFigure 3 in Supplement 1).
Additional subgroup analyses were conducted for articles reporting on infants born before 26 weeks of gestation, mode of ventilation at time of randomization (noninvasive vs invasive), HOL at treatment (≤72 HOL vs >72 HOL), and medication type (ibuprofen vs indomethacin vs >1 agent) for the composite outcome of death at 36 weeks or at discharge (whichever occurred later) or moderate to severe BPD. No statistically significant differences were observed (eFigure 4 in Supplement 1).
Seven studies (n = 1741) reported a nonsignificant increase in death at 36 weeks or moderate to severe BPD in the active treatment group vs in the expectant management group (486 of 874 [55.6%] vs 434 of 867 [50.1%]; RD, 5.9% [95% CI, 1.2%-10.6%]; RR, 1.10 [95% CI, 1.00-1.21]; P = .06) (Table 1 and eFigure 2 in Supplement 1).
Nine studies (n = 1943) reported significantly more deaths at 36 weeks in the active treatment group vs the expectant management group (139 of 974 [14.3%] vs 109 of 969 [11.2%]; RD, 2.8% [95% CI, 0.1%-5.6%]; RR, 1.27 [95% CI, 1.01-1.61]; P = .04) (Table 1 and eFigure 5 in Supplement 1). All 10 studies (n = 2035) reported more deaths at 36 weeks or at discharge (whichever occurred later) in the active treatment group vs the expectant management group (158 of 1018 [15.5%] vs 126 of 1017 [12.4%]; RD, 2.8% [95% CI, -0.2% to 5.7%]; RR, 1.25 [95% CI, 1.01-1.56]; P = .04) (Table 1 and Figure 2B). For subgroups of infants with a GA less than 29 weeks, deaths (both definitions) were also significantly higher in the active treatment group compared with the expectant management group (Table 1 and eFigures 6 and 7 in Supplement 1).
Eight studies (n = 1833) reported on moderate to severe BPD, which showed a nonsignificant increase in the active treatment group vs the expectant management group (372 of 785 [47.4%] vs 349 of 806 [43.3%]; RD, 4.5% [95% CI, -0.8% to 9.8%]; RR, 1.08 [95% CI, 0.95-1.23]; P = .25) (Table 1 and Figure 2C). The effect size was similar for the subgroup analysis (6 studies [n = 1520]) of infants with a GA less than 29 weeks (354 of 647 [54.7%] vs 340 of 671 [50.7%]; RD, 3.1% [95% CI, -4.2% to 10.4%]; RR, 1.07 [95% CI, 0.93-1.22]; P = .36) (Table 1 and eFigure 8 in Supplement 1).
Results of the analysis of secondary outcomes in RR and RD with the relevant P values are presented in Table 2. Eight studies (n = 1821) showed a nonsignificant increase in PVL in the active treatment group vs the expectant management group (52 of 913 [5.7%] vs 32 of 908 [3.5%]; RR, 1.50 [95% CI, 0.98-2.30]; P = .06). However, there was a significant risk difference for PVL between the 2 groups, favoring expectant management (RD, 1.8% [95% CI, 0.4%-3.2%]; P = .01). There were no statistically significant differences between groups for the other listed secondary outcomes. Not all trials in this meta-analysis reported cause of death. There was a slight preponderance of sepsis and gastrointestinal causes of death in the active treatment group. The reported causes of death are presented in eTables 7 and 8 in Supplement 1.
eFigures 2, 9, and 10 in Supplement 1 are forest plots for the other primary or secondary outcomes for which a meta-analysis was conducted. eTable 9 in Supplement 1 presents the secondary outcomes for which only a review was conducted due to insufficient data to perform a meta-analysis.
The findings of this meta-analysis suggest that active closure of PDA during the first 2 weeks of life may potentially be harmful for preterm infants with a GA less than 33 weeks. Death at 36 weeks or later was significantly higher in the active treatment group vs the expectant management group (158 of 1018 [15.5%] vs 126 of 1017 [12.4%]; P = .04) and for the composite outcome of death at 36 weeks or later or moderate to severe BPD (516 of 918 [56.2%] vs 465 of 915 [50.8%]; P = .02). For the subgroup of infants with a GA less than 29 weeks, poorer outcomes of similar magnitudes were seen. There was a nonsignificant increase in moderate to severe BPD in the active treatment group. In a European population-based cohort study, PDA treatment was associated with a notably higher adjusted risk of BPD or death. In a previous study that used Finnish national registry data for preterm infants, Härkin et al found an association between mortality and PDA treatment after adjustment for known risk factors. In an observational study on extremely preterm infants, lower rates of PDA treatment were associated with a decrease in mortality.
Not all trials reported cause of death. Available data for cause of death are presented in eTables 7 and 8 in Supplement 1. There was a slight preponderance of sepsis and gastrointestinal causes of death in the active treatment group. Based on our findings, it was not possible to explain the cause of the higher mortality rates seen in the active treatment group.
There could be 2 reasons that explain the better outcomes seen in the expectant management group: (1) PDA could confer physiologic advantages and survival protection for the preterm infant and/or (2) the pharmacologic agents used may have deleterious effects of their own. In preterm infants with respiratory distress syndrome (RDS), pulmonary blood flow (PBF) was substantially lower in those with more severe lung disease. As lung disease worsened, the PDA augmented PBF, thereby improving left ventricular output (LVO), resulting in better cerebral blood flow (CBF). Impaired CBF could explain our observation that PVL incidence was higher in the active treatment group. Ductal patency resulted in higher systemic blood flow compared with when the ductus was ligated. Previous studies have reported that ductal ligation led to left ventricular dysfunction and lower LVO. Arterial partial pressure of oxygen (Pao) was substantially higher and partial pressure of carbon dioxide was notably lower when the ductus was open. Tooley listed pulmonary ischemia as a major factor in the pathogenesis of hyaline membrane disease. In that study, pulmonary ischemia caused impaired cellular metabolism and surfactant synthesis, and it reduced lung volume and compliance. Pulmonary vasodilation with acetylcholine improved arterial Pao and lung compliance. With lung ischemia, the ensuing inflammatory cascade worsened lung injury. Thus, PDA could be a beneficial adaptive response in preterm infants with lung disease to increase pulmonary and CBF and improve oxygenation. Moreover, in preterm infants with lung disease and pulmonary hypertension, PDA may serve as a pop-off valve to reduce right ventricular afterload, minimizing risk of cor pulmonale.
PDA is more common in infants with lung disease, both in RDS and in BPD. It is likely that the ductus was patent secondary to the presence of lung disease, rather than the PDA itself causing more severe lung disease. In RDS, PG and inflammatory cytokine levels were high. BPD severity was also directly correlated with the degree of inflammation. Proinflammatory cytokines in turn stimulate cyclooxygenase COX-2 and PG synthesis. Potent vasodilatory cytokines (eg, tumor necrosis factor-α, interleukin-6) maintain ductal patency independent of PG pathways. COX inhibitors, peroxidase inhibitors, and corticosteroids constrict the ductus by reducing PG and inflammatory cytokine levels. COX inhibitor success rates of only 13% to 56% are likely due to the unopposed vasodilatory effects of cytokines. In previous studies, the observation that PDAs closed as BPD improved in severity, or after an infant completed a course of postnatal steroids, is consistent with the role of inflammation in maintaining ductal patency.
Indomethacin, ibuprofen, and paracetamol are commonly used to close PDA. Indomethacin reduces CBF and oxygenation in preterm infants. In the preterm lung, indomethacin has been reported to impair surfactant production, reduce lymphangiogenesis, and increase the fraction of inspired oxygen requirement. Previous studies have reported that ibuprofen had substantial pulmonary vasoconstrictive effects and inhibited newborn lung angiogenesis. Paracetamol metabolism via the cytochrome P450 2E1 pathway produces the toxic metabolite N-acetyl-p-benzoquinone imine, which covalently binds to intracellular proteins, causing cellular dysfunction and death. Paracetamol increases pulmonary vascular resistance, and its use in preterm infants for PDA treatment has been associated with increased mortality. All 3 medications affected the preterm lung, which could lead to increased hypoxemia.
Controversy exists regarding the definition of hemodynamically significant PDA. The commonly used criterion is PDA diameter. In a previous study, a PDA diameter larger than 1.5 mm at days 1 to 2 of life was associated with the need for later treatment of symptomatic PDA. However, there was no reduction in PDA-associated morbidities, despite pharmacologic treatment targeted at infants with hemodynamically significant PDAs using diameter as a proxy for PDA severity. A PDA severity score has been devised in an attempt to better select infants with truly hemodynamically significant PDA. In previous studies, the proportion of infants meeting treatment criteria using the PDA severity score (approximately 50%) was similar to that when PDA diameter was the criterion for hemodynamically significant PDA (approximately 33%). Although PDA is the result of substantial preterm lung disease and not the cause of it, its presence still needs to be managed to avoid pulmonary overperfusion. Management can be achieved without nonsteroidal anti-inflammatory drugs (NSAIDs) or paracetamol by optimal ventilation, judicious fluids, and prudent use of diuretics.
The sample size of the RCTs included in this meta-analysis was large, with 2035 infants total. There was also minimal study heterogeneity, suggesting insignificant variation in the true treatment effect size.
This meta-analysis has some limitations. Recruitment included infants from a wide range of GAs, and studies used different PDA treatment strategies with different pharmacologic agents. Despite this, the benefits of expectant management were also seen in the subgroup of more immature infants (<29 weeks of GA) regardless of treatment protocol. Because PDA treatment was started during the first 2 weeks of life (7 trials started within the first 72 HOL), our results may not be generalizable to infants who received later treatment. Open-label medical treatment was greater in the expectant management group (297 [29.2%]) vs the active treatment group (164 [16.1%]). Overall, lesser use of NSAIDs or paracetamol and more prolonged PDA exposure with expectant management was associated with improved outcomes. BPD definitions varied among the RCTs, although most carried out the molecular oxygen reduction test devised by Walsh et al (defining moderate to severe BPD). Subgroup analyses of other clinical points of interest (eFigure 4 in Supplement 1) did not show any significant differences for the composite outcome of death at 36 weeks or at discharge (whichever occurred later) or moderate to severe BPD. However, we were unable to rule out a type II error (false negative) due to the small number of studies reporting these data.
The findings of this meta-analysis appear to challenge the time-honored conventional wisdom in the management of PDA in preterm infants. Expectant management in the first 2 weeks of life was associated with significant reductions in mortality and in the composite outcome of death at 36 weeks or at discharge or moderate to severe BPD. Our results signal a need to review existing protocols for aggressively closing PDA. Until new data emerge from ongoing trials, the use of NSAIDs or paracetamol during the first 2 weeks of life for PDA closure must be undertaken cautiously, balancing presumed benefits and potential risks.
Corresponding Author: Jiun Lee, MMed, Department of Neonatology, Khoo Teck Puat National University Children's Medical Institute, National University Hospital, National University Health System, 5 Lower Kent Ridge Rd, Singapore 119074, Singapore ([email protected]).
Author Contributions: Dr Lee had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.
Concept and design: Buvaneswarran, Wong, Quek, Lee.
Acquisition, analysis, or interpretation of data: Buvaneswarran, Wong, Liang, Lee.
Drafting of the manuscript: Buvaneswarran, Wong, Liang, Lee.
Critical review of the manuscript for important intellectual content: All authors.
Additional Contributions: We acknowledge Sheena Nishanti Ramasamy, MPH (Department of Paediatrics, National University of Singapore), for her assistance with manuscript formatting, editing, and submission. No financial compensation was provided for this contribution.