Bronchopulmonary dysplasia

Normal lung development.

Respiration after birth is entirely dependent on the architecture of the peripheral lung saccules and alveoli. The close interface between alveolar epithelial cells and endothelial cells of the pulmonary microvasculature enables efficient exchange of oxygen and carbon dioxide (FIG. 3). Lung formation begins between week 3 and week 6 of gestation. Subsets of endodermal cells from the anterior foregut form the initial tracheal and pulmonary lung buds that invade the splanchnic mesenchyme. Branching morphogenesis⁵² forms the conducting airways and peripheral acinar buds that will ultimately form the alveoli after birth (FIG. 4). Complex paracrine interactions among various mesenchymal cells, including pericytes, smooth muscle cells, endothelial cells and various types of fibroblasts, and respiratory epithelial cells and endothelial progenitor cells are involved in the formation of the pulmonary parenchyma^53–55. Acinar structures dilate later in gestation, creating the peripheral saccules (the primitive gas exchange portion of the lung) and alveoli that will accommodate postnatal adaptation to breathing. During morphogenesis, the lung is transformed from a solid gland-like tissue to an expanded open alveolar structure in which gas exchange can occur (FIG. 5). During the last trimester (28–40 weeks of gestation), peripheral saccules undergo a process of septation that further divides the airspaces, creating and increasing the number of alveoli and thereby increasing the surface area for gas exchange. Peripheral lung epithelial cells mature into alveolar type 1 (AT1) cells and AT2 cells, the latter producing pulmonary surfactant (FIG. 3). Thus, preterm infants at the edge of viability (22–23 weeks of gestation) begin ventilation at the canalicular–saccular stage of lung formation, well before morphogenesis and alveolar differentiation are completed.

Pulmonary surfactant.

Pulmonary surfactant is a complex mixture of phospholipids, primarily phosphatidylcholine, and the surfactant proteins SP-A, SP-B, SP-C and SP-D, which together define the physical structure, function and metabolism of surfactant in the alveolus (FIG. 3). Synthesis of surfactant lipids and proteins depends on the differentiation of AT2 cells, which occurs fairly late in gestation. Consequently, a lack of pulmonary surfactant as a result of incomplete differentiation of AT2 cells causes RDS in preterm infants⁵⁶. Assisted ventilation, continuous airway pressure and supplemental oxygen can initially support ventilation after birth. Furthermore, administration of exogenous pulmonary surfactant has greatly facilitated the transition to air breathing, decreasing the requirement for oxygen and assisted ventilation and thereby improving survival but secondarily increasing the number of infants at risk of BPD.

In conclusion, postnatal adaptation of the preterm lungs to breathing at birth is challenged by initial lung injury owing to surfactant deficiency, exposure to increased oxygen, mechanical ventilation, inadequate nutrition, infection and inflammation, which together provide considerable hurdles to normal pulmonary growth and repair and result in a loss of alveolar surface area (FIG. 5) that has lifelong consequences.

Chorioamnionitis and inflammation.

Inflammation is the common pathway that initiates the lung injury that can progress to BPD. Neonatal respiratory support at delivery can easily injure the premature lung during the transition from fluid-filled lungs to air breathing⁶⁶. Subsequent ventilatory support and oxygen exposure can amplify injury. Some injury may be unavoidable, even with gentle assisted ventilation and surfactant treatment in infants <30 weeks gestational age. Subsequent inflammation from sepsis or necrotizing enterocolitis can increase the risk of developing BPD. These post natal inflammatory exposures are often preceded by two antenatal exposures that are potent effectors of delayed lung development and lung injury. More than 50% of pregnancies that result in very preterm births (<28 weeks gestational age) have histological evidence of chorioamnionitis⁶⁷ — inflammation of the fetal membranes, maternal decidua and, often, the amniotic fluid — which may be the proximal cause of the preterm labour. Concurrently, in ~90% of pregnancies that are at risk of preterm delivery, maternal antenatal corticosteroid treatment is prescribed to treat RDS and decrease infant mortality⁶⁸. In the complex context of prematurity and BPD, these exposures confound simple interpretations of the causality of BPD.

Chorioamnionitis simply indicates an intrauterine exposure to inflammation, most often associated with Ureaplasma sp., but multiple vaginal commensals and pathogens have also been identified⁶⁷. Chorioamnionitis-associated deliveries may also have ‘sterile’ inflammation. In clinical practice, the identity of the multiple microorganisms, the duration of fetal exposure and the location of the infection (decidua, amniotic fluid or fetus) are usually unknown. Studies in multiple animal models have demonstrated that fetal breathing exposes the fetal lung to infection and/or inflammation in amniotic fluid and causes lung inflammation and often a fetal inflammatory response⁶⁹. Low-grade lung inflammation induces lung maturation in experimental models. In clinical practice, histologically-confirmed chorioamnionitis is associated with decreased RDS incidence⁷⁰, whereas the presence of Ureaplasma sp. in cord blood is associated with increased incidence of BPD but no change in the incidence of RDS⁷¹. In another clinical series, increased severity of BPD correlated with increased severity of chorioamnionitis⁷². However, a summary meta-analysis of 59 studies could not definitively associate chorioamnionitis with increased BPD risk²⁸, perhaps reflecting issues with standardizing the diagnoses of chorioamnionitis and the complex interactions of multiple perinatal risk factors.

Most very preterm fetuses exposed to chorioamnionitis are also exposed to antenatal corticosteroids, which in animal models suppress the inflammation associated with chorioamnionitis, cause an arrest of alveolarization and also induce greater lung maturation than either exposure alone⁷³. Tolerance or preconditioning (modulation of the response to a fetal or neonatal exposure by a previous fetal exposure) is a possible reason why single fetal or neonatal exposures in animal models poorly replicate outcomes in preterm infants. Depending on timing and the exposure, a second similar or different exposure can result in a blunted or exaggerated lung injury response. In fetal sheep, an intra-amniotic injection of endotoxin causes fetal lung inflammation. However, prior chronic exposure of pregnant ewes to live Ureaplasma sp. completely blocks the fetal lung response to endotoxins⁷⁴. Similarly, fetal exposure to intra-amniotic lipopolysaccharide (LPS) can protect the newborn from hypotension from intravascular LPS administration⁷⁵. Very preterm infants are exposed to multiple potent antenatal and postnatal inflammatory modulators that modify injury and repair pathways, which is expressed as the clinical variability and severity of BPD.

Alterations of the lung microbiome have been recognized in multiple respiratory disorders. It is conceivable that, similar to chorioamnionitis, alterations in the lung microbiome prime the developing lung to injury⁷⁶. The increasing evidence linking the microbiome in preterm infants to BPD has been analysed in a systematic review⁷⁷, which reported substantial heterogeneity in the six included studies. Microbial dysbiosis may be associated with BPD progression and severity. Most of these infants were born by caesarean section and were exposed to postnatal antibiotics. Novel insights into systems biology based on ‘omics’ approaches may enable the temporal analysis of the fetal and neonatal lung and gut microbiomes in preterm infants, and ultimately permit manipulation to correct dysbiosis and restore a ‘healthy’ microbial environment.

Vascular abnormalities.

Pulmonary hypertension and related pulmonary vascular disease (PVD) have long been recognized as strong contributors to poor survival in preterm infants with BPD¹. A diagnosis of pulmonary hypertension that persists beyond the first few months of life has been associated consistently with mortality rates as high as 40–50%, ranging from a report in 1980 (REF.⁷⁸) to a study in 2007 (REFS^79,80). Prospective studies have provided echocardiographic evidence of pulmonary hypertension in 14–25% of preterm infants at 36 weeks post-menstrual age (PMA), with pulmonary hypertension especially prevalent (29–58%) in infants with severe BPD^81–83. Thus, despite major advances in perinatal and neonatal intensive care unit (NICU) care with improved survival and changes in the nature of BPD over the decades, pulmonary hypertension continues to be a prominent clinical challenge in the pathobiology and outcomes of infants with BPD.

There has been growing awareness in the past decade that pulmonary hypertension in premature infants encompasses several PVD phenotypes (BOX 2), including the fairly high incidence of sustained elevation of pulmonary vascular resistance (PVR) after birth that can cause profound hypoxaemia due to extrapulmonary right-to-left shunting across the foramen ovale and/or ductus arteriosus (persistent pulmonary hypertension of the newborn (PPHN)) with a slow adaptation or ‘delayed vascular transition’^75,84–86. Although more often considered in near-term or term infants with hypoxaemia at birth, PPHN also occurs in preterm infants and is negatively associated with gestational age (FIG. 6a). Severe parenchymal lung disease can contribute to elevated PVR; however, abnormalities of the pulmonary vasculature can also contribute to high PVR, especially in the setting of antenatal stresses, such as oligohydramnios and prolonged premature rupture of membranes^84–87. Importantly, echocardiography-confirmed delayed pulmonary vascular transition⁸⁶ and early pulmonary hypertension⁸⁸ are associated with a higher risk of the subsequent development of BPD, late pulmonary hypertension and higher mortality^86,88. Experimental studies support epidemiological findings that adverse intrauterine stimuli are sufficient to impair vascular growth and induce longstanding and severe pulmonary hypertension independent of postnatal factors, such as hyperoxia and ventilator-induced injury⁸⁹.

In addition to the important effect of pulmonary hypertension, preclinical and clinical studies suggest that early disruption of lung vascular growth and function can impair growth of the distal airspace (the ‘vascular hypothesis’ of BPD)^90–93. Furthermore, strong clinical studies have shown that abnormal placental vascular structure and placental hypoperfusion on histopathology are strikingly associated with neonatal outcomes of IUGR and susceptibility to subsequent development of BPD and pulmonary hypertension^94,95. Prospective studies have shown that antenatal factors assessed on the first day of life^30,96, cord blood biomarkers of impaired angiogenesis^97–99 and early echocardiography findings of pulmonary hypertension shortly after birth are strongly linked to the risk of BPD, pulmonary hypertension, prolonged NICU hospitalization and late respiratory outcomes in childhood^83,100 (FIG. 6b–d). Thus, antenatal determinants not only cause abnormalities of short-term respiratory function, as reflected by NICU course, but early injury in utero causes sustained disruption of lung and pulmonary vascular structure throughout infancy, reflecting the adverse effect of fetal programming^89,101 (FIG. 1). Antenatal stresses, including chorioamnionitis and pre-eclampsia (with or without IUGR), in preterm neonates contribute to BPD risk^102–104, although data from animal models of mechanisms linking antenatal stress to BPD pathogenesis and late cardiorespiratory outcomes are limited.

Current challenges in defining BPD.

For most diseases, diagnosis is an end in itself. For preterm infants, a diagnosis of BPD is more akin to a functional lung assessment at the time a preterm infant is expected to transition out of hospital (similar to how forced expiratory volume in 1 second (FEV1) is used in conditions such as cystic fibrosis)¹⁰⁵. More than 30 years after BPD was defined by Shennan et al.¹⁰⁶, how well a BPD diagnosis can predict pulmonary outcomes in infancy and childhood, and how those long-term pulmonary outcomes should be defined, remain primary research and clinical questions¹⁰⁷.

The populations at risk and respiratory treatment strategies have also evolved since publication of the 2001 National Heart, Lung, and Blood Institute (NHLBI) workshop report¹⁰⁸. The number and survival of the smallest and most preterm infants has increased, and some NICUs now routinely care for babies born at 22 weeks of gestation^24,26. Respiratory care practices have also changed, as recently documented by the Prematurity and Respiratory Outcomes Program (PROP) cohort of 765 surviving infants of <29 weeks of gestation¹⁰⁹. In the PROP cohort, 359 infants (47%) were treated with nasal cannula flow at 36 weeks PMA, including 95 infants (12%) on flow with room air. This creates new problems for BPD classification, as it is not possible to differentiate whether flow is needed for control of breathing or for lung parenchymal injury. Data capture outside a rigorous research environment is also challenging. In the database maintained by the National Perinatal Registry of the Netherlands, only 67% of preterm infants were correctly categorized according to the NHLBI workshop definition¹¹⁰, owing predominantly to a false-negative diagnosis rate of 31%, including 9% with severe BPD. A bedside analysis of the ‘shift’ of the oxyhaemoglobin dissociation curve (accomplished through simultaneous measurement of FiO₂ and peripheral saturation at a single time point) was proposed as an objective test for the presence and severity of BPD^111,112. This measurement would provide more accurate assessment of oxygen need than the Walsh ‘physiological test’, but would be difficult to accomplish in most infants outside a research protocol. A retrospective analysis of infants from the NICHD Neonatal Research Network has led to a proposal to modify the 2018 NICHD definition³ to a severity scale that is based on the use of positive pressure at 36 weeks PMA instead of supplemental oxygen. Infants were classified as no BPD (no support), grade 1 (nasal cannula ≤2 l/min), grade 2 (nasal cannula >2 l/min or non-invasive positive airway pressure), or grade 3 (invasive mechanical ventilation). These criteria predicted death or serious respiratory morbidity at 18–26 months corrected age in 81% of infants in the study¹¹³.

Although BPD is a ‘yes/no’ diagnosis at 36 weeks PMA, the term has become shorthand for the presence of respiratory morbidity in a preterm infant before and after discharge from hospital. It is increasingly recognized that defining BPD at 36 weeks PMA by the use of oxygen alone is inadequate. Data from the Canadian Neonatal Network revealed that oxygen use and/or respiratory support is a better indicator of chronic respiratory insufficiency than oxygen alone, and that assessment at term equivalent (40 weeks PMA) increases the predictive value¹⁴. Newer proposed definitions have suggested quantifying oxygenation through assessment of oxygen dissociation curves, or alternatively relying solely on the amount of positive pressure an infant receives at 36 weeks PMA^112,113. One of the greatest needs is a definition and severity scale to systematically describe the continuum of disease severity for at least the first year of life. One promising example is the Chronic Lung Disease of Infancy (CLDi) score¹⁰⁷, although, like BPD, it is based on details of clinical management, such as supplemental oxygen, bronchodilators and the need for rehospitalization. More objective measures, such as pulmonary function testing, require sedation and are not yet practical outside the research setting. A challenge for any classification system beyond the birth hospitalization is to account for other environmental confounders, such as smoking and viral infections.

The 2001 NHLBI workshop definition took the important step of including a BPD severity scale¹⁰⁸, and infants with severe BPD were subsequently found to have higher mortality and rates of adverse outcomes after discharge than those with mild or moderate disease¹¹⁴. However, the definition of severe BPD was overly broad, and combined infants receiving 31% oxygen by low flow nasal cannula with infants on high ventilator settings and receiving drugs for pulmonary hypertension. This issue was partly addressed by the NHLBI 2018 revision, which reserved grade 3 BPD for infants receiving positive pressure or nasal cannula flow >3 l/min in addition to oxygen³. To better refine the approach to defining disease severity for clinical care and research, the BPD Collaboration Group recommended that severe BPD be further subclassified, with infants receiving only nasal cannula oxygen at 36 weeks PMA defined as type 1 severe BPD and infants requiring mechanical ventilation defined as type 2 severe BPD¹¹⁵. The Canadian Neonatal Network added weight to this idea by showing that a more severe BPD diagnosis (that is, at 36 weeks PMA requiring oxygen plus respiratory support versus oxygen only) better predicted serious respiratory morbidity at 18–22 months¹⁴. However, because of the low incidence of type 2 severe BPD in any single centre, long-term outcomes remain poorly understood for this population. In addition, although pulmonary hypertension is a common comorbidity and probably contributes to the pathophysiology of severe BPD, it has not been incorporated in any existing definition¹¹⁶.

The most severe BPD leads to early mortality prior to 36 weeks PMA. Although this outcome was estimated to occur in 3% of the PROP cohort¹¹⁷, these especially high-risk infants have been inconsistently classified using existing definitions. The 2018 NICHD workshop definition proposed a grade 3A BPD for infants with early death between 14 days postnatal age and 36 weeks PMA from persistent parenchymal lung disease and respiratory failure that is not attributable to other neonatal morbidities. Further work and validation of the NICHD BPD risk calculator⁴⁹ is needed to better identify this population of infants as early as possible in their clinical course (ideally before 14 days), as a particularly high-risk group for novel therapies.

Finally, future BPD definitions should provide strong specificity in addition to sensitivity. In the PROP cohort, about 50% of infants with persistent respiratory disease at 1 year of life had no or mild BPD at 36 weeks PMA⁹⁶. Improving predictive value will require better objective measurements and biomarkers of lung injury and BPD and the ability to incorporate antenatal factors, such as IUGR, maternal hypertension, maternal smoking and male sex.

Biomarkers.

Prevention of BPD and long-term cardiorespiratory impairment remains elusive, making identification of early predictive BPD biomarkers to guide new therapies a laudable goal. Preterm infants with differing risks of developing BPD have been enrolled in early prevention trials, sometimes with adverse consequences. A prime example is the trials of high-dose dexamethasone, which was later found to be associated with a higher risk of cerebral palsy and intestinal perforation¹¹⁸. BPD is a heterogeneous disease resulting from multiple pathophysiological processes, with varying magnitudes of impairment of alveolar septation, lung fibrosis and abnormal vascular development and remodelling¹¹⁹. Failure to identify subpopulations at highest risk and with distinct mechanisms of disease (endotypes) contributes to exposure to therapies that are unlikely to benefit or may cause harm to individual patients, as well as scepticism about biologically plausible therapies that may benefit a subset of at-risk infants^120,121. Identifying infants at high risk of BPD at birth is possible, as was shown for the PROP cohort⁹⁶. Infants who have a persistent need for respiratory support beyond 10–14 days of age are at the highest risk of a BPD diagnosis at 36 weeks^96,122. Although persistent oxygen requirement in the first weeks of life may be a clinical predictor of later BPD, it provides little information about disease endotypes or therapeutic interventions that may benefit individual patients.

A good BPD biomarker should be easily measurable, be affordable, use a reliable assay with high analytical sensitivity and specificity, predict later lung disease or serve as a surrogate for long-term respiratory outcomes, elucidate underlying pathophysiological mechanisms (thereby refining BPD risk by endotypes) and be validated in diverse replication cohorts. A validated BPD biomarker offers the potential to reduce health-care costs and the risks of broadly applying ineffective therapies. Biomarkers may be valuable for earlier diagnosis of BPD, enabling initiation of therapies when they may be more effective or avoidance of therapies and their potential hazards¹¹⁹. A validated biomarker may also facilitate the development of tailored, personalized therapeutic approaches and improve patient endotyping prior to study randomization, thereby enhancing safety and efficacy of randomized controlled trials of novel therapies. Conversely, biomarker testing may increase screening costs and result in fewer infants who are eligible for trials, thus requiring larger multicentre collaborative efforts¹²⁰.

Despite their promise, there are many challenges to identifying BPD biomarkers. An important consideration is the optimal biomarker source or specimen type, which could include DNA, microRNA, amniotic fluid, tracheal aspirate, blood, urine and stool. For example, the levels of IL-1β, IL-6, CXCL10 and uric acid are substantially higher in tracheal aspirate but not in plasma of patients with moderate and severe BPD than in that of infants with no BPD¹²³. The best approach to normalizing a sample for volume collected or dilution factors in some specimen types, such as tracheal aspirate and urine, must be determined. The optimal timing for measurement of the biomarker and therefore therapeutic intervention (at birth and/or the first hours, days or weeks of life) must also be ascertained. Normal patterns of biomarker expression must be determined across a range of gestational and postnatal ages to identify patterns that predict disease or deviation from normal development. Determining the appropriate control group is a unique challenge for research in the preterm population. Healthy term infants may not represent an appropriate control when biomarker expression changes with developmental maturation. It is not always clear whether the absolute value or a change in biomarker level over time will yield the most predictive signal or whether to apply cut-off values or quartiles for risk stratification. Last, interactions among various biomarkers must be taken into account; for example, the balance of pro-inflammatory and anti-inflammatory cytokines or pro-oxidants and antioxidants may be a better predictor of BPD and later respiratory morbidity than a single biomarker of inflammation or oxidative stress.

Various biomarkers have been studied in association with BPD but have yet to be validated as predictors of important long-term outcomes (see TABLE 2 for a non-exhaustive list of BPD biomarkers). Ideal biomarkers for BPD will _target causal pathways that have been implicated in BPD pathogenesis^120,121, such as biomarkers of endothelial injury and dysfunction^124,125; growth factors and biomarkers of altered angiogenesis^6,126–131; biomarkers of epithelial injury and fibrosis^132–135; biomarkers of nitric oxide deficiency or insufficiency^{120,136–140}; biomarkers of oxidative stress, oxidant and antioxidant capacity, and redox status^141–143; oxyradical biomarkers^141–144; biomarkers of inflammation and host immune responses^129,145,146; genomic, epigenomic, metabolomic¹⁴⁷, microbiomic¹⁴⁸, methylomic and transcriptomic biomarkers^119,149; lung function measures^119,150; MRI of the lungs for measurement of perfusion, blood flow, ventilation, gas exchange, mechanics and early changes in alveolar structure¹¹⁹; and biochemical¹⁵¹ and physiological biomarkers of vascular injury and pulmonary hypertension^{80,82,123,140,152}. A 2013 systematic review identified serum KL6, CC16 and neutrophil gelatinase-associated lipocalin, and end-tidal carbon monoxide (etCO), as performing well in predicting BPD in preterm infants¹⁵³. Identification of early markers of endothelial or vascular injury may lead to mechanistic interventions to support normal vascular and alveolar growth and prevent BPD.

Rare single-gene diseases with a BPD-like phenotype probably occur in preterm infants. In the absence of a high suspicion of such a single-gene pulmonary disease, genomic analysis is unlikely to be useful as a screening biomarker. To date, only a few underpowered genome-wide association studies (GWAS) on BPD have been published^149,154 (reviewed in REF.³⁷). Very large genomic studies with international or multicentre participation and including multiple ethnicities, antenatal risk factors and careful definition of phenotypes, may provide important clues about pathogenetic pathways and BPD endotypes. Although identification of biomarker-defined BPD endotypes holds promise for a new era of BPD prevention research, none of the biomarkers discussed above have been effectively used. Harnessing the promise of biomarkers will depend on identification and validation of affordable, reproducible, point-of-care or rapid biomarker tests and advances in computational biology and artificial intelligence¹⁵⁵.

Avoiding intubation and ventilation.

Attempts to avoid lung injury in extremely preterm infants have led to recommendations to avoid intubation and ventilation in the first minutes of life and a preference for non-invasive respiratory support¹⁵⁹. Besides nasal continuous positive airway pressure (NCPAP), nasal intermittent positive pressure ventilation (NIPPV) and high-flow nasal cannulae have become more popular, but the efficacy of these respiratory support methods in reducing the rates of BPD has been modest^160–163. Exogenous surfactant therapy improves neonatal outcomes, including reducing mortality, air leak and the combined outcome of death or BPD^164,165. Traditionally, surfactant has been administered using an endotracheal tube. A number of techniques have been developed for the delivery of surfactant to infants who are managed with NCPAP, including less-invasive surfactant administration (LISA)¹⁶⁶, which uses McGill’s forceps to guide a feeding tube between the vocal cords. Alternatively, minimally invasive surfactant therapy (MIST)¹⁶⁷ uses a firmer vascular catheter and does not require forceps. A meta-analysis of six randomized trials in 895 infants showed that surfactant delivered using a thin catheter reduced the composite outcome of death or BPD compared with conventional intubation for surfactant delivery¹⁶⁸. Some NICUs have responded to this evidence enthusiastically and implemented routine administration of surfactant to infants on NCPAP. Others have been more cautious and await the results of ongoing larger randomized controlled trials to determine the place of this therapy in modern neonatal intensive care¹⁶⁹. Alternative approaches to surfactant administration, such as nebulization¹⁷⁰ and administration into the pharynx or via a laryngeal mask airway¹⁷¹, have potential but require further rigorous assessment before general use.

Based on evidence that extremely preterm infants exhibit persistent surfactant dysfunction, late or repeated surfactant administration has been explored. Although the TOLSURF trial showed no difference in survival at 36 weeks between infants without BPD receiving late surfactant administration versus no surfactant, there was a trend toward improvement at 40 weeks (with >20% decrease in the number infants requiring respiratory support at 40 weeks) and a substantial decrease in the number of infants who required home respiratory support during the first year^172,173, an issue that is very important for parents because of the level of stress they experience.

Caffeine.

Treatment with respiratory stimulants, such as caffeine, is the best-evaluated treatment for reducing BPD risk. The Caffeine for Apnea of Prematurity (CAP) trial demonstrated that caffeine administration reduces the risk of BPD and shortens the duration of ventilation with an endotracheal tube and exposure to supplemental oxygen¹⁷⁴. Importantly, these short-term benefits are followed by an improvement in cognitive and motor outcomes in infants at 18 months of age¹⁷⁵. Important questions about caffeine remain, including whether a higher dose and prolonged treatment lead to further improvements in important outcomes.

Corticosteroids.

Antenatal steroid therapy represents a major advance in perinatal care, which has resulted in reduced mortality, respiratory distress syndrome, intraventricular haemorrhage and necrotizing enterocolitis, although it does not reduce rates of BPD¹⁷⁶.

Corticosteroids are powerful downregulators of inflammation and have been widely used postnatally to prevent and treat BPD. Concerns about adverse effects of these agents, such as neurodevelopmental impairment, mostly with dexamethasone, in the early 2000s led to recommendations against their use¹⁷⁷. Further evidence has emerged suggesting that the benefits may outweigh the dangers when the risk of BPD is high (>50%)¹⁷⁸. In a small randomized controlled trial, a short course of dexamethasone delivering a lower initial and total dosage than used previously led to a reduction in the duration of ventilation and was not associated with long-term neurodevelopmental impairment^179,180. The current decade has seen the pendulum swing back in favour of the prescription of corticosteroids. A European survey showed that 13.9% of infants born between 24 and 29 weeks of gestation were prescribed postnatal corticosteroids¹⁸¹. However, substantial variation in prescription occurred in practice (3.1–49.4%) across the regions surveyed, which was not explained by neonatal characteristics. The authors suggest that this may reflect different interpretations of the evidence, in particular for infants at highest risk of BPD.

Other attempts to improve the risk/benefit ratio of steroid therapy have involved evaluation of other corticosteroids and routes of administration. Based on extensive studies by Watterberg et al.^{104,182–184}, the PREMILOC study compared the early use of 10 days of hydrocortisone therapy to placebo in 523 infants <28 weeks of gestation¹⁸⁵. Unfortunately, the trial was stopped at 66% of planned recruitment because of financial considerations. The primary outcome, survival without BPD, was higher with hydrocortisone treatment than with placebo (60% versus 51%, P = 0.04). Follow-up of survivors at 2 years of age revealed no important differences in rates of neurodevelopmental impairment¹⁸⁶. A smaller trial evaluated the effect of hydrocortisone administered for 22 days in 372 infants born <30 weeks of gestation who were ventilator-dependent between day 7 and day 14 of life¹⁸⁷. There were no important differences in rates of death or BPD between the hydrocortisone and placebo groups. The most recent individual patient data (IPD) meta-analysis including four randomized controlled trials (982 infants) suggests that early low-dose hydrocortisone therapy increases survival without BPD in very preterm infants, despite an increased risk of intestinal perforation (when given in association with indomethacin) and late-onset sepsis¹⁸⁸.

Avoidance of systemic administration underpinned the NEuroSIS trial of inhaled budesonide^189,190. Although budesonide reduced the rate of BPD, a follow-up study showed no improvement in rate of death or disability and an increased risk of mortality compared with placebo. In a randomized controlled trial of 265 infants weighing <1,500 g who were treated with combination budesonide and surfactant (Survanta) or surfactant alone administered intratracheally, rates of death or BPD were 66% with surfactant alone and 42% with the combination therapy¹⁹¹. Follow-up into infancy revealed no differences in rates of neurodevelopmental impairment. Although promising, uncertainty about the optimal dose persists¹⁹² and this intervention needs to be tested further in larger, multicentre trials.

Inhaled nitric oxide.

Inhaled nitric oxide (iNO) is an effective, widely used treatment for neonatal pulmonary hypertension. Studies in animal models showed that NO is required for normal lung development and that replacement therapy may benefit the injured lung^193,194. However, these benefits have not been replicated in human infants. The authors of an IPD meta-analysis that included 14 randomized controlled trials and 3,430 preterm infants suggested that routine use of iNO in this population could not be recommended¹⁹⁵. Observations that the effectiveness of iNO might vary by ethnicity led to an IPD meta-analysis examining this question. A pooled analysis of three randomized controlled trials showed an interaction between ethnicity and the outcome death or BPD. A reduction in rate of this outcome was found in the African American population but not the white or Hispanic groups¹⁹⁶. This observation may be regarded as hypothesis-generating and an appropriate basis for further prospective randomized controlled trials¹⁹⁷.

Oxygen saturation _targeting.

Balancing the risks and benefits of oxygen therapy has been a long-term, unrealized goal of neonatal intensive care. Currently, oxygen administration to preterm infants is titrated to achieve certain oxygen saturation ranges. An IPD meta-analysis of five high-quality randomized controlled trials comparing high (91–95%) and low (85–89%) saturation ranges provides useful guidance for clinicians¹⁹⁸. The higher range was associated with a reduced risk of death and necrotizing enterocolitis but no difference in the primary outcome, the combined rate of death or major disability at a corrected age of 18–24 months. Treatment for retinopathy of prematurity was more common in infants in the high than in the low saturation _target group but there was no difference in the rates of blindness. Importantly, there was no important difference in the rates of BPD between the two treatment groups.