Publication

• Title: Erythropoietin for Neonatal Hypoxic-Ischemic Encephalopathy: A Randomized Clinical Trial
• Acronym: PAEAN — Preventing Adverse Outcomes of Neonatal Hypoxic Ischaemic Encephalopathy with Erythropoietin
• Year: 2026
• Journal published in: JAMA Pediatrics
• Citation: Liley HG, Hunt RW, O’Connell RL, et al. Erythropoietin for neonatal hypoxic-ischemic encephalopathy: a randomized clinical trial. JAMA Pediatr. Published online July 27, 2026.

Context & Rationale

• Background

  • Moderate or severe neonatal hypoxic-ischaemic encephalopathy is an important cause of neonatal death, cerebral palsy, epilepsy, cognitive impairment and lifelong disability.
  • Therapeutic hypothermia initiated within 6 hours of birth and continued for 72 hours improves survival without major neurodevelopmental disability, but a substantial residual burden remains.
  • Erythropoietin has erythropoietic, anti-apoptotic, anti-inflammatory, antioxidant, angiogenic and neuroregenerative effects, with erythropoietin receptors expressed within the developing central nervous system.
  • Extensive animal work and several small early clinical studies suggested that repeated high-dose erythropoietin might reduce brain injury and improve neurodevelopment.
  • Those encouraging findings required confirmation in a sufficiently large trial in which all infants received contemporary standard therapeutic hypothermia.
• Research Question/Hypothesis

  • Does repeated high-dose intravenous erythropoietin, added to therapeutic hypothermia, reduce death or moderate or severe motor or cognitive developmental deficit at 2 years in term or near-term infants with moderate or severe hypoxic-ischaemic encephalopathy?
  • The investigators hypothesised that erythropoietin would provide neuroprotection additional to that achieved by hypothermia, without causing important adverse effects.
• Why This Matters

  • There is no routinely recommended pharmacological adjunct to therapeutic hypothermia for neonatal hypoxic-ischaemic encephalopathy.
  • The outcome of death or important neurodevelopmental disability is directly relevant to children, families and healthcare systems.
  • PAEAN was deliberately developed in collaboration with the HEAL investigators, using closely aligned eligibility criteria and treatment regimens, thereby providing an independent phase 3 test of a biologically compelling therapy.

Design & Methods

• Research Question: In infants with moderate or severe hypoxic-ischaemic encephalopathy receiving therapeutic hypothermia, does repeated high-dose erythropoietin improve survival without moderate or severe neurodevelopmental deficit at 2 years compared with placebo?
• Study Type: Investigator-initiated, international, phase 3, multicentre, parallel-group, superiority, randomised, double-blind, placebo-controlled trial conducted in 24 neonatal intensive care units: 18 in Australia, 5 in New Zealand and 1 in Singapore.
• Population:
  • A total of 313 infants were randomised between May 2016 and March 2021: 156 to erythropoietin and 157 to placebo.
  • Infants were eligible if born at ≥35+0 weeks’ gestation and able to be randomised before 23 hours of age.
  • Perinatal depression required at least one of the following: 10-minute Apgar score ≤5; continuing positive-pressure ventilation, continuous positive airway pressure or chest compressions at 10 minutes; or cord or infant blood obtained within 60 minutes showing pH <7.00 or base deficit ≥12 mmol/L.
  • Moderate or severe encephalopathy was defined between 1 and 6 hours after birth by at least 3 of 6 modified Sarnat criteria in the moderate or severe range, or by 2 criteria plus seizures requiring anticonvulsant treatment before randomisation.
  • Controlled whole-body therapeutic hypothermia had to begin by 6 hours of age and be planned for 72 hours, with controlled rewarming.
  • Important exclusions were birth weight <1800 g, suspected major chromosomal or congenital abnormality, head circumference below the 3rd centile, contraindication to erythropoietin, planned use of another erythropoiesis-stimulating agent, or consideration of imminent withdrawal of life-sustaining treatment.
• Intervention:
  • Intravenous epoetin alfa 1000 IU/kg, capped at 4000 IU, infused over 5 minutes.
  • Five doses were scheduled: the first within the first day after birth, followed as closely as possible by doses 24, 48, 96 and 144 hours after the first dose.
  • Treatment ceased following death, a decision to withdraw life-sustaining treatment, or discharge before completion of the five-dose course.
  • A protocol amendment allowed the final dose to be administered at least 24 hours early when removal of intravenous access, transfer or discharge was imminent.
  • All infants received 72 hours of whole-body hypothermia targeting 33.5°C, followed by controlled rewarming.
• Comparison:
  • An equivalent volume of intravenous 0.9% sodium chloride, packaged in identical numbered vials and administered according to the same five-dose schedule.
  • Infants received the same therapeutic hypothermia strategy and other neonatal intensive care determined by the treating clinical team.
  • Off-protocol erythropoietin or another erythropoiesis-stimulating agent was prohibited during the first 2 weeks of life.
• Blinding: Parents, treating clinicians, research personnel and investigators were blinded to allocation. Erythropoietin and placebo were repackaged in identical numbered vials, and allocation remained concealed until treatment and follow-up data had been retrieved. The committee adjudicating incompletely assessed neurodevelopmental outcomes was also blinded.
• Statistics: A total of 300 infants, 150 per group, was required to detect a 19-percentage-point absolute reduction in death or moderate or severe developmental deficit, from 46% to 27%, with 90% power, a two-sided alpha of 0.05 and allowance for 10% non-adherence or loss to follow-up. The primary comparison used unadjusted log-binomial regression to estimate a relative risk. Efficacy analyses were described as intention-to-treat but were necessarily available-case analyses because infants without ascertainable outcomes were excluded; safety analyses included infants who received at least one study dose. Secondary and tertiary analyses were exploratory and were not adjusted for multiple comparisons.
• Follow-Up Period: The primary outcome was assessed at 2 years. The original 22–26-month assessment window was extended to 36 months because of the COVID-19 pandemic; when formal assessments were unavailable, blinded adjudication could use clinical and parent-reported information collected up to 43 months.

Key Results

This trial was not stopped early. Predefined reviews of 30-day safety data were undertaken after approximately 25%, 50% and 75% of planned recruitment; no stopping boundary was crossed.

• The prespecified 19-percentage-point absolute improvement was not observed; the primary point estimate favoured placebo and its confidence interval remained compatible with both modest benefit and clinically important harm.
• Mortality, cerebral palsy, motor deficit, cognitive deficit, continuous Bayley scores, early encephalopathy and the ordinal developmental outcome were all concordant with the primary finding.
• No common erythropoietin-specific toxicity was identified, but the trial had insufficient power to determine the effect on uncommon events such as major-vessel thrombosis.

Internal Validity

• Randomisation and Allocation: Randomisation was central, allocation was 1:1 and balancing incorporated study site and baseline modified Sarnat severity. Identical numbered vials were issued only after baseline information had been entered, providing strong allocation concealment.
• Dropout or Exclusions: One infant assigned to erythropoietin did not receive study treatment following study withdrawal. Two-year mortality status was confirmed for 146/156 infants assigned to erythropoietin and 149/157 assigned to placebo. The primary outcome was available for 138/156 and 143/157, leaving 18 and 14 infants respectively without a primary outcome. Overall missingness of 10.2% was balanced, but the available-case primary analysis cannot completely eliminate attrition bias.
• Performance/Detection Bias: Parents, clinicians, research personnel and investigators remained blinded, and placebo administration closely reproduced active treatment. Death was objective, formal developmental testing used standardised instruments and cases requiring clinical adjudication were reviewed by a blinded committee.
• Protocol Adherence: The first dose was given to 154/156 infants assigned to erythropoietin and 157/157 assigned to placebo. Corresponding treatment completion on days 2, 3, 5 and 7 was 96.8% versus 99.4%, 93.5% versus 97.4%, 85.7% versus 89.3%, and 56.3% versus 56.2%. Only 76/156 (48.7%) and 81/157 (51.6%) received all five doses, although more than 99% of administered doses were within 90% of the specified dose.
• Baseline Characteristics: Groups were well matched. Mean gestational age was 39.4 ± 1.7 versus 39.3 ± 1.6 weeks; moderate encephalopathy was present in 77.6% versus 77.1%; severe encephalopathy in 22.4% versus 22.9%; mean early blood pH was 6.94 ± 0.18 versus 6.92 ± 0.18; and mean 10-minute Apgar score was 4.6 ± 2.4 versus 4.8 ± 2.2.
• Heterogeneity: Multicentre recruitment increased clinical diversity, but all participants met structured encephalopathy criteria and received a standardised hypothermia strategy. Randomisation was balanced by severity and site. Recruitment nevertheless varied markedly between centres, and only 5 infants were enrolled in Singapore.
• Timing: Therapeutic hypothermia began at a median of 1.5 hours after birth in the erythropoietin group and 1.8 hours in the placebo group. The first study dose was administered much later, at a mean of 18.3 versus 18.7 hours; observed ranges extended to 24.5 and 28.0 hours. This met the intended pragmatic first-day strategy for most infants, but may have missed an earlier acute neuroprotective window.
• Dose: The 1000-IU/kg regimen was selected from neonatal pharmacokinetic work intended to reproduce plasma exposure associated with benefit in animal models. Administered doses were highly accurate, although the optimal central nervous system exposure, the effect of the 4000-IU cap and the importance of later regenerative dosing remain uncertain.
• Separation of the Variable of Interest: Active erythropoietin was prohibited in the placebo group and no crossover was reported. Pharmacological separation was therefore clear for administered doses, although exposure to the complete five-dose course occurred in only 48.7% versus 51.6% of participants.
• Key Delivery Aspects: Whole-body hypothermia targeting 33.5°C for 72 hours was used throughout. Median time from starting cooling to target temperature was 2.0 hours in the erythropoietin group and 1.5 hours in the placebo group; servo-controlled devices were used in 85.3% and 81.5% respectively.
• Adjunctive Therapy Use: No clinically important between-group differences were reported in respiratory support, inspired oxygen, anticonvulsant therapy, inotropic support or relevant haematological, renal and coagulation measurements during treatment. There was therefore no evidence that differential co-intervention obscured the treatment effect.
• Outcome Assessment: The composite combined the competing outcome of death with clinically important motor and cognitive deficits. Standardised neurological and Bayley assessments were appropriate, but pandemic-related extension of the assessment window and blinded adjudication from alternative clinical information introduced some measurement heterogeneity.
• Statistical Rigor: The principal model was appropriate for estimating a relative risk. The primary test retained a two-sided alpha of 0.05, while secondary analyses were explicitly exploratory and unadjusted for multiplicity. The lower-than-anticipated control event rate reduced precision, and the principal publication did not report the planned adjusted and subgroup analyses.

Conclusion on Internal Validity: Internal validity is moderate to strong. Randomisation, concealment, blinding, baseline balance and concordant outcomes were robust, but confidence is tempered by 10.2% missing primary outcomes, incomplete delivery of the five-dose course, late initiation relative to the acute injury, changes in outcome definitions and an available-case rather than complete intention-to-treat analysis.

External Validity

• Population Representativeness: The cohort represents term and near-term infants with clearly defined moderate or severe hypoxic-ischaemic encephalopathy receiving contemporary therapeutic hypothermia in specialist neonatal intensive care. Approximately 58% were transferred from another hospital, supporting applicability to regional referral systems.
• Severity: Approximately 77% had moderate and 23% severe encephalopathy. The placebo primary-event rate of 28.7% suggests better outcomes than those assumed from older hypothermia trials and may reflect earlier recognition, broader identification of moderate encephalopathy and improvements in neonatal intensive care.
• Excluded Populations: The findings do not directly apply to mild encephalopathy, gestation below 35 weeks, birth weight below 1800 g, major congenital or chromosomal abnormalities, microcephaly, infants presenting too late for hypothermia, or infants for whom imminent withdrawal of intensive care was being considered.
• Healthcare Setting: Australia and New Zealand contributed 98% of participants. Results are highly applicable to high-resource neonatal systems with rapid retrieval, servo-controlled cooling, neuroimaging and formal developmental follow-up, but less directly applicable to low-resource settings.
• Intervention Feasibility: Five intravenous doses over 7 days require ongoing vascular access. The low proportion receiving every dose illustrates the practical difficulty of preserving intravenous access in recovering infants solely for an investigational treatment.
• Applicability: The trial directly addresses adjunctive erythropoietin in cooled infants. It does not determine whether erythropoietin monotherapy might have a different effect where safe therapeutic hypothermia is unavailable.

Conclusion on External Validity: External validity is strong for term and near-term infants with moderate or severe hypoxic-ischaemic encephalopathy treated with therapeutic hypothermia in high-resource neonatal intensive care systems. Generalisability is limited for mild disease, preterm infants, late presenters, uncooled infants and low-resource settings.

Strengths & Limitations

• Strengths:
  • International, multicentre, phase 3 design.
  • Central randomisation, concealed allocation, identical placebo and comprehensive blinding.
  • Clinically meaningful primary outcome assessed after sufficient time for neurodevelopmental deficits to become apparent.
  • Structured eligibility criteria and balancing by site and encephalopathy severity.
  • High accuracy of administered doses and comparable co-interventions.
  • Independent safety monitoring with prespecified stopping rules.
  • Blinded adjudication salvaged outcome information when pandemic restrictions prevented standard assessments.
  • Close alignment with the independent HEAL trial permits unusually strong replication of the principal efficacy finding.
• Limitations:
  • The observed placebo event rate was 28.7%, substantially below the assumed 46%, reducing power to detect smaller but potentially relevant effects.
  • Primary outcome data were unavailable for 32/313 infants, and the principal analysis excluded these participants.
  • Only approximately half of each group received all five scheduled doses.
  • The first dose was administered at a mean of approximately 18 hours, potentially after the most responsive phase of secondary energy failure.
  • The primary motor and cognitive thresholds changed between the original protocol and the final statistical analysis plan.
  • Formal assessment ages and methods became more heterogeneous because of the COVID-19 pandemic.
  • The sample was inadequate to establish or exclude uncommon adverse effects.
  • Planned adjusted and subgroup analyses were not presented in the principal report or supplied supplemental results.
  • Almost all participants were recruited in Australia and New Zealand.
  • Magnetic resonance imaging, amplitude-integrated electroencephalography, pharmacokinetic, biomarker and General Movements substudies were not reported with the principal clinical results.

Interpretation & Why It Matters

• Clinical Practice

  High-dose erythropoietin should not be routinely added to therapeutic hypothermia for term or near-term infants with moderate or severe hypoxic-ischaemic encephalopathy.
• What PAEAN Resolves

  PAEAN provides a rigorous phase 3 test of a therapy supported by extensive biological rationale, animal studies and encouraging early clinical data. It found no evidence that erythropoietin adds clinically meaningful neuroprotection to modern therapeutic hypothermia.
• Consistency of Evidence

  The absence of benefit across mortality, cerebral palsy, cognitive impairment, motor impairment, continuous Bayley scores and early neurological outcomes makes an isolated failure of the composite endpoint unlikely.
• Safety

  PAEAN did not demonstrate a definite safety problem, but the absence of efficacy means that even uncertain uncommon harms, additional intravenous access and treatment burden weigh against clinical use.
• What Remains Unanswered

  The trial does not determine whether substantially earlier administration, alternative dosing, biomarker-selected treatment, use in mild encephalopathy, or erythropoietin monotherapy where hypothermia is unavailable could have different effects.
• Translational Lesson

  Mechanistic plausibility and animal neuroprotection were insufficient predictors of incremental clinical benefit once an effective background therapy was present. Adjunctive therapies must demonstrate added benefit over therapeutic hypothermia rather than benefit over untreated experimental injury.

Controversies & Other Evidence

• Evolution of the primary outcome definition: The 2016 protocol defined moderate or severe motor deficit as cerebral palsy with Gross Motor Function Classification Scale level ≥1 and cognitive deficit as a Bayley-III cognitive score ≤80. The final 2024 statistical analysis plan instead used cerebral palsy with Gross Motor Function Classification Scale level ≥2 and a cognitive score <85, which became the published definitions.¹² The motor change made that component less inclusive, while the cognitive change made it more inclusive. The net effect on the primary result was not reported, and no analysis using the original thresholds was presented.
• Intention-to-treat terminology and missing outcomes: The article described efficacy analyses as intention-to-treat, whereas the statistical analysis plan specified a modified intention-to-treat population containing participants with available outcomes. Consequently, 32 randomised infants were absent from the primary comparison.²³ The post hoc tipping-point analysis was reassuring: even assigning every missing erythropoietin outcome as favourable and every missing placebo outcome as unfavourable produced RR 0.86; 95% CI 0.62 to 1.18; P=0.356. A statistically significant estimate of harm required imputing adverse outcomes to 15/18 missing erythropoietin participants while assuming placebo outcomes were missing at random.
• Power and the contemporary control event rate: The study was designed around a very large improvement from 46% to 27%. The actual placebo event rate was 28.7%, close to the originally hypothesised erythropoietin rate. PAEAN therefore convincingly rejects the anticipated large effect but is less precise for smaller benefit, modest harm or uncommon adverse events.
• Independent replication by HEAL: HEAL randomised 501 cooled infants with moderate or severe hypoxic-ischaemic encephalopathy to a closely related five-dose erythropoietin regimen or placebo. Death or neurodevelopmental impairment occurred in 52.5% versus 49.5%; RR 1.03; 95% CI 0.86 to 1.24; P=0.74. The mean number of serious adverse events per infant was higher with erythropoietin, 0.86 versus 0.67; RR 1.26; 95% CI 1.01 to 1.57.⁴ Differences in developmental thresholds prevent a simple comparison of absolute event rates, but two independent, double-blind phase 3 trials provide compelling evidence against clinically important adjunctive benefit.
• Was inadequate exposure responsible? A pharmacokinetic analysis of 89 erythropoietin-treated HEAL participants found that more than 95% achieved the prespecified 48-hour and 7-day plasma exposure targets; higher exposure was not associated with serious adverse events.⁵ This makes inadequate systemic exposure an unlikely explanation, although neither trial established the optimal drug concentration within injured human brain tissue.
• Timing remains biologically debatable: The mean first dose in both PAEAN and HEAL was administered more than 16 hours after birth. Earlier dosing might affect acute excitotoxic, oxidative or inflammatory injury differently. However, erythropoietin was also proposed to promote delayed repair, and the absence of any signal across repeated doses and multiple outcomes substantially weakens the timing explanation.
• Additivity with hypothermia was never assured: Near-term fetal sheep experiments found that erythropoietin and hypothermia each affected white-matter injury pathways but produced overlapping anti-inflammatory and anti-apoptotic effects, with little additive protection from the combination.⁶ Therapeutic hypothermia may therefore already modify much of the pathway targeted by erythropoietin.
• Unreported planned analyses: The final statistical analysis plan prespecified adjusted sensitivity analyses incorporating encephalopathy severity and geographical site or region, together with treatment-interaction analyses by baseline Sarnat severity and site.² These analyses were not presented in the principal article or supplied supplemental results. Given the modest sample and small severe-HIE subgroup, such analyses would remain exploratory rather than overturning the overall result.
• Rare thrombosis remains uncertain: PAEAN recorded a numerical excess of major-vessel thrombosis. Its post hoc fixed-effect meta-analysis of PAEAN, HEAL and NEATO found 10/435 events with erythropoietin and 5/425 with placebo; RR 1.96; 95% CI 0.68 to 5.70; P=0.21; I²=0%.³ The estimate is too imprecise to establish harm, but it cannot provide reassurance about a potentially doubled risk.
• Meta-analytic conclusions have depended on study selection: A 2024 review of 7 studies and 903 infants found no reduction in death or neurodisability, RR 0.68; 95% CI 0.43 to 1.09; P=0.11, and no reduction in cerebral palsy, RR 0.68; 95% CI 0.33 to 1.40; P=0.30.⁷ A separate review of 12 studies and 1562 infants reported lower mortality with high-dose erythropoietin plus hypothermia, OR 0.65; 95% CI 0.47 to 0.91; P=0.01, but its search ended in May 2024 and therefore did not include PAEAN; it also pooled smaller and methodologically heterogeneous studies with the definitive HEAL trial.⁸ PAEAN materially shifts the highest-quality evidence towards no adjunctive benefit.
• Current standard care remains therapeutic hypothermia: The 2026 American Academy of Pediatrics clinical report recommends therapeutic hypothermia at 33.5–34.5°C, initiated within 6 hours and continued for 72 hours, for eligible infants born at ≥36 weeks.⁹ The report preceded publication of PAEAN; PAEAN strengthens the case that erythropoietin should not be incorporated as a routine adjunct.
• Erythropoietin without hypothermia is a separate question: A review of 5 small studies involving 348 infants in low- and middle-income countries reported death or neurodisability in 27.6% with erythropoietin monotherapy versus 49.7% with comparison treatment; RR 0.56; 95% CI 0.42 to 0.75.¹⁰ Only 3 studies contributed to that outcome, sample sizes were small and risk of bias was substantial. Those data neither contradict PAEAN nor justify extrapolating PAEAN to settings in which therapeutic hypothermia is unavailable.

Summary

• PAEAN randomised 313 infants with moderate or severe hypoxic-ischaemic encephalopathy receiving therapeutic hypothermia to five doses of intravenous erythropoietin 1000 IU/kg or placebo.
• Death or moderate or severe developmental deficit occurred in 34.1% with erythropoietin and 28.7% with placebo; RR 1.19; 95% CI 0.84 to 1.68; P=0.33.
• There was no evidence of benefit for mortality, cerebral palsy, motor deficit, cognitive deficit, Bayley scores, early encephalopathy or the overall ordinal distribution of outcomes.
• No definite common toxicity was demonstrated, although uncommon harms, including major-vessel thrombosis, were too infrequent to assess reliably.
• Despite some missing outcomes, incomplete five-dose treatment, relatively late first dosing and changes in endpoint definitions, PAEAN and HEAL together provide strong evidence against routine adjunctive erythropoietin in cooled infants.

Overall Takeaway

PAEAN is an important, practice-defining confirmatory phase 3 trial showing that repeated high-dose erythropoietin does not improve survival or neurodevelopment when added to therapeutic hypothermia for moderate or severe neonatal hypoxic-ischaemic encephalopathy. Its concordance with HEAL closes a major therapeutic avenue and reinforces the need to test plausible neuroprotective therapies against, rather than in the absence of, effective contemporary standard care.

Bibliography

• 1.Liley HG, O’Connell RL, PAEAN Study Investigators. Preventing Adverse Outcomes of Neonatal Hypoxic Ischaemic Encephalopathy with Erythropoietin: a phase III randomised placebo-controlled multicentre clinical trial. Protocol version 2.0. February 15, 2016. Supplement 1 to: Liley HG, et al. JAMA Pediatr. Published online July 27, 2026.
• 2.O’Connell RL, Liley HG. PAEAN statistical analysis plan. Version 3.0. September 6, 2024. Supplement 2 to: Liley HG, et al. JAMA Pediatr. Published online July 27, 2026.
• 3.Liley HG, Hunt RW, O’Connell RL, et al. Supplemental online content for: Erythropoietin for neonatal hypoxic-ischemic encephalopathy: a randomized clinical trial. JAMA Pediatr. Published online July 27, 2026.
• 4.Wu YW, Comstock BA, Gonzalez FF, et al; HEAL Consortium. Trial of erythropoietin for hypoxic-ischemic encephalopathy in newborns. N Engl J Med. 2022;387(2):148-159.
• 5.Frymoyer A, Vasconcelos AG, Juul SE, Comstock BA, Heagerty PJ, Wu YW. On target dosing: erythropoietin exposure in neonates with hypoxic-ischemic encephalopathy in the HEAL trial. Pediatr Res. 2025;98(1):218-223.
• 6.Wassink G, Davidson JO, Crisostomo A, et al. Recombinant erythropoietin does not augment hypothermic white matter protection after global cerebral ischaemia in near-term fetal sheep. Brain Commun. 2021;3(3):fcab172.
• 7.Marsia S, Kumar D, Raheel H, et al. Evaluating the safety and efficacy of erythropoietin therapy for neonatal hypoxic-ischemic encephalopathy: a systematic review and meta-analysis. Pediatr Neurol. 2024;152:4-10.
• 8.He M, Zhao Y, Jiang J, Mo W, Yao Q. Efficacy and safety of high-dose erythropoietin combined with hypothermia therapy in neonatal hypoxic-ischemic encephalopathy: a systematic review and meta-analysis. Mol Biotechnol. 2026;68:987-997.
• 9.Zanelli SA, Wusthoff CJ, Lucke AM, Kaufman DA; Committee on Fetus and Newborn; Section on Neurology. Therapeutic hypothermia for neonatal hypoxic-ischemic encephalopathy: clinical report. Pediatrics. 2026;157(2):e2025073627.
• 10.Ivain P, Montaldo P, Khan A, et al. Erythropoietin monotherapy for neuroprotection after neonatal encephalopathy in low-to-middle income countries: a systematic review and meta-analysis. J Perinatol. 2021;41:2134-2140.