Authors: Matthew J. Kielt (1Comprehensive Center for Bronchopulmonary Dysplasia, Nationwide Children’s Hospital and Department of Pediatrics, The Ohio State University, Columbus, OH, USA.), Timothy D. Nelin (2Division of Neonatology, Department of Pediatrics, Children’s Hospital of Philadelphia and University of Pennsylvania Perelman School of Medicine, Philadelphia, PA, USA.), Steven H. Abman (3Section of Pulmonary and Sleep Medicine and the Pediatric Heart Lung Center, Department of Pediatrics, University of Colorado Anschutz Campus, Aurora, CO, USA.), Leif D. Nelin (1Comprehensive Center for Bronchopulmonary Dysplasia, Nationwide Children’s Hospital and Department of Pediatrics, The Ohio State University, Columbus, OH, USA.)
Categories: Article
Source: Journal of perinatology : official journal of the California Perinatal Association
Authors: Matthew J. Kielt, Timothy D. Nelin, Steven H. Abman, Leif D. Nelin
Bronchopulmonary dysplasia (BPD) remains a prevalent and complex complication of preterm birth, yet current definitions fail to capture the heterogeneous and evolving nature of the disease. Longitudinal studies reveal persistent pulmonary disease throughout the lifespan. Despite progress in identifying phenotypes, biomarkers to characterize disease endotypes to guide effective and precise therapies remain elusive. Precision medicine approaches, including multicenter deep phenotyping and integration of multi-omic data, are essential to identify meaningful disease subtypes that inform individualized care. This perspective traces the evolution of BPD definitions, outlines their limitations, and presents a path forward focused on collaborative data networks, enriched trial designs, and longitudinal outcome measures. Recognizing BPD and subsequent cardiopulmonary disease related to prematurity as a lifelong disease, not just a NICU outcome, is critical to improving long-term care and developing targeted interventions for this vulnerable population.
Northway and colleagues were the first to describe bronchopulmonary dysplasia (BPD) in 1967, with the diagnosis made in preterm infants treated with invasive mechanical ventilation on the basis of serial changes in clinical course and radiologic stages, as well as with post-mortem histologic findings [1]. In this landmark study, BPD was defined in a very different patient population than the extremely preterm infants who are currently managed in our neonatal intensive care units (NICU). In the “pre-surfactant era”, early preterm survivors had a median birthweight and gestational age at birth of 2234 g and 34 weeks, respectively, and patients that subsequently developed BPD required prolonged invasive mechanical ventilation with high supplemental oxygen (FiO2 > 0.80) [1]. Today, we still use the term “BPD” to characterize persistent and late respiratory findings, and now with modern therapies in the “post-surfactant era”, the diagnosis of BPD is most commonly observed in extremely preterm infants [2–4] and is generally applied to infants born at <32 weeks gestation [2, 5].
As the respiratory care of extremely premature infants has changed dramatically [6], there have been on-going modifications to the definition of BPD to reflect the disease’s changing epidemiology. These modified definitions have largely been based on treatment given at different time points to operatively, not pathophysiologically, define BPD. In 1978 the NHLBI hosted a workshop which defined BPD based on supplemental oxygen dependence at 28 days of life and abnormal radiological findings at 30 days of life [7]. In 1988 Shennan [8] proposed that supplemental oxygen requirement at 36 weeks post menstrual age (PMA) was a more useful predictor of abnormal pulmonary outcome in the first 2 years of life than earlier timepoints. In 2001, an NIH Workshop was convened, and the definition of BPD was again revised to account for disease severity based on levels of respiratory support at 36 weeks PMA in infants that required at least 28 days of supplemental oxygen therapy [9]. This definition was widely used for the next 2 decades, while the debate over definitions of BPD continued. An NICHD Workshop in 2016 proposed to add radiological criteria back into the definition, to capture infants with severe chronic lung disease who died prior to reaching 36 weeks PMA, and to include high-flow nasal cannula (HFNC) which came into relatively widespread use in NICUs in this century [5]. In 2017, the BPD Collaborative recognizing that the 2001 NIH Workshop definition of severe BPD [9] was too broad, recommended characterizing severe BPD as type 1 and type 2; where type 1 corresponds to need for non-invasive positive pressure respiratory support, while type 2 corresponds to need for invasive mechanical ventilation at 36 weeks PMA [10]. In 2019 the NICHD Neonatal Research Network published data examining 18 different potential definitions of BPD using respiratory support at 36 weeks PMA to predict death or serious respiratory morbidity through 18–26 months corrected age [2]. Jensen et al. [2] found that a definition using 3 grades based on the receipt of low-flow nasal cannula (grade 1), non-invasive positive pressure (grade 2), or invasive mechanical ventilation (grade 3) at 36 weeks PMA had the best predictive value, although the c-statistic was only 0.79. This definition has become widely used in the literature based on its unambiguous criteria and relatively good predictive value. Nonetheless, whether these changes will sufficiently enhance clinical care, research, and the development of precise therapies are still being explored.
We focus on infants diagnosed with BPD in this review, however given the issues with current definitions and data suggesting that preterm birth compromises lung function later in childhood and adolescence we need to consider that even in former preterm infants without the diagnosis of BPD there may be long term pulmonary sequalae [11]. Vollsaeter et al. [12] reported that young adults born extremely preterm had lower FEV1 and FEF25–75 at 18–25 years of age than did term controls. Simpson, et al. [13] found in a cohort of preterm infants without BPD that lung function deteriorated between 4 and 12 years of age with a decrease in z-score for FEF25–75 of −0.06/year and 73% of these children reported wheezing during their lifetime. In a single center study from the UK it was reported that survivors born extremely preterm had significantly impaired pulmonary function compared to term controls at 19 years of age. A recent Australian study using the Tasmanian Longitudinal Health Study cohort with lung function measured at 53 years of age found that subjects born very preterm (≤28 weeks to 32 weeks gestation) had lower FEV1 and FEF25–75 and more were diagnosed with COPD than in subjects born at term [14]. Recently, investigators in the UK have utilized spirometry from the Respiratory Health Outcomes in Neonates cohort to define an obstructive phenotype (which they termed prematurity-associated obstructive lung disease or pOLD), a prematurity-associated preserved ratio impaired spirometry phenotype (which they termed pPRISm) and a preterm/term control phenotype (FEV1 ≥ lower limits of normal) [15]. Thus, preterm birth puts people at risk for altered pulmonary function well into adulthood without having had the diagnosis of BPD in infancy. This altered pulmonary function found in preterm infants has been termed prematurity associated lung disease (PLD). The European Respiratory Society has implemented a network called PELICAN (Prematurity’s Effects on the Lungs in Children and Adults Network) to provide clinical guidelines and to study these patients [16]. Previously, the American Thoracic Society published an official clinical practice guideline for these patients [17]. In cohorts in the UK, spirometry has been used to begin to delineate phenotypes of PLD including pOLD, pPRISM, and prematurity-associated dysanapsis (pDysanapis) [15, 18]. There is a real need to better understand and continue longitudinal respiratory care to fully capture this pattern of PLD that exists even outside of the formal diagnosis of BPD in order to improve pulmonary outcomes in preterm infants over their entire life course.
A unifying feature of the many BPD definitions proposed since Northway’s original description in 1967 is the reliance on clinical data alone to formally make a diagnosis of BPD. For example, the current most widely used definition [2] is still based solely on respiratory support prescribed for preterm infants born at <32 weeks gestation. Consequently, important radiographic, histologic, and objective assessments of lung function are not considered when diagnosing BPD. Indeed, to our knowledge, BPD is the only human lung disease that is defined by treatment rather than by objective biologic or physiologic criteria. There is a lack of national guidelines for initiating or weaning respiratory support in this patient population. The use of therapies applied in a non-standardized fashion has led to a great deal of variability in diagnosing BPD across providers and institutions [19, 20]. An additional important problem with the current definitions is that they are used at only one point in time (i.e., 36 weeks PMA), while the post-36-week PMA disease trajectories can be highly variable [21, 22]. Some patients with grade 3 BPD are extubated early and sent home within 4–6 weeks of 36 weeks PMA, while others receive continued mechanical ventilation via a tracheostomy tube at discharge from the NICU [21].
BPD has a highly variable presentation at 36 weeks PMA with signs ranging from very mild to extremely severe, and this heterogeneity of disease presentation has led to some identifiable clinical phenotypes as demonstrated in definitions with severity grading [10, 23, 24]. However, a weakness of current definitions is that the spectrum of BPD clinical heterogeneity is not fully captured by the 3 broad severity grades and is not sufficient to capture the variability of long-term respiratory outcomes that follows preterm birth (Table 1). Recently, there has been interest in identifying potentially clinically relevant phenotypes within the “umbrella” diagnosis of BPD. For example, Wu et al. [25] reported that within a single center cohort of patients all diagnosed with grade 3 BPD there were 3 distinct clinical phenotypes, parenchymal lung disease, pulmonary hypertension, and large airway disease. However, in this cohort the most common presentation was the overlap of all 3 clinical phenotypes, while the risk for adverse outcomes increased with the number of phenotypes present [25]. McAnany et al. [26] replicated these findings in another single center cohort of patients with grade 3 BPD. However, these phenotypes were not inclusive of small airway disease, chest wall mechanics, control of breathing, and other clinical factors that impact outcomes.
Importantly, Shepherd et al. [27] found that infant pulmonary function testing (iPFT) identified 3 distinct iPFT phenotypes (obstructive, restrictive, and mixed), and that these iPFT phenotypes were associated with in-hospital outcomes such as tracheostomy and days on mechanical ventilation. A recent study of serial lung function studies of preterm infants that included assessments of airway and distal lung function and lung diffusion capacity (DLCO) during the first year of life demonstrated distinct physiologic phenotypes that included predominantly airways or parenchymal disease, or a variable combination of both features [28]. Higano et al. [29] developed an MRI score for infants with BPD and found that it could predict some short-term outcomes and had real potential to phenotype disease. Dysanapsis has also been described in some infants and children with a history of preterm birth with or without the diagnosis of BPD, which may represent patterns of differential growth of airways and parenchyma over time [30, 31]. Further, another important phenotype within BPD is pulmonary vascular disease. Pulmonary hypertension complicates up to 40% of severe BPD cases and markedly increases the risk for adverse outcomes [10, 32]. Furthermore, there is evidence that antenatal perturbations in pulmonary vascular development also result in the alveolar simplification typical of BPD, suggesting that not only is the vasculature important both as a phenotype in established BPD, but also involved in the pathogenesis of at least some forms of BPD [33–35]. Thus, the various phenotypes found within the diagnosis of BPD re-enforce the need for multidimensional phenotyping to fully understand disease trajectories and optimally apply therapies. It is also reasonable to assume that understanding these phenotypes will lead to important insights into the heterogenous pathogenesis of BPD.
BPD should probably not be considered a single disease entity but rather a collection of multiple, overlapping disorders. These likely include, but are not necessarily limited to, abnormalities in one or more of the airway structure and function, pulmonary vasculature, lung parenchymal growth, and control of breathing. As discussed above clinical and physiologic studies have identified distinct but often coexisting phenotypes [25–28]. A more precise early characterization of clinical BPD or BPD sub-phenotypes could help with developing precision therapeutics for established disease, focusing on “treatable traits”, with the potential to identify novel preventive strategies that impact lung disease over the life course.
There is an increasing recognition that preterm infants with, or even without, the diagnosis of BPD can exhibit early changes in lung function and structure that can persist throughout childhood and follow an altered trajectory throughout the lifespan [36–38]. Northway, et al. [39] provided one of the first descriptions of persistent altered pulmonary function abnormalities consistent with obstructive lung disease and often with bronchial hyperreactivity in a small cohort of adolescent and young adults who had been diagnosed with BPD in the NICU. There have since been multiple reports of impaired pulmonary function in older children, adolescents, and adults who had BPD diagnosed as infants. These reductions in pulmonary function are consistent with a persistent obstructive pattern, with lower forced expiratory volume at 1 second (FEV1), forced vital capacity (FVC), and forced expiratory flow 25–75% (FEF25–75) [40–44]. In addition, there have been several small reports of persistent alterations in lung structure using various imaging modalities in young adults diagnosed with BPD during their NICU stay with most demonstrating emphysematous changes [45–48].
In healthy individuals, lung function improves with age and growth during childhood, peaks in early adulthood, and then slowly declines with age but never reaches a level that results in symptomatology later in life [49]. In contrast, in individuals with BPD there is mounting evidence that lung function will be lower throughout childhood and never reach the young adulthood peak seen in healthy individuals [49–51]. Thus, as lung function declines with age, individuals with BPD start from a lower peak lung function and are thus more likely to reach symptomatic thresholds in adulthood, particularly when compounded by exposures such as smoking or other adverse stimuli [36, 50]. Martinez [51] reported that there was a significant relationship of the quartile of maximal flow at functional residual capacity (VmaxFRC) measured shortly after birth with the post-bronchodilator FEV1:FVC ratio in the same subjects at 26 to 32 years of age. Moschino et al. [43] followed VmaxFRC, FEV1, and FEF25–75 in 17 preterm infants diagnosed with moderate-to-severe of BPD from 2 years of age until 24 years of age and found that the z-scores for VmaxFRC demonstrated significant obstruction at 2 years of age and the FEV1 and FEF25–75 showed no improvement in z-score over the study period. Thus, it is highly likely that at least a subset of infants diagnosed with BPD in the NICU will go on to develop symptomatic chronic obstructive pulmonary disease (COPD) in adulthood.
Although, the current definitions of BPD have moved the field forward, it is becoming increasingly clear that BPD is a complex lung disease with a highly variable presentation, a multi-factorial pathogenesis, and a time-varying disease progression. The only unifying feature shared by every patient diagnosed with BPD is preterm birth at <32 weeks gestation with subsequent need for some form of respiratory support at 36 weeks PMA. As such, BPD is too often considered an outcome rather than a life-long pulmonary disease that requires continued care and therapy after the initial NICU stay. Current definitions are NICU focused using non-standardized treatment approaches to define and diagnose BPD at only one time-point early in the patient’s life course. These definitions have worked relatively well for identifying many of the infants with the most severe forms of the disease during their NICU admissions, which has contributed to improvements in survival rates for infants with established severe BPD prior to NICU discharge. However, as we have moved the field forward it has become clear that current definitions are not adequate for optimal follow-up and continued care of BPD patients throughout life. To meaningfully advance outcomes in BPD, there is an urgent need for precision approaches grounded in the objective analysis of large-scale clinical and multi-omic datasets to identify phenotypes, define endotypes, and uncover biomarkers of disease progression and therapeutic response across the life course.
Using the current definitions of BPD the diagnosis cannot technically be made until 36 weeks PMA. However, in preterm infants it often becomes clear to the clinicians that the patient will have BPD well before they reach 36 weeks PMA. In fact, several prediction models demonstrate that BPD risk can be estimated relatively early based on clinical and demographic characteristics [52, 53]. While these models highlight that many of the determinants of BPD are well defined in preterm infants, they also reveal a critical despite fairly accurate prediction of BPD, there are no tools to predict disease trajectory in BPD. Hence, developing criteria that can identify those patients that have BPD prior to 36 weeks PMA is a key element for improving long-term outcomes while these patients are still hospitalized in the NICU. For example, Sammour et al. [54] described that the initiation of “BPD” care in extremely preterm infants who remain on invasive mechanical ventilation for the first 28 days of life (long before 36 weeks PMA) was associated with shorter length of stay and reduced need for tracheostomy.
As it has become increasingly clear that prematurity with or without the diagnosis of BPD can lead to life-long deficits in lung function, BPD definitions need to account for how the disease improves (or worsens) over time, particularly once infants are discharged from the NICU. The diagnosis of BPD is often lost as the patient ages and older BPD patients with persistent obstruction often end up with a diagnosis of asthma, which in many cases may be a misclassification. Many BPD patients have a form of chronic obstructive lung disease throughout childhood, which differs from asthma [42, 55]. Thus, developing an understanding of post-NICU disease trajectories, phenotypes, and endotypes in established BPD over the life course is imperative to improving the life-long outcomes of patients with BPD.
The current definitions of BPD can be highly arbitrary and, as such, may lack biologic relevance. Consequently, there exists a pressing need for objective analysis of phenotypes and endotypes to develop biomarkers that can be used in defining BPD and for defining sub-phenotypes within the current umbrella diagnosis of BPD. There have been reports, as discussed above, of potential clinical phenotypes within the current umbrella of established BPD. However, the largest phenotypic groups were those that exhibited all the proposed clinical phenotypes in these studies [25, 26]. Implementing objective clinical phenotyping is necessary to begin the process of precision therapies in established BPD.
While identifying objectively defined clinical phenotypes is a critical first step toward precision therapy in established BPD, clinical phenotypes alone do not reveal underlying mechanisms or therapeutic targets capable of restoring lung function or reversing arrested lung development. To achieve this, there is a need to define biologically-informed endotypes or genotypes in established BPD that associate with specific disease trajectories, clinical outcomes, and/or responses to therapy. Although numerous studies have explored early-life endotypes associated with BPD development in preterm infants (reviewed in 56–59) and genetic studies have provided conflicting results in terms of the heritability of BPD susceptibility [60–68], there are no endotypes identified that can be used either as predictive molecular biomarkers or as a target for preventative therapies at this time. More importantly in terms of established BPD there have been few, if any, studies done trying to identify specific endotypes in established BPD to predict disease course or responses to therapies [69]. This constitutes a major knowledge gap in the field.
Even in the early phases of discovery, the identification of potential clinical phenotypes within the umbrella diagnosis of BPD demonstrates that heterogeneity is a hallmark of the disease. BPD, by definition, occurs exclusively in preterm infants, therefore it is important to consider the heterogenous, and still poorly understood, causes of preterm birth [70]. Some of the underlying causes of preterm birth may directly impair lung development, contributing to the variability in the presentation of BPD and the progression of established BPD. Moreover, studies by Li et al. [65] and Hamvas et al. [68] found that rare genetic variants were implicated in BPD pathogenesis in some preterm infants highlighting the complex and multifactorial nature of the disease.
Given the heterogeneity of established BPD, continued reliance on small, single-center studies is unlikely to meaningfully advance the field. As is true for most pediatric chronic diseases, BPD is a relatively rare occurrence and will require coordinated, large-scale efforts. To move beyond incremental gains, multi-center research networks are needed to support clinical data gathering, such that clinical deep phenotyping on a large scale can occur. There have been several such multi-center networks developed within neonatology and pediatric pulmonology. Some examples include the Neonatal Research Network, the Children’s Hospitals Neonatal Consortium (CHNC), the Pediatric Pulmonary Hypertension Network, and the US National Registry for Childhood Interstitial and Diffuse Lung Disease. These types of networks need to be fostered and maintained for the purpose of identifying novel therapeutic strategies for established BPD. Recently the CHNC developed a severe BPD focus group that uses the clinical data from the NICU stay available in the CHNC database for observational studies in established BPD [21, 32, 71]. Similarly, the BPD Collaborative, a collaboration between neonatologists and pediatric pulmonologists, has developed a Registry that spans childhood for patients with grade 2 and 3 BPD [20, 72–76]. Large and rigorous observational studies in cohorts of patients with established BPD are essential for developing an understanding of the clinical spectrum of disease trajectories found within the current diagnosis of BPD. In addition to understanding the clinical spectrum of disease trajectories, there are numerous therapies that are currently prescribed to infants with established BPD and care practices that lack evidence from multicenter randomized controlled trials. Networks such as the Neonatal Research Network, the BPD Collaborative, and CHNC offer platforms to conduct pragmatic clinical trials of existing therapies (such as the use of diuretics and bronchodilators) and common care practices (such as ventilator approaches and enteral feedings).
To reach the goal of precision therapy for established BPD, clinical trials networks dedicated to BPD must be established. Disease specific clinical trials networks have been highly successful in developing specific therapies that have improved outcomes in other diseases. An excellent example of this in a chronic pediatric disease is the Cystic Fibrosis Foundation Therapeutic Development Network (CFF TDN), which has expediated the development and implementation of a series of pivotal clinical trials. The CFF TDN has now partnered globally with the European CF Society Clinical Trials Network and the CF Canada Accelerating Clinical Trials to establish a unified scientific review process that enables prioritization of studies across the networks in response to the growing therapeutic pipeline in CF [77]. While such networks have been a feature of adult-oriented lung research for decades, BPD-centered research networks remain nascent in their stages of development. Since BPD is a rare disease, clinical trials in established BPD are limited by small sample sizes, often requiring multi-center randomized controlled trials (RCTs) that span many years. This has substantially slowed the development of precision therapeutic approaches.
To accelerate progress, the field must consider the adoption of innovative clinical trial designs that extend beyond the classic RCT examining one intervention in equal sized control and intervention groups. The COVID 19 pandemic was associated with the rapid adoption of various types of innovative clinical trial designs across large networks which resulted in an ability to quickly move potential COVID therapies from study to clinical use, examples of these include the WHO Solidarity Trial Consortium [78] and the REMAP-CAP trial [79]. Furthermore, innovative trial designs within large networks are being used in adult asthma to accelerate development of therapies for severe and exacerbation-prone asthma. For example, the PrecISE Asthma Network uses a platform trial approach wherein a control group can be used for more than one intervention group [80, 81].
Another important aspect of innovative trial design is the incorporation of enrichment strategies to optimize targeting of therapeutic interventions to those phenotypes or endotypes that are most likely to benefit. Enrichment refers to the selection of a patient cohort that is most likely to benefit from a given study intervention rather than using an inclusive cohort. Enrichment studies can use prognostic and/or predictive enrichment. Prognostic enrichment refers to choosing a cohort that is most likely to experience an adverse outcome. Prognostic enrichment decreases the numbers needed to detect intervention-induced differences in the selected outcome but does not address likely inherent differences among subjects in therapeutic responses. As discussed above, end-points in established BPD need to move beyond NICU outcomes and consider longer term outcomes throughout childhood. However, designing a trial where the end-point is an outcome at school age or beyond (i.e., FEV1) is exceedingly difficult for a variety of reasons, and may not even be relevant upon study completion due to the very long study period required. This problem can only be overcome by developing early objective predictive biomarkers of later outcomes that can then be used as surrogate outcomes for prognostic enrichment of clinical trials studying longer term consequences of BPD. Predictive enrichment addresses sample size in terms of selecting patients that are most likely to respond to a given trial intervention. However, our current ability to identify subjects most likely to respond to a given treatment or therapy in established BPD is very rudimentary. Thus, the development of biomarkers predictive of therapeutic response as discussed earlier in this review are required for developing predictive enrichment strategies for clinical trials in established BPD.
Beyond this framework, structured long-term follow-up is essential to address the persistent morbidities associated with BPD. Dedicated BPD follow-up clinics, typically staffed by both neonatologists and pediatric pulmonologists, provide surveillance for growth, lung function, pulmonary hypertension, and neurodevelopment [36]. However, considerable variation exists in clinic availability, scope, and frequency, reflecting a critical need for consensus approaches to longitudinal monitoring. Ultimately, interdisciplinary, research-guided consensus is needed to define what assessments are performed, how often children are followed, and which providers are involved to ensure comprehensive long-term care for those with BPD or PLD.
Decades of preclinical and clinical research have identified multifactorial causes, both antenatal and postnatal, that increase the risk for BPD. Although a great deal has been learned regarding the pathogenesis of BPD, this knowledge has not yet translated into effective preventive therapies for the disease. The treatment of established BPD has only relatively recently received increasing attention by both neonatologists and pediatric pulmonologists. There are many barriers to developing much needed precision approaches to the management and treatment of established BPD relatively small numbers of patients, lack of evidence for current standard of care practices, large therapeutic variability in approaches to established BPD, lack of clear clinical phenotypes and how they affect disease trajectories, and an almost complete lack of biomarkers for either outcomes or responses to therapies.
To develop precision therapies with the potential to substantially improve long-term outcomes, we, as a field, need to move beyond our current definitions and develop objective diagnostic criteria that capture the life-long disease trajectories associated with established BPD. The first step toward precision medicine for established BPD is to use multi-omic approaches in deeply phenotyped large clinical cohorts to identify relevant and objectively identifiable phenotypes and their related endotypes. The second step, which should follow quickly after the first step begins to yield data, is to design multi-center clinical trials using innovative design strategies to optimize sample size to rapidly assess potential precision interventions in appropriately enriched study cohorts. These two steps are eminently achievable in our age of innovative data analytics and clinical translational science. To drive this vital initiative forward in an efficient manner requires a close collaboration between multi-site networks, funding agencies, parent groups, and the pharmaceutical industry.