OBJECTIVE: Bronchopulmonary dysplasia (BPD) is the focus of many intervention trials, yet the outcome measure when based solely on oxygen administration may be confounded by differing criteria for oxygen administration between physicians. Thus, we wished to define BPD by a standardized oxygen saturation monitoring at 36 weeks corrected age, and compare this physiologic definition with the standard clinical definition of BPD based solely on oxygen administration.
METHODOLOGY: A total of 199 consecutive very low birthweight infants (VLBW, 501 to 1500 g birthweight) were assessed prospectively at 36±1 weeks corrected age. Neonates on positive pressure support or receiving >30% supplemental oxygen were assigned the outcome BPD. Those receiving ≤30% oxygen underwent a stepwise 2% reduction in supplemental oxygen to room air while under continuous observation and oxygen saturation monitoring. Outcomes of the test were “no BPD” (saturations ≥88% for 60 minutes) or “BPD” (saturation <88%). At the conclusion of the test, all infants were returned to their baseline oxygen. Safety (apnea, bradycardia, increased oxygen use), inter-rater reliability, test–retest reliability, and validity of the physiologic definition vs the clinical definition were assessed.
RESULTS: A total of 199 VLBW were assessed, of whom 45 (36%) were diagnosed with BPD by the clinical definition of oxygen use at 36 weeks corrected age. The physiologic definition identified 15 infants treated with oxygen who successfully passed the saturation monitoring test in room air. The physiologic definition diagnosed BPD in 30 (24%) of the cohort. All infants were safely studied. The test was highly reliable (inter-rater reliability, κ=1.0; test–retest reliability, κ=0.83) and highly correlated with discharge home in oxygen, length of hospital stay, and hospital readmissions in the first year of life.
CONCLUSIONS: The physiologic definition of BPD is safe, feasible, reliable, and valid and improves the precision of the diagnosis of BPD. This may be of benefit in future multicenter clinical trials.
Survival in very low birth weight neonates (VLBW, 501 to 1500 g) has steadily improved with 84% of all VLBW neonates surviving.1 While most of these survivors are healthy, chronic lung injury remains a significant health burden.2 In the NICHD Neonatal Research Network, 23% of the VLBW neonates born in 1995 to 96 developed bronchopulmonary dysplasia (BPD).1
The lack of a precise definition of BPD is a confounder to any study where BPD is an outcome variable. Most neonates with BPD do not undergo lung biopsy or any physiologic test and thus their pulmonary disease is defined clinically, on the basis of the sustained need for supplemental oxygen at 36 weeks postmenstrual age. The validity of this definition is supported by evidence that oxygen dependence at 36 weeks is predictive of long-term impairment in pulmonary function.3 An inherent limitation of defining BPD by the need for supplemental oxygen is that the need for oxygen is determined by individual physicians, rather than on the basis of a physiologic assessment. The assumption that the criteria on which the decision to administer oxygen is uniform and applied similarly across institutions is erroneous. As there is no consensus in the literature, neonatologists have widely divergent practices regarding oxygen saturation. Indeed, published literature cites acceptable saturation ranges from 88 to 98%.4,5,6,7,8,9 At least one study suggests that VLBW neonates do just as well, and may actually have improved outcomes, when managed with lower oxygenation saturation (70 to 90%).10
For the past 10 years, federal payers have required documentation of desaturation in room air in adults to satisfy the diagnosis for chronic obstructive pulmonary disease.11 While such a room air test is now routine in adults, it has not been studied in premature infants. Research suggests that a room air test can be accomplished safely in selected infants receiving low amounts of oxygen with higher saturations.5,12
There is a need for a standardized test that can simply and safely be used across different institutions to define BPD. We hypothesized that a structured, short period of oxygen saturation monitoring coupled with gradual oxygen weaning to room air in selected infants receiving low amounts of oxygen would reduce the number diagnosed with BPD.
From October 1999 to February 2001, 199 consecutive VLBW neonates (501 to 1500 g birthweight) were followed prospectively. Neonates with congenital anomalies were excluded. Birthweight was 969±403 g (mean±SD) and gestational age 27.3±2.6 weeks. At 36±1 weeks corrected age, respiratory outcomes were defined in survivors as BPD or no BPD using two different definitions. The first definition of BPD used was a clinical definition based on oxygen administration at 36 weeks gestation. The second definition was a new physiologic definition based on combined measurement of oxygen saturation and oxygen administration. The outcomes of this cohort were assessed in an existing longitudinal follow-up program at 8 and 20 months of corrected age.
A comparison group of 10 neonates who had never had lung disease or who had recovered from minor amounts of lung disease and had been without oxygen treatment for at least 48 h were also studied. These infants represented the only premature infants free of lung disease in our institution within the study period.
Definition of BPD
For this study, BPD was defined on a physiologic basis that combined oxygen and ventilation support with an assessment of saturation. Infants treated with mechanical ventilation, continuous positive airway pressure, or with supplemental oxygen concentration exceeding 0.30 were diagnosed with BPD without additional testing. Infants in supplemental oxygen <0.30 underwent a timed stepwise reduction to room air. Those who failed the reduction were diagnosed with BPD. No BPD was defined as treatment with room air with an oxygen saturation ≥88%, or passing a timed, continuously monitored oxygen reduction test.
Failure was defined as oxygen saturation 80 to 87% for 5 minutes, or <80% for 1 minute. There were two methods for passing: a rapid pass and pass after full monitoring. Rapid pass criteria were met by successfully weaning to room air with all saturation ≥96% for 15 minutes. If instead the saturations were 88 to 95%, the infant was monitored for a full 60 minutes in room air and defined a pass when all saturations exceeded 88% in that 60-minute period. In this study, no infants met the fast pass criteria and all were studied for the full monitoring period.
We selected the saturation cutoff value of 88% after surveying the literature for a “gold standard” for oxygen saturation in preterm neonates, reviewing published literature on saturation in preterm infants who have recovered from respiratory distress syndrome (RDS). Existing literature did not demonstrate a consistent standard for oxygen saturation targets; acceptable saturations ranged from 70 to 98%.10,13,14 We found few existing data on saturation values in neonates who had recovered from RDS. Therefore, we utilized the methodology recommended by health service researchers when no gold standard exists: the expert consensus panel.15 The panel consisted of the 14 principal investigators of the National Institutes of Child Health and Human Development Neonatal Research Network. Members completed a survey on current oxygen prescription practices in their NICUs. The lowest acceptable saturation that the expert panel cited was 88%. Therefore, the cutoff value of a passing test was defined as 88%. The impact of varying cutoff values from 88 to 92% was assessed.
Delivered oxygen concentration was directly measured in infant's receiving oxygen by hood using an oxygen analyzer (MiniOx III, Catalyst Research, Maryland) placed directly above the infant's head inside the hood. In infant's receiving oxygen by nasal cannulae, the delivered oxygen concentration (termed the effective oxygen concentration) was calculated using the technique described by Benaron and Benitz as modified for the STOP-ROP trial.13,16
The attending physician individualized the neonate's management. Oxygen was given either by oxygen hood or by nasal cannulae at the discretion of the care team. Infants cared for in nasal cannulae, received blended oxygen at a fixed liter flow, typically with flows of 1 l or less. Oxygen saturation at our center is generally maintained between 88 and 96%.
Oxygen Reduction Test
Infants receiving supplemental oxygen ≤0.30 at 36±1 week corrected age were eligible for a timed oxygen reduction test. Neonates were continuously monitored with a cardiorespiratory monitor, and pulse oximeter (Nellcor N200, Malickrodt Inc.), and directly observed by a trained neonatal research nurse throughout the reduction test. Baseline clinical data on the prior 12 h of heart rate, respiratory rate, oxygen saturation, and frequency of apnea and bradycardia were abstracted from the clinical record to determine baseline eligibility. Infants were studied in the supine position with a pulse oximeter placed on a limb in their usual baseline oxygen 30 minutes after a feeding. No formal assessment of sleep state was made. The test was performed in three parts: baseline, reduction phase, and return to usual oxygen. Values for heart rate, respiratory rate, oxygen saturation, and frequency of apnea (cessation of breathing for 20 seconds) and bradycardia (heart rate <80 beat/minute for ≥10 seconds) were recorded every 60 seconds for a 15-minute baseline period. All occurrences of movement artifact (defined as visible motion of the infant together with loss of pleythsmograph signal from the monitor) were recorded and corresponding saturation values were deleted. If all baseline saturation values exceeded 88% in ≤0.30 supplemental oxygen, the infant was eligible for a stepwise oxygen reduction test. In the oxygen reduction phase, oxygen by hood was reduced by 2% steps every 10 minutes to room air with continuous monitoring of the infant and oxygen saturation. In infants receiving oxygen by nasal cannulae, the flow was maintained constant, and oxygen concentration was reduced in 2% increments by blender until the concentration was 21%, flow was then reduced in 0.1 l increments to zero flow. The nasal cannulae was removed from the nares but left affixed to the face, to not disturb the infant.
A comparison group of 10 convalescent premature infants was monitored in the same fashion for 60 minutes in room air.
Safety, Reliability, and Validity Evaluation
Adverse events that might be associated with an oxygen reduction challenge, defined before the study began, were assessed in every infant who underwent the reduction. These consisted of: apnea >20 seconds, bradycardia (heart rate <80 beat/min for >10 seconds), and increase in baseline FiO2 of 10% persisting for >1 hour after the reduction test. Reliability of the test was assessed in two ways: inter-rater reliability and test–retest reliability. Inter-rater reliability was assessed in five patients who were tested simultaneously by two observers; when agreement in this group was found to be 100%, no additional patients were tested. Test–retest reliability was assessed in 12 patients who were studied twice within the same 24-hour period. Validity was assessed by correlating the results of the physiologic definition of BPD with three potential measures of ongoing lung disease: discharge home in oxygen, length of hospital stay, and readmissions to hospital in the first year of life.
Human Subject Protections
The study was approved by the Institutional Review Board of University Hospitals of Cleveland. Infants were studied after written informed consent was obtained from the parent or guardian. The risk of study participation was minimized by the continuous presence of an experienced neonatal research nurse. All infants were returned to their baseline supplemental oxygen at the end of the reduction test. The primary physician caring for the infant was given the results of the test. Further modifications in treatment, if any, were made at their discretion.
Group differences on continuous measures were compared with Student's t-test. Categorical variables were assessed with the χ2 test. Significance was set at p< 0.05. Reliability of the test was assessed in two ways by measuring inter-rater reliability in five patients who were tested simultaneously by two observers. Test–retest reliability was assessed in 12 patients who were studied twice within 24 hours. Test–retest reliability was measured with the κ statistic, with a value of 0.75 representing excellent agreement as suggested by Landis and Koch.14 The impact of varying saturation cutoff values on the numbers of infants diagnosed with BPD was assessed formally.
The outcomes of the neonates at 36 weeks corrected age are shown in Figure 1. Of the 33 (17% of the total population) eligible for the oxygen reduction test, consent was not obtained in nine (missed=3, no legal guardian for consent=2, parent refused=3, attending refused=1) leaving 24 studied with the reduction test. At the time of the oxygen reduction test, nine of the infants were receiving oxygen by hood, and 15 by nasal cannulae with flow rates at a median of 1.0 l/minute (range 0.03–2.0 l/minute). A total of 15–24 (63%) infants tested passed the oxygen reduction phase and successfully weaned to room air with saturations greater than or equal to 88%. The mean duration of the oxygen reduction test was 67.3±25.6 minutes. As expected, the duration of the study was significantly shorter in those who failed compared to those who passed (45.3±11.4 minutes vs 82.5±21.2 minutes, p<0.001).
The physiologic definition diagnosed BPD in 30 (24%) of 124 neonates. Of these 30 patients, 21 were diagnosed with BPD because of their continued use of a high level of ventilator or oxygen support, and nine because of a failed oxygen reduction test. In contrast, the clinical definition diagnosed BPD in 45 (36%) of these 124 infants. Overall, using the physiologic definition reduced the number of infants diagnosed with BPD from 36 to 24%. The difference is statistically significant (p=0.011).
To determine if we could identify the group of infants who failed the reduction test on the basis of demographic characteristics, we compared those who passed and those who failed the oxygen reduction test (Table 1). There were no statistically significant differences in mean birthweight, mean gestational age, level of oxygen support at the time of the challenge, or baseline heart rate, respiratory rate, and saturation identified. More of the neonates who failed were born at less than 26 weeks.
Safety of the Oxygen Reduction Test
No infants experienced the predefined adverse safety events of bradycardia, apnea, or increase in baseline oxygen saturation. One infant who failed the challenge required a 3% increase in oxygen for <5 minutes. In the nine infants who failed the challenge, respiratory rate and saturation returned to baseline within 10 minutes of a return to the usual oxygen. The clinical care team identified no adverse effects of the oxygen reduction test.
Reliability and Validity of the Physiologic Definition
In the development of a new measurement tool, it is important to systematically assess the performance of the tool. In assessing the performance of the physiologic definition, we studied both validity (the extent to which the tool measures the desired variable) and reliability (the ability of a tool to return consistent results on repeated administration). Validity was assessed by correlating the results of the physiologic definition with discharge home in oxygen, length of hospital stay, and hospital readmission in the first year of life. Ultimately, five of 126 (3.1%) infants who were classified as no BPD and seven of 26 (26.9%) infants classified as BPD were discharged home in oxygen (p<0.001). In the five infants ultimately classified as no BPD yet discharged home in oxygen the infants passed the test with a cutoff of 88% but would have failed if the saturation cutoff had been raised to 90%. Infants classified as BPD had significantly longer hospital stays compared to those with no BPD (121.3±31.0 days (95% confidence interval 109.4, 133.2) vs 59.7±26.8 (95% confidence interval 55.2, 64.3); p<0.0001). Patients diagnosed with BPD also had significantly more rehospitalizations in the first year of life (0.78±1.05 (95% confidence interval 0.65, 1.45) vs 0.23±0.75 (95% confidence interval 0.62, 0.88); p=0.0022).
The oxygen reduction test also had high reliability. Reliability was assessed in two ways: inter-rater reliability and test–retest reliability. Two observers simultaneously and independently assessed five patients. These assessments agreed in all five cases, yielding a 100% concordance rate. A total of 12 patients had the test repeated within 24 hours. All five patients who passed the test the first time also passed the second time. Of the seven patients who failed the test, six failed on the retest. Concordance between the two tests was excellent, with 92% of the tests yielding the same result (κ statistic=0.83).
Impact of Varying the Saturation cutoff
We assessed the impact of varying oxygen saturation cutoff values on the outcome of the test (Table 2). As anticipated, raising the saturation cutoff required to pass the test increased the numbers of infants who failed. If pass was defined with a saturation cutoff value of ≥88%, 63% of all infants tested passed this standard. If instead a saturation value of ≥92% was required to pass, then 39% of tested infants passed. This represents a shift of nine infants from pass to fail. Even at the higher saturation cutoff of 92%, we identified infants treated with oxygen who were able to maintain that saturation in room air.
Saturation Values in Comparison Group of Neonates Receiving Room Air
A total of 10 neonates who were receiving room air at 36 weeks corrected age were studied for 60 minutes under the same physiologic definition protocol. The saturation values in these neonates had a mean and median of 98% (range 91 to 100).
The physiologic definition described in this report standardized the definition of BPD in this single center study. The physiologic definition diagnosed BPD in 30 (24%) of 124 neonates, while the clinical definition diagnosed BPD in 45 (36%) of these 124 infants. The test proved to be safe and had excellent reliability. In addition, the test demonstrated high validity, as the results were highly correlated with clinical measures of ongoing pulmonary disease including discharge home in oxygen, length of initial hospitalization, and readmissions to hospital in the first year of life. The disparate results between this test standardized by oxygen saturation and the clinical definition of BPD may indicate that oxygen saturation goals differed between different attending physicians.
In this evaluation, the test proved to be safe and feasible. In fact, one could argue that neonates monitored under this protocol with continuous direct nursing observation were safer than infants who are weaned in routine clinical practice when direct observation is performed only intermittently. The test was required in 17% of this cohort and thus is feasible to perform within a clinical trial. We arbitrarily chose the length of each weaning step and the observation period in room air as 10 and 60 minutes, respectively, to allow sufficient time at each step for the saturation to equilibrate. With these time parameters, the test required no more than 2 hours to perform and thus was feasible. In this cohort, all infants reached saturation equilibrium within 5 minutes of each wean. Therefore, the protocol may be modified to 5 minutes at each weaning step, which will reduce the time needed for the study.
The magnitude of the change in the definition of BPD is comparable to the effect size seen in clinical trials of therapeutic modalities such as vitamin A that demonstrated an 11% reduction in BPD.17 We speculate that the use of the physiologic definition may reduce the variation in the diagnosis of BPD between centers. Such variation may have minimized treatment effects of efficacious therapies in the past.
We were unable to identify any demographic characteristics of infants who failed the test that distinguished them from those who passed. The mean birthweight of those who failed was 100 g less than those who passed; although this difference was not statistically significant, it may be clinically important. In addition, significantly more of those who failed the test were born at less than 26 weeks gestation compared to those who passed. The relatively small number of infants studied under this protocol may have limited our statistical ability to detect real differences.
Our study may be criticized for the selection of the saturation cutoff point of <88% as the failure definition. The oxygen saturation cutoff point has merit on several levels. First, there is no consensus on an acceptable saturation for infants with BPD. Second, the STOP ROP study, the only published trial in which saturation levels have been randomly assigned in preterm infants, randomized patients to be studied at similar saturation levels of 89 to 94 or 96 to 99%. The study demonstrated no meaningful differences in outcome between the groups managed with saturations in these ranges, except for a higher rate of BPD in the group managed with higher saturations.13 Third, at least one study has suggested that infants managed with lower saturations in the range of 70 to 90% had reduced retinopathy and BPD with similar acute neurologic outcomes.10 The analysis of the impact of the saturation cutoff point indicates that there is very little impact on the passing rate with a cutoff of 88, 89, or 90% saturation. Some may find the cutoff of 90% more acceptable. The test can be modified easily to use different levels of saturation as a cutoff point for the definition of failure should researchers desire. A second potential criticism is the relatively brief observation period in room air. All infants who failed the test did so with in the first 30 minutes of the room air observation period. If one intended to use this test to guide clinical decisions about an individual patient's care, a longer observation period might be advisable.
The ideal saturation level for the management of preterm infants is not known. We monitored a small group of convalescent premature infants in room air and found the median saturation values to be 98%. These data are similar to those reported by Ng18 and colleagues in 33 preterm infants studied in room air. Two radically different approaches to oxygen saturation targets have been proposed: the first approach suggests that oxygen saturation should mimic the lower saturation values seen in utero and thus potentially optimize the body's environment to developmentally regulated cellular processes. The second approach argues that preterm infants should be managed with oxygen goals that match the higher saturation values achieved in healthy preterm and term newborns. This debate can only be resolved by clinical trials that randomize infants in oxygen to different saturation values, and assess both short- term markers of oxidant injury and longer-term outcomes.
The study is intended to standardize the outcome of BPD as an endpoint in clinical trials. It may be difficult to achieve consensus on various practices, such as oxygen administration, between physicians and centers participating in multicenter trials. This physiologic definition leaves clinical practice the same, and uses a brief window of evaluation where all babies are assessed similarly. This definition will minimize practice variation between centers for the short period of the study. Minimizing variation may allow trials to more accurately detect real treatment differences that were previously masked by such variation. The physiologic definition is not intended to be used as a guide for clinical management. The infants studied, including those who passed the challenge, were all returned to their baseline oxygen and managed at the discretion of their attending physician. It was not the intent of this study to evaluate the safety of allowing infants who passed to remain in room air, and thus we cannot comment on the safety of that practice nor would we recommend it based on these data. Additional work is needed to determine if this timed oxygen reduction would be of benefit in weaning oxygen in clinical practice.
The physiologic definition of BPD proposed improves the definition of BPD by minimizing variations introduced by oxygen prescription practices. However, better tools are needed to identify these infants. Such tools may include biomarkers of lung injury, or simplified tests of infant pulmonary function and may ultimately replace this simple clinical standard. Further work is needed to expand this study to multiple institutions and insure that the test is generalizable and feasible. Such a multicenter evaluation is in progress.
Lemons JA, Oh W, Korones SB et al. Very low birthweight outcomes of the National Institute of Child Health and Human Development Neonatal Research Network, January 1995 to December 1996. Pediatrics 2001;107:e1–e6.
Furman L, Borawski-Clark E, Aucott S, Hack M . Hospitalization as a measure of morbidity among very low birth weight infants with chronic lung disease. J Pediatr 1996;128:447–452.
Shennan AT, Ohlsson A, Lennox K, Hoskins EM . Abnormal pulmonary outcomes in premature infants: prediction from oxygen requirements in the neonatal period. Pediatrics 1988;82:527–532.
Garg M, Bautista D, Keens TG . Clinically unsuspected hypoxia during sleep and feeding in infants with BPD. Pediatrics 1988;81:635–642.
Martin RJ, DeFiore J, Jana L et al. Persistence of the biphasic ventilatory response to hypoxia in preterm infants. J Pediatr 1998;132:960–964.
Moyer-Mileur LJ, Nielson D, Pfeffer KD, Witte MK, Chapman DL . Eliminating sleep-associated hypoxemia improves growth in infants with bronchopulmonary dysplasia. Pediatrics 1996;98:779–783.
Singer L, Baley J, Hawkins S et al. Oxygen desaturation complicates feeding in infants with bronchopulmonary dysplasia after discharge. Pediatrics 1992;90:380–384.
Sekar KC, Duke JC . Sleep apnea and hypoxemia in recently weaned premature infants with and without bronchopulmonary dysplasia. Pediatr Pulmonol 1991;10:112–116.
Zanardo V, Trevisanuto D, Dani C, Bottos M, Guglielmi A, Cantarutti F . Oxygen saturation in premature neonates with bronchopulmonary dysplasia in a hammock. Biol Neonate 1995;67:54–58.
Tin WMD, Pennefather P, Hey E . Pulse oximetry, severe retinopathy, and outcome at one year in babies of less than 28 weeks gestation. Arch Dis Child Fetal Neonatal Ed 2001;84:F106–F110.
HCFA. Scheduling and Documenting Recertification for Medical Necessity for Oxygen. HCFA Manual of Procedures. Vol. 5. Washington, DC: HCFA; 1998. Section 6.3.
Barrington K, Finer N, Li D . Pre-discharge respiratory recordings in VLBW infants. J Pediatr 1996;129:934–940.
STOP-ROP Investigators. Supplemental Therapeutic Oxygen for Prethreshold Retinopathy of Prematurity. (STOP-ROP) A randomized controlled trial. I: Primary outcomes. Pediatrics 2000;105:295–310.
Landis JR,, Koch DG . The measurement of observer agreement for categorical data. Biometrica 1977;33:159–164.
Aday LA . Defining and clarifying survey variables. In: Designing and Conducting Health Surveys. 2nd ed. San Francisco: Jossey-Bass Publishers; 1996.
Benaron D, Benitz W . Maximizing the stability of oxygen delivered by nasal cannula. Arch Pediatr Adolesc Med 1994;148:294–300.
Tyson JW, Wright LL, Oh W et al. Vitamin A supplementation for extremely low-birth-weight infants. N Engl J Med 1999;24:1962–1968.
Ng ASN, Primhak RA, Shaw NJ . Arterial oxygen saturation in healthy preterm infants. Arch Dis Child Fetal Neonatal Ed 1998;790:F64–F66.
We wish to thank the infants and their families who participated in the study. We also acknowledge the helpful critiques and suggestions of our colleagues Maureen Hack and Richard Martin, and the helpful suggestions of an anonymous reviewer.
Supported in part by a Specialized Clinical Investigator Award from the National Institutes of Child Health and Development, HD21364-15SI.
About this article
Cite this article
Walsh, M., Wilson-Costello, D., Zadell, A. et al. Safety, Reliability, and Validity of a Physiologic Definition of Bronchopulmonary Dysplasia. J Perinatol 23, 451–456 (2003). https://doi.org/10.1038/sj.jp.7210963
Pediatric Pulmonology (2021)
Discontinuing Nasal Continuous Positive Airway Pressure in Infants ≤32 Weeks of Gestational Age: A Randomized Control Trial
The Journal of Pediatrics (2021)
A Randomized Placebo-Controlled Pilot Trial of Early Targeted Nonsteroidal Anti-Inflammatory Drugs in Preterm Infants with a Patent Ductus Arteriosus
The Journal of Pediatrics (2021)
Effect of Early Targeted Treatment of Ductus Arteriosus with Ibuprofen on Survival Without Cerebral Palsy at 2 Years in Infants with Extreme Prematurity: A Randomized Clinical Trial
The Journal of Pediatrics (2021)
Respiratory distress syndrome and bronchopulmonary dysplasia after fetal growth restriction: Lessons from a natural experiment in identical twins