Hyperinflation and its association with successful transition to home ventilator devices in infants with chronic respiratory failure and severe bronchopulmonary dysplasia

To estimate the association between lung hyperinflation and the time to successful transition to home ventilators in infants with sBPD and chronic respiratory failure. Infants with sBPD <32 weeks’ gestation who received tracheostomies were identified. Hyperinflation was the main exposure. Time from tracheostomy to successful transition to the home ventilator was the main outcome. Kaplan-Meier and multivariable Cox proportional hazards were used to estimate the relationships between hyperinflation and the main outcome. Sixty-two infants were included; 26 (42%) were hyperinflated. Eleven died before transition, and 51 successfully transitioned. Hyperinflation was associated with both mortality (31% vs 8.3%, p = 0.02) and an increased duration (72 vs. 56 days) to successful transition (hazard ratio (HR) = 0.38, 95% CI: 0.19, 0.76, p = 0.006). Growth velocity was similar after tracheostomy placement. In infants with chronic respiratory failure and sBPD <32 weeks’ gestation, hyperinflation is related to mortality and inpatient morbidities.


INTRODUCTION
Severe bronchopulmonary dysplasia (sBPD) affects 20% of preterm infants <32 weeks' gestation [1,2], and confers risks of mortality, prolonged hospital stays, cardiopulmonary morbidities, and neurodevelopmental impairment in early childhood [3]. Approximately 5% of infants with sBPD develop chronic respiratory failure and require chronic mechanical ventilation via tracheostomy. Together, these therapies permit maintaining physiologic stability to follow a trajectory focusing on enhancing growth and physical, cognitive, and social development [4]. The ability to transition to an outpatient, or home, mechanical ventilator (MV) in these infants is a pre-requisite to facilitate safe discharge from the hospital [5,6]. However, the timing and success, as well as patient-level predictors of successful transition to home ventilators, may be uncertain in a subset of these infants.
Chronic respiratory failure with sBPD is often characterized by combinations of heterogeneous lung development, architecture, and dysfunction with cystic and hyperinflated airspaces, shifting atelectasis, and airway malacia [7,8]. With these pathologies, lung hyperinflation may impede flow-triggering on MVs. Though the mechanisms are not clearly delineated, impaired flow-triggering with hyperinflation may lead to inefficient, if not paradoxical, diaphragmatic excursion and asynchronous air-flows with MVs [9]. Despite adjusting the sensitivities of MV's flow-triggering, we have noticed that transitions to home ventilators can frequently be unsuccessful among those infants with significant asynchrony, and many of these infants are also hyperinflated on their chest radiograph (CXR) [10]. In the current project, we seek to identify and understand whether hyperinflation is related to a delayed transition to home ventilators based on the presumed mechanism outlined. Understanding this relationship would lead to more accurate parental counseling and perhaps more targeted preventive and management strategies and that may improve both hyperinflation and perhaps alter the course of affected infants after tracheostomy placement.
We hypothesize that infants with sBPD and hyperinflation have delayed transition to home ventilators after tracheostomy. Our main objective is to estimate the association between time to successful transition to home ventilators and hyperinflation among patients with sBPD <32 weeks' gestation and chronic respiratory failure.

METHODS
Using our center's data in the Children's Hospital Neonatal Database, we reviewed medical records of infants with sBPD <32 weeks' gestation that underwent a tracheostomy between 2010 and 2018 at our center. SBPD was defined as infants receiving >2 L/min via nasal cannula or positivepressure ventilation at 36 weeks' post-menstrual age (PMA) as prior studied [11,12]. We excluded infants who received a tracheostomy prior to transfer to our institution, were born >32 weeks' gestation, had significant congenital anomalies, or who recovered sufficiently to avoid mechanical ventilation post-tracheostomy.
The primary outcome was the time to successful transition to home ventilators, and this was defined as stability for 7 consecutive days after transition from the inpatient MV device. Reversion back to an inpatient MV prior to this time was considered unsuccessful. However minor adjustments in settings (rate, PEEP) were not characterized as failure. At the time of this study, we exclusively transitioned patients with sBPD to an LTV™ (Vyaire Medical, Inc™; Chicago, IL) for their outpatient care. In general, stability was defined as acceptable gas exchange (P v CO2~50 mmHg, F i O 2 < 0.35); respiratory rate (<~60/min while awake); acceptable respiratory effort (minimal dyspnea); absence of intravenous sedative or analgesic medications; peak inspiratory pressures <35 cm H 2 0, positive end expiratory pressure generally <12 cm H 2 0; and inspiratory time around 0.7 s. Lack of success was coded when patient dyspnea, distress, or impairment in gas exchange were noted. Though some of these are subjective criteria, these have been consistently established in our clinical practice prior to and after transition.
The primary exposure was collected from the portable CXRs most adjacent to the time tracheostomies were placed: the presence of hyperinflation was defined as 10 or more posterior ribs expanded above the diaphragm on the right side. The CXR was generally reviewed by two neonatal clinicians (authors ML, RC, MR) to minimize variability. Radiologists' written interpretations were considered though sometimes comments on hyperinflation were missing as they had been written in prior interpretations of radiographs. Adjudication was made by consensus between these radiologists' and clinicians' interpretations. At the time of this study, CXRs after tracheostomy were not systematically ascertained or examined, and thus, the time to resolution of hyperinflation was not measured. Anecdotally, hyperinflation was most often static. When improvement occurred, it took many months (>3-6) after tracheostomy.
Maternal characteristics and infant variables related to prematurity-or BPD-associated morbidities were ascertained through medical record review. Also, because affected infants may develop pulmonary hypertension concurrent with their respiratory failure, we quantified whether infants had right ventricular dilation (yes/no) based on echocardiography done prior to their tracheostomy insertion. Stratified by hyperinflation (≥10 ribs), we observed and calculated differences in these characteristics using bivariate analyses, and correspondingly, student t-test, Chi-square, or nonparametric testing, as appropriate.
To observe the timing of transition, we created Kaplan-Meier curves also stratified by hyperinflation. Then, multivariable Cox proportional hazards models were created to determine whether hyperinflation was independently associated with the main outcome. Candidate variables were those hypothesized to influence this duration that were present and known at the time of tracheostomy, and those that were significantly associated with the primary outcome in bivariable analyses (p < 0.2) were considered. Backward selection was used, and variables were retained if they were significantly associated with the main outcome and the association between hyperinflation and the primary outcome changed by 20%.   One secondary outcome was inpatient mortality. Also, as somatic growth is often either indicative or a by-product of respiratory stability, we analyzed the relationship between hyperinflation and growth velocity after tracheostomy. Analyses were performed using STATA v14 (College Station, TX), and institutional review board oversight was obtained prior to medical record chart review (IRB 2020-3773) and prior to establishing or accessing CHND (IRB 2009-13982).

RESULTS
Eighty-one patients with tracheostomy and a diagnosis of sBPD were identified. After excluding infants born >32 weeks' gestation, those who did not require home ventilators or received tracheostomy prior to transfer and those with significant congenital anomalies, 62 patients and their data were analyzed (Fig. 1).
Of these 62 infants, 26 (42%) were noted to be hyperinflated by CXR (Table 1). Infants with hyperinflation were less likely to be multiple gestation. Otherwise, no other significant differences were observed between infants with and without hyperinflation.
Secondary outcomes demonstrated that infants with hyperinflation had a higher risk of mortality after tracheostomy placement without conversion to home ventilators (31% vs 8.3%, p = 0.02). Length of NICU stay was significantly longer in infants who had observed hyperinflation (125 vs 83 days, p = 0.018). However, weight-gain velocity after tracheostomy was similar (26.8 vs 30 grams/day, p = 0.31).

DISCUSSION
In our population of preterm infants with sBPD and chronic respiratory failure, radiographic lung hyperinflation at the time of tracheostomy was associated with a longer time to transition to home ventilators. This finding was persistent after adjusting for a marker of pulmonary hypertension (right ventricle dilation) and was unaffected when including traditional markers of confounding variables (Table 1). Furthermore, affected infants with hyperinflation had an increased risk of mortality posttracheostomy and longer NICU length of stay. Thus, hyperinflation appears to be a marker of disease severity in sBPD and chronic respiratory failure [13]. It may also reflect effects of the disease interacting with the type, mode, and/or duration of ventilation prior to tracheostomy. We have observed that hyperinflation may be transient early in the evolution of BPD, but it is typically fixed by the time tracheostomy placement is performed at or beyond 44 weeks PMA.
It has been suggested that hyperinflation may be modifiable by adjustments in ventilation (i.e., higher PEEP, lower rate, increased bronchodilator therapy) [7,8,14,15]. These studies uniformly call for further evidence-based strategies and interventions to prevent hyperinflation earlier during preterm infants' hospitalizations. Our results further support this need. In addition, given our observations that hyperinflation frequently co-exists with patientventilator asynchrony, these results call for a need to understand the relationship between these two and, if verified, develop and utilize alternative or more sensitive triggering of MVs used for affected infants. Finally, these findings may assist clinicians in setting parental expectations for the trajectory of hospitalization, risk of mortality, and process of transitioning to achieve medical stability after tracheostomies are placed among infants with sBPD and chronic respiratory failure.
Air trapping and intrinsic positive end expiratory pressure (PEEP) are common in sBPD [16,17], and prior studies have demonstrated links between hyperinflation, intrinsic PEEP and ventilator asynchronies [8,15,18]. We have observed similar findings in our clinical practice. These asynchronies can lead to ventilator-induced lung injury and can contribute to further ventilator-induced diaphragmatic dysfunction [19]. Furthermore, asynchrony has a known association with increased risk for mortality [20]. Detecting asynchrony is often difficult, requiring direct observations of the patient's respiratory pattern and the timing of the MV's delivery of each breath [8]. Asynchrony may be intermittent and related to patients' sleep-wake cycles and other clinical states (e.g., inadequate sedation, post-operative pain) [20]. Sometimes, "BPD spells" and asynchrony co-exist and appear to contribute to cardiopulmonary instability. Thus, the burdens of MV asynchrony are important for these infants. Though we cannot establish a causal link to whether hyperinflation causes asynchrony, our clinical observations suggest that hyperinflation and its ramifications deserve further exploration.
CXRs demonstrating hyperinflation can alert a clinician to the increased risk for impaired ventilator synchrony and allow earlier modifications in management strategies. During the time period of this retrospective review, we did not review CXRs after tracheostomies systematically. However, in our experience, hyperinflation persisted anecdotally in CXRs for many months after tracheostomy placement. Identifying effective strategies to prevent, treat, or mitigate concurrent co-morbidities (e.g., airway  malacia) requires rigorous clinical trials to reduce burdens of chronic respiratory failure on infants with sBPD. RV dilation was also a significant risk factors for delayed time to transition. Prior studies have recognized that a diagnosis of pulmonary hypertension increases risk of morbidity and mortality among infants with a diagnosis of BPD [21]. In fact, this diagnosis was associated with longer length of stay, increased risk for mortality, increased supplemental oxygen requirement and higher respiratory severity score, increased length of time for mechanical ventilation, need for tracheostomy and increased rate of readmissions [22,23]. It can be surmised that the severity of illness for patients with pulmonary hypertension does not lessen at the same rate as for those without. As patients must be considered stable at the time of transition to home ventilators, it is not surprising that this diagnosis may also lead to a longer time to transition.
Our results show that mortality risks are greater among those who exhibited radiographic hyperinflation, which underscores the necessity to understand more about its pathophysiology so preventive strategies can be undertaken, if possible. Our findings may help to explain why, around the time of tracheostomy, ventilator settings may need substantial adjustment to more chronic settings based on the underlying pathophysiologic findings (e.g., longer time constants, larger dead space, ventilation-perfusion mismatching). Given the result in this study that more patients with hyperinflation died than those without, earlier adjustment of support to mitigate pathologic air trapping may be helpful, though further study and examinations are clearly needed.
This study has important limitations. Hyperinflation was defined subjectively and arbitrarily, and CXR ascertainment during exhalation or inhalation was not controlled. The CXRs used were anterior-posterior views without lateral views and were assessed at a single point in time. Although it would have been inconsistent with our clinical observations, it is possible that infants were misclassified if hyperinflationor acceptable inflationwas transiently present. Examining longitudinal CXRs may mitigate this limitation. However, our retrospective design is particularly subject to ascertainment biases: stable infants were less likely to have CXRs taken relative to unstable infants. A future studywith standardized or prospective CXR frequencywould be necessary to address these issues.
Second, these results are subject to confounding by indication, namely that the CXR appearance may have altered the management of these infants which then changed the association between hyperinflation and the time-to-successful transition to the home ventilators. Also, this study uses the measurement of hyperinflation as a surrogate marker for patient-MV asynchrony. Some hyperinflated infants may not have exhibited MV asynchrony, and conversely, other infants without hyperinflation might have struggled with synchronizing with an MV. Still, our clinical experience and observation suggested that hyperinflation was tightly associated with asynchrony. Further, our results may not be generalizable to others' practice/hospitals or other MVs where hyperinflation may be less impactful as many newer home ventilators have more sensitive flow triggers. Lastly, this study does not discriminate between the potential (synergistic) mechanisms of hyperinflation including air trapping, intrinsic endexpiratory pressures, ventilatory management, heterogenous lung compliance, and/or diaphragmatic functions. The inflation on CXR reflects the intersection of these mechanisms, and evaluating them separately was not possible retrospectively and remains a clinical challenge prospectively.
Understanding the time course, and its determinants, from tracheostomy through achieving medical stability to transition to home ventilators has value in managing these infants and in counselling their families. This study describes the possible influence of hyperinflation as a marker of illness severity on this time course during the hospitalization. Studies examining the relationship of hyperinflation to ventilator asynchronies and methods to mitigate both are needed in order to discover novel mechanisms and to yield evidence-based strategies that can improve these infants' outcomes.

DATA AVAILABILITY
All data generated or analyzed during this study are included in this published article.