Inhaled Corticosteroids for Bronchopulmonary Dysplasia: A Meta-analysis
CONTEXT: Bronchopulmonary dysplasia (BPD) in preterm infants remains a major health burden despite many therapeutic interventions. Inhaled corticosteroids (IC) may be a safe and effective therapy.
OBJECTIVE: To assess the safety and efficacy of IC for prevention or treatment of BPD or death in preterm infants.
DATA SOURCES: PubMed, the Cochrane Library, Embase, and CINAHL from their inception until November 2015 together with other relevant sources.
STUDY SELECTION: Randomized controlled trials of ICs versus placebo for either prevention or treatment of BPD.
DATA EXTRACTION: This meta-analysis used a random-effects model with assessment of quality of evidence using the Grading of Recommendations Assessment, Development and Evaluation (GRADE) system.
RESULTS: Thirty-eight trials were identified, and 16 met inclusion criteria. ICs were associated with a significant reduction in death or BPD at 36 weeks’ postmenstrual age (risk ratio [RR] = 0.86, 95% confidence interval [CI] 0.75 to 0.99, I2 = 0%, P = .03; 6 trials, n = 1285). BPD was significantly reduced (RR = 0.77, 95% CI 0.65 to 0.91, I2 = 0%, 7 trials, n = 1168), although there was no effect on death (RR = 0.97, 95% CI 0.42 to 2.2, I2 = 50%, 7 trials, n = 1270). No difference was found for death or BPD at 28 days’ postnatal age. The use of systemic steroids was significantly reduced in treated infants (13 trials, n = 1537, RR = 0.87, 95% CI 0.76 to 0.98 I2 = 3%,). No significant differences were found in neonatal morbidities and other adverse events.
LIMITATIONS: Long-term follow-up data are awaited from a recent large randomized controlled trial.
CONCLUSIONS: Very preterm infants appear to benefit from ICs with reduced risk for BPD and no effect on death, other morbidities, or adverse events. Data on long-term respiratory, growth, and developmental outcomes are eagerly awaited.
- BPD —
- bronchopulmonary dysplasia
- CI —
- confidence interval
- IC —
- inhaled corticosteroids
- IVH —
- intraventricular hemorrhage
- PDA —
- patent ductus arteriosus
- PMA —
- postmenstrual age
- PNA —
- postnatal age
- PRISMA —
- Preferred Reporting Items for Systematic Reviews and Meta-analyses
- RCT —
- randomized controlled trial
- RD —
- risk difference
- RR —
- risk ratio
Bronchopulmonary dysplasia (BPD) remains a major cause of mortality and early morbidity in extremely low birth weight infants, with a concomitant increase in later neurodevelopmental impairment.1 The pathogenesis of BPD includes, but is not limited to, inflammatory processes within the immature lung.2–4 Because of their antiinflammatory properties, corticosteroids have been and still are widely used for both the prevention and treatment of BPD in preterm infants. Early systemic use of corticosteroids leads to a significant reduction in BPD but is unfortunately also associated with significant adverse effects on growth and neurodevelopmental outcome with increased incidence of cerebral palsy.5 Inhaled administration is an attractive alternative that may offer clinical efficacy without incurring adverse effects. Over more than 2 decades, this intervention has been studied in a number of randomized controlled trials (RCTs).6,7 These studies used a variety of drugs, at a wide range of doses and that were administered over assorted periods of time, representing any 1 of early “prevention,” later “treatment,” or administration over a prolonged period covering both options. In addition, the combined samples have been insufficient to establish either safety or clear efficacy. A large recent multicenter trial may significantly alter the findings of the previous published meta-analyses.8
Moreover, previous meta-analyses have attempted to separate studies of prevention and treatment, despite significant overlap in ages at administration of the inhaled corticosteroid (IC). Shah et al reported a meta-analysis of studies of “early” postnatal ICs, defined as administration that was started before age 2 weeks. This analysis included 7 trials with 492 infants, although the primary and secondary outcome variables were only reported in 5 of the 7 trials including 429 infants. The duration of the study intervention varied from 10 days to 4 weeks. Onland et al reported a meta-analysis of “late” studies defined as treatment starting after age 7 days. Because the entry criteria of these 2 analyses overlapped and both included studies that started between age 1 and 2 weeks, 1 study met inclusion criteria for both analyses. The “late” analysis included 8 studies and 232 infants. However, most of the main outcome variables were reported in only a small number of the studies with few infants providing data.
In addition to these 2 meta-analyses, there are 2 additional analyses comparing inhaled and systemic steroids for either prevention or treatment of BPD.9,10 However, it is not possible to include the inhaled arm of these studies in a meta-analysis of early or late inhaled steroids as there is no appropriate control group.
Accordingly, this up-to-date systematic review and meta-analysis offers useful information for practitioners considering the role of inhaled steroids for all preterm infants at risk for BPD. In fact, this systematic review adds a single, large study that includes more infants than all the previous studies together, and by combining both approaches, prevention and treatment, and including preterm infants from birth onward, the larger sample size helps answer this important question.
This systematic review and meta-analysis was performed in accordance with our published protocol that was prepared according to the 2015 Preferred Reporting Items for Systematic Reviews and Meta-Analyses—Protocol (PRISMA-P) guideline of 2015 and reported according to PRISMA guidelines (2009).11–13 The objective was to assess the efficacy and safety of ICs administered to preterm infants for either the prevention or treatment of BPD while including all relevant RCTs.
Criteria for inclusion of an RCT were as follows: (1) preterm infants of gestational age 22 0/7 to 36 6/7 weeks considered to be at risk for BPD, including both ventilated and nonventilated infants and (2) an RCT comparing any IC versus control (placebo or no treatment) at any dose and any duration of treatment and administered either by a metered-dose inhaler or by nebulization. Excluded were trials either of systemic corticosteroids or corticosteroids in liquid form administered by direct tracheal instillation.
The primary outcomes included the composite outcome of mortality or BPD defined as requirement for supplemental oxygen at 28 days PNA or 36 weeks’ postmenstrual age (PMA), together with the individual components of the composite outcome. Secondary outcomes included surrogate measures of respiratory insufficiency, such as duration of mechanical ventilation or supplemental oxygen and rescue administration of systemic steroids, in addition to neonatal morbidities and adverse events attributable to the interventions (Fig 1).
The literature search (December 1, 2015) was conducted by using PubMed.gov of the US National Library of Medicine (Medline from 1966 onwards), the Cochrane Library, Embase (from 1974), and CINAHL (from 1982) (see Supplemental Information). In addition, we searched the trial registers www.clinicaltrials.gov, www.controlled-trials.com, and www.who.int/trialsearch, as well as lists of references from relevant studies and abstracts from the proceedings of relevant academic meetings, including the Pediatric Academic Societies and the European Society for Pediatric Research. Search strategies are shown in Supplemental Information.
References identified via the literature search were screened by 2 of the authors (I.P., T.K.), and data were extracted independently to standardized data extraction forms, entered into Review Manager (RevMan 5.3) by 1 of the reviewers, and checked by a second reviewer. Discrepancies were resolved by group discussion. Extracted data included the characteristics of the study and its population, description of the intervention and comparisons, outcome measures and measurements tools and results. The risk of bias of included studies was assessed using the Cochrane risk of bias tool, including sequence generation, allocation concealment, blinding (of participants, personnel, and outcome assessors), incomplete outcome data, selective outcome reporting, and other sources of bias for the RCTs. Accordingly, these items are described as having a “low,” “high,” or “unclear” risk of bias.14
The quality of the evidence was assessed according to the Grading of Recommendations Assessment, Development and Evaluation (GRADE) approach for the suggested 5 criteria for downrating our confidence in effects estimates (risk of bias, inconsistency, imprecision, indirectness, and publication bias) and the 3 criteria for uprating our confidence (large effect, dose-response gradient, and opposing confounding). On the basis of these criteria, the quality of evidence judgment can range from very low (+) to high (++++).15
Measures of Treatment Effect
The meta-analysis used a random effects model as primary analysis to estimate treatment effects. The treatment effects for dichotomous outcomes are expressed as a risk ratio (RR) with 95% confidence intervals (CI). Continuous outcomes, such as duration of oxygen therapy, mechanical ventilation, and hospitalization, were reported in different formats across the studies, and therefore meta-analysis is restricted to a subset of studies, and additional information is provided regarding studies that could not be included in the meta-analysis. Supplemental information was obtained from authors where required. Certain authors (n = 2) could not be contacted, despite approaches via assorted media, and in these cases, previously published information from the Cochrane analyses was used.6,7
The authors assessed the likelihood of publication bias as low after an extensive literature search process that included clinical trials registries, together with visual assessment that did not reveal marked asymmetry of funnel plots.
Thirty-eight trials were initially identified with 16 trials meeting inclusion criteria (see Table 1 and Supplemental Figure 3 and Supplemental Table 3).8,16–30 Thirteen studies were excluded because they used a nonrandomized design. Nine studies did not meet inclusion criteria for the following reasons: randomization after 36 weeks’ PMA in 2, comparison of systemic versus inhaled steroid treatment in 5, direct intratracheal steroid instillation in 1, and publication of substudy from an included study in 1. Overall, the 16 studies included 1596 infants, 804 in the active intervention groups, and 792 in the control groups. More than half of the infants were studied in 1 RCT,8 and together 4 of the 16 RCTs comprised 79% of the overall sample.8,19,23,28
Description of Studies
All 16 studies were RCTs, 13 were published as full reports, and 3 were published only as abstracts.16,21,25 The ICs studied included beclomethasone (6), budesonide (4), fluticasone (3), flunisolide (2), and dexamethasone (1) (see Supplemental Information). The start of the intervention varied from day 1 of life to age 60 days, although all but 2 of the studies began therapy within the first 2 weeks of life. In addition, the duration of the study interventions ranged from 1 week to a potential maximum of 9 weeks in 1 study that included infants from 23 weeks’ gestational age and continued therapy until 32 weeks or weaning from oxygen and ventilation. However, in practice, the mean duration of the intervention in this study was 33 days. Eleven of the 16 studies included only ventilated infants, 4 added infants on nasal ventilator support, and 1 study also included infants receiving only supplemental oxygen. Full details of the studies are to be found in the supplementary online material (Supplemental Table 4).
The risk of bias was assessed as low in 9 studies and as low/unclear in 7, mostly because of limited information regarding randomization of study infants (see Fig 2; study data available in Supplemental Table 4). The Grading of Recommendations Assessment, Development, and Evaluation assessment of the quality of evidence is shown in Table 2. The data for the primary outcome variable and the secondary variable, BPD, were assessed as moderate quality in view of a degree of inconsistency in the results. The quality was low for mortality data because of both inconsistency and imprecision.
Primary Outcome Variables
Death or BPD at 36 weeks’ PMA, the primary outcome variable, was significantly reduced in the treated group (6 trials, n = 1285, RR = 0.86, 95% CI 0.75 to 0.99, I2 = 0%, P = .03). Five smaller studies reported the alternative composite variable of death or BPD at age 28 days of life and found no significant effect (5 trials, n = 429, RR = 0.98, 95% CI 0.88 to 1.11, I2 = 0%) (see Supplemental Tables 5 and 6).
The incidence of BPD alone at 36 weeks’ PMA fell more markedly than the composite variable (7 trials, n = 1168, RR = 0.77, 95% CI 0.65 to 0.91, I2 = 0%, P = .003). However, the incidence of BPD at 28 days PNA was not found to be significantly altered by the study intervention (7 trials, n = 528, RR = 0.92, 95% CI 0.76 to 1.11, I2 = 46%). By comparison, the incidence of death alone seems unaffected by the study interventions either at age 28 days or at 36 weeks PMA (28 days: 6 trials, n = 480, RR = 0.69, 95% CI 0.30 to 1.56, I2 = 46%; 36 weeks: 7 trials, n = 1270, RR = 0.97, 95% CI 0.42 to 2.20, I2 = 50%).
Secondary Outcome Variables
A variety of measures were used across the studies to attempt to gauge the effect of inhaled steroids on the severity or duration of BPD to expand the assessment of the overall treatment effect. Because many of the studies initiated therapy in ventilated infants, failure to successfully extubate was considered a measure of interest at various time points that included 7, 14, and 21 to 28 days and at the latest reported time point. Failure to extubate was significantly reduced at day 14 (6 trials, n = 232, risk difference (RD) –0.21, 95% CI –0.41 to 0.00; P = .05, I2 = 67%) and at the latest reported time point (1 trial, n = 14, RD –0.46, 95% CI –0.91 to –0.01; P = .05). In addition, there was a reduction of borderline significance at day 7 (3 trials, n = 92, RD –0.19, 95% CI –0.48 to 0.10; P = .2, I2 = 73%) and at 21 to 28 days (3 trials, n = 294, RD –0.14, 95% CI –0.30 to 0.03, I2 = 58%, P = .11). The Forest plots are available in the online supplemental information (Supplemental Figures 4A, 4B, 4C, 4D, 4E, 4F, 4G, and 4H).
Duration of mechanical ventilation or oxygen supplementation was reported as mean or median or total days, and in the absence of raw data, these data were not combined. In studies reporting mean, no significant effect was found on the duration of mechanical ventilation (3 trials, n = 113, mean difference [days] –3.91, 95% CI –15.42 to 7.61, I2 = 78%). Likewise, no significant effect was found on duration of oxygen supplementation (3 trials, n = 89, mean difference [days] –2.15, 95% CI –9.59 to 5.28, I2 = 0%). However, these outcome measures were available only in a small proportion of the overall sample.
The use of systemic corticosteroids as a rescue intervention, presumably because of perceived failure of other therapies to improve the infant’s respiratory status, is another commonly used intermediate measure of the effect of the ICs. A significant reduction was found in administration at any time point (13 trials, n = 1537, RR = 0.87, 95% CI 0.76 to 0.98, I2 = 3%). No significant effect was found in the subset of studies reporting specifically at 36 weeks (3 trials, n = 352, RR = 0.87, 95% CI 0.47 to 1.61).
ICs showed no significant effect on the occurrence of major neonatal morbidities. The incidence of sepsis seems unaffected (12 trials, n = 1282, RR = 1.12, 95% CI 0.96 to 1.3, I2 = 0%). No significant effect was found for the interventions on central nervous system injury that was reported in various forms. Any grade of intraventricular hemorrhage (IVH) was reported in 5 trials (n = 391, RR = 0.96, 95% CI 0.85 to 1.09, I2 = 0%) while severe IVH (grades 3 and 4) was reported as a separate variable in 3 trials (n = 362, RR = 1.43, 95% CI 0.76 to 2.69, I2 = 0%). Periventricular leukomalacia was reported in 3 trials (n = 362, RR = 1.17, 95% CI 0.55 to 2.48, I2 = 0%). The study of Bassler et al found no effect on a composite variable, termed “brain injury” that included IVH, periventricular leukomalacia, and ventriculomegaly (n = 838, RR = 1.25, 95% CI 0.94 to 1.65). Similarly, no effect was found on the incidence of patent ductus arteriosus (PDA) as reported in various forms (any PDA: 4 trials, n = 128, RR = 0.93, 95% CI 0.51 to 1.73, I2 = 57%; PDA requiring drug treatment: 2 trials, n = 909, RR = 0.79, 95% CI 0.59 to 1.07, I2 = 50%; PDA requiring surgery: 2 trials, n = 909, RR = 1.09, 95% CI 0.18 to 678, I2 = 68%). The incidence of air leak was unaffected by the interventions (2 trials, n = 83, RR = 0.93, 95% CI 0.14 to 6.04, I2 = 0%). No effect was found for the intervention on ophthalmic morbidities (retinopathy of prematurity stage 2 or higher: 4 trials, n = 1086, RR = 1.12, 95% CI 0.93 to 1.35, I2 = 0%; retinopathy of prematurity requiring treatment: 2 trials, n = 977, RR = 1.18, 95% CI 0.70 to 1.99; I2 = 36%); Cataract: 1 study, n = 253, RR = 0.35, 95% CI 0.01 to 8.56). Necrotizing enterocolitis (NEC) was similarly unaffected by the study interventions (4 trials, n = 1192, RR = 0.76, 95% CI 0.54 to 1.06, I2 = 0%).
Growth in the neonatal period was assessed by a variety of measures in 6 studies. These included weight, length, head circumference, and rate of growth, and the measures could not be combined for review. However, no significant effects were detected in the individual studies. Assessment of a potential effect of ICs on adrenal suppression was studied with either cortisol levels or an adrenocorticotropic hormone test in 4 studies with no adverse effect being detected.
Long-term health, growth, and neurodevelopment were reported in a single study at 3 years of age.28 No significant difference was found between the groups in incidence of respiratory readmissions, cerebral palsy, developmental delay, blindness, or deafness.
No significant effect was found on the incidence of either any adverse event reported (1 trial, n = 856, RR = 0.93, 95% CI 0.73 to 1.19) or specific adverse events, such as hyperglycemia (5 trials, n = 1002, RR = 0.96, 95% CI 0.74 to 1.24, I2 = 0%), hypertension (5 trials, n = 1002, RR = 0.80, 95% CI 0.39 to 1.62, I2 = 0%) or gastrointestinal hemorrhage or perforation (3 trials, n = 1139, RR = 0.97, 95% CI 0.63 to 1.51, I2 = 0%) (see Supplemental Table 7).
This systematic review and meta-analysis demonstrates a beneficial effect of ICs on the incidence of BPD at 36 weeks’ PMA with a number needed to treat of 14. In addition, the intervention was associated with a reduction in the administration of systemic steroids as a “rescue therapy.” In contrast, there was no evidence of an effect on the incidence of death, neonatal morbidity, or other adverse effects. Limited long-term neurodevelopmental outcome data are available but are expected to be markedly expanded in the near future with publication of data from the Bassler et al study.
The absence of evidence for an effect on mortality in the meta-analysis is of particular importance in view of the results of the Bassler et al study that showed a significant reduction in the incidence of BPD at 36 weeks. However, the composite outcome of death or BPD was of borderline significance as a result of a nonsignificant trend to increased mortality in the treated group. Although no other studies of ICs have suggested an increase in mortality, it is reassuring that the combination of all 16 relevant trials showed no effect on mortality.
The apparent discrepancy between the effect of ICs on BPD at 36 weeks and the lack of effect at 28 days may relate to the markedly higher sample size available at 36 weeks and also to the relative insensitivity of BPD at 28 days as a predictor of BPD at 36 weeks.
After >2 decades of intermittent investigation, our data suggest that ICs may be considered for the prevention or treatment of BPD in preterm infants. However, the marked heterogeneity of the studies included in this analysis precludes any unambiguous observations on a number of important questions. As described, clinical heterogeneity was observed with regard to the drug, dosage, timing, and duration. Accordingly, no recommendations may be offered on these issues. However, it is worth noting that more than half of the sample of this analysis was derived from a single study that demonstrated a statistically significant reduction in the incidence of BPD using budesonide administered from birth until a maximum age of 32 weeks’ PMA. This is therefore the most robust evidence for a recommendation. To solidify this recommendation, it will be necessary to update this analysis when the neurodevelopmental follow-up of this study is published.
Prevention of BPD remains a major challenge. Currently accepted approaches include attempts at minimizing injury associated with mechanical ventilation and oxygen, and these varied techniques have met with some success.31–33 Caffeine is effective and widely used.34 Likewise, systemic steroids are known to be effective, and thus, in the most challenging cases, clinicians and families frequently need to balance risks and benefits in decision-making.5,35 However, despite all of this research activity, BPD remains unacceptably frequent, and a new potent therapeutic intervention such as routine ICs for infants at risk may significantly improve outcome for these infants.1
The implication of this analysis is to establish the place of ICs, possibly budesonide, in particular, as a potentially efficacious and safe therapy for the prevention or treatment of BPD in preterm infants. This analysis will require an update with long-term follow-up data in the future.
- Accepted September 22, 2016.
- Address correspondence to Eric Shinwell, MD, Department of Neonatology, Ziv Medical Center, Rambam St, Tsfat 13100, Israel. E-mail:
This trial has been registered with the International Prospective Register of Systematic Reviews (PROSPERO identifier CRD42015019628).
FINANCIAL DISCLOSURE: The authors have indicated they have no financial relationships relevant to this article to disclose.
FUNDING: No external funding.
POTENTIAL CONFLICT OF INTEREST: Drs Shinwell and Bassler are authors of a study that is included in the systematic review. The other authors have indicated they have no potential conflicts of interest to disclose.
- Doyle LW,
- Halliday HL,
- Ehrenkranz RA,
- Davis PG,
- Sinclair JC
- Shah SS,
- Ohlsson A,
- Halliday HL,
- Shah VS
- Shah SS,
- Ohlsson A,
- Halliday HL,
- Shah VS
- Shamseer L,
- Moher D,
- Clarke M, et al; PRISMA-P Group
- Higgins JPT,
- Altman DG,
- Sterne JAC
- Pokriefka E,
- Mehdizadeh B,
- Rabbani A
- Townsend SF,
- Hale KA,
- Thilo EH
- Fok TF,
- Lam K,
- Dolovich M, et al
- Yong WSC,
- Carney S,
- Pearse RG,
- Gibson AT
- Dargaville PA,
- Aiyappan A,
- De Paoli AG, et al
- Shinwell ES,
- Karplus M,
- Bader D, et al
- Copyright © 2016 by the American Academy of Pediatrics