Introduction The incidence of chronic lung disease (CLD) varies among groups defined by their early pattern of respiratory disease.
Methods The study examined data collected prospectively on 1204 of the 1506 infants born in 2002–2004 at 23–27 weeks gestation who survived to 36 weeks post-menstrual age. Based on their initial respiratory presentation and need for supplemental oxygen during the first 2 weeks, infants were classified as having early and persistent pulmonary dysfunction (EPPD), early recovery of pulmonary function followed by deterioration (PD) or consistently good pulmonary function characterised by low FiO2 (Low FiO2).
Results CLD was diagnosed in 69% of infants with EPPD, 52% with PD, and 17% in the Low FiO2 group. Birth weight z score <−1 conveyed information about CLD risk in all three groups and was the major risk factor for infants in the Low FiO2 group (OR 27; 95% CI 7 to 95). Mechanical ventilation at 7 days was associated with increased risk in the PD (OR 4.2, 95% CI 2.5 to 6.9) and EPPD groups (OR 2.7, 95% CI 1.5 to 4.7), but not the Low FiO2 group (OR 1.5, 95% CI 0.5 to 3.9).
Conclusion The likelihood of a very preterm infant developing CLD and the profile of risk factors linked with CLD are related to the infant's pattern of respiratory disease during the first 2 postnatal weeks. Among infants with little exposure to oxygen during this period, fetal growth restriction, not mechanical ventilation, is the factor with the strongest association with CLD.
Statistics from Altmetric.com
Early pulmonary dysfunction in extremely low gestational age newborns (ELGANs) can be characterised by three distinct patterns, based on the fraction of inspired oxygen (FiO2) they require in the first 2 postnatal weeks.1 A minority of ELGANs have relatively normal pulmonary function throughout the first 2 postnatal weeks. Another group has pulmonary deterioration (PD), characterised by resolving lung disease during the first postnatal week, followed in the second week by a requirement for increased supplemental oxygen and, in some cases, mechanical ventilation. A third group has early and persistent pulmonary dysfunction (EPPD) requiring mechanical ventilation and high concentrations of supplemental oxygen throughout this time.
What is already known on this topic
▶ Pulmonary dysfunction in the first 2 postnatal weeks in preterm infants can be divided into three types according to the fraction of inspired oxygen the infants require.
▶ The risk of chronic lung disease (CLD) varies among infant groups defined by their pulmonary function during the first 2 postnatal weeks.
What this study adds
▶ Among infants with little exposure to oxygen during this period, fetal growth restriction (FGR) is the factor most strongly associated with CLD.
▶ Among infants with pulmonary deterioration or early and persistent pulmonary dysfunction, other factors, such as gestational age, male gender and mechanical ventilation, convey a similar amount of information about CLD risk as FGR.
The incidence of chronic lung disease (CLD), also known as bronchopulmonary dysplasia (BPD), varies among groups defined by their early respiratory function. Among infants with EPPD, approximately two-thirds develop CLD, and the oxygen and ventilation exposures in this group most resemble historical antecedents of BPD.1 Infants with PD are at moderate risk of CLD and have less exposure to oxygen and ventilation than EPPD infants. Although almost one-fifth of infants with relatively normal lung function throughout the first 2 postnatal weeks develop CLD, they have virtually no exposure to supplemental oxygen and ventilation during that period. These observations suggest that several pathophysiological pathways play a role in the development of CLD. Understanding the relative contribution of CLD antecedents within these groups might provide clues to the mechanisms of injury that lead to CLD.
The objective of this study was to identify clinical and demographic antecedents and modifiers of CLD risk in three groups of infants defined by their pattern of early postnatal respiratory function.
The ELGAN study
The infants included in this analysis are a subset of infants enrolled in a multi-centre epidemiological study to identify characteristics and exposures that increase the risk of structural and functional neurological disorders in ELGANs (the ELGAN Study).1 2 From March 2002 to August 2004, women delivering before the 28th week of gestation at one of 14 participating institutions were asked to enrol in the study. Individual institutional review boards at each of the institutions approved the enrolment and consent processes (see Acknowledgements for the list of institutions that approved the study). Of 1506 infants enrolled, 1204 had both the information necessary for early respiratory status assignment and survived to 36 weeks post-menstrual age (PMA) when a CLD diagnosis was made.
Patterns of early respiratory function and CLD
ELGANs were classified into three mutually exclusive groups: those with relatively normal pulmonary function (Low FiO2 group: FiO2 consistently <0.23 on all days between 3 and 7 postnatal days and receiving FiO2 ≤0.25 on day 14), those with PD during the second week of life after a period of normal lung function (PD group: FiO2 <0.23 on any days between 3 and 7 days and receiving FiO2 >0.25 on day 14), and those with early and persistent pulmonary dysfunction (EPPD group: FiO2 consistently ≥0.23 on all days between 3 and 7 postnatal days and receiving FiO2 >0.25 on day 14).1 There were no patients with an FiO2 >0.23 on any day between 3 and 7 postnatal days and receiving <0.25 on day 14, so this group was not included. The diagnosis of CLD was based on whether or not the child was receiving supplemental oxygenation at 36 weeks PMA.
Demographic, pregnancy and neonatal variables
Information on pregnancy characteristics and data describing newborns at the time of delivery included maternal race, receipt of antenatal steroids, chorioamnionitis, pregnancy complications, multi-fetal pregnancy, gender, gestational age and birth weight. The birth weight z score is the number of standard deviations the infant's birth weight is above or below the median weight of infants at the same gestational age in a standard data set.3
The SNAP-II (Score for Neonatal Acute Physiology-II)4 was calculated from measurements taken during the first 12 h of life. Mode of ventilation was defined as the highest level of support on each day and ranged from no support, supplemental oxygen by hood or nasal cannula, nasal continuous positive airway pressure (CPAP) and conventional mechanical ventilation to high frequency ventilation and was recorded on days 0–7, 14, 21 and 28, and at 36 weeks PMA. We also recorded the number of days each infant received supplemental oxygen, CPAP and conventional mechanical ventilation (including high frequency ventilation). Diagnoses of pneumothorax, pulmonary interstitial emphysema (PIE) and pulmonary haemorrhage were those made by the clinicians caring for the infant.
Confirmed early bacteraemia was defined as recovery of an organism from blood drawn during the first postnatal week, and confirmed late bactaeremia as recovery of an organism from blood drawn during the second, third or fourth week. Confirmed tracheal colonisation required the recovery of a pathogen from tracheal aspirate.
The diagnosis of patent ductus arteriosus (PDA) was made by clinicians using their own operational definitions, which might or might not have included echocardiographic findings. We did not record the day of diagnosis. We recorded whether the PDA was ligated and whether indomethacin was offered as medical therapy. If a child had surgical ligation of the PDA, the infant was assigned to the surgical therapy group only, even if the infant first received medical therapy. Infants classified as having received medical therapy received indomethacin and did not have the PDA ligated.
The presence of chorioamnionitis and funisitis was determined by an ELGAN study pathologist at each institution who first engaged in training procedures to minimise interobserver variability, was masked to maternal history, and used predefined operational definitions; 1126 placentas were examined.5 6
Medications were recorded if given on any day during the first week through the fourth week, and included surfactant, analgesics (ie, morphine, fentanyl or methadone), sedatives (ie, lorazepam, midazolam or chloral hydrate), vitamin A and steroids (ie, hydrocortisone and dexamethasone). Indications were not recorded.
We evaluated whether groups of antecedents of CLD differed among the three early respiratory pattern groups. First we calculated risks of CLD among infants classified by their early respiratory pattern and the presence or absence of other characteristics and exposures. The characteristics that most clearly distinguished infants at highest risk of CLD from their peers were then included in logistic regression models to assess the strength of association of each characteristic/exposure with the risk of CLD within each early respiratory pattern group, while adjusting for other factors included in the regression. Gestational age categories (23–24, 25–26, 27 weeks) and birth weight z score groups (<−2, ≥−2 but <−1, ≥−1)7 8 were included in every multivariable model.
Because postnatal phenomena, such as the need for ventilatory assistance, can be influenced by antepartum phenomena, we created logistic regression models in which risk factors are ordered in a temporal pattern, so that the earliest occurring predictors/covariates of an outcome (eg, CLD) are entered first and are not displaced by later occurring covariates.9,–,14 For these time-oriented risk models (TORMs), we categorise sets of antecedents/covariates by the time they occur or are identified. We grouped prenatal and birth characteristics and exposures into the antenatal epoch, all exposures and characteristics during the first week into the early neonatal epoch, and exposures and characteristics occurring or reported between weeks 2 and 4 into the late neonatal epoch. We included in the antenatal epoch a hospital stratum (group) term to account for the possibility that infants born at a particular hospital are more like each other than like infants born at other hospitals.15 Because the risk of CLD among infants born with a sibling did not differ from that of singletons, we did not adjust for number of fetuses.
We used a step down procedure within each epoch, seeking a parsimonious solution without interaction terms. After the antenatal epoch variables were identified, the early neonatal variables were added. We then dropped non-significant neonatal variables but did not permit displacement of antenatal variables. Finally, the late neonatal epoch variables were added to the reduced antenatal/early neonatal set and non-significant late neonatal variables dropped. The contributions of relevant variables in the final model are presented as ORs with 95% CIs. We created three TORMs of CLD, one for each pattern of early respiratory function.
Among 1506 infants enrolled in the ELGAN Study, 1204 survived to 36 weeks PMA and constitute the cohort for this study (figure 1). Early respiratory function was categorised as EPPD in 42% (n=508) of the infants, as PD in 38% (n=456) and as Low FiO2 in 20% (n=240). The incidence of CLD varied among groups defined by these respiratory patterns during the first 2 postnatal weeks. CLD was diagnosed in 69% of infants in the EPPD group, 52% in the PD group and 17% in the Low FiO2 group.
Lower gestational age and lower birth weight were associated with an increased risk of CLD in all respiratory-pattern groups. Lower birth weight z scores were associated with an increased risk of CLD in all groups, particularly the Low FiO2 group. In this group, 38% of infants with a birth weight z score between −2 and −1 and 70% of infants with a birth weight z score of <−2 developed CLD, compared to 13% among appropriate for gestational age birthweight infants. Among infants in the Low FiO2 group, those delivered for maternal indications (preeclampsia) were at an increased risk of CLD (40%, compared to 10–19% among other indications for delivery). In the PD group, male infants were more likely to develop CLD (56% vs 47% among female infants). Aside from gestational age, birth weight and birth weight z score, no antenatal characteristic or exposure among infants in the EPPD group conveyed information about the risk of CLD (table 1).
Early neonatal factors
A SNAP-II greater than 30 and receipt of mechanical ventilation on postnatal day 7 were associated with an increased risk of CLD in all groups. In the Low FiO2 group, receipt of dexamethasone, vitamin A, analgesics and sedation was associated with an increased risk of CLD. In the PD group, receipt of surfactant, hydrocortisone, dexamethasone and analgesics, as well as confirmed bactaeremia and tracheal infection, were associated with an increased risk of CLD. In the EPPD group, receipt of surfactant was associated with an increased risk of CLD and receipt of dexamethasone, sedation and vitamin A was associated with a decreased risk of CLD (table 2).
Late neonatal factors
In all groups, infants who received dexamethasone, analgesics or sedation, as well as those who were ventilated on day 14 or 21, had confirmed tracheal infection, developed pneumothorax or developed isolated intestinal perforation, were at increased risk of CLD. In the Low FiO2 group, receipt of hydrocortisone and vitamin A, or surgery for necrotising enterocolitis (NEC) was associated with an increased risk of CLD. Infants in the PD group were at increased risk of CLD if they received hydrocortisone, had confirmed bactaeremia, or required surgery for NEC. In the EPPD group, confirmed bactaeremia and PDA were associated with an increased risk, and receipt of vitamin A was associated with a decreased risk of CLD (table 3).
Placenta histology and microbiology
Thrombosis of fetal stem vessels, infarcts and increased syncytial knots were associated with an increased risk of CLD, and this was most pronounced in the Low FiO2 group (table 4). In all groups, no organism or group of organisms recovered from the placenta conveyed information about CLD risk (data not shown).
Time-oriented risk models
Low FiO2 group
In the Low FiO2 group, birth weight z score <−1 was associated with an increased risk of CLD. In the early neonatal epoch, only SNAP-II >30 was associated with an increased risk of CLD, while receipt of surfactant was associated with a reduced risk. In the late neonatal epoch, receipt of analgesics was associated with an increased risk of CLD. The presence of PIE was also associated with increased risk; however, this diagnosis was assigned to only six infants (table 5).
Infants in the PD group were at increased risk of CLD if their gestational age was <27 weeks, their birth weight z score was <−1, and they were male. In the early neonatal epoch, SNAP-II >30, definite bactaeremia and mechanical ventilation on day 7 were associated with an increased risk of CLD, while a diagnosis of pneumothorax in the late neonatal epoch was associated with an increased risk of CLD (table 5).
Among infants in the EPPD group, those with a gestational age of 23–24 weeks and a birth weight z score of <−1 were at increased risk of CLD. Delivery for preeclampsia or a fetal indication approached nominal statistical significance. Mechanical ventilation on day 7 was the only risk factor for CLD in the early neonatal epoch. Two late neonatal variables, PIE and receipt of hydrocortisone, provided additional information about an increased risk of CLD (table 5).
In this study, we evaluated CLD risk factors in three groups of infants characterised by their pulmonary function during the first 2 weeks after birth. These groups differed in their likelihood of developing CLD. Differences in early exposure to oxygen might not only have defined these three groups but also influenced CLD risk. However, we hypothesised that each group had its own risk profile for CLD that included factors not directly related to pulmonary function or therapies. We were particularly interested in antecedents of CLD among infants with little exposure to increased concentrations of oxygen during early life (the Low FiO2 group). Although multiple factors were significant in univariate analysis (eg, vitamin A), once adjustments were made in multivariate analysis, many of these were no longer significant.
In a previous report, we demonstrated that fetal growth restriction (FGR) was the antenatal factor that best predicted CLD.16 In the current study, FGR was associated with increased CLD risk in all groups defined by early pulmonary function, even after adjustment for early and late postnatal neonatal exposures and other morbidities. The observation that this effect was most pronounced in the Low FiO2 group suggests that processes that limit fetal growth might predispose to abnormal lung growth before and after birth, ultimately resulting in pulmonary dysfunction. This cascade of events appears to occur even in the absence of exposures likely to result in lung injury.
Severe growth restriction might contribute to CLD risk in several ways. First, factors that impair fetal somatic growth might also impair lung development, resulting in abnormal development of terminal air sacs and alveoli, or abnormal pulmonary angiogenesis.17 This might be similar to the changes characteristic of the ‘new BPD’.18 19
Second, an imbalance between angiogenic and antiangiogenic factors might disrupt normal placental angiogenesis and result in abnormal fetal angiogenesis, including the vasculature of the fetal lung. Preeclampsia is the disorder most closely associated with FGR and is also associated with disturbed angiogenesis.20 Our observation of an increased risk of CLD among those whose placenta had increased syncytial knots, a histological abnormality characteristic of preeclampsia, supports this possibility.
Third, chronic fetal hypoxia, sometimes identified as a factor that impairs growth,21 might be accompanied by impaired lung development. This possibility is supported by observations from animal studies. Impaired alveolar and pulmonary artery development occurs in neonatal mice exposed to chronic hypoxia during the first 2 weeks of life, a period of lung development that corresponds to human fetal lung development during the third trimester.22 Regardless of the mechanism, the interaction between factors that control fetal somatic growth and lung maturation is complicated.
FGR also was associated with CLD in the PD group and to a lesser extent in the EPPD group. In the PD group, characteristics associated with gestational age and male gender appear to have influenced CLD risk. In addition, among neonatal exposures, mechanical ventilation was most influential. Although gestational age and FGR were important in the EPPD group, mechanical ventilation at 7 days and dexamethasone treatment in the late neonatal period were the factors most strongly associated with CLD. The latter association probably reflects selective use of this therapy in infants likely to develop CLD. PIE was also associated with increased risk. These observations suggest that even in the absence of vulnerability imparted by growth restriction, infants with PD are particularly vulnerable to CLD due to extreme immaturity or exposures that injure the lung (eg, high oxygen concentrations and mechanical ventilation).
In the Low FiO2 and EPPD groups, exposure to analgesics and dexamethasone, respectively, during the late neonatal period was associated with an increased risk of CLD. For the Low FiO2 group, one possibility is that analgesic use itself renders an infant more likely to develop CLD, directly or through secondary effects. Narcotics, such as morphine, depress the respiratory drive and might therefore prolong the need for mechanical ventilation and, thus, increase the risk of ventilator-induced lung injury. In a randomised controlled trial,23 infants treated with mechanical ventilation who were allocated to receive continuous morphine infusions were ventilated for 1 week longer than infants allocated to placebo.24 Another explanation for this observation is that exposure to analgesics or dexamethasone is indicative of greater illness severity or longer duration of mechanical ventilation. A more valid conclusion concerning dexamethasone is available from meta-analyses of randomised, placebo controlled trials,25 26 which indicate that this treatment reduces the risk of CLD. The decision to treat an infant with an analgesic or to ventilate an infant might reflect the physician's perception that the infant is sufficiently ill to require these therapies, rather than analgesics or ventilation contributing to CLD risk. Alternatively, many infants treated with mechanical ventilation receive analgesics for sedation. This is an example of confounding by indication, a problem common to many epidemiological studies incorporating clinical decisions as variables, outcomes or exposures.27 28 For example, later-occurring conditions or events, such as NEC or PDA ligation, might both require treatment with narcotic analgesics and increase the risk of CLD through pathogenic mechanisms not involving the medication.
Intra-amniotic inflammation has been implicated in the development of so-called atypical CLD (ie, CLD not preceded by respiratory distress syndrome).29 All infants in our Low FiO2 group and many infants in our PD group would have been classified as having atypical CLD using this definition. We found no association between CLD and either histological markers of placental inflammation or microbiological evidence of placental infection and CLD in groups defined by their early lung function. Our data suggest that intrauterine infection is unlikely to be a risk factor for CLD, regardless of the presence or absence of early respiratory disease. Chorioamnionitis has been inconsistently reported as affecting respiratory outcomes (respiratory distress syndrome and CLD),30 31 most likely because the diagnosis of chorioamnionitis is challenging and does not provide information about the organism, duration or extent of fetal involvement. One possibility why we did not find an association is that respiratory care practices in ELGAN centres might have attenuated the putative pathway to CLD that involves intrauterine inflammation followed by postnatal ventilator-induced lung injury.31
We did not observe a relationship between severe NEC and CLD risk, as has been reported previously.32 33 One explanation for the difference in these findings is that our sample size permitted adjustment for a large number of confounders and antecedents. Another is that antenatal phenomena contributed to both CLD and NEC, but our use of time oriented regression modelling reduced the probability of perceiving an epiphenomenon (such as NEC) as a risk factor.
The risk of CLD varies among infant groups defined by their pulmonary function during the first 2 postnatal weeks. Among infants with little exposure to oxygen during this period, FGR is the factor most strongly associated with CLD. Among infants with PD or EPPD, other factors, such as gestational age, male gender and mechanical ventilation, also convey nearly as much or more information about CLD risk. Therefore, CLD among infants with little exposure to oxygen early in life might result almost exclusively from fetal phenomena related to lung growth. These observations have the potential to inform future investigations of the biological mechanisms underlying the association between patterns of early pulmonary function and the development of CLD.
The authors wish to acknowledge their ELGAN study colleagues.
ELGAN Study Investigators Olaf Dammann, Tufts Medical Center, Boston, Massachusetts, USA; Bhavesh L Shah, Baystate Medical Center, Springfield, Massachusetts, USA; Camilia Martin, Beth Israel Deaconess Medical Center, Boston, Massachusetts, USA; Robert Insoft, Brigham and Women's Hospital, Boston, Massachusetts, USA; Karl Kuban, Boston Medical Center, Boston, Massachusetts, USA; Francis Bednarek, U Mass Memorial Health Center, Worcester, Massachusetts, USA; John Fiascone, Tufts Medical Center, Boston, Massachusetts, USA; Richard A Ehrenkranz, Yale University School of Medicine, New Haven, Connecticut, USA; T Michael O'Shea, Wake Forest University/Baptist Medical Center, Winston-Salem, North Carolina, USA; Stephen C Engelke, University Health Systems of Eastern Carolina, Greenville, North Carolina, USA; Carl Bose, The University of North Carolina at Chapel Hill, Chapel Hill, North Carolina, USA; Mariel Poortenga, Ed Beaumont, DeVos Children's Hospital, Grand Rapids, Michigan, USA; Nigel Paneth, Sparrow Hospital, Lansing, Michigan, USA; Michael D Schreiber, University of Chicago Hospital, Chicago, Illinois, USA; Daniel Batton, William Beaumont Hospital, Royal Oak, Michigan, USA; Greg Pavlov, Frontier Science and Technology Research Foundation, Amherst, New York, USA; and our project officer, Deborah Hirtz.
Funding This study was supported by a cooperative agreement with NINDS 5U01NS040069-04. CB was supported by the Thrasher Research Fund.
Competing interests None.
Ethics approval This study was conducted with the approval of the individual IRBs at participating institutions.
Provenance and peer review Not commissioned; externally peer reviewed.
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.