Association between Intermittent Hypoxemia and Severe Bronchopulmonary Dysplasia in Preterm Infants

Abstract

Rationale: Bronchopulmonary dysplasia increases the risk of disability in extremely preterm infants. Although the pathophysiology remains uncertain, prior exposure to intermittent hypoxemia may play a role in this relationship.

Objectives: To determine the association between prolonged episodes of intermittent hypoxemia and severe bronchopulmonary dysplasia.

Methods: A post hoc analysis of extremely preterm infants in the Canadian Oxygen Trial who survived to 36 weeks’ postmenstrual age was performed. Oxygen saturations <80% for ⩾1 minute and the proportion of time per day with hypoxemia were quantified using continuous pulse oximetry data that had been sampled every 10 seconds from within 24 hours of birth until 36 weeks’ postmenstrual age. The study outcome was severe bronchopulmonary dysplasia as defined in the 2001 NIH Workshop Summary.

Measurements and Main Results: Of 1,018 infants, 332 (32.6%) developed severe bronchopulmonary dysplasia. The median number of hypoxemic episodes ranged from 0.8/day (interquartile range, 0.2–1.1) to 60.2/day (interquartile range, 51.4–70.3) among the least and most affected 10% of infants. Compared with the lowest decile of exposure to hypoxemic episodes, the adjusted relative risk of severe bronchopulmonary dysplasia increased progressively from 1.72 (95% confidence interval, 1.55–1.90) at the 2nd decile to 20.40 (95% confidence interval, 12.88–32.32) at the 10th decile. Similar risk gradients were observed for time in hypoxemia. Significant differences in the rates of hypoxemia between infants with and without severe bronchopulmonary dysplasia emerged within the first week after birth.

Conclusions: Prolonged intermittent hypoxemia beginning in the first week after birth was associated with an increased risk of developing severe bronchopulmonary dysplasia among extremely preterm infants.

Clinical trial registered with www.isrctn.com (ISRCTN62491227) and www.clinicaltrials.gov (NCT00637169).

At a Glance Commentary

Scientific Knowledge on the Subject

Severe bronchopulmonary dysplasia (BPD) and greater exposure to prolonged episodes of intermittent hypoxemia (oxygen saturations <80% for ⩾1 min) during the first 2–3 months after birth are both associated with an increased risk of disability in extremely preterm infants. Whether greater exposure to prolonged episodes of intermittent hypoxemia increases the risk of severe BPD is uncertain.

What This Study Adds to the Field

This post hoc observational study of 1,018 extremely preterm infants who were enrolled in the Canadian Oxygen Trial examined the relationship between exposure to prolonged hypoxemic episodes and the risk of severe BPD using continuous pulse oximetry data recorded from within 24 hours after birth until 36 weeks’ postmenstrual age. The risk-adjusted probability of developing severe BPD increased progressively from 4.2% (95% confidence interval, 2.5–5.9%) among infants in the lowest decile of exposure to prolonged hypoxemic episodes to 85.2% (95% confidence interval, 79.7–90.8%) among infants in the highest exposure decile. Significant differences in hypoxemic exposure among infants who developed severe BPD, compared with those who did not, emerged within the first week after birth and increased in magnitude over the first 4 weeks of life.

Bronchopulmonary dysplasia (BPD), or infant chronic lung disease, is a common complication of prematurity and a well-established risk factor for childhood neurodevelopmental impairment (1–5). It is uncertain whether BPD plays a causal role in this association and, if it does, what mechanisms may be involved. One potential contributor is intermittent hypoxemia. Almost all extremely preterm infants, including those receiving supplementary respiratory support and invasive mechanical ventilation, experience intermittent hypoxemia during the first months after birth owing to apnea of prematurity and cardiorespiratory instability (6, 7). An examination of oxygen saturation values measured by continuous pulse oximetry (Sp_O2) in extremely preterm infants who were enrolled in the multicenter COT (Canadian Oxygen Trial) demonstrated a significant association between exposure to prolonged hypoxemic episodes (Sp_O2 <80% for at least 1 min) during the first 2–3 months after birth and the risk of late death or disability at 18 months’ corrected age (8). Two single-center studies found that very preterm infants who developed BPD more frequently exhibited episodes of brief hypoxemia (events lasting at least 10 s) during the first month after birth than infants who did not develop BPD (9, 10). Whether infants who develop BPD experience greater exposure to prolonged hypoxemic episodes during the newborn period is unknown. This question is of particular importance because short hypoxemic episodes, events lasting <1 minute, may not predispose extremely preterm infants to increased risk of childhood neurodevelopmental impairment (8). Demonstration of an association between recurrent and prolonged episodes of hypoxemia during the first months after birth and the development of BPD may identify an important and potentially causal factor that links BPD to poor neurologic outcomes. Therefore, we conducted the present post hoc analysis to measure the strength of the association between exposure to prolonged intermittent hypoxemia between birth and 36 weeks’ postmenstrual age (PMA) and the risk of severe BPD.

Methods

Design and Population

We conducted a post hoc, observational study using prospectively collected clinical and oximeter data from the COT (11). This trial enrolled infants of 23–27 weeks’ gestation within 24 hours after birth in 25 hospitals in Canada, the United States, Argentina, Finland, Germany, and Israel. Study infants were born between December 2006 and August 2010. Follow-up assessments occurred between October 2008 and December 2012. The original trial cohort excluded infants considered not viable; those with severe congenital malformations, cyanotic heart disease, or pulmonary hypertension; and those unlikely to be available for long-term follow-up (11). The present study was limited to participants who survived to 36 weeks’ PMA and had suitable oximeter and in-hospital respiratory outcome data. This analysis was completed in 2021 according to a protocol that had been approved by the COT steering committee. The research ethics boards of all clinical centers in the COT approved the trial protocol, and written informed consent was obtained from a parent or guardian of every study infant.

Study Oximetry and Analysis of Oxygen Saturation Data

The implementation of study oximetry and details of the oximeter data processing in the COT were described previously (8, 11). Briefly, continuous pulse oximetry was conducted beginning within 24 hours after birth and continued until at least 36 weeks’ PMA in all study participants, except those discharged to home at an earlier time point. All study oximeters were set to a 16-second averaging time with data sampled and stored every 10 seconds. Downloaded oximeter data were electronically transferred to the trial data coordinating center and screened for validity. Invalid measures, defined as those with a 0 value for heart rate or Sp_O2, or any of the oximeter exception flags for displaced sensor, ambient light, interference, or low perfusion were discarded. Only oximeter data recorded up until 36 weeks’ PMA were considered in the present study.

This analysis categorized hypoxemia in two ways. First, we quantified the daily frequency of prolonged hypoxemic episodes lasting approximately 1 minute or longer (six or more consecutive 10-second Sp_O2 sample values <80%). Second, we calculated the total daily proportion of time in hypoxemia (100 × total duration of recordings with Sp_O2 values <80%/total duration of oximeter recording). These categorizations were chosen because a prior analysis of COT data had demonstrated an association between these specific measures of hypoxemia and the risk of adverse neurodevelopment in early childhood (8). The same relationship could not be confirmed for exposure to short intermittent hypoxemic episodes (8).

Outcome

The primary outcome in this analysis was severe BPD defined according to the 2001 NIH Workshop Summary (12). Severe BPD was a prespecified secondary outcome in the COT and diagnosed in infants who received supplemental oxygen for more than 12 hours per day on at least 28 days before 36 weeks’ PMA and at least 30% supplemental oxygen and/or positive airway pressure within 2 days before or after 36 weeks’ PMA (11). The outcome of late death or disability, defined as death between 36 weeks’ PMA and follow-up at 18 months’ corrected age or survival with neurodevelopmental impairment, was used to confirm an association between severe BPD and adverse 18-month outcomes in the present study cohort (8).

Statistical Analysis

The number of hypoxemic episodes recorded on each study participant was determined for each study day between COT enrollment and 36 weeks’ PMA. For the primary analyses, these data were summarized into a single median value for each infant. The strength of the association between the median daily number of hypoxemic episodes and severe BPD was assessed using multivariable logistic regression. The model was fitted with severe BPD as the dependent variable. Square root–transformed values for the median number of hypoxemic episodes per day, expressed as a continuous variable, were used in the model to stabilize the skewed distribution of the observed hypoxemia data. The model was adjusted for other prespecified baseline independent risk factors by including the following covariates: gestational age, birth weight, sex, treatment with antenatal corticosteroids, birth at a study center, study center, and the assigned COT treatment group. Although the model was fitted to individual-patient data, model fit and risk gradients have been computed by subdividing the cohort into deciles according to the median number of hypoxemic episodes per day. Adjusted relative risks were calculated from the logistic regression models by computing the ratio of the average modeled probability of developing severe BPD in each decile relative to the lowest decile. Equivalent analyses were performed to measure the strength of the association between the median proportion of time per day in hypoxemia and the development of severe BPD.

To assess the longitudinal association between intermittent hypoxemia and severe BPD, the daily frequencies of prolonged hypoxemic episodes were summarized as median values for each infant for each postnatal week between birth and 2 months of age. Owing to an observed parabolic rise and fall in the frequency of hypoxemic events, the relationship between hypoxemia and postnatal age was first assessed using a quadratic mixed-effects model with robust variance estimation. The model was fitted with the weekly frequency of hypoxemic episodes as a continuous dependent variable and a quadratic product of severe BPD × (week)² as an independent term. The model was adjusted for the same baseline covariates used in the logistic regression models. Random-effects terms for the patient identifier and weekly time point variables were used, as was a second-order autoregressive correlation structure, to account for within-subject correlation of the weekly frequency of hypoxemic episodes. To estimate the rate of rise in hypoxemic episodes, a linear mixed-effects model using the same parameters as described above, with the exception of a severe BPD × week product term, was constructed to compare the linear change in the weekly frequency of hypoxemic events over the first 4 weeks after birth between infants who developed severe BPD and those who did not. Equivalent methods were used to model the longitudinal patterns in the proportion of recorded time per day in hypoxemia.

To facilitate clinical interpretation of the longitudinal data, the mixed-effects models were constructed and depicted graphically using the untransformed (observed) measures of hypoxemia. However, to determine whether the longitudinal patterns in the exposure to hypoxemia differed over the first 8 weeks of age among infants who developed severe BPD and those who did not, the coefficients associated with the product terms (interactions) of severe BPD × week or (week)² and their corresponding P values were estimated from models that used the square root–transformed hypoxemia data.

The association between severe BPD and late death or disability assessed at 18 months’ corrected age was evaluated using multivariable Poisson regression with robust variance estimation. All statistical analyses were conducted using Stata/SE 15.1 (StataCorp LLC).

Results

Of the 1,201 COT participants, 1,035 survived to 36 weeks’ PMA and were eligible for this study. A total of 17 infants had missing outcome data, valid oximeter data, or both, leaving 1,018 (98.4%) for this analysis cohort. Table 1 summarizes the clinical characteristics and respiratory outcomes of the study population. Severe BPD was diagnosed in 332 infants (32.6%). Late death or disability occurred more frequently among infants who developed severe BPD than among those who did not (56.3% vs. 36.0%; adjusted relative risk 1.47; 95% confidence interval [CI], 1.27–1.71). See Table E1 in the online supplement for additional 18-month outcome data.

Table 1.

Characteristics of the Study Participants

Characteristic	Study Cohort (n = 1,018)*
Gestational age, mean (SD), wk	25.8 (1.1)
Birth weight, mean (SD), g	856 (189)
Maternal race
White	675 (66.3)
Black	174 (17.1)
Asian	106 (10.4)
Other/unknown	63 (6.2)
Antenatal corticosteroids	920 (90.4)
Sex, F	479 (47.1)
Born at study center	939 (92.2)
Status at enrollment
Age at enrollment, median (IQR), h	17.8 (11.8–22.1)
Supplemental oxygen	365 (35.9)
Any use of positive airway pressure	986 (96.9)
Endotracheal tube in situ	770 (78.1)
Received surfactant	879 (86.4)
Severe bronchopulmonary dysplasia	332 (32.6)

Definition of abbreviations: IQR = interquartile range; n = number of participants.

^*Data are expressed as n (%) unless otherwise indicated. Data are for 1,018 Canadian Oxygen Trial participants who survived to 36 weeks’ postmenstrual age and had the necessary pulse oximeter and in-hospital respiratory outcome data to perform the study analyses.

Valid oximetry recordings were available for a median duration of 59.5 days (interquartile range [IQR], 50.3–67.0) between COT enrollment within the first 24 hours after birth and 36 weeks’ PMA. A total of 0.2% of oximeter data had been discarded because of an exception code generated by the study oximeter. The median number of prolonged intermittent hypoxemic episodes per day in the analysis cohort was 13.2 (IQR, 4.9–27.3). The median proportion of time per day with an Sp_O2 <80% was 3.6% (IQR, 1.7–6.8%).

Figure 1 displays the relationship between the measures of hypoxemia and the modeled probability of developing severe BPD. Both the frequency of intermittent hypoxemic episodes and the proportion of time per day with Sp_O2 <80% between birth and 36 weeks’ PMA demonstrated a “dose–response” association with the risk of developing severe BPD. Compared with the lowest decile of exposure to hypoxemic episodes, the adjusted relative risk of severe BPD increased progressively from 1.72 (95% CI, 1.55–1.90) at the 2nd decile to 20.40 (95% CI, 12.88–32.32) at the 10th decile (Table 2). Similar results were obtained for the proportion of time per day with Sp_O2 <80%, although the estimated relative risks were stronger with episodes per day (Table 3).

Figure 1.

Association between exposure to prolonged hypoxemia and the probability of developing severe bronchopulmonary dysplasia (BPD). The left panel shows the relationship between the median number of hypoxemic episodes lasting six or more consecutive 10-second readings or approximately 1 minute or longer and the probability of developing severe BPD. The right panel shows the relationship between the median proportion of time per day with the amount of oxygen saturation as measured by pulse oximetry (Sp_O2) <80%. Both measures of hypoxemia are summarized for the period of time beginning within the first 24 hours after birth and ending at 36 weeks’ postmenstrual age or at discharge to home if it occurred earlier. The solid curves and shaded areas display the probabilities and 95% confidence intervals derived from an adjusted logistic regression model fitted with severe BPD as the dependent variable and the square root–transformed hypoxemia data as the continuous independent variable. The dashed curves display the probabilities obtained from the unadjusted logistic regression model. To enable clinical interpretation, the scales of the x-axes quantify the untransformed rates of hypoxemia. To visualize model fit, the study cohort was subdivided into deciles according to the amount of exposure to each measure of hypoxemia. The observed (unadjusted) rates of severe BPD for each decile, plotted against the mean, square root–transformed rates of hypoxemia in each decile, are superimposed on the graphical displays. The C statistic values for the adjusted logistic regression models were 0.861 for prolonged hypoxemic episodes per day and 0.844 for the proportion of time per day with Sp_O2 <80%, indicating strong model fit.

Table 2.

Association between Prolonged Intermittent Hypoxemic Episodes and Severe BPD

Exposure Decile	Hypoxemic Episodes per Day [Median (IQR)]	Observed Rates of Severe BPD [n/N (%)]	Modeled Probability of Developing Severe BPD [% (95% CI)]*	Adjusted RR (95% CI) for Developing Severe BPD*†
1	0.8 (0.3–1.1)	0/102 (0)	4.2 (2.5–5.9)	Reference
2	2.6 (2.1–3.1)	4/102 (3.9)	7.2 (4.9–9.5)	1.72 (1.55–1.90)
3	4.9 (4.3–5.7)	9/102 (8.8)	10.8 (8.1–13.5)	2.58 (2.16–3.09)
4	7.9 (7.1–8.6)	19/102 (18.6)	15.5 (12.6–18.4)	3.71 (2.89–4.75)
5	11.4 (10.7–12.1)	24/101 (23.8)	21.2 (18.2–24.2)	5.08 (3.73–6.91)
6	15.5 (14.2–16.5)	30/102 (29.4)	27.9 (24.8–31.1)	6.69 (4.68–9.56)
7	20.1 (18.6–21.6)	42/102 (41.2)	36.1 (32.7–39.6)	8.65 (5.80–12.90)
8	27.3 (24.7–29.5)	59/102 (57.8)	47.7 (43.3–52.2)	11.43 (7.37–17.72)
9	36.7 (33.6–40.2)	65/102 (63.7)	61.8 (56.0–67.5)	14.78 (9.31–23.49)
10	60.2 (51.4–70.3)	80/101 (79.2)	85.2 (79.7–90.8)	20.40 (12.88–32.32)

Definition of abbreviations: BPD = bronchopulmonary dysplasia; CI = confidence interval; COT = Canadian Oxygen Trial; IQR = interquartile range; n = number of infants with severe BPD; N = total number of infants in the specified decile group; RR = relative risk.

^*Model-estimated probabilities of developing severe BPD and relative risks comparing individual hypoxemia exposure deciles with the lowest decile were calculated from a logistic regression model that used a square root transformation of the median number of daily hypoxemic episodes as a continuous independent variable. The model was adjusted for gestational age, birth weight, sex, treatment with antenatal steroids, birth at a study center, study center, and COT treatment group. The relative risks were estimated at the mean observed value for the square root–transformed measure of hypoxemia in each decile.

^† P < 0.001 for the modeled risk gradients comparing each exposure decile with the lowest decile.

Table 3.

Association between the Proportion of Time in Hypoxemia and Severe BPD

Exposure Decile	Proportion of Time per Day with SpO2 <80% [% Median (IQR)]	Observed Rates of Severe BPD [n/N (%)]	Modeled Probability of Developing Severe BPD [% (95% CI)]*	Adjusted RR (95% CI) for Developing Severe BPD*†
1	0.4 (0.2–0.5)	0/102 (0)	5.4 (3.3–7.5)	Reference
2	1.0 (0.8–1.2)	7/102 (6.9)	9.1 (6.4–11.7)	1.69 (1.52–1.88)
3	1.7 (1.6–1.9)	12/102 (11.8)	13.3 (10.4–16.2)	2.48 (2.06–2.98)
4	2.4 (2.2–2.6)	21/102 (20.6)	17.4 (14.4–20.4)	3.25 (2.55–4.13)
5	3.1 (3.1–3.4)	24/101 (23.8)	22.5 (19.5–25.6)	4.20 (3.14–5.63)
6	4.1 (3.9–4.4)	33/102 (32.4)	28.6 (25.6–31.7)	5.34 (3.81–7.49)
7	5.3 (5.0–5.6)	44/102 (43.1)	36.2 (32.8–39.6)	6.76 (4.62–9.88)
8	6.8 (6.4–7.3)	51/102 (50.0)	45.8 (41.5–50.1)	8.54 (5.64–12.95)
9	9.1 (8.4–9.7)	62/102 (60.8)	58.6 (52.9–64.2)	10.93 (7.01–17.03)
10	13.7 (12.2–16.3)	78/101 (77.2)	79.9 (73.5–86.3)	14.90 (9.51–23.36)

Definition of abbreviations: BPD = bronchopulmonary dysplasia; CI = confidence interval; COT = Canadian Oxygen Trial; IQR = interquartile range; n = number of infants with severe BPD; N = total number of infants in the specified decile group; RR = relative risk; Sp_O2 = oxygen saturation as measured by pulse oximetry.

^*Model-estimated probabilities of developing severe BPD and relative risks comparing individual hypoxemia exposure deciles with the lowest decile were calculated from a logistic regression model that used a square root transformation of the median proportion of time in hypoxemia per day as a continuous independent variable. The model was adjusted for gestational age, birth weight, sex, treatment with antenatal steroids, birth at a study center, study center, and COT treatment group. The relative risks were estimated at the mean observed value for the square root–transformed measure of hypoxemia in each decile.

^† P < 0.001 for the modeled risk gradients comparing each exposure decile to the lowest decile.

The longitudinal patterns of hypoxemia during the first 8 weeks after birth were notable for the following findings: All infants had rising rates of hypoxemia during the first 4–5 weeks after birth, followed by gradual declines in hypoxemia thereafter (Figure 2, left panel, and Tables E4 and E5). Despite this shared pattern, however, the amounts of exposure to hypoxemia differed significantly between infants who had a diagnosis of severe BPD at 36 weeks’ PMA and those who did not. During the first 4 weeks after birth, the frequency of intermittent hypoxemia increased at over twice the rate among infants who developed severe BPD (model-adjusted weekly increase in the mean number of hypoxemic events per day: 11.03; 95% CI, 9.95–12.11) compared with those who did not (weekly increase: 4.53; 95% CI, 4.03–5.02; P < 0.001 for interaction) (Figure 2, right panel). Significant differences in the rates of hypoxemia were already observed by the end of the first week after birth. The linear mixed-effects model estimated that infants who developed severe BPD experienced, on average, 11.1 (95% CI, 9.9–12.2) prolonged intermittent hypoxemic events per day during the first week compared with 8.3 (95% CI, 7.7–9.0) events per day among infants who did not develop severe BPD (P < 0.001). The corresponding rates for Week 4 after birth were 44.2 events per day (95% CI, 40.8–47.6) among infants with severe BPD and 21.9 (95% CI, 20.3–23.5) among infants without severe BPD. Comparisons of the modeled values obtained using the square root–transformed rate data confirmed significant differences (P < 0.001) between the two groups at each weekly time point over the first 4 weeks. Similar results were obtained when hypoxemia was expressed as the proportion of time per day with Sp_O2 <80% (Figure 2).

Figure 2.

Exposure to prolonged hypoxemia by postnatal age stratified by severe bronchopulmonary dysplasia (BPD) status. (A) The top panels show the association between the median number of prolonged hypoxemic episodes per day by postnatal age in weeks for infants who did and those who did not develop severe BPD. (B) The bottom panels show the corresponding results for the median proportion of time per day with an oxygen saturation as measured by pulse oximetry (Sp_O2) <80%. The plots on the left depict the rates of hypoxemia over the first 8 postnatal weeks, and the plots on the right depict the rates of hypoxemia over the first 4 weeks. The solid curves and shaded areas display the estimated rates of hypoxemia and 95% confidence intervals derived from the adjusted mixed-effects models. The dashed curves display rates from the unadjusted models. To visualize model fit, the average observed (unadjusted) rates of hypoxemia at each weekly time point are plotted. The P values reported in the plots correspond to the severe BPD × (week)² (plots on the left) and the severe BPD × week (plots on the right) interaction terms included in the mixed-effects models using the square root–transformed hypoxemia rate data. These P values of <0.001 indicate significant differences in Sp_O2 trajectories between infants who developed severe BPD and those who did not. The model-derived rates of prolonged intermittent hypoxemia and the proportion of time per day with Sp_O2 <80% differed significantly (P < 0.001) at each weekly time point over the first 4 postnatal weeks between infants who developed severe BPD and those who did not. Tables E2 and E3 show the model-estimated weekly rates of hypoxemia for the first 4 postnatal weeks. The observed weekly rates of hypoxemia for the first 8 postnatal weeks are reported in Tables E4 and E5. CI = confidence interval.

Discussion

This post hoc study examined the association between prolonged intermittent hypoxemia during the neonatal period and the diagnosis of severe BPD in extremely preterm infants who participated in the COT and survived to 36 weeks’ PMA. The risk of severe BPD increased with increasing exposure to intermittent hypoxemia. Significant differences in hypoxemic exposure between infants who developed severe BPD, compared with those who did not, emerged within the first week after birth and increased in magnitude over the first 4 weeks. We previously reported a strong association between prolonged intermittent hypoxemia and late death or childhood disability in the same multicenter cohort (8). In the present analysis, we confirmed an association between severe BPD and late death or adverse neurodevelopment. Collectively, these results indicate that early and prolonged intermittent hypoxemia may contribute to the well-known risk of poor developmental outcomes among infants with severe BPD (2–5).

Because of its post hoc and observational design, our study cannot determine whether prolonged intermittent hypoxemia causes lung injury and the subsequent development of BPD or whether it is simply a marker of greater severity in lung disease. Data from animal models indicate that intermittent hypoxemia may promote inflammation and oxidative stress, both of which are implicated in the pathophysiology of BPD (13–16). Despite these shared mechanistic features, murine studies suggest intermittent hypoxemia alone may not be sufficient to generate the parenchymal lung damage that is characteristic of BPD (17, 18). However, recurrent hypoxemia may contribute to lung tissue injury when it occurs in conjunction with reoxygenation and subsequent hyperoxemia during administration of oxygen therapy or at times of metabolic stress, such as infection (15, 17–19). Hypoxemia may also prompt clinicians to administer greater amounts of respiratory support or oxygen therapy, and these interventions may, in turn, exacerbate lung damage (20). Lastly, it is unknown whether distinct mechanisms of intermittent hypoxemia, such as immature respiratory control, loss of lung volume, or increased pulmonary vascular resistance, may differentially predispose extremely preterm infants to lung injury and the development of BPD (7).

Previous studies of intermittent hypoxemic episodes during the first months after birth in extremely premature infants observed that the frequency of hypoxemic events increased over the first 4 weeks after birth and then gradually declined (8, 15, 21). The same pattern was observed in the present study, regardless of subsequent severe BPD status. However, the infants who developed severe BPD demonstrated significantly greater exposure to prolonged hypoxemic episodes within the first week after birth and exhibited a greater than twofold higher rate of rise in the frequency of events over the first month. These findings suggest that real-time recording and examination of continuous pulse oximetry data may aid early identification of extremely preterm infants who are at high risk for poor respiratory outcomes.

The rates of prolonged hypoxemic events remained higher through 8 weeks of age in infants who developed severe BPD compared with those who did not. It is uncertain when, after 36 weeks’ PMA, intermittent hypoxemia resolves in the majority of extremely preterm infants with severe BPD. However, postdischarge follow-up studies show that episodic hypoxemia can persist throughout childhood in preterm-born individuals who developed the most-extreme forms of BPD (22–24).

The present analysis focused on the outcome of severe BPD as defined in the 2001 NIH Workshop Summary (12). This definition was chosen because we analyzed the data of participants in the COT for whom severe BPD had been recorded as a prespecified secondary outcome (11). Two previous single-center studies demonstrated an association between greater exposure to intermittent hypoxemia, defined as events lasting at least 10 seconds during the first 4 weeks after birth, and the use of supplemental oxygen at 36 weeks’ PMA (9, 10). It is uncertain whether such short hypoxemic episodes are associated with severe BPD. Importantly, only prolonged hypoxemic episodes lasting at least 1 minute have been shown to be associated with neurodevelopmental disability in early childhood (8). Nonetheless, prior studies combined with the present results provide consistent evidence that intermittent hypoxemia is an early, reliable predictor of the development of chronic respiratory illness in extremely preterm infants (9, 10).

The post hoc and observational design of this study is its main limitation. We have identified a plausible association between prolonged intermittent hypoxemia and the subsequent diagnosis of severe BPD. However, future, prospective studies will be needed to determine whether intermittent hypoxemia plays a causal role in the development of BPD. One possible approach would entail the random assignment of extremely preterm infants to different strategies aimed at reducing the frequency of prolonged hypoxemic episodes. A second limitation is the fact that the study cohort was enrolled in a clinical trial of oxygen saturation targeting that used modified study oximeters (11). Saturation values <80% were well below both of the trial target ranges and were not affected by the oximeter masking feature (11). In addition, the randomly assigned treatment in the COT did not affect the risk of severe BPD, and all present analyses were adjusted for the COT treatment group (11). Finally, it is possible that a future study using alternative oximeter averaging times and sampling rates may yield modestly different results. However, the shared oximeter parameters between the present study and our prior post hoc analysis of COT data provide strong evidence that prolonged hypoxemic episodes, as characterized here, may contribute to the risk of adverse neurodevelopment associated with BPD (8).

To conclude, prolonged intermittent hypoxemia beginning in the first week after birth was associated with an increased risk of developing severe BPD in this large and multicenter cohort of extremely preterm infants. This finding, combined with the previous observation that prolonged hypoxemic episodes may predispose extremely preterm infants to later adverse neurodevelopmental outcomes, suggests that intermittent hypoxemia may contribute to the high rates of developmental disability among survivors with BPD (8). Confirmation of these observations in future studies would support research that targets prevention of these episodes as a means to improve long-term respiratory and neurologic outcomes among extremely preterm infants.