Discussion
Many of the variables that predict home oxygen in CNLD such as early gestational age, low birth weight, invasive ventilation time or total duration of respiratory support did not predict oxygen weaning within 12 months CGA.6,7  Our finding that infants who were discharged on higher oxygen flow rates required home oxygen therapy for longer is similar to what has been reported previously.12  Variables typically associated with severity of CNLD like high frequency oscillation ventilation, early gestational age, or hypercapnia at term CGA did not correlate with the duration of home oxygen therapy.  Therefore, infants should not be presumed to need longer oxygen therapy based on ventilation requirements in NICU.
 
In our cohort, prolonged neonatal hospital stay was associated with increased duration of home oxygen therapy.  The reported duration of home oxygen therapy in CNLD worldwide over the last 40 years is between 2 and 13.5 months (Table 5).12-24  When home oxygen was initially introduced, weaning was based on a target partial pressure of arterial oxyhemoglobin (PaO2) of 55-60 mmHg.13-15  Continuous overnight pulse oximetry has become more commonly used to guide oxygen weaning with expert consensus suggesting targeting a mean peripheral arterial hemoglobin saturation (SpO2) > 93-95% and ≤ 5% of the overnight oximetry time spent with SpO2 ≤ 90% or SpO2 ≤ 93%.25-28  However, there is a significant lack of evidence supporting these recommendations and thus home oxygen weaning in CNLD remains highly variable between health services.29,30 
 
Several service provision variables likely contribute duration of oxygen therapy for these infants.  This includes who oversees oxygen weaning following discharge from neonatal care and the frequency of follow up visits in the first year of life.  Models of care differ across centers with neonatology, general pediatrics, general practitioners, respiratory, and sleep specialists all involved in patient care to varying degrees.  Some centers may even have oxygen weaning unsupervised by a medical provider.12,30  At our center, infants are reviewed every 6-8 weeks post-discharge (Appendix 1) and the median age of home oxygen cessation was 6.8 months CGA (IQR 4.4), with 51% of infants discharged on oxygen flow rates of 200-250 mL/min (Table 1).  In contrast, centers where a clinical nurse undertakes weaning in the community may wean oxygen slower and be affected by the follow up frequency. 
 
Centers that rely on polysomnography (PSG) assessment for oxygen weaning report prolonged duration of home oxygen therapy possibly due to PSG waiting times (Table 5).12,31  In our experience, the majority of infants with CNLD can be safely weaned using overnight oximetry monitoring and PSG is reserved for infants requiring more than 500 mL/min of supplemental oxygen at time of discharge, infants with a clinical suspicion of comorbid obstructive sleep apnea (OSA), or when there is concern that an additional etiology contributes to hypoxemia.  
 
The current lack of standardization of home oxygen initiation and weaning makes it difficult to assess outcomes in these infants.  Currently, no evidence exists to suggest that one method of weaning supplemental oxygen therapy is superior to another and to our knowledge no studies have been able to evaluate long-term outcomes following a standardized weaning protocol.  This highlights the need for prospective longitudinal studies in view of the continued increase in the number of preterm infants requiring home oxygen.32
 
Infants with CNLD are known to be at higher risk of lower respiratory tract infections (especially respiratory syncytial virus) and those receiving home oxygen therapy are 10% more likely to be readmitted for a respiratory illness with 49-58% of infants with severe CNLD requiring readmission within the first two years of life.5,33,34  In our study, infants discharged on higher oxygen flow levels required more admissions (including PICU) which is in keeping with findings in other studies.5  It is important for clinicians to monitor these higher risk infants closely during this early period and ensuring families are aware of the increased vulnerability if an infant is discharged on ≥ 500 mL/min of oxygen.  The number of ED visits within the first two years did not correlate with discharge oxygen flow rates, perhaps indicating that home oxygen may help reduce severity of subsequent respiratory illnesses.
 
Wheeze and bronchial hyper-responsivity are well described following preterm birth35-38 and infants with CNLD are more likely to receive additional oxygen, bronchodilators and inhaled steroids within the first year of life.23  In our study, doctor documented wheeze and corticosteroid use (inhaled or oral) in the first 2 years of life were not significantly associated with duration of oxygen therapy or oxygen flow rates at discharge.  Further longitudinal follow up of our cohort would be required to determine if such presentations occur later in genetically predisposed individuals, with prematurity and CNLD increasing risk during pre-school years.  All infants with CNLD requiring home oxygen at our center receive respiratory syncytial virus immunoglobulin prophylaxis during the winter months which may have reduced wheezy hospital presentations compared to other studies.
 
Infants were more likely to be discharge on higher levels of oxygen therapy (≥ 200 mL/min) if they had higher pCO2 measured at term CGA and increased CGA at hospital discharge.  In our cohort, 30.9% (n=46) had moderate hypercapnia (pCO2 55-65 mmHg) and 6% (n=9) had severe hypercapnia (pCO2 > 65 mmHg) measured at term CGA (Table 1).  Moderate to severe hypercapnia was not predictive of home oxygen cessation but was significantly associated with requiring >250mL/min oxygen at discharge (Table 3).  In a sub-analysis, infants with pCO2 ≤ 55 mmHg had a median time to oxygen weaning of 6.3 months (95% CI, 5.7-7.0) whereas patients with pCO2 > 55 mmHg had a median time of 8.4 months (95% CI, 7.7-9.1) (Appendix 4).  The differences were statistically significant at our time on oxygen cut-off points of CGA 6 months (p=0.001), 9 months (p=0.035), and 12 months (p=0.034).  This information may be useful clinically to guide parent’s expectations of a longer duration of oxygen therapy in those who have hypercarbia at term CGA.  
 
It is unclear whether alternative or additional respiratory support would be beneficial in infants with hypercarbia as there are no studies assessing the impact of hypercarbia and level of respiratory support on outcomes in CNLD.  Centers may consider PSG to evaluate the need for pressure support in infants who have evidence of alveolar hypoventilation (more than 25% of the total sleep time with pCO2 > 50 mmHg)41 but it is unclear which infants with moderate to severe hypercapnia benefit from pressure support over supplemental oxygen.  A Canadian retrospective cohort study identified that of infants born at GA < 29 weeks, only 0.9% and 0.6% were discharged on non-invasive ventilation (NIV) or tracheostomy ventilation respectively.42  The indications for initiating NIV were not stated.  After laryngomalacia, CNLD was the second most common indication for home CPAP over a period of 10 years at Westmead Children’s Hospital.43  No specific gas exchange or PSG criteria for initiating CPAP was mentioned.  Future work to explore practices for NIV therapy in infants with CNLD across centers would be a first step in determining whether this is an option that should be offered more readily in this population and whether this impacts longer term outcomes.
 
Our study is limited by its retrospective nature but is strengthened by the inclusion of a reasonable sample size.  The multinomial logistic regression model to predict oxygen flow rates at discharge would need to be validated on a larger cohort of patients as the lower accuracy was affected by the smaller sample size in some comparative groups.  There is selection bias as only infants managed at a single center were included for analysis.  We recognize that length of oxygen therapy and weaning practices are often based on the anecdotal experience of individual clinicians due to the lack of evidence in this area.  This is consistent with practice in other centers as there are currently no standardized protocols to guide weaning of home oxygen therapy in infants with CNLD.  Recall bias is minimized by our multidisciplinary oxygen clinic following a consistent follow up and weaning process (Appendix 1).  Although our study only describes associations between a priori maternal and infant variable with home oxygen duration and oxygen flow rates, it provides insightful information about which infants may require prolonged oxygen therapy or experience more hospital readmissions.