Association of the Respiratory Severity Score with Bronchopulmonary Dysplasia-Associated Pulmonary Hypertension in Infants Born Extremely Preterm

Objective: To test the hypothesis that elevations in the respiratory severity score (RSS) are associated with increased risk of bronchopulmonary dysplasia-associated pulmonary hypertension (BPD-PH). Study Design: Retrospective cohort study of infants born extremely preterm admitted to a BPD center between 2010–2018. Echocardiograms obtained ≥36 weeks’ post-menstrual age (PMA) were independently adjudicated by two blinded cardiologists to determine the presence/absence of BPD-PH. Multivariable logistic regression estimated the association between RSS with BPD-PH. Result: BPD-PH was observed in 68/223 (36%) of subjects. The median RSS at time of echocardiography was 3.04 (Range 0–18.3). A one-point increase in RSS was associated with BPD-PH, aOR 1.3 (95% CI 1.2–1.4), after adjustment for gestational age and PMA at time of echocardiography. Conclusion: Elevations in the RSS were associated with a greater risk of BPD-PH. Prospective studies are needed to determine the validity and performance of RSS as a clinical susceptibility/risk biomarker for BPD-PH.


Introduction
Bronchopulmonary dysplasia (BPD) is the most common morbidity experienced by infants born extremely preterm. 1 Approximately 20-40% of infants with BPD develop pulmonary vascular disease, which manifests clinically as BPD-associated pulmonary hypertension (BPD-PH). [2][3][4][5] The risk of BPD-PH is inversely proportional to gestational age at birth and a diagnosis of BPD-PH is associated with shortand long-term morbidity with mortality estimates ranging from 12-38%. 2,4,[6][7][8] Given the risk of morbidity and mortality associated with BPD-PH, 9 consensus guidelines recommend obtaining a screening echocardiogram at 36 weeks' post-menstrual age (PMA), which corresponds to the time of formal BPD diagnosis. 10,11 Nonetheless, echocardiogram utilization for the diagnosis of BPD-PH varies from 26-86% between United States Children's Hospitals neonatal intensive care units (NICUs) 4 and the optimal timing of repeat BPD-PH echocardiography after 36 weeks' PMA remains unknown.
Underutilization of echocardiograms to evaluate for BPD-PH may result in delayed diagnosis, delayed initiation of therapy, and increased morbidity for patients. However, the clinical utility of echocardiographic assessment of BPD-PH may be limited 12 and overutilization of this diagnostic tool may result in unnecessary exposure to invasive diagnostics, including cardiac catheterization, 12 and increased costs related to clinician-driven tests. 13 Consequently, there is a real need for improved precision regarding the utilization of echocardiography for the evaluation of BPD-PH. Identifying pragmatic BPD-PH susceptibility/risk biomarkers that can be measured non-invasively and at no cost to the patient may plausibly improve BPD-PH risk strati cation and reduce variability in echocardiogram utilization in this population.
The respiratory severity score (RSS) is a non-invasive biomarker that is equal to the product of mean airway pressure (MAP) and the fraction of inspired oxygen (FiO 2 ). 14 An elevated RSS (modi ed to account for patients on non-invasive modes of respiratory support) has previously been shown to predict adverse in-hospital outcomes for infants with established severe BPD. 15 Whether an elevated RSS is associated with an increased risk for BPD-PH remains unknown. Therefore, in this study, our objective was to test the hypothesis that elevations in the RSS, measured concurrent with echocardiography, are associated with increased risk of BPD-PH in infants born extremely preterm.

Materials/Subjects and Methods
We performed this retrospective cohort study in accordance with Strengthening the Reporting of Observational Studies in Epidemiology guidelines. 16 Included patients were born < 28 weeks' gestation who were admitted to the 24-bed BPD unit at Nationwide Children's Hospital (NCH) between January 1, 2010 and December 31, 2018. Included patients had diagnosis of severe BPD, as de ned by 2001 National Institutes of Health Consensus criteria, 10 which were the criteria for enrollment into our BPD program during the study period. Included subjects had at least one echocardiogram performed at or after 36 weeks' PMA (+/-1 week). Excluded patients were those with congenital heart disease (excluding atrial septal defect, ventricular septal defect, and patent ductus arteriosus), a genetic syndrome or aneuploidy, a major congenital malformation, and those patients with missing clinical data needed to calculate the RSS. Clinical characteristics of included patients were gathered from the medical record.
The Institutional Review Board at NCH approved this study with a waiver of consent.
We calculated the RSS using respiratory support data that were readily available at the time of echocardiography. As described in prior studies, we calculated the RSS on subjects treated with invasive mechanical ventilation, non-invasive positive airway pressure, low-ow nasal cannula, and room air. 17 For subjects treated with invasive mechanical ventilation, the RSS is equal to the product of MAP and FiO 2 where MAP was calculated based on prescribed ventilator settings based on the following equation: MAP = PEEP + [(PIP-PEEP) x (t i /t i + t e )] and PEEP is the positive end-expiratory pressure, PIP is the peak inspiratory pressure, t i is the inspiratory time, and t e is the expiratory time. Brie y, to calculate the RSS for patients on non-invasive positive pressure ventilation, we assumed MAP = PEEP+((PIP-PEEP)×(t i /(t i +t e )), for patients on continuous positive airway pressure (CPAP), we assumed MAP = CPAP, for patients on high-ow nasal cannula ≥ 2 liters per minute (LPM) we assumed MAP = ow in LPM, for patients on lowow nasal cannula < 2 LPM, we assumed MAP = 1 and for patients on room air, we assumed MAP = 0. 17 The primary outcome for this study was echocardiographic evidence of BPD-PH ≥ 36 weeks' PMA, de ned dichotomously (yes/no). At our center, infants born extremely preterm with severe BPD were screened for BPD-PH at approximately 36 weeks' PMA. For infants with no prior history of BPD-PH, repeat screening echocardiograms were performed every 1-2 months until hospital discharge. For infants with BPD-PH, echocardiograms were typically obtained every 1-2 weeks until resolution of BPD-PH, de ned as 2 consecutive echocardiograms without evidence of BPD-PH. Consistent with previous studies, 8, 17 BPD-PH was determined as a binary variable using the following hierarchy: 1) estimated right ventricular systolic pressure (RVSP) greater than 40 mmHg; 2) RVSP/systemic systolic blood pressure greater than 0.5; 3) any cardiac shunt with bidirectional or right-to-left ow; 4) or any degree of ventricular septal wall attening. RVSP was determined from the tricuspid regurgitation jet velocity using the modi ed Bernoulli equation and assuming a right atrial pressure of 5 mm Hg. 8, 18 During the study period, brain natriuretic peptide (BNP) and N-terminal pro b-type natriuretic peptide (NT-pro-BNP) were not routinely assayed at the time of BPD-PH evaluation. Determination of BPD-PH was made following independent adjudication of clinical echocardiograms by 2 blinded pediatric cardiologists. Disagreements regarding echocardiographic evidence of BPD-PH were settled through discussion, with involvement of a third reviewer, as necessary.

Statistical Analysis
Continuous data had skewed distributions and are summarized using median and interquartile range (IQR). Categorical data are presented as number and percent. We compared clinical characteristics for the cohort, strati ed by the study outcome of at least one echocardiogram demonstrating BPD-PH after 36 weeks' PMA using Mann-Whitney U and chi-square tests, for continuous and categorical data, respectively. Generalized linear mixed models with logit link and infant as a random effect were used to estimate the association of RSS with BPD-PH. Odds ratios and predicted probabilities of BPD-PH are presented for both unadjusted and adjusted models. Adjusted estimates included gestational age and PMA at the time of echocardiography as xed effects. BPD severity was not included as a covariate in adjusted models because it is a collider rather than a confounder for the association between RSS and BPD-PH 19 and prior studies have demonstrated the presence of BPD-PH across the spectrum of BPD disease severity. 7 Statistical analyses were performed using R version 4.2.0 (R Foundation for Statistical Computing, Vienna, Austria). Odds ratios were considered statistically signi cant if the 95% con dence interval did not cross 1.
The clinical characteristics of the cohort are presented in Table 1. In general, members of the cohort were born at extremely low gestational ages, born at extremely low birth weights, and were male. Mortality in this cohort was 4%. As compared to subjects without BPD-PH, subjects with BPD-PH were more likely to be born to mothers with a hypertensive disorder of pregnancy, be born at lower birth weights, SGA, be female, have grade III or IV IVH, have echocardiographic evidence of pulmonary hypertension prior to 36 weeks' PMA, have longer lengths of stay, and die (Table 1).  Fig. 3. The precision of the estimated probability decreased as the RSS increased due to fewer cohort subjects with elevated levels of the RSS (Fig. 3).

Discussion
In this study, we observed that an increased RSS, measured concurrent with echocardiography, was associated with an increased risk of BPD-PH in a cohort of infants born extremely preterm. This knowledge is important because infants with BPD-PH are at a 5-fold increased risk of death compared to infants without BPD-PH 9 and it is critical that infants with BPD are screened for BPD-PH in order to promptly diagnosis and treat the disease. 4,8,18,20 Echocardiograms are the most frequently used screening assessment for BPD-PH yet echocardiograms are expensive, 13 may lack speci city for BPD-PH, 12 and are variably utilized across centers. 4 Cardiac catheterization is the gold standard for BPD-PH diagnosis 21 yet this procedure carries procedural risks for infants born extremely preterm with major comorbidities such as BPD. 22 Additionally, there are no reliable biomarkers of BPD-PH. 23 As such, the optimal timing of interval screening for BPD-PH after 36 weeks' PMA remains unknown. Identifying clinical risk factors that can be measured concurrently with echocardiography and identify infants born extremely preterm who are at the highest risk of BPD-PH may increase precision utilization of echocardiography for the evaluation of BPD-PH and facilitate timely initiation of pulmonary vasodilator therapy.
To our knowledge, this is the rst study to identify a pragmatic clinical risk score, measured noninvasively and at no cost to the patient, that is associated with concurrent BPD-PH in infants born  25,26 and necrotizing enterocolitis. 8,27,28 To identify a clinical BPD-PH risk factor that could be measured in close proximity to BPD-PH screening, Gentle and colleagues recently compared the frequency and duration of intermittent hypoxemia events in the week prior to BPD-PH assessment for 80 infants born extremely preterm with and without BPD-PH. In this well-designed, prospective case-control study, the authors observed that the duration of intermittent hypoxemic events with oxygen saturations < 70% predicted BPD-PH with an area under the receiver operator characteristic curve of 0.71. 29 Nonetheless, on adjusted analyses, the authors found that intermittent hypoxemic event characteristics did not reliably discriminate infants born extremely preterm with and without BPD-PH, 29 suggesting a limited utility for these measures as independent predictors for BPD-PH in heterogeneous cohorts of infants born extremely preterm. Notably, in our study, the estimated change in probability of BPD-PH for a one-point increase in the RSS retained signi cance even after adjustment for gestational age and PMA at the time of echocardiogram ( Table 2).
Given that exposure to mechanical ventilation and supplemental oxygen have both been implicated in the pathogenesis of BPD-PH, we speculate that the observed association between elevations in the RSS and an increased risk of BPD-PH has plausible criterion validity. Though speci c mechanisms remain unknown, exposure of the developing preterm lung to a combination of ventilator-induced lung injury and oxidative stress likely contributes to impaired angiogenic signaling, endothelial progenitor cell dysfunction, and disrupted vascular growth in infants born extremely preterm who develop BPD-PH. [30][31][32] Observational studies have identi ed chronic exposure to mechanical ventilation 4 and supplemental oxygen 33 as independent risk factors for the development of BPD-PH. Additionally, for those infants with established BPD-PH, the likelihood of discontinuation of pulmonary vasodilator therapy following NICU discharge is associated with a fewer number of days of invasive respiratory support, suggesting an earlier resolution of disease in those patients with earlier discontinuation of respiratory support. 34 The results of these observational studies support clinical translational ndings that demonstrate an imbalance of anti-angiogenic and pro-angiogenic factors in the developing lungs of infants born preterm who are exposed to mechanical ventilation and supplemental oxygen. [35][36][37] While the associations observed in our study do not support a causal relationship between chronic exposure to mechanical ventilation/supplemental oxygen and BPD-PH, our results suggest that the RSS, which quanti es these exposures, may plausibly serve as a clinical risk score that can be measured longitudinally and used to monitor the risk of BPD-PH across the stages and spectrum of clinical disease. 7, 34, 38 A particular strength of the RSS is that it can be measured across all levels of respiratory support that are used in the treatment of infants with BPD. Consequently, the score can plausibly be used to stratify BPD-PH risk over time as patients gradually wean from respiratory support or worsen in the setting of clinical deterioration.
Though there has been longstanding interest in identifying biochemical biomarkers that are associated with increased risk of BPD-PH, the diagnostic utility of BPD-PH biomarkers remains uncertain. 39 Serum levels of BNP and NT-pro-BNP have been used to risk stratify adults and children with pulmonary arterial hypertension 40, 41 as these peptides are released from the myocardium following wall stress and elevated levels of BNP and/or NT-pro-BNP may plausibly indicate right ventricular strain in the setting of pressure or volume overload. 42 A limited number of retrospective, single-center studies have evaluated BNP and/or NT-pro-BNP as potential biomarkers for BPD-PH in infants born preterm with BPD. [43][44][45][46][47][48] While these studies have observed signi cant elevations in BNP and/or NT-pro-BNP in infants with BPD-PH as compared to infants without BPD-PH, [44][45][46]48 interpretation of the diagnostic accuracy of these studies is limited due to their single-center retrospective designs, relatively small number of study subjects, and limited evaluation of BNP and/or NT-pro-BNP after 36 weeks' PMA at timepoints concurrent with echocardiography. 39,[44][45][46]49 In the largest study evaluating the diagnostic accuracy of BNP for BPD-PH in infants with established BPD (i.e., after 36 weeks' PMA) to date, Avitabile and colleagues identi ed a robust speci city but poor sensitivity for BNP relative to concurrent echocardiography suggesting a modest utility of BNP for the diagnosis and ongoing evaluation of BPD-PH. 48 Published clinical practice guidelines recommend universal echocardiography screening for infants with BPD treated with positive airway pressure at 36 weeks' PMA. 18,50,51 Despite these recommendations, Lagatta and colleagues observed signi cant inter-center variation in echocardiogram utilization for the purpose of BPD-PH diagnosis among infants enrolled in the Children's Hospital Neonatal Database with echocardiogram use rates ranging between 26-86% across the 23 centers that participated in this study. 4 Echocardiograms are among the most utilized and expensive clinician-driven tests performed in the NICU 13 and may be a potential target for expenditure reduction in healthcare spending associated with care provided to infants born extremely preterm. 52,53 To date, the cost-effectiveness of interval screening echocardiograms for infants with BPD-PH has not been studied. Though further studies are needed, we speculate that incorporation of a clinical BPD-PH risk score, such as the RSS, into BPD-PH screening recommendations may safely reduce the utilization of routine screening echocardiograms in low-risk infants (i.e., those with no prior history of BPD-PH and low RSS measurements) and precisely inform targeted BPD-PH screening in those infants at the highest risk of disease.
Our study has important limitations that should be considered. As this was a single-center study performed at a referral center with a dedicated BPD unit, it is possible that our results may not be generalizable. Given the retrospective observational design of the study, there is also the potential for incomplete confounding. Though we included model covariates known to be associated with both respiratory disease severity and BPD-PH, we acknowledge that unmeasured confounders could plausibly modify model effect estimates. Additionally, restriction of the cohort to infants born < 28 weeks' gestation may contribute to selection bias in this study as we potentially excluded a relatively small proportion of infants born between 28-31 weeks' gestation with BPD who develop BPD-PH after 36 weeks' PMA. 4 Lastly, as we de ned BPD-PH dichotomously, we could not evaluate the association of RSS with BPD-PH severity. Our institution does not pursue cardiac catheterization to diagnose BPD-PH and so hemodynamic diagnostics that could be used to stratify BPD-PH severity were not available in this cohort. Given that the accuracy of echocardiographic assessments of BPD-PH severity is limited, 12,18 further studies, potentially in conjunction with advanced diagnostics including cardiac catheterization 12 and cardiac magnetic resonance imaging, 54 are needed to more precisely determine the association of RSS with BPD-PH severity.
In conclusion, our study demonstrated that elevations in the RSS were associated with an increased risk of BPD-PH. Measurement of the RSS approximate to the time of echocardiography may improve risk strati cation for infants born extremely preterm who are evaluated for BPD-PH. Prospective studies that validate the RSS as a clinical susceptibility/risk biomarker are needed to identify the optimal timing of echocardiographic assessments for the evaluation and management of BPD-PH in infants born extremely preterm.  Distribution of timing of echocardiography in the cohort, strati ed by BPD-PH status at the time of assessment.