Evaluation of the association between patent ductus arteriosus approach and neurodevelopment in extremely preterm infants

Objective Assess if unit-level PDA management correlates with neurodevelopmental impairment (NDI) at 18–24 months corrected gestational age (cGA) in extremely preterm infants. Study design: Retrospective analysis of infants born at < 29 weeks (2014–2017) across two units having distinct PDA strategies. Site 1 utilized an echocardiography-based treatment strategy aiming accelerated closure (control). Site 2 followed a conservative approach. Primary endpoint: NDI, characterized by cerebral palsy, any Bayley-III composite score < 85, sensorineural/mixed hearing loss, or at least unilateral visual impairment. Results


Introduction
In extremely preterm infants, the patent ductus arteriosus (PDA) can spontaneously close, but this closure may take several weeks after birth, especially in the most immature neonates.This delay in closure can be associated with alterations in the distribution of blood ow between the systemic (such as the cerebral) and the pulmonary circulation (1,2).In the context of underdeveloped organs, a signi cant leftto-right shunt via the PDA may result in a phenomenon known as "diastolic steal," where blood ow is diverted from the systemic circulation into the pulmonary circulation (3).Numerous early neonatal morbidities have been associated with the persistence of the left to right ductus, such as bronchopulmonary dysplasia (BPD), necrotizing enterocolitis (NEC), retinopathy of prematurity (ROP), and cerebral intraventricular hemorrhage (IVH) (4).Patency of the ductus is also known to decrease cerebral blood ow velocity and alter its patterns by Doppler ultrasonography in preterm infants (5).These mechanisms have been suggested as potential risk factors for conditions such as periventricular leukomalacia (PVL), white matter damage, reduced cerebral volume, and cerebral palsy (CP) (6).
Controversy exists regarding the optimal approach for PDA management in the extremely preterm infants (7)(8)(9).Indeed, medications like non-steroidal anti-in ammatory medications (NSAIDs) accelerate ductal closure without clear improvements in neonatal outcomes outlined by trials, and these medications have side effects that can impact neurodevelopmental outcomes, such as survival without CP (3,6).Concerns have been raised regarding cerebral vasoconstriction and ischemia following exposure to NSAIDs (5,10,11).Surgical ligation comes with procedural risks, post-ligation hemodynamic instability and postoperative concerns, which may contribute to long-term adverse outcomes.Considering the absence of clear bene ts to systematic PDA closure, recent recommendations have advocated for the avoidance of NSAIDs in the rst two weeks of life (12).Therefore, in recent years, a conservative approach has been increasingly used (13,14), leading to con icting results: while some reports describe improved outcomes (1,15,16), others outline potential concerns in the most immature infants (17).
Few studies have evaluated the impact of a strict conservative PDA approach policy with the approach of early targeted treatment regarding neurodevelopmental outcomes of extremely preterm infants.As such, this study aims to compare the neurodevelopmental outcomes at 18 to 24 months of corrected gestational age (cGA) in preterm infants born at < 29 weeks of gestational age (GA) and exposed to each of these approaches.We hypothesized that preterm infants exposed to a conservative PDA policy would have an increase rate of adverse neurodevelopmental pro le compared to those exposed to a targetedtreatment policy.

Methods
This is a retrospective observational cohort study based on the data from two level-4 neonatal intensive care units (NICUs) in Montreal, Canada.Site 1 (targeted-treatment approach) and Site 2 (conservative approach) admitted both inborn and outborn extremely premature newborns.A conservative approach regarding the PDA was adopted at Site 2 at the end of 2013, and its uptake progressively increased as outlined in Table 1.Both institutions share many comparable characteristics in terms of level of care, patient populations (socioeconomic status and risk factors), presence of neonatal nurse practitioners and of a neonatal transport team.Both institutions moved to a modern single-room unit in 2015 (Site 2) and in 2016 (Site 1).In addition, both hospitals are university-a liated institutions and have a neonatal follow-up program, where extremely preterm infants are evaluated by an interdisciplinary team (pediatrician, occupational therapist, physical therapist and psychologist) specialized in their developmental assessment, with standardized methodologies for evaluation, as part of the ongoing Canadian Neonatal Follow-Up (CNFUN) registry.
Perinatal and neonatal data for this study were retrieved from the local neonatal databases maintained in the context of the Canadian Neonatal Network (CNN) collaboration, while information on neurodevelopment was obtained from the local CNFUN databases.Both registries have patient information entered electronically by trained abstractors into a data-entry program at each site with builtin error checking that has shown high reliability and internal consistency (18).
Inclusion and exclusion criteria.We included infants born less than 29 completed weeks of estimated GA, between January 2014 and December 2017, provided they were admitted to the respective institution within the rst 24 hours after birth (as babies admitted beyond this period are often coming from other Level-3 centers and transferred for surgical or subspecialty assessments).We restricted the analysis sample to the infants who survived to their appointment and for whom follow-up evaluations were available (not lost to follow-up).We excluded infants with signi cant congenital/genetic anomalies.The study was approved by the research ethics board of both institutions.
De nitions of neonatal outcomes.Bronchopulmonary dysplasia (BPD) was de ned as respiratory support or oxygen supplementation at 36 weeks cGA (19).Signi cant intraventricular hemorrhage (IVH) was classi ed as grade 3 or more according to the Papile's classi cation (20), and/or presence of periventricular leukomalacia (PVL) (21).Necrotizing enterocolitis (NEC) was determined as a stage 2 and above according to the Bell's classi cation (22,23).Prolonged rupture of membranes (PROM) (24) was de ned when it occurred for more than 24 hours before delivery.Chorioamnionitis was de ned as maternal fever >38.4 °C in the 24 hours prior to the delivery (irrespective of whether the mother received epidural analgesia), uterine tenderness if mentioned in the chart, or leucocytosis >15,000 per mL (25).Retinopathy of prematurity (ROP) was de ned as stage 3 or more, according to the International Classi cation for Retinopathy of Prematurity (26), or requiring treatment with laser photocoagulation or intra-ocular injection of bevacizumab.Small for gestational age (SGA) was de ned as a birth weight <10th percentile for GA (27).Late onset sepsis was de ned by a positive bacterial or fungal culture in blood and/or cerebrospinal uid beyond two days of life (28).
Follow-up evaluation at 18-24 months cGA.Alongside a complete neurological examination, audiological and ophthalmological assessments, both follow-up programs administer the Bayley Scales of Infant and Toddler Development -Third Edition (Bayley-III) to assess the participant's development.For this study, at the 18 to 24 months visit, data from the Bayley-III was collected, as well as follow-up data regarding the presence of CP, hearing and/or visual impairment(s).When CP was diagnosed, the gross motor functional classi cation score (GMFCS) was collected (29).Only the motor and cognitive scores were collected for the Bayley-III assessment due to high heterogeneity in ethnicities attending both programs (with various rst languages other than French and English; this was different between sites as per previously reported ( 30)).Accordingly, the assessment of language scores was challenging.Any neurodevelopmental impairment (NDI) was de ned as any degree of CP, a Bayley-III composite score (motor or cognitive) <85, a sensorineural/mixed hearing loss, or at least a unilateral visual impairment.Signi cant NDI referred to CP (GMFCS score >3), Bayley-III composite score (motor or cognitive) <70, a hearing loss requiring an aid/implant, or a bilateral visual impairment (19).The outcomes of interest were rates of any NDI (primary outcome) or signi cant NDI (secondary outcome).
Statistical analyses.Results are described as means with standard deviation (SD) or medians with interquartile range (IQR) for continuous variables, and with counts and percentages for categorical variables.Fisher's exact test was used to assess differences in categorical variables.The student's t-test or Wilcoxon-Mann-Whitney test were used to assess differences in continuous variables with a parametric or non-parametric distribution, respectively.We estimated adjusted odds ratio and 95% con dence interval using multivariable logistic regression models to evaluate the association between the exposure (PDA management policy) and the outcomes of interest.Main models were adjusted for confounders/effect modi ers a-priori identi ed, that included: presence of bronchopulmonary dysplasia (as an effect modi er), gestational age at birth, exposure to chorioamnionitis and presence of signi cant IVH/PVL.As secondary analysis, we rede ned the exposure as PDA treatment, regardless of site.Further, models were re-run with the adjustments taking into account birth variables that may be associated with site and with the outcomes: SNAP-II score >20, chorioamnionitis and sex.Statistical analyses were done using Stata SE Program (Version 14.2, College Station, Texas).

Results
During the study period, 586 infants were admitted to both NICUs.Of these, 209 were excluded (108 died before 18-24 months cGA, 91 lacked follow-up data, and 10 had a major congenital anomaly), leaving 377 infants (n=234 at Site 1 and n=143 at Site 2) for analysis (Figure 1).Mortality was higher in Site 1 compared to Site 2 (Site 1: 28% vs Site 2: 15%).Demographics, clinical characteristics, and neonatal outcomes are shown in Table 1.Birth weight, GA at birth, sex, mode of delivery, SGA status, outborn status, singleton status, SNAP-II score, antenatal steroid exposure and postnatal steroid exposure were similar between centers.Site 1 had a higher number of BPD (n=123 [53%] vs n=50 [35%]).Site  2 shows the evolution of the rate of PDA treatment exposure and the neurodevelopmental outcomes per year and per site.Characteristics regarding the cohort of infants lost to follow-up are detailed in (Supplementary Table 1).
Neurodevelopmental outcomes at follow-up are shown in Table 3.We saw no difference in terms of Bayley-III scores or other neurodevelopmental outcomes between the groups, apart from the Bayley-III absolute motor composite score, which was lower at Site 2 (85±18, vs 92±13).However, there was no signi cant difference for the incidence of motor composite score <85 or <70.The head circumference and weight at 18 months of cGA were similar.Further, in the birth year of 2017 (rate of PDA treatment was 0% at Site 2 [full adoption of conservative policy], compared to 53% at Site 1, Table 2), rate of any NDI and signi cant NDI were similar between sites (any NDI: 38% vs 36%; signi cant NDI: 13% vs 10% -for Site 1 and Site 2, respectively).After adjustments for confounders (Table 3 and Supplementary Table 2), no association was detected between site (as a surrogate to PDA management policy) and the outcome of "any NDI" or "signi cant NDI".Also, no association was detected between exposure to "PDA treatment" (both sites combined) and the outcomes of interest (any NDI or signi cant NDI).We also evaluated the proportion of infants with death or signi cant NDI, as well as death or any NDI, and found no differences by center.However, this data did exclude missing information regarding those infants lost to follow-up.Further, we evaluated these composite outcomes (death or any NDI /death or signi cant NDI) within a multiple logistic regression model adjusting for the above confounders and effect modi ers, and found no association between them and the site exposure.

Discussion
In this cohort, unit-level PDA management policy (conservative vs targeted-exposure) in extremely preterm infants was not associated with signi cant NDI or any NDI.Our results indicated overall higher motor scores averages in the interventional site, although this did not reach clinical signi cance based on cutoff thresholds of 85 and 70.Although the rate of mortality at Site 1 was higher, the composite outcome of death or NDI was similar between sites.
Controversies regarding optimal PDA management in the preterm population exist, and whether neonatal outcomes are in uenced by accelerating ductal closure (31).The most commonly administered medications to accelerate ductal closure remain NSAIDs, although acetaminophen has been increasingly used in the past years (32).In our study, we could not explore the impact of acetaminophen exposure, as this medication had not been implemented during the years of the study.Exposure to NSAIDs has also been associated with side effects, such as oliguria, platelet dysfunction, cerebral vasoconstriction ( 5) and pulmonary anti-angiogenesis phenomenon (33,34).Recently, a randomized controlled trial has shown that an expectant PDA management (de ned as no intentional treatment to accelerate ductal closure with ibuprofen, within 72 hours of postnatal life in preterm infants with a left to right shunting PDA > 1.5 mm) was non-inferior to an early-ibuprofen strategy regarding the combined outcome of NEC, moderate-tosevere BPD or death (35).Authors postulated that early ibuprofen exposure may be harmful, as the incidence of moderate-to-severe BPD was signi cantly higher in their early-ibuprofen exposed group (7).
With the motivation to nd a better strategy to accelerate ductal constriction and mitigate its theoretical effects on systemic and pulmonary circulations, acetaminophen has been proposed but has shown variable e cacy regarding ductal constriction (36).PDA ligation has been associated with adverse neurodevelopmental outcomes, as well as surgical and post-surgical complications (37).Studies are currently ongoing regarding early shunt closure with a catheterized-based intervention (38).
A recent randomized multi-center controlled trial reported on the outcome of survival without CP at 24 months of cGA in < 28 weeks with a large PDA diagnosed between 6 to 12 hours of life and who were exposed to either ibuprofen or placebo (6).Early echocardiography-targeted ibuprofen did not impact the primary outcome.However, in the placebo group, 62% received open-label rescue therapy, raising doubts about the applicability of the results to a strict non-interventional approach.In another retrospective Japanese cohort study, preterm infants born at < 29 weeks and exposed to a conservative policy regarding the PDA were evaluated at 5 years cGA, using the Kaufman Assessment Battery for Children.No increased risk of neurodevelopmental disorders was found at 5 years of corrected age (39).Those with a closed ductus within the rst 14 days of life were found to have similar neurodevelopmental scores when compared to those who achieved ductal closure after 14 days, indicating that early ductal patency duration did not signi cantly impact their outcomes.Furthermore, in a multicenter observational study of 1090 preterm low-birthweight infants, the presence of a PDA was found in 135 infants and was not associated with differences in cognitive or behavioral competences at age 3, 8, and 18 years (40).A report from Spain also evaluated the neurodevelopmental outcomes of premature infants < 29 weeks at birth after their adoption of a conservative PDA policy, in a pre-post study design (41).Although they found a signi cant drop in the exposure to PDA-related management strategies, infants within their conservative epoch were still exposed to medical (32%) or surgical (7%) related therapies.Neurodevelopmental outcomes at 2 years were similar between the two epochs.In their study, a longer PDA duration (> 28 days) was not associated with worst outcomes in terms of survival without neurodevelopmental impairment.Therefore, there is a growing body of evidence in the literature that supports our ndings.Nonetheless, there is still a lack of comparisons between a conservative approach and the exposure to acetaminophen or to a catheter-based closure in terms of long-term follow-up and subsequent outcomes.
When comparing various strategies to accelerate ductal constriction, Oncel et al. did not describe any difference of Bayley II score at 18-24 months cGA of preterm infants born at < 30 weeks exposed to acetaminophen versus ibuprofen (42).Several studies have demonstrated increased rates of neurodevelopmental impairments in preterm infants who underwent PDA ligation (37).One study adjusting for perinatal characteristics and pre-ligation morbidities found no difference in mortality or neurodevelopmental impairment at 18-24 months cGA.This suggests some residual confounding by indication within the previous studies associating ligation to later adverse outcomes (43).Also, Robinson et al. reported, in a retrospective study, that medical and/or surgical PDA treatment in infants born at < 25 weeks GA was an independent risk factor for adverse neurodevelopment at 2 to 3 years of age (44).In addition, a Swedish study evaluating infants born at < 27 weeks described an increased risk of moderate to severe neurodevelopmental impairment and a lower full-scale intelligence quotient at 6.5 years of age in those exposed to a primary PDA surgery approach (45).Finally, a pre-post design retrospective study evaluated the impact of an "early ligation" versus a "selective ligation" approach in infants < 28 weeks exposed to prophylactic indomethacin (46).Infants in the selective ligation group had similar neurodevelopmental outcomes at 18-36 months compared to the historical more "aggressive" approach, outlining that introduction of a progressively increasing selection of patients undergoing ductal closure was not associated with a drastic change in their developmental outcomes.
Considering that studies on PDA treatment in the preterm population were not able to demonstrate meaningful short or long-term bene ts of neonatal outcomes (35), many institutions have opted to progressively adopt a more conservative approach (9,39,47).However, most studies have underrepresented the most immature population of preterm infants, which are those born at < 26 weeks and even those born at < 24 weeks.Recently, Giesinger et al. described improved early outcomes in infants born at 22-23 weeks of GA after the introduction of a selective early PDA treatment (within 24 hours of postnatal life), using an initial approach with acetaminophen (48).Future studies within this vulnerable subgroup should evaluate whether these ndings translate into improved neurodevelopmental outcomes within that high-risk population.
Our study is observational and with a high number of children lost to follow-up or who did not survive to follow-up (n = 91 infants; Supplemental Table ).Differential mortality rates between sites (Site 1: 28% vs Site 2: 15%) may also have brought a survival bias in the analysis.Although we found no association between sites and the composite outcomes of death or any NDI, as well as death or signi cant NDI, these analyses are limited by the lack of information regarding the patients lost to follow-up.In addition, language scores for the Bayley-III assessment were not included, due to a substantial number of unilingual families whose primary language at home was neither French nor English.Moreover, socioeconomic factors may signi cantly in uence the neurodevelopment during infancy, which were not accounted for.Data regarding the duration of ductal patency and magnitude of the left to right shunt were not available.Our database lacked other important information, such as antenatal magnesium sulfate and delayed cord clamping, which have been associated with neuroprotective properties in the preterm population.Also, follow-up beyond 18-24 months may also be important to detect subtle, longerterm impacts on neurodevelopment.Despite these limitations, our study provides some real-context implementation of a conservative approach regarding the PDA management, with a progressive drop to 0% of treatment exposure in the conservative site, with reassuring evidence of the implementation of such policy pertaining to neurodevelopmental outcomes.The neurodevelopmental evaluation was performed by trained personnel, using validated neurodevelopmental tools.Our study was carried in two comparable sites in terms of clinical practices, socioeconomic status of targeted population (same city), organization of care and training/experiences of medical teams.Finally, all data were collected through the local CNN and CNFUN registries, which are both dataset with standardized de nitions across sites (49).
In conclusion, progressive adoption of a conservative PDA policy was not associated with different neurodevelopmental motor and cognitive outcomes in infants born at < 29 weeks, compared with a targeted PDA treatment strategy using mainly NSAIDs.Although our ndings are reassuring, they need to be corroborated in larger studies including infants born before < 26 weeks, as well as at later ages at follow-up.Tables Table 1: Characteristics by Site (2014 to 2017 cohort)

Supplementary Files
This is a list of supplementary les associated with this preprint.Click to download.

Funding
/Support: This project was supported by the Neonatology divisions of Centre Hospitalier Universitaire Sainte-Justine and McGill University Health Centre.The NeoCardioLab and its platforms were funded by the Department of Pediatrics of McGill University, the Just for Kids Foundation, the Foundation of Stars, as well as the Grand De Pierre Lavoie.Role Funder/Sponsor: No Funder/Sponsor for this study.Presentations: The poster relating to this study was presented at: Canadian National Perinatal Research Meeting in 2022 and the 89e Congrès d'ACFAS (Association canadienne-française pour l'avancement des sciences) Data availability: Data for this study cannot be shared publicly as it is bound by con dentiality agreements governed by hospital o ces.Aggregate data may be made available from the corresponding author (GA) on reasonable request.49.Li Ching Ng L, Patel S, Plourde H, Besner ME, Lapointe A, Bizgu V, et al.The association between BMI trajectories and bronchopulmonary dysplasia among very preterm infants.Pediatr Res.2022 Nov;1-7.