Bronchopulmonary dysplasia (BPD) is a severe chronic respiratory illness that develops in over half of infants born extremely premature.¹ BPD predisposes affected infants to earlier mortality and deficits in cardiopulmonary health, neurodevelopment, and quality of life that can persist through childhood and into adulthood.¹ Systemic corticosteroids are one of the few drugs shown in randomized clinical trials or meta-analyses to significantly reduce the risk of developing BPD.² However, use of these potent medications in neonatology is highly controversial due in part to concerns regarding drug safety.

Randomized clinical trials published in the 1980s demonstrated that administration of dexamethasone to preterm infants led to earlier weaning from invasive ventilation and shorter exposure to supplemental oxygen.^3,4 These data prompted enthusiastic use of postnatal corticosteroids to reduce respiratory morbidity and prevent BPD in preterm infants born during the 1990s.⁵ However, a rapid reversal in practice soon followed after publication of postdischarge outcome data from randomized clinical trials showing that dexamethasone increased the risk of neuromotor dysfunction and cerebral palsy (CP).^5,6 Since that time, investigators have sought to identify corticosteroid formulations and treatment strategies that are effective for reducing the risk of BPD and safe for use in very preterm infants. Over the past 20 years, Doyle and colleagues have been responsible for one of the most important milestones towards advancing this goal. In 2005, Doyle and colleagues published an innovative meta-analysis and meta-regression of data from 14 randomized trials of corticosteroids⁷ that included assessment of neurodevelopmental outcomes in early childhood. In that study, they showed for the first time that an infant’s underlying risk of developing BPD at 36 weeks’ postmenstrual age may significantly influence the balance of risks and benefits of systemic corticosteroids. Using the rate of BPD in the trial control group as a measure of baseline disease risk, Doyle and colleagues⁷ demonstrated net harm (ie, increased risk of death or CP) in trials that recruited infants with low average risks of BPD and net benefit (ie, reduced risk of death or CP) in trials that recruited infants with high average risks of BPD. This analysis, which was subsequently updated in 2014,⁸ played an important role in ending the near-embargo on the use of corticosteroids to prevent BPD and provided a route for clinicians to selectively administer corticosteroids to high-risk infants.

In JAMA Pediatrics, Doyle and colleagues⁹ have provided a second update of their original meta-regression. This analysis uses data from 26 randomized clinical trials that include a combined total of 3700 preterm infants. Of these 26 studies, 18 investigated dexamethasone (including 2 studies that are new relative to the 2014 publication) and 8 investigated hydrocortisone (including 6 post-2014 studies). Compared to the prior 2 meta-regressions, this new analysis evaluates the composite outcome of survival without CP, rather than the composite of death or CP, and examines for the first time whether the previously observed relationship between systemic corticosteroids and CP varies between trials that studied dexamethasone and those that studied hydrocortisone. Analogous to Doyle and colleagues’ prior work, this study⁹ plots the rate of BPD as defined in the original trials against the risk difference for survival without CP, calculated as the observed outcome rate in corticosteroid-treated infants subtracted from the observed rate in the untreated controls. Infants who were discharged alive but were not assessed at follow-up were assumed to have survived without CP, a limitation acknowledged by the authors. Individual trials were weighted based on their contribution to the overall analysis, with larger trials providing greater weight in general.

The primary analysis using data from all 26 trials⁹ produced no clear evidence of a relationship between the risk of BPD in the trial control groups and the effect of corticosteroids on survival without CP. However, when stratified by drug type, there was a significant positive relationship between the risk of BPD and the effect of dexamethasone on survival free of CP. For every 10% increase in the risk of BPD, the risk difference for survival without CP increased by 3.8% (95% CI, 1.5%-5.9%). The estimated treatment effect transitioned from potential harm to benefit when the risk of BPD in the trial control groups exceeded 50% (95% CI, 30%-70%), indicating that infants with baseline risks of BPD greater than 50% to 70% may benefit from dexamethasone, while infants with baseline risks below 30% to 50% may be harmed. This treatment effect did not differ significantly among trials of dexamethasone that initiated corticosteroid therapy during vs after the first postnatal week. Hydrocortisone was associated with weak evidence of benefit in infants at low risk of BPD, but unlike with dexamethasone, the differential treatment effect was not statistically significant.

These findings strengthen our understanding that the use of dexamethasone in preterm infants at high risk of BPD may be a safe and effective intervention to improve long-term neurologic outcomes. However, important gaps remain that prevent full evidence-based use of this medication. None of the available randomized clinical trials of dexamethasone were adequately powered to assess long-term neurologic safety. Only 3 of the 18 trials of dexamethasone included in the analysis by Doyle and colleagues⁹ provided data on CP in more than 100 infants, and across all 18 studies, the presence or absence of CP was known in only 979 total infants. The wide range of evaluated doses and treatment durations further complicates clinical application. The reported cumulative doses of dexamethasone administered in the included trials ranged from 0.89 to 7.9 mg per kg, with even higher doses potentially prescribed to some infants who received prolonged invasive ventilation.^3,4 Treatment durations ranged from 1 to 42 days.^3,4 Some experts recommend that the dose and duration of corticosteroids used to prevent BPD should be minimized to mitigate the risks of these drugs.^3,10,11 Yet recent analyses demonstrate that cumulative doses of dexamethasone that exceed 2 mg per kg may be required to significantly reduce the risk of BPD.¹² An important unanswered question is whether prevention of BPD, which may require higher doses of dexamethasone than are commonly used in current practice, is necessary to improve neurologic outcomes with corticosteroids or if more subtle respiratory benefits (eg, earlier extubation, shorter duration of positive airway pressure) observed in trials of lower-dose dexamethasone will result in long-term developmental gains.

Meta-analyses of randomized clinical trials demonstrate that initiation of dexamethasone during the first 7 days after birth in preterm infants increases the risk of CP, but that initiation after the first week does not.^3,4 As such, the finding by Doyle and colleagues⁹ that selective administration of dexamethasone during the first postnatal week to infants at high risk of BPD may improve survival without CP is particularly notable. This novel result should not reverse the long-standing recommendation against the use of early dexamethasone in preterm infants but should encourage equipoise and enthusiasm for a contemporary randomized clinical trial of dexamethasone that is appropriately powered to assess for small but meaningful differences in neurodevelopment.

The average treatment effect observed in randomized clinical trials indicates that hydrocortisone does not alter the risks of BPD or adverse neurodevelopment in preterm infants.^3,4,13 The results from Doyle and colleagues, which are similar to other recent reports, further clarify that hydrocortisone is unlikely to significantly increase survival without CP if selectively administered to preterm infants with either high or low risk of BPD.^14,15 However, treating extremely preterm infants with hydrocortisone beginning in the first 24 to 48 postnatal hours may decrease mortality.¹⁶ This potential benefit has prompted calls for additional trials of prophylactic hydrocortisone in this vulnerable population.

Systemic corticosteroids are one of the most widely used drugs in neonatal medicine. Although their true risks and benefits to prevent BPD remain uncertain, the work by Doyle and colleagues⁹ continues to encourage critical appraisal of how the safety and efficacy of these drugs and other therapies may depend on a patient’s underlying risks for important clinical outcomes. Such efforts will hopefully lead to better patient outcomes in neonatology through more accurate prediction of treatment harms and benefits and development of evidence-based individualized care plans.

Corresponding Author: Erik A. Jensen, MD, MSCE, Division of Neonatal and Developmental Medicine, Department of Pediatrics, Stanford University School of Medicine, 453 Quarry Rd, MC 5560, Palo Alto, CA 94304 (jensene@stanford.edu).

Published Online: November 18, 2024. doi:10.1001/jamapediatrics.2024.4572

Conflict of Interest Disclosures: None reported.