Targeting Inflammation to Prevent Bronchopulmonary Dysplasia: Can New Insights Be Translated Into Therapies?

Abstract

Bronchopulmonary dysplasia (BPD) frequently complicates preterm birth and leads to significant long-term morbidity. Unfortunately, few therapies are known to effectively prevent or treat BPD. Ongoing research has been focusing on potential therapies to limit inflammation in the preterm lung. In this review we highlight recent bench and clinical research aimed at understanding the role of inflammation in the pathogenesis of BPD. We also critically assess currently used therapies and promising developments in the field.

Preterm birth affects 12.5% of pregnancies in the United States, and this rate continues to increase.¹ The common corollary, bronchopulmonary dysplasia (BPD), affects up to 43% of infants born at <1500 g.² BPD has long-lasting effects including poor neurodevelopmental outcomes and long-term pulmonary dysfunction.^3,4 Unfortunately, few interventions currently used to prevent or treat BPD do so with certain benefit that outweighs harm. Only caffeine has a narrow confidence interval (CI) around estimates of efficacy, whereas those for postnatal steroids and vitamin A are wide.⁵ Because inflammation is central to the pathogenesis of BPD, it is disappointing that this understanding has not translated into useful therapies. Here we review recent concepts on inflammation that might help identify potential new therapeutic targets and highlight specific mediators with human correlates in the pathogenesis of BPD. The transcription factor nuclear factor κB (NF-κB) is a central cellular mediator of inflammation and is linked to the pathogenesis of many pulmonary diseases including acute respiratory distress syndrome, asthma, and chronic obstructive pulmonary disease.⁶ Here we discuss its potential pathogenic role in BPD. Finally, we critically evaluate whether common clinical interventions, including mechanical ventilation, administration of glucocorticoids, and emerging therapies, affect the impact of inflammation on the preterm lung. Despite an enormous body of bench work that has identified key molecular components of the inflammatory cascade, we conclude that much of this work has not yet been translated into evidence-based therapies.^5,7

ADVANCES IN SCIENCE AND TECHNOLOGY: INFLAMMATION AND BPD

General and Methodologic Considerations

In 1975, Philip⁸ proposed that the etiology of BPD was multifactorial, largely composed of external forces: the duration of exposure to oxygen and pressure. As inflammation entered this paradigm, it included external sources (chorioamnionitis, postnatal infections), iatrogenic sources (ventilation, oxygen), and the internal host response.^9–15 In 1999, Jobe¹⁶ amended Philip's model to incorporate multiple dimensions of inflammation and created a unified model of “new BPD.” However, even as this paradigm was confirmed by experimental data, few innovative therapies have proven efficacious. Is it useful to ask “why not?”

Several methodologic issues complicate moving potential therapies from bench to bedside for the treatment of BPD. One issue is the obvious difficulty of extrapolating animal data to human preterm infants. This issue is especially evident when the animal studies use 1 insult (eg, hyperoxia, lipopolysaccharide [LPS]) of limited duration, which is an infrequent occurrence in human newborns. However, single-hit models do carry explanatory power and generate hypotheses relevant to human disease (Table 1). However, the molecular redundancy within the complex inflammatory process complicates the translation of experimental interventions into treatments. The multiple stimuli and pathways that lead to NF-κB activation illustrate this complexity (Fig 1). Finally, human studies remain of small size. For example, of the nearly 30 studies that have attempted to predict BPD from proinflammatory biomarkers in tracheal aspirate, blood, and urine samples,^17–19 only 2 were of “reasonable size” to address population risks.^20,21 Ambavalan et al²⁰ examined 1067 preterm infants in a prospective cohort study, of which 606 infants developed BPD. An early (at <3 days of life) increase in serum levels of interleukin 8 (IL-8) and IL-10 or early decreases in levels of RANTES (regulated on activation normal t-cell expressed and secreted) protein and (at days of life 14–21) increases in the level of IL-6 predicted BPD (Table 2). Bose et al²¹ studied 1506 infants but excluded 38% of the original cohort secondary to death or lack of follow-up.²¹ They found that the protective effect of an elevated RANTES level (at <7 days of life) was lost after adjusting for mechanical ventilation. These 2 large studies need replication to better characterize timing and the importance of such markers.

TABLE 1.

Inflammatory Mediators With Animal and Human Data That Suggest a Role in the Pathogenesis of BPD

Factor	Animal Data				Human Data, Tracheal Aspirate Levels in Infants Who Develop BPD
	Regulated by NF-κB	Intervention	Insult	Effect
CINC-1 (rat)	Yes228	Neutralizing antibody	O2	Blocks pulmonary PMN influx229	↑20,230,231
IL-8 (human)
IL-1β	Yes232	IL-1R antagonist	O2	Inhibits PMN influx and improved alveolar number233	↑20,233
IL-6	Yes234	Pulmonary overexpression	O2	Increases mortality rate235	↑20,235,237
MMP-9	Yes238	MMP-9−/− mice	O2	Improves lung morphology238	↑239
MCP-1	Yes240	Neutralizing antibody	O2	Blocks pulmonary PMN influx241	↑242
CCSP	Unknown	CCSP−/− mice	O2	Increases mortality rate243	↓244
					Intratracheal administration of recombinant protein ↓ lung PMN245
MIF	Unknown	MIF−/− mice	Preterm delivery	Increases mortality rate246	↓246

CINC-1 indicates cytokine-induced neutrophil chemoattractant 1; ↑, increase; ↓, decrease; MMP-9, matrix metalloproteinase 9; MCP-1, monocyte chemoattractant protein 1; CCSP, Clara cell secretory protein; MIF, macrophage migration inhibitory factor; PMN, polymorphonuclear leukocyte.

FIGURE 1.

The transcription factor NF-κB regulates the cellular response to inflammatory and oxidant stress, as well as stretch. Normally, NF-κB remains sequestered in the cytoplasm bound to members of a family of inhibitory proteins: IκB. The canonical pathway, active after inflammatory stress (TNFα, LPS, etc) results in activation of IKK (IκB kinase), which phosphorylates (p) the inhibitory protein IκBα on 2 N-terminal serine residues. With hyperoxia, a tyrosine kinase phosphorylates (p) the inhibitory protein IκBα on tyrosine 42. The pathway that leads to NF-κB activation after stretch has not been described. Phosphorylation of IκBα results in the ubiquitination (u) and subsequent degradation of IκBα, which allows for dissociation and nuclear translocation of the active NF-κB dimer (the p65 and p50 subunits are used as examples here), binding to specific nucleotide sequences on promoter genes and transcriptional activation or repression of gene expression. Known inhibitors of NF-κB activation, including NO, dexamethasone, azithromycin, and pentoxifylline, are shown where they affect the activation cascade. Multiple genes with a purported role in the pathogenesis of BPD are controlled by NF-κB. IFNγ indicates interferon γ; MIP-2, macrophage inflammatory protein 2; RANTES, regulated on activation, normal T-cell expressed and secreted; ELAM-1, endothelial-leukocyte adhesion molecule 1; ICAM-1, intracellular adhesion molecule 1; VCAM, vascular cell adhesion molecule; HGF, hepatocyte growth factor; G-CSF, granulocyte colony-stimulating factor; GM-CSF, granulocyte-macrophage colony-stimulating factor; SOD, superoxide dismutase; MMP-9, matrix metallopeptidase 9; iNOS, inducible NO synthase; EGR-1, early growth response 1.

TABLE 2.

Estimates of Risk Conferred by Selected Predisposing Factors for BPD

Factor Associated With Development of BPD	BPD Definition (28 Days of Life vs 36 wk postmenstrual age)	OR or RR (95% CI)
Serum cytokines
IL-8, early20	36 wk	14.98 (1.29–173.90)a
IL-10, early20	36 wk	2.03 (1.11–3.72)a
RANTES, early20	36 wk	0.72 (0.56–0.91)a
IL-6, late20	36 wk	2.42 (1.11–5.29)a
SNP association
TNF-30866	36 wk	1.03 (0.85–1.25)b
U Urealyticum (tracheal)79	28 d	2.834 (2.288–3.512)b
	36 wk	1.620 (1.129–2.325)b

RANTES indicates regulated on activation normal T-cell expressed and secreted.

^aOR.

^bRR.

NF-κB: A Potential Link Between Oxidant and Inflammatory Stress and BPD

The transcription factor NF-κB regulates the cellular response to inflammatory and oxidant stress.²² Activation of this ubiquitous transcription factor potentially determines the net cellular response to inflammatory and oxidant insult in the lung. In the “oxygen radical disease” model, Saugstad^23–25 proposed an intracellular abundance of reactive oxygen species contributed to the pathogenesis of BPD. Inflammatory and oxidant insults stimulate NF-κB via discrete signaling pathways, which fine-tune the cellular response.²² In a quiescent cell, NF-κB remains sequestered in the cytoplasm bound to a member of the IκB family of inhibitory proteins (α, β, ε).²⁶ After phosphorylation and degradation of the inhibitory proteins, NF-κB translocates to the nucleus. The dimeric NF-κB complex is composed of different combinations of 5 subunits: p50, p52, p65, c-Rel, and RelB. Once in the nucleus, specific subunit dimer combinations bind to unique DNA oligonucleotide sequences.²⁷ Adding further control, some dimeric complexes contain transactivation domains (p65–p50 heterodimers) that increase gene transcription, whereas others (p50–p50 homodimers) repress gene transcription.²⁸ Many proinflammatory mediators associated with BPD are direct targets of NF-κB (Fig 1). After activation, NF-κB increases expression of its inhibitory protein IκBα, which shuttles NF-κB dimers out of the nucleus and results in a tightly regulated negative feedback loop.²⁹ Finally, each of the 3 inhibitory proteins (IκBα, IκBβ, and IκBε) have unique characteristics, and their presence determines a complex oscillatory pattern of NF-κB–regulated gene expression.³⁰ Together, this complexity enables NF-κB to tightly control the transcription of genes.

NF-κB displays maturational differences in response to oxidant and inflammatory stress. For example, neonatal lymphocytes show increased NF-κB activation in response to various stimuli when compared with their adult counterparts.^31,32 Similarly, fetal lung fibroblasts, in contrast to adult cells, demonstrate hyperoxia-induced NF-κB activation.³³ In vivo, hyperoxia-induced NF-κB activation is enhanced in alveolar epithelium and endothelium of neonatal mice but not in adults.³⁴ Both inflammatory and oxidant stress-induced activation of NF-κB impairs branching morphogenesis in the developing lung,^35,36 which suggests that NF-κB not only controls the expression of proinflammatory genes but also controls the expression of growth factors and proapoptotic and antiapoptotic proteins.³⁷ Therefore, there may be unintended consequences of modulating NF-κB activation in the developing lung.

Some human data link NF-κB to BPD. It is unclear yet whether the presence of activated NF-κB indicates its pathologic role or merely represents a response to injury. Nevertheless, if tracheal aspirates from preterm infants contain leukocytes demonstrating NF-κB activation, there is an increased risk of developing BPD³⁸ and an association with severity of RDS,³⁹ duration of mechanical ventilation, Ureaplasma urealyticum colonization, and exposure to chorioamnionitis.⁴⁰ Agents that inhibit NF-κB activation have shown promise in clinical trials aimed at preventing BPD. These agents include dexamethasone, azithromycin, nitric oxide (NO), and pentoxifylline.^41–44

Prenatal and Fetal Modulators of Inflammation

Genetics of the Host Responses to Inflammation

The genetic predisposition for BPD was recently reviewed comprehensively.^12,45–47 Parker et al⁴⁸ first proposed a genetic susceptibility to BPD when they found that the BPD status of 1 twin predicted BPD in the second twin. Subsequent studies of 450 and 318 preterm twins characterized a risk for BPD from both genetic and environmental factors.^49,50 Beyond these twin-birth association studies, specific nucleotide polymorphisms (SNPs) have been investigated. However, the excitement that this has generated is tempered by the methodologic constraints on the validity of some studies, which sometimes include less-than-stringent levels of statistical significance, given the issue of multiple testing.^51–53 Thirty-three studies have linked BPD to specific SNPs.^12,53–72 These studies enrolled between 33 and 1209 patients. Many of these studies focused on SNPs in proinflammatory and anti-inflammatory mediators including tumor necrosis factor α (TNFα), IL-4, IL-10, IL-12, monocyte chemoattractant protein 1 (MCP-1), surfactant protein A (SPA), surfactant protein D (SPD), transforming growth factor β (TGFβ), mannose-binding lectin, matrix metalloproteinase 16 (MMP-16), and interferon γ (IFNγ). Note that although small studies have suggested a link between TNF-308 SNPs and the risk of developing BPD, a recent meta-analysis that included a total of 804 infants failed to show statistical significance for this relationship (Table 2).⁶⁶ Although small studies are hypothesis-generating, only larger studies can address the methodologic and statistical criteria outlined by Attia et al^51–53 to provide robust validation of previous findings.

Ureaplasma Infection and Chorioamnionitis: Causal Agents or Prevalent Bystanders?

The old neonatal obsession with the potential role of U urealyticum was reviewed recently⁷³ but with new twists. Although U urealyticum colonization in preterm sheep does not result in BPD, in nonhuman primate models it does.^74–78 One meta-analysis of 23 studies that included 751 infants revealed a significant association between U urealyticum colonization and BPD at 36 weeks (Table 2).⁷⁹ However, the authors urged caution, because the included studies demonstrated large heterogeneity, and the greatest association was present in the smallest studies. Results of more recent research have been conflicting; some studies have shown an association of U urealyticum colonization with BPD,^80,81 whereas others have shown no association.^82,83 Because U urealyticum causes intra-amniotic bacterial infection, its role in BPD may have been exaggerated.

Similar considerations apply to chorioamnionitis. Although clinical chorioamnionitis (defined as maternal fever and uterine-abdominal wall tenderness) occurs in 40% of preterm pregnancies at <28 weeks,⁸⁴ histologic chorioamnionitis occurs in up to 80% of these pregnancies.⁸⁵ In 1996, Watterberg et al⁸⁶ proposed a causal link between histologic chorioamnionitis and BPD. However, receipt of antenatal steroids was an exclusion criterion in that study. Studies performed since the widespread administration of antenatal steroids to pregnant mothers threatening preterm birth have shown either a protective effect or no association between chorioamnionitis and BPD.^87–100 In a large population-based study of 798 premature infants with a 90% rate of exposure to antenatal steroids, histologic chorioamnionitis protected against BPD.¹⁰¹ Moreover, evidence of a fetal response to inflammation, evidenced by umbilical vasculitis, conferred more protection than chorioamnionitis alone.⁹⁹ Future studies should not only discern the presence of chorioamnionitis but also the fetal response to it.

Because NF-κB has a central role in regulating the cellular response to inflammation, does it play a role in the fetal response to chorioamnionitis? Animal models of the fetal inflammatory response syndrome (FIRS) suggest that it does. Exposure to intra-amniotic LPS increases NF-κB activation in bronchoalveolar lavage–derived neutrophils and monocytes of lambs.¹⁰² In a murine model of LPS chorioamnionitis, NF-κB activation led to an enhanced type II cell maturation.¹⁰³ Furthermore, human preterm amnion cells show a more pronounced NF-κB response to LPS compared with term controls.¹⁰⁴ Similarly, NF-κB activation is seen in fetal capillaries of human infants with funisitis and chorioamnionitis.¹⁰⁵ These findings suggest that NF-κB may mediate inflammation in the fetal lung.

Postnatal Modulators

Bacterial Sepsis in the Neonate

The term systemic inflammatory response syndrome (SIRS) has been adapted to children and newborns.¹⁰⁶ However, it may not be sensitive and, thus, may miss bacterial infections¹⁰⁷; here we discuss data only in which positive growth identifies an organism. Stoll et al¹⁰⁸ demonstrated that rates of BPD increased from 35% to 62% in a cohort of 5447 very low birth weight infants from the Neonatal Research Network after early-onset sepsis (odds ratio [OR]: 2.4 [95% CI: 1.2–4.7]). This relationship was confirmed recently in a population study from Israel of 15 839 infants (OR: 1.74 [95% CI: 1.24–2.43]).¹⁰⁹ It is interesting to note that the protective effect of chorioamnionitis and funisitis on the development of BPD was lost if the infant experienced early-onset sepsis (OR: 1.98 [95% CI: 1.15–3.39]).¹⁰¹ In fact, Lahra et al¹⁰¹ found that the infants at highest risk for BPD were born to mothers without histologic chorioamnionitis but who had experienced sepsis (OR: 2.71 [95% CI: 1.64–4.51]). Late-onset sepsis also increases the risk of BPD (relative risk [RR]: 2.32 [95% CI: 1.95–2.77]).^110,111 These data suggest that irrespective of the timing, inflammatory exposure from sepsis plays an important role in the development of BPD.

Oxygen Toxicity

Hyperoxia is a powerful proinflammatory stimulus, and its role in the pathogenesis of BPD was reviewed recently.^112,113 Although a full discussion of hyperoxia-induced pulmonary inflammation is beyond the scope of this review, recent clinical studies are relevant. Even short-term exposure to hyperoxia affects the developing lung. When infants born at 24 to 28 weeks' gestation were randomly assigned to resuscitation in the delivery room with either 90% or 30% oxygen, the incidence of BPD at 36 weeks' gestation was reduced from 31.7% to 15.4% (RR: 0.51 [95% CI: 0.21–1.21]).¹¹⁴ Infants exposed to 90% oxygen had significantly elevated serum TNFα and IL-8 levels. In addition, a recent meta-analysis revealed that limiting oxygen exposure in the NICU by adopting lower pulse-oximetry goals could reduce the incidence of BPD in premature infants from 40.8% to 29.7% (OR: 0.73 [95% CI: 0.63–0.86]).¹¹⁵ This meta-analysis was validated by the Surfactant, Positive Pressure, and Pulse Oximetry Randomized Trial (SUPPORT) which showed that infants who were randomly assigned to lower pulse-oximetry goals less frequently developed BPD (RR: 0.82 [95% CI: 0.72–0.93]) and retinopathy of prematurity (RR: 0.52 [95% CI: 0.37–0.73]).¹¹⁶ However, concerns about lowering oxygen-saturation ranges have arisen. Specifically, infants enrolled in the SUPPORT-NICHD trial and randomly assigned to lower pulse-oximetry goals had a higher mortality rate by discharge (number needed to harm: 27) (RR: 1.27 [95% CI: 1.01–1.60]).¹¹⁶ However, this was only 1 of 4 separate ways of assessing mortality (at 7 days, 14 days, 36 weeks' postmenstrual age, or by discharge) that was statistically significant. Results from 3 similar large, randomized controlled trials (RCTs) (Canadian Oxygenation Trial [COT] and Benefits of Oxygen Saturation Targeting II [BOOST II], and BOOST-UK) are pending.^117,118 We advise that neonatologists retain equipoise while these trials answer whether adopting lower pulse-oximetry goals will improve outcomes at 18 to 22 months, which is the a priori primary outcome of all 4 of these trials.

Molecular Mechanisms Signaling Stretch: Understanding Barotrauma/Volutrauma in Animal Models

Excessive lung stretching results in barotrauma/volutrauma and is a powerful proinflammatory force.¹¹⁹ Multiple signaling pathways, including NF-κB, translate stretch into a proinflammatory signal,^120,125 and the degree of stretch determines unique cytokine-expression profiles. Preterm lambs subjected to large-tidal-volume ventilation show upregulation of multiple proinflammatory mediators including IL-1β, IL-6, IL-8, and Toll-like receptors 2 (TLR-2) and 4 (TLR-4).¹²⁶ In addition, systemic inflammation occurs, indicated by a hepatic acute-phase response.¹²² There are strong developmental differences in the pulmonary cytokine response to excessive stretch.^127,128 For example, acute exposure to high-tidal-volume ventilation and hyperoxia induces a pulmonary cytokine response (IL-1β, IL-6, and TNFα) in adult mice, which is attenuated in neonates. Even then, chronic exposure to hyperoxia and high-tidal-volume ventilation will induce pulmonary cytokine release (TNFα and IL-6) in the newborn lung.¹²⁸

Animal data suggest that the inflammatory response to stretch can be prevented. Lung injury induced in mice exposed to hyperoxia and high-tidal-volume ventilation is reversed by NF-κB inhibition.¹²⁹ Dexamethasone inhibits NF-κB activation and prevents lung cytokine expression in mice exposed to high-tidal-volume ventilation.¹³⁰ It is significant that IL-6 elevation in ventilated preterm lambs is attenuated by gentle ventilation (lower tidal volumes).¹³¹ The mode of ventilation also plays a role, as indicated by the fact that intubated piglets had markedly differing cytokine responses compared with animals ventilated with high-frequency nasal ventilation.¹³² These data suggest that modification of current practices could decrease inflammation and injury in the preterm lung.

CRITICAL ASSESSMENT: ANTI-INFLAMMATORY THERAPIES, OLD AND NEW

Therapies that may decrease inflammation in the preterm lung are vitiated by uncertainty (wide CIs around estimates of efficacy or harm) and fraught with potential undesired serious adverse effects (eg, cerebral palsy after postnatal steroids). Because the search for efficacious anti-inflammatory agents continues, we highlight emerging therapies for preventing or treating BPD.

Interruption of Key Components of the Inflammatory Cascade

The commonest paradigm for inflammation is not BPD but, rather, severe sepsis.¹³³ Superficially, it might be logical to ask whether cytokine cascades integral to inflammation could be blocked by antibody therapy. Several large adult trials have investigated this avenue for treating severe sepsis. By 2000, 60 trials that used various monoclonal antibodies directed at TNFα had recruited 4197 patients and showed a cumulative reduction in 28-day mortality rates (OR: 0.87 [95% CI: 0.76–0.98]).¹³³ This modest benefit has led to consideration of polyclonal TNFα antibodies.¹³⁴ Other trials have evaluated the efficacy of IL-1 receptor antagonist and platelet-activating factor receptor antagonist with similar cumulative ORs.¹³⁵ Such modest reductions in mortality have not yet passed into clinical practice because of continued uncertainty.

These limited benefits to date may reflect the redundancy of the inflammatory system, which has led to attempts to broaden the inflammatory target. Because of its pivotal role in microthrombi formation in sepsis, protein C has been the focus of much attention. However, despite initial excitement (PROWESS [Recombinant Human Activated Protein C Worldwide Evaluation in Severe Sepsis]),¹³⁶ the promise of recombinant activated protein C has not been borne out in adults. The meta-analysis of 4911 participants with severe sepsis revealed no reduction in 28-day mortality rates (RR: 0.92 [95% CI: 0.72–1.18]).¹³⁷ There have been no trials limited to newborns, and this group may be at increased risk for bleeding complications.^138,139 Similar to neonatologists who lack useful treatments for BPD, adult intensivists are revisiting the use of low-dose corticosteroids for the treatment of sepsis.¹⁴⁰

The Anti-inflammatory Component of Stem Cell Therapy for Preventing BPD

Research using stem cells to prevent or treat developing BPD has burgeoned over the past decade.^141–144 There are several different stem cells (embryonic stem cells, bone marrow–derived stem cells, and tissue progenitor cells), but most work in neonatal lung injury has focused on bone marrow–derived mesenchymal stem cells (MSCs). Controversy remains as to whether these cells can actually engraft in the lung and differentiate into lung epithelial cells.¹⁴⁵ If they do, these cells may protect and repair the damaged lung by several mechanisms: physical repair by adopting a native cell phenotype or repair of existing cells by exerting immunomodulatory, anti-inflammatory, and antiapoptotic effects. Some of the protective effect of MSC administration are conferred by paracrine mediators, termed the “MSC secretome.”¹⁴³ Results of preclinical studies indicate a role of MSCs in the treatment of acute lung injury in adults¹⁴⁶; however, their role in the treatment of BPD remains to be defined.

In newborn rodents, systemic administration of stem cells obtained from either bone marrow or cord blood attenuates hyperoxia-induced lung inflammation.^147–150 However, administration of conditioned medium from mesenchymal cells can provide similar levels of protection.¹⁴⁹ Furthermore, newborn rats treated with cord blood MSCs display attenuated pulmonary myeloperoxidase, IL-6, TNFα, and transforming growth factor β expression.¹⁴⁷ More data are needed to determine if anti-inflammatory effects help explain the protection seen with stem cell administration.

Gentle Ventilation: Limiting the Damage We Cause

Inflammation is a major component of ventilator-induced lung injury in adults¹⁵¹ and newborn infants.^152,153 Three ventilatory strategies impact inflammatory changes in the lung: noninvasive approaches; low-tidal-volume ventilation; and the use of positive end-expiratory pressure (PEEP).

The “gentle-ventilation” approach is increasingly taken with the preterm infant to avoid intubation with noninvasive ventilator support. It is true that individual trials of aggressive early continuous positive airway pressure (CPAP) therapy versus intubated ventilation in the delivery room have not resulted in a reduced rate of BPD.^154–156 Nonetheless, pooling data on combined mortality and BPD at 36 weeks' corrected age has suggested a benefit (Table 3). Other strategies of gentle ventilation include intubation to deliver surfactant and early extubation.^157,158 Nasal intermittent mandatory ventilation (NIMV) holds promise,¹⁵⁹ but larger trial results are pending.¹⁶⁰

TABLE 3.

Efficacy of Continuous Positive Airway Pressure for Prevention of BPD or Death

The primary outcome of the studies depicted was pooled and analyzed by using RevMan 5 software (Cochrane Collection). M-H indicates Mantel-Haenszel odds ratio. Heterogeneity: χ² = 1.32, df = 2 (P = .52); I² = 0%. Test for overall effect: z = 1.97 (P = .05).

An extension of gentle ventilation is a low-tidal-volume strategy. An adult RCT that demonstrated that low-tidal-volume ventilation improved mortality rates in adults with acute respiratory distress syndrome¹⁶¹ sparked much work in newborns. Mechanistically, only sparse RCT data have shown effects of differing modes of ventilation on inflammatory mediators. Lista et al¹⁶² randomly assigned preterm infants to either high-frequency oscillatory ventilation or low-tidal-volume guarantee and found reduced inflammatory markers in tracheal aspirates in those who were assigned to low-tidal-volume guarantee. A Cochrane analysis confirmed a statistically significant reduction of death and/or BPD (number needed to treat: 8) (RR: 0.73 [95% CI: 0.57–0.93]) with targeted low-tidal-volume ventilation.¹⁶³ Together with animal studies, these data suggest that using gentle ventilation may result in decreased pulmonary inflammation in preterm neonates who need respiratory support.

For adult disease, Gattinoni et al¹⁶⁴ urged PEEP to recruit lung volume. Muscedere et al¹⁶⁵ showed that in an in vitro model, setting PEEP above the lower inflection point preserved the lung's mechanical properties and attenuated proinflammatory cytokine expression.¹⁶⁶ Using appropriate lung-opening pressure with an adequate lower inflection point in newborn piglets exposed to mechanical ventilation reduces the influx of activated leukocytes into the lungs.¹⁶⁷ However, finding the appropriate opening pressures in adult humans is tricky¹⁶⁸ and is impractical in neonates because it requires paralysis. This may explain why use of an “appropriate PEEP” was never implemented clinically or tested in trials despite observed benefits in infants.¹⁶⁹ Studies that define empirical levels of PEEP that should be set in newborns have been sparse.^170,171 However, several adult trials of high-PEEP versus low-PEEP strategies have been completed.¹⁷² In general, an empirical oxygen grid against varying PEEP levels was used to set PEEP, rather than pulmonary function tests. An individual patient meta-analysis revealed an overall reduction of the end point of days in hospital and days on respiratory support,¹⁷³ which suggests that simply identifying and using ideal PEEP may reduce inflammatory changes in the preterm lung.

Azithromycin

Macrolides have both antimicrobial and anti-inflammatory properties.¹⁷⁴ Azithromycin decreased IL-6 expression and improved lung morphology and mortality rates in neonatal rats exposed to hyperoxia.¹⁷⁵ Furthermore, azithromycin inhibited inflammatory stress–induced NF-κB activation in tracheal aspirate cells taken from premature infants.⁴³ A pilot study that evaluated the safety and effectiveness of azithromycin in extremely low birth weight infants showed that the treatment group received fewer days of mechanical ventilation, but the study was underpowered to find a difference in the rate of BPD.¹⁷⁶ A phase 2 study is currently underway to determine the effectiveness of this therapy in decreasing the incidence of BPD.¹⁷⁷

NO and Its Role as an Anti-inflammatory Agent

Data from several large RCTs performed to determine if NO can prevent BPD in preterm infants are still being combined into a meta-analysis from individual patient data.¹⁷⁸ However, the National Institutes of Health Consensus for Inhaled Nitric Oxide Therapy for Premature Infants mandates new trials before it can be considered a standard of care.¹⁷⁹ Here we briefly discuss the anti-inflammatory properties of NO.^180,181 Many of its anti-inflammatory properties are mediated through the inhibition of canonical, inflammatory stress–induced, and atypical, oxidant stress–induced NF-κB activation.^44,182–193 This is seen in healthy adult human subjects who have a higher endogenous NO production and associated NF-κB inhibition when compared with asthmatic subjects and those with pulmonary hypertension.¹⁹⁴ To date, no data exist to answer whether NO affects NF-κB signaling in the preterm lung.

Antioxidants

The role of antioxidants in preventing BPD was reviewed recently.¹⁹⁵ The largest RCT evaluated intratracheal copper zinc superoxide dismutase to prevent BPD in infants who weighed <1200 g.¹⁹⁶ This treatment did not alter the incidence of BPD, but treated infants had significantly better pulmonary outcomes at 1 year of age. Some have voiced a concern about the potential untoward effect of scavenging reactive oxygen species given their role in intracellular signaling in the developing lung, brain, and retina.¹⁹⁷ The role of antioxidants for the prevention of BPD remains unclear.

PERCEPTION: THE LACK OF USEFUL THERAPIES FORCES US TO REVISIT AN OLD NEMESIS—CORTICOSTEROIDS

Neonatologists and corticosteroids have had a long and unstable relationship.^198–201 Systemic glucocorticoids decrease inflammation and increase both surfactant synthesis and lung epithelial differentiation in the developing lung.^202,203 Irrespective of the precise mechanism, corticosteroids seem to have some benefit in treating ventilator-dependent infants at high risk for BPD. Efficacy of postnatal dexamethasone for treating ventilator dependency in BPD was first shown in 1983.²⁰⁴ As postnatal corticosteroid use became routine, infants were treated prophylactically with longer courses and higher doses. This treatment practice dominated the 1990s. When Yeh et al²⁰⁵ showed an increased risk of cerebral palsy in infants exposed to corticosteroids early, practices abruptly changed. A meta-analysis of controlled trials revealed a relationship between early dexamethasone exposure and cerebral palsy.²⁰⁶ A major outcry ensued against steroids that limited their use, even for late disease.^207,208 Unfortunately, no distinction was made between the early, indiscriminate use of steroids and late, targeted use of this therapy. The influential statements of the American Academy of Pediatrics made it virtually impermissible to use steroids,²⁰⁹ although there were occasional voices urging caution over the interpretation of the data.^210,211 This climate sabotaged an RCT that was designed to address the impact of postnatal corticosteroids on the primary outcome of neurodevelopmental outcome, which was stopped early because of a lack of equipoise.²¹² Consequently, clinicians are left with broad confidence estimates for all efficacy or harm outcomes (Table 4).

TABLE 4.

Efficacy of Selected Treatments for the Prevention of BPD

Treatments to Prevent BPD	Control		BPD		RR (95% CI)
	n/N	%	n/N	%
Caffeine247	447/954	46.9	350/963	36.3	0.63 (0.52–0.76)
Vitamin A248	193/347	55.6	163/346	47.1	0.89 (0.80–0.99)
Early corticosteroids, <8 d of age215	535/1638	32.7	423/1648	25.7	0.79 (0.71–0.88)
Late corticosteroids, >7 d of age217	146/230	63.5	108/241	44.8	0.72 (0.61–0.85)
Superoxide dismutase196	36/154	23.4	37/148	25.0	1.02 (0.89–1.16a
Azithromycin176	10/16	83.3	9/19	64.3	0.71 (0.33–1.53a
Continuous positive airway pressure (unpublished data)	533/1172	45.4	495/1193	41.5	0.91 (0.83–1.00)a

^aCalculated by authors using published data.

The limited number of useful therapies available to prevent BPD, along with a decrease in steroid use, seemed to result in a rising incidence of BPD.^212–214 The recent meta-regression that demonstrated that corticosteroids will decrease the risk of poor neurodevelopmental outcome if an infant's baseline risk of developing BPD is >55%, along with recent updates of the Cochrane reviews, have affected our thinking.^215–217 They argue that the widespread use of steroids to prevent BPD is contraindicated but that therapy for ventilator dependency or early BPD is warranted.²¹⁸ Thus, determining an infant's risk of developing BPD becomes even more clinically important. Simple lung mechanics are unlikely to be helpful.²¹⁹ Although exhaled NO has been proposed as a marker of inflammation, whether it is a better predictor of BPD over simple clinical predictors (eg, birth weight) remains unclear.²²⁰ However, end-tidal carbon monoxide on day-of-life 14 does predict BPD well (OR: 15.17 [95% CI: 2.02–113.8]).²²¹ Confirmation of this and other new predictive tools are needed.

Hence, the dexamethasone pendulum is beginning to swing back, as a recent statement from the American Academy of Pediatrics confirmed.²²² Concerns about dexamethasone have led some investigators to evaluate the use of hydrocortisone for preventing BPD.²²³ A systematic review of available RCTs revealed no effect of hydrocortisone on preventing BPD.²⁰¹ However, most trials have used very low doses of hydrocortisone, especially when compared with the doses of dexamethasone used to prevent BPD. Others have advocated even lower doses of dexamethasone.²²⁴ It remains eminently arguable that given the limited treatment options for the prevention of BPD, and its serious consequences,^225,226 the use of glucocorticoids is appropriate for specific patients at high risk of developing BPD.^203,227

CONCLUSIONS: WHERE ARE WE HEADED?

The role of inflammation in the pathogenesis of BPD is firmly established. Unfortunately, clinicians have few therapeutic interventions for limiting inflammation and preventing BPD. Because the etiology of BPD is multifactorial, anti-inflammatory therapies may represent only part of the solution. Only by combining bench translational studies and rigorous trials will practice at the bedside result in limiting lung injury in the preterm infant.