Article Text

Download PDFPDF

Ventilatory control and supplemental oxygen in premature infants with apparent chronic lung disease
  1. Ferdinand Coste1,
  2. Thomas Ferkol1,2,
  3. Aaron Hamvas1,
  4. Claudia Cleveland3,
  5. Laura Linneman1,
  6. Julie Hoffman1,
  7. James Kemp1,3
  1. 1Departments of Pediatrics, Washington University School of Medicine, Saint Louis, Missouri, USA
  2. 2Department of Cell Biology and Physiology, Washington University School of Medicine, Saint Louis, Missouri, USA
  3. 3Sleep Medicine Laboratory, Saint Louis Children's Hospital, Saint Louis, Missouri, USA
  1. Correspondence to Dr James S Kemp, Division of Pediatric Allergy, Immunology, and Pulmonary Medicine, Department of Pediatrics, 660 South Euclid Avenue, Mailbox 8116, Saint Louis, MO 63110, USA; kemp_j{at}


Objectives Our goal was to evaluate changes in respiratory pattern among premature infants born at <29 weeks gestation who underwent a physiological challenge at 36 weeks postmenstrual age with systematic reductions in supplemental oxygen and inspired airflow.

Study design Subjects were all infants enrolled in the Prematurity and Respiratory Outcomes Project at St. Louis Children's Hospital and eligible for a physiological challenge protocol because they were receiving supplemental oxygen or augmented airflow alone as part of their routine care. Continuous recording of rib cage and abdominal excursion and haemoglobin oxygen saturation (SpO2%) were made in the newborn intensive care unit.

Results 37 of 49 infants (75.5%) failed the challenge, with severe or sustained falls in SpO2%. Also, 16 of 37 infants (43.2%) who failed had marked increases in the amount of periodic breathing at the time of challenge failure.

Conclusions An unstable respiratory pattern is unmasked with a decrease in inspired oxygen or airflow support in many premature infants. Although infants with significant chronic lung disease may also be predisposed to more periodic breathing, these data suggest that the classification of chronic lung disease of prematurity based solely on clinical requirements for supplemental oxygen or airflow do not account for multiple mechanisms that are likely contributing to the need for respiratory support.

  • Neonatology
  • Respiratory

Statistics from

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.

What is already known on this topic

  • Infants born prematurely are said to have chronic lung disease if they require supplemental oxygen at 36 weeks postmenstrual age. These infants also sometimes have unstable respiratory patterns that can cause hypoxaemia and that can be made more stable by treating them with oxygen.

What this study adds

  • This study quantifies the prevalence, in a small cohort, of the potential contribution of an unstable respiratory pattern to the need for supplemental oxygen. It shows that a substantial minority of infants with presumed chronic lung disease who cannot be weaned to room air during a physiological challenge fail the challenge after developing periodic breathing.


Clinical manifestations of chronic lung disease (CLD) of prematurity, also known as bronchopulmonary dysplasia (BPD), have changed over the last 45 years.1 ,2 CLD has been defined by persistence of a supplemental oxygen requirement at 36 weeks postmenstrual age (PMA) among infants born before 32 weeks gestation who usually have persistent respiratory symptoms and chest radiograph changes. Historically, CLD of prematurity has been considered a consequence of lung immaturity and lung injury caused by immature surfactant production, incomplete alveolarisation, oxygen toxicity, barotrauma and infection.3 Despite important advances in treatment, many infants still require supplemental oxygen at 36 weeks PMA.

Infants with CLD have long-term respiratory morbidity with higher rates of re-hospitalisation and are more likely to require respiratory medications during the first year of life. The cost for treatment of CLD in the USA in 2005 exceeded $2 billion.4 ,5

Although supplemental oxygen use has come to be synonymous with CLD,6 this conventional definition has limitations. While a persistent supplemental oxygen requirement may indicate that a newborn has only airway or alveolar disease, unstable ventilatory control can lead to hypoxaemia that will also respond to supplemental oxygen support.7–11 In some studies, lengthy periods of spontaneous periodic breathing (PB) were recorded in more than half of premature subjects.12 ,13

This paper describes a single-centre study within the Prematurity and Respiratory Outcomes Project (PROP), a multicentre study intended to identify genetic modifiers and biomarkers that will lead to targeted therapies for CLD. We examined breathing patterns during a systematic reduction in supplemental oxygen and flow in 49 consecutive preterm infants prescribed supplemental oxygen or airflow support for hypoxaemia, and who met current clinical criteria for CLD. We assessed the prevalence of unstable ventilatory control to determine how often it might contribute to the child's hypoxaemia. These findings suggest a more complete phenotypic description of CLD is needed when gas exchange is altered, a step that is critical to identifying mechanistic factors and understanding the degree of involvement of alveolar and airway disease, or immature ventilatory control.

Experimental methods


From 12 August 2011 to 30 July 2013, 124 consecutive neonates born between 24 and 28 weeks gestation were enrolled prospectively at Saint Louis Children's Hospital, one of the seven centres participating in the National Institutes of Health-supported PROP. At 36 weeks PMA, the age at which the diagnosis of CLD is typically assigned, infants receiving supplemental oxygen or airflow support via nasal cannula were eligible to undergo a physiological challenge, to confirm that without supplemental oxygen or augmented airflow their SpO2% would be unacceptably low and consistent with a diagnosis of CLD. Infants requiring mechanical ventilation or those deemed unstable by the clinical care team were excluded. The type and degree of support were chosen by the clinical care team without predetermined criteria (figure 1).

Figure 1

Subject enrolment and outcomes. PMA, postmenstrual age.

Respiratory outcome measures

At 36 weeks PMA, infants receiving supplemental oxygen or airflow support via nasal cannula at fraction of inhaled oxygen (FiO2) of 0.21 underwent a physiological challenge protocol according to the PROP Manual of Procedures.14

Infants were studied in the supine position, and the study was initiated with infants in behaviourally defined quiet sleep.15 ,16 During a 5 min run-in period, the infant received the FiO2 and airflow rate prescribed by the clinicians supervising care. SpO2% was recorded continuously using a NONIN pulse oximeter (NONIN Medical, Minneapolis, Minnesota, USA). In response to a change in arterial oxyhaemoglobin saturation (SaO2%), the oximeter has an effective averaging time in newborns of 1.5–3.0 s, depending on the pulse rate.

Respiratory inductance plethysmography (RIP) was recorded (Biocapture, Cleveland Medical Devices, Cleveland, Ohio, USA) by placing elastic bands around the infant at the nipple line and the umbilicus17 ,18 during the run-in period, when the infant received prescribe FiO2 and flow, and during the challenge period. Data describing SpO2%, per cent of quiet sleep time in periodic breathing (%PB) and minute ventilation during quiet sleep from quantitative diagnostic calibration (QDC)-calibrated RIP18 were compared before and during the challenge.

RIP was also used to provide an index of the severity of respiratory system disease. The degree of rib cage versus abdomen dyssynchrony over 60 breaths for each infant was described by a phase angle (φ) calculated from the plot of rib cage versus abdominal excursion.17 ,19 Healthy term infants have φ near 12° during quiet sleep.20

After the 5 min run-in, FiO2 was reduced to 0.21 in steps by 0.20 or less, followed by wean of airflow provided by nasal cannula. Each step down in FiO2 was observed for 5 min, each step down in flow for 10 min. Failure at any step in the weaning protocol resulted in termination of the challenge with resumption of the prescribed support. If the infants tolerated all decrements of support during the weaning period, an observation period of up to 60 min ensued with the infant breathing only ambient room air. Physiological challenge failure was defined as SpO2 <90% for five consecutive minutes within any step in the challenge, or SpO2 <80% for 15 consecutive seconds. If the infant did not worsen to meet criteria for failure during the 60 min observation, he or she was deemed to have passed the room air challenge (RAC). In most instances, we refer to the challenge as a ‘physiological’ challenge, rather than RAC, because many infants failed before breathing room air alone.

Measurements during physiological challenge

Relative changes in minute ventilation were calculated from QDC-calibrated RIP.18 PB was defined as at least three central apnoeas lasting at least 3 s separated by <20 s of regular breathing.21

Minute ventilation and percentage of time with PB for the entire run-in period were compared with the final minutes of the challenge. For minute ventilation, the 5 min run-in period was compared with the last minute of the challenge (failure or completion). For PB, the per cent of time in PB for the 5 min run-in was compared with the per cent during the last three minutes before failure or completion of challenge. Some infants were still receiving caffeine as prescribed by the clinical team, and its potential impact on PB and challenge failure was also analysed. Caffeine levels were not routinely measured in these infants or in our neonatal intensive care unit, in general.

Statistical measurement and comparisons

Kruskal–Wallis test was used for comparison of caffeine effect among each of the three groups. Categorical comparisons based on passing or failing the challenge were done by χ2 test. Statistical significance was defined as p<0.05.


A total of 124 infants were enrolled (figure 1). Three infants failed during the run-in period when continuous recordings showed that the prescribed support did not maintain SpO2% in the passing range. Forty-nine infants underwent the challenge (table 1). Twelve infants (24.5%) maintained oxygen saturations for 60 min of observation breathing room air alone and passed RAC. Five infants had PB during the run-in period while on prescribed support before the challenge (range 5.1–102.0 s of PB, 1.7–34.0% of 5 min period). Two of these did not maintain adequate SpO2% during the run-in period and were not tested further. Seventeen infants displayed more PB during the physiological challenge, and 16 of these infants failed (figure 2). These 16 represented 43.2% of infants failing the challenge. The increase in per cent of time with PB during the challenge, for those showing an increase, was 48.6%±21.6% (range 15.0–100.0%). That is, a hypothetical infant with 1 min of PB during the 5 min run-in period (20% of recording time) and 2 min of PB during the last three minutes of the challenge (67% of the time) would have had an increase by 47 in the per cent of recording time with PB.

Table 1

Descriptive statistics for infants undergoing physiological challenge: birth weight, gestational age, postmenstrual age, gender, FiO2, flow rate for air and O2 delivered by nasal cannula, days receiving mechanical ventilation and phase angle (ϕ) as a measure for rib cage abdominal dyssynchrony

Figure 2

Diagram showing a 5 min recording from a patient who failed the challenge. The upper two tracings show respiratory excursions. The lowest tracing is SpO2% averaged every 3–5 s. Note the stable SpO2% >90% when respirations are regular. Note the intermittent hypoxaemia and characteristic oscillations in SpO2% seen with the brief run of periodic breathing (PB) early in the tracings, and the continuous oscillations when PB is established. This pattern occurred in 43.2% of challenge failures. For this subject, continuous periodic pattern was associated with failure range falls in SpO2% to <80%. First vertical mark: challenge begun, followed by brief run of PB. Second vertical mark: quiet sleep. Sixth vertical mark: room air challenge failure was declared. The smaller box is an enlarged 30 s excerpt from between vertical marks 5 and 6.

Because caffeine treatment may limit PB, we assessed the impact on PB of caffeine use, retrospectively. Six infants were treated with caffeine at the time of the challenge (12.2% of subjects); four of six passed RAC without any PB; the two failing while on caffeine had 33% and 41% increases in the per cent of time with PB, but the caffeine concentrations were not measured. Thus, while only 12.2% of subjects were receiving caffeine at the time of the challenge, they accounted for 33.0% of those passing (4 of 12).

The number of days since caffeine had been discontinued was not different among infants failing without PB, passing RAC or failing with increased PB (Kruskal–Wallis, p=0.73, medians, 11.2, 11.6 and 10.7 days without caffeine at the time of challenge).

A total of 7 of 37 infants failing the challenge had an increase in QDC-calibrated minute ventilation during challenge, with 3 of 12 infants passing the challenge also showing increased minute ventilation (table 2, figure 3).

Table 2

Comparison of changes in minute ventilation and changes in per cent of sleep time with periodic breathing between those passing and those failing the physiological challenge

Figure 3

Prevalence of changes in minute ventilation and respiratory pattern among infants receiving the physiological challenge. The majority of infants (73.6%), whether they passed or failed, either developed periodic breathing (PB) or did not increase their minute ventilation, or both, in response to the challenge. Potentially beneficial response defined as increase in minute ventilation and no PB. Potentially detrimental, or dysfunctional, response is no change or reduction in minute ventilation, much PB or combination of failure to increase minute ventilation while having PB.

Analysis of the impact of birth weight, gestational age, PMA, gender and level of support on rates of passing and failing the challenge is best left to evaluation of the much larger sample from the multicentre PROP cohort. The size of our single-centre sample did not permit sufficient statistical power for subgroup comparisons. In our sample of only 49 subjects, however, those passing did not differ statistically from those failing by birth weight (p=0.09), gestational age (p=0.51), the diagnosis of intraventricular haemorrhage using ultrasonography (χ2 test with Fisher's exact test, p=0.32), PMA (p=0.60), gender (χ2 test, p=0.62) or FiO2 (p=0.30). Those failing had higher flow rates prescribed before the challenge started (0.4 vs 1.1 L/min, p=0.01). Sixteen infants failed with increased PB; the subgroup failing at 36 weeks PMA with more PB had slightly lower mean gestational age than those 12 infants who passed RAC, but this did not reach statistical significance (p=0.10, 95% confidence limits of difference, −1.8 to 0.2 weeks). To the extent that PB leads to clinically recognised apnoea, these findings are consistent with a later resolution of apnoea among infants born earlier.22

Although none of the infants was intubated or receiving mechanical ventilatory support at 36 weeks PMA, they are likely at considerable risk for CLD (table 1). The median duration for mechanical ventilation was 12.0 days. Many also had marked RC-ABD (Rib cage – Abdominal) dyssynchrony with a mean φ in the range described for preterms with BPD.20


In this study, we show that a large proportion of premature infants who meet clinical criteria for CLD, defined by supplemental oxygen or augmented airflow use at 36 weeks PMA, had unstable control of breathing that was unmasked during the physiological challenge. Unstable ventilatory control may contribute to the inability of the clinical teams to further decrease the flow or FiO2. Specifically, we show that 32.7% of the infants studied treated with supplemental oxygen and airflow support at 36 weeks PMA need nasal cannula support at least in part due to PB and periodic apnoea, and only 16.0% of infants who failed the challenge demonstrated a potentially beneficial response to hypoxaemia by increasing minute ventilation. Furthermore, we show that 24.5% of infants receiving supplemental oxygen and airflow via nasal cannula at the time of the challenge were able to tolerate room air in the short term using PROP criteria.14 ,23

These results have important implications for study design and subject selection for future investigations that attempt to define the pathophysiological bases of CLD phenotypes. Within PROP, which is designed to identify clinical risk factors and candidate biomarkers for CLD among infants born between 24 and 28 weeks PMA, we suspect that the current clinical definition of CLD will complicate our studies because the need for supplemental oxygen support at 36 weeks PMA may be multifactorial and reflect an interaction between parenchymal dysfunction due to lung disease and immature respiratory control. Infants with significant CLD may also be particularly susceptible to developing PB when they become hypoxaemic with even brief central apnoeas.24 ,25 Some of the infants studied had marked chest wall dyssynchrony (table 1) in addition to requiring supplemental O2, suggesting that significant respiratory system compromise contributed to their propensity for desaturation during PB. Some infants meeting clinical criteria for CLD did in fact tolerate room air when evaluated using a formal protocol, while others showed an immature respiratory pattern such as PB that worsened during hypoxaemia but that can be corrected with supplemental oxygen. A more physiological definition of CLD is needed to ensure that mechanistic biomarkers being evaluated are actually linked to parenchymal lung disease and to distinguish these infants from those that are otherwise well or have hypoxaemia for different reasons.2

Correction of respiratory pattern with supplemental O2 may be particularly applicable to infants born at the earliest gestational ages who otherwise would be expected to have higher rates of CLD.

Based on published reports,23 we expected that some subjects would pass RAC and thus not have CLD as currently defined, which was the case (24.5% passed RAC). However, the high frequency of challenge failures in infants who developed periodic apnoea with characteristic oscillations in oxyhaemoglobin saturations was not anticipated, even though supplemental oxygen therapy has been shown to stabilise breathing patterns in preterm infants who have PB.26 ,27 Thus, although the association of PB and hypoxaemia is established, the frequency with which PB might contribute to the need for supplemental O2 among a group of infants at risk for CLD has not been previously quantified.

In this regard, it is likely (figure 2 with oscillations) that observations made at the crib side, even by experienced nurses and clinicians, do not reliably identify infants that have altered breathing pattern causing hypoxaemia, leading to under-recognition of this phenomenon in the intensive care nursery. The frequent occurrence of intermittent hypoxaemia during PB, but before physiological challenge criteria for sustained hypoxaemia were met, is demonstrated in the first minutes of the tracing in figure 2. RIP scalar tracing allowed us to continuously capture respiratory patterns clearly over many minutes and provided a record of respiratory pattern that can be quickly reviewed.

A total of 5 of our 52 subjects potentially eligible for the challenge study displayed PB while on their prescribed support, and 2 of these 5 infants failed to maintain passing SpO2% during the run-in period. A limitation of this study is that we did not record respiratory pattern for longer periods before reducing supplemental oxygen support, and it is possible that other infants in our cohort may have substantial PB on prescribed support that was not captured before the challenge.

We speculate that our findings may be relevant to the thus far unexplained observation from a large study showing that prior therapy with caffeine reduced the rate of BPD among treated premature infants.28 ,29 A plausible mechanism for the reduced need for supplemental oxygen in the group randomised to caffeine therapy before 36 weeks PMA would be that caffeine reduced rates of PB and apnoea in these subjects, perhaps by accelerating the development of stable respiratory control.30–33

Other sites within the PROP collaborative did not routinely record respiratory pattern using RIP during the challenge, but it would be surprising if our findings were not generalisable and applicable to the larger cohort, given the clear propensity of premature infants to develop PB.10–13 ,21 ,22

In conclusion, we want to emphasise that unstable control of breathing is common in premature infants who met clinical criteria for CLD. Unstable respiratory patterns may occur alone or in conjunction with airways and airspace diseases in many infants born between 24 and 28 weeks gestation, and contribute to the supplemental oxygen requirement at 36 weeks PMA. The role of PB and immature breathing patterns must be considered if we are to better understand the real contribution of lung disease to hypoxaemia in premature infants.



  • Contributors FC, wrote the first draft of the manuscript and revised subsequent drafts. He also had the primary role in planning and carrying out data analysis. AH and TF are the Washington University Principle Investigators for the PROP grant. They both participated in planning the study, data analysis and manuscript critique. CC participated in planning and performed oximetry and inductance plethysmography recordings. JH and LL were the site coordinators for the PROP study. They recruited subjects and obtained informed consent. They were primarily involved with recording and labelling patient clinical and demographical data. JK is the principal co-investigator at Washington University for the PROP study involving respiratory physiology. He participated in planning the study, data analysis and worked closely with FC in manuscript review.

  • Funding The authors were supported by the National Institutes of Health (NIH) award HL101465, which supports the Prematurity and Respiratory Outcomes Project (PROP).

  • Competing interests None.

  • Patient consent Obtained.

  • Ethics approval Washington University Human Research Protection Office and the PROP Observational Study Monitoring Board.

  • Provenance and peer review Not commissioned; externally peer reviewed.