Chronic Oxygen Dependency in Infants Born at Less Than 32 Weeks' Gestation: Incidence and Risk Factors
Objective. To assess incidence and clinical risk factors of chronic oxygen dependency (COD) among survivors who were born at or before 31 weeks' gestation.
Methods. This prospective, multicenter study enrolled 802 infants who were born at or before 31 weeks' gestation and admitted to 8 level III neonatal intensive care units in northern and eastern France from January 1 through December 31, 1997. Need for oxygen to maintain oxygen saturation between 92% and 96% was assessed at 28 days of life and at 36 and 42 weeks' postconceptional age (PCA). Stepwise logistic regression analysis was used to identify the incidence of COD and the risk factors related to its occurrence.
Results. The mortality rate was 14%. Antenatal corticotherapy was administered to 51% of patients, surfactant therapy to 76% of the ventilated patients, and high-frequency oscillatory ventilation at day 1 to 32%. At 28 days and 36 and 42 weeks' PCA, respectively, 25%, 15%, and 6% of survivors had COD. After adjustment for intercenter variations, we identified the significant risk factors for COD at these dates: a low gestational age, a high score on the Clinical Risk Index for Infants, intrauterine growth restriction, and surfactant treatment.
Conclusion. COD incidence was high at 28 days of life but decreased dramatically by 42 weeks' PCA. This study confirmed previously reported risk factors and underlined the importance of intrauterine growth restriction and the Clinical Risk Index for Infants as significant risk factors.
Recent advances in perinatal care have improved the survival rates of very preterm infants. New therapies introduced during the past decade have reduced mortality, but their effectiveness in reducing the occurrence of bronchopulmonary dysplasia (BPD) remains controversial.1 BPD incidence varies in published studies from 4.6% to 72%.2,,3 This broad range may be explained by the heterogeneity of the studied populations, of management practices, or of disease definitions. The initial description of chronic oxygen dependency (COD) by Northway et al4 and later by Bancalari et al5 defined it according to oxygen requirements at 28 days of age. As newborns of lower gestational age (GA) survived, Shennan et al6 proposed a more open-ended approach to COD: chronic lung disease (CLD). CLD, defined as oxygen dependency at 36 weeks' postconceptional age (PCA), seemed to be a better marker of abnormal pulmonary outcome. Most studies about COD incidence have been either retrospective or performed in the presurfactant era.
We conducted this prospective, multicenter study to learn more about COD today. Its purpose was to estimate the incidence and risk factors for COD, at various PCAs, in premature infants who are born at or before 31 weeks' gestation in the surfactant era.
Eight level III neonatal intensive care units in northern and eastern France collaborated in a prospective study from January 1 to December 31, 1997. They all are located near sea level. The number of neonatal intensive care cots ranges from 6 to 16 among centers. All infants who were born at 31 weeks' GA or younger and admitted to participating units were enrolled.
The following variables were studied: gender, GA (ascertained by fetal ultrasonography before 14 weeks' gestation or the date of the mother's last menstrual period), and birth weight. Antenatal history included such disorders as pregnancy-related hypertension or prolonged premature rupture of the membranes (for >24 hours), complete antenatal corticotherapy (2 doses at least 12 hours before delivery), mode of delivery (vaginal delivery, cesarean section before or during labor), and inborn status. The Clinical Risk Index for Infants (CRIB) score7 was calculated for each patient. Concerning outborn patients, data for CRIB score calculation were collected from outborn unit charts by the neonatal transfer team. For home deliveries, data were recorded continuously during transport. Respiratory data included need for surfactant therapy (Curosurf, Chiesi Pharmaceutical, Parma, Italy, or Exosurf, Burroughs Wellcome, Research Triangle Park, NC), modes of ventilation, and need for supplemental oxygen at 28 days of life and at 36 and 42 weeks' PCA. Oxygen therapy was defined as justified if needed to achieve an oxygen saturation between 92% and 96% during quiet sleep (Nellcor oximeter pulse N-200; Nellcor Inc, Haywood, CA). Neonatal infection was defined as clinical signs of illness associated with increased inflammatory markers and a positive culture of blood, urine, or tracheal samplings. Nosocomial infection was considered for clinical deterioration with a positive blood culture. We also recorded data about symptomatic treated patent ductus arteriosus (indomethacin or ligature), need for diuretic and corticosteroid therapy, intracerebral hemorrhage8 or leukomalacia, retinopathy of prematurity, and survival or death at discharge. The protocol did not impose any standard policies (eg, fluid intake, ventilation, indomethacin, corticosteroids) on the centers for treatment of these infants.
Intrauterine growth restriction (IUGR) was defined according to Lucas charts.9 Incidence of COD was stratified by GA and by birth weight. GA was coded as bands because of small numbers in the youngest GA. Centers were coded as categorical data and monitored with the partial method. The center with the lowest incidence of COD was used as the reference group. CRIB scores were coded in 3 bands (0–5, 6–10, and 11–20) because of a small number of patients in the band 16 to 20.
To assess the reliability of collected data, we listed different questions about the same item (eg, duration of oxygen supplementation and PCA at oxygen weaning). Moreover, in each center, an extra observer checked the validity of collected data for 5% of included infants.
The association between recorded variables and COD was tested by univariate analysis. Variables were selected for logistic regression when they met the following criteria recommended by Greenland10: 1) P < .20, and 2) they had been observed previously to have effects on COD incidence in premature infants. Statistical significance was determined for categorical data by χ2 test or Fischer exact test when appropriate. Multivariate analysis was conducted by stepwise logistic regression (BMDP Software Inc, Los Angeles, CA).
This 12-month prospective, multicenter study enrolled 802 infants. None of them had any congenital malformations or chromosomal abnormalities. No one was subsequently excluded for missing data. Concerning validity of collected data, no major error was observed among the 5% controlled files.
Distribution according to GA was as follows: 54 (7%) born at 23 to 25 weeks, 136 (17%) at 26 to 27 weeks, 259 (32%) at 28 to 29 weeks, and 353 (44%) at 30 to 31 weeks. Univariate analysis did not show any significant difference in the distribution of GA among centers.
Table 1 shows demographic and obstetric characteristics of infants. Rate of antenatal steroid therapy was low (51%), with a range from 33% to 77% among the 8 centers. Sixty-eight percent of the newborns were inborn. Eight deliveries occurred at home. Rate of deliveries were 11%, 20%, and 68% in level 1, 2, and 3 units, respectively. Table 2 illustrates the initial severity of illness, assessed according to need for ventilation on first day of life, surfactant therapy in the first 24 hours of life, and CRIB score.
Mortality rate in the overall cohort was 14%. Figure 1 summarizes incidence of death and COD according to GA. Of 706 infants who survived to 28 days, 175 (25%) developed BPD, defined as requiring oxygen at that point. At 36 weeks' PCA, 106 (15%) of 695 survivors required oxygen; 18 of these were not oxygen dependent at 28 days of life. At 42 weeks' PCA, 41 infants (6%) were oxygen dependent; 39 were discharged from the hospital with oxygen.
The incidence of COD at the different endpoints varied widely among the 8 centers (Table 3). Incidence at 28 days of life and 36 weeks' PCA, respectively, ranged from 9% and 2.5% for center A to 46% and 32% for center F. The centers did not differ significantly as to GA, pregnancy-related disorders, prenatal steroids, mode of delivery, or surfactant therapy.
The univariate analysis found the following risk factors for COD: center, low GA, IUGR, high CRIB score, surfactant treatment, treated patent ductus arteriosus, and grades 3 and 4 intraventricular hemorrhage (Table 3). Birth weight was significantly lower in every COD group: 966 ± 250 g versus 1301 ± 312, 915 ± 217 versus 1278 ± 316, and 933 ± 282 versus 1218 ± 325 at 28 days of life and 36 and 42 weeks' PCA, respectively.
Table 4 shows the adjusted odds ratios obtained from the multiple regression analysis. At 36 weeks' PCA, the following factors were significantly predictive of COD: centers, low GA, IUGR, high CRIB score, and surfactant replacement therapy. As a high colinearity was observed among GA, IUGR, and birth weight, this last variable was not introduced in the multivariate model.
To our knowledge, this is the first prospective, multicenter study to describe incidence and risk factors of COD in a cohort of very premature survivors in the surfactant era. The mortality rate of 14% is consistent with other published data.11,,12
COD incidence varies widely in the published data.2,,313–15 Some of this difference can be explained by the various combinations of numerators and denominators used to compute this incidence. The denominator may be based on figures from the hospital,13 multiple center,14 or the general population.15 Unlike population-based data, our denominator does not include infants who were recorded as liveborn but who died before reaching the nursery. In the population-based study by Darlow et al,14 these newborns represented 12.5% of the very low birth weight population. As we expressed COD incidence among survivors, this difference does not affect the interpretation of our results.
The standard for oxygen was the same in all 8 centers: an oxygen saturation between 92% and 96% during quiet sleep, with the same precalibrated oximeter. Differences in management (eg, fluid intake, ventilation settings, postnatal corticosteroids) might explain variations in the incidence of COD among centers. In an 8-center survey about BPD, Avery et al16 pointed out a similar degree of variability in nursery policies among centers, especially for respiratory care.
After multiple regression analysis that adjusted for center, other risk factors remained important. The most significant was GA.
Few studies have focused on populations defined by GA; they most often reported data stratified by birth weight. Fenton et al15reported BPD rates in a population-based study of 2 regions, Trent and British Columbia, in 1990 and 1991. BPD incidence in survivors of 32 weeks or less or of 750 to 1500 g was 16.3% and 24.5%, respectively. The CLD rate was 6.4% and 18.5%.
Lenoir et al17 described the short- and mid-term outcome of 1157 neonates with respiratory distress syndrome (RDS), born during a 15-month period in 1990 to 1991. BPD occurred in 72% of the infants who were born before 26 weeks' GA, 50% at 26 to 27 weeks' GA, 28% at 28 to 29 weeks' GA, and 13% at 30 to 31 weeks' GA.
In a recent 1-year prospective multicenter study, Rubaltelli et al18 enrolled 1048 infants who were born before 32 weeks' gestation. The infants were treated with surfactant and managed with oxygen, continuous positive airway pressure, and conventional ventilation. Incidence of BPD was low in every GA group: 10% for infants born at 23 to 25 weeks, 13% for those born at 26 to 27 weeks, 9% for those born at 28 to 29 weeks, and 2% for those born at 30 to 31 weeks. The author explained this lower incidence of BPD by a high rate of antenatal steroid treatment (71% of infants with respiratory disorders vs 51% in our study). This difference also might be explained by mortality: mortality data were not provided, and the BPD incidence was calculated with the entire population, rather than simply survivors, as the denominator.
It seems to us that GA is a reliable predictor of COD. It is related to the morphologic and functional immaturity of the respiratory system, and this immaturity is one of the major determinants of COD in very preterm infants.
We also found the CRIB score to be related to COD. As others have reported,7,,19 CRIB was correlated with the risk of death in our population, a finding indicating that the score was valid. Our results show that CRIB also is an accurate index of initial neonatal risk of COD at 28 days of life and at 36 and 42 weeks' PCA.
CRIB initially was defined as “a tool for assessing initial neonatal risk and comparing performance of neonatal intensive care units” for infants with a birth weight of 1500 g or less or a GA younger than 31 weeks.19 In 1996, Fowlie et al20 reported that this scoring system also could contribute to more comprehensive predictive models of death and nosocomial bacteremia. More recently, these authors21 showed that throughout the first week of life, CRIB score was correlated with the risk of death, disability, and prolonged treatment with oxygen at 28 days and 36 weeks' PCA. Only 1% of the high-risk neonates included in Fowlie's study, however, were treated with surfactant replacement. Accordingly, it has been suggested that “this relationship may not be present in the surfactant era” and that “surfactant use has aged the CRIB score faster than other scores.”22 As surfactant replacement therapy has a minimal impact on BPD,23 we do not think that this conclusion necessarily follows. It certainly is not confirmed by the results of our multicenter prospective study in which 76% of the ventilated infants received surfactant.
Moreover, our results are logically consistent, because 4 of the 6 variables considered in CRIB are confirmed BPD risk factors (low GA, birth weight, and minimum and maximum inspired fraction of oxygen in the first 12 hours of life, which can be considered as markers of severe RDS). Our findings thus demonstrate that CRIB is not yet at risk of being “washed away by clinical advance.”22 It remains an early and reliable score that helps to predict the occurrence and severity of COD in infants who are born at less than 32 weeks' gestation.
IUGR, present in 25% of our population, also was a risk factor for COD. Distribution of IUGR was homogeneous among the 8 centers. Conflicting results in the early 1990s reported that IUGR was a protective factor24 or a risk factor for COD.17 More recently, Todd et al,13 using multivariate analysis, also reported a strong correlation between IUGR and COD at 36 weeks' PCA in a cohort of 1454 infants who were born at 24 to 32 weeks' gestation. These clinical results are supported by experimental data about malnutrition-induced adverse pulmonary effects during the perinatal period.25,,26
Antenatal steroid therapy was not correlated with a lower incidence of COD in univariate analysis. We believe that the beneficial effects of antenatal steroid therapy on COD would have been demonstrated if the rate of treatment has been clearly higher.
Incidence of surfactant therapy was 57% in the overall population but 76% of ventilated infants. This indicates an early therapeutic approach to RDS management and not a rescue strategy, which would have selected the most severe patients. The correlation between surfactant replacement and COD in our population suggests that surfactant therapy could not prevent RDS from developing into BPD. Two meta-analyses1,,27 failed to find that surfactant treatment, regardless of its type or mode, has any direct preventive effect on BPD that seems to be determined multifactorially and is not linked to RDS alone.
It is unexpected that although inflammation is thought to be associated closely with CLD,28,,29 none of the relevant factors recorded in our study (premature rupture of membranes, neonatal or nosocomial sepsis, corticosteroid therapy) was found to be a significant risk factor. Because no data were available concerning chorioamnionitis, we could not test its relation with CLD.
Although the introduction of surfactant therapy has improved the survival rates of very preterm infants, COD incidence still remains high at 28 days of life. It decreases dramatically after 42 weeks' PCA, however. Its incidence is highest for the lowest GA survivors, a finding that emphasizes the importance of immaturity as the main pathogenic factor. This study, while confirming the previously reported risk factors and the variability of incidence among centers, underlines the importance of IUGR and initial illness severity as measured by CRIB score as strongly significant risk factors.
We thank Serono Company, which provided pulse oximeters to each center for this study.
Members of Collaborative Northern and Eastern France Study Group of Neonatal Intensive Care Units: B. Romeo, G. Krim (Amiens); J.Y. Pauchard, A. Menget (Besancon); B. Geneste, D. Semama, J.-B. Gouyon (Dijon); P. Truffert, A. Liska, P. Lequien (Lille); F. Cneude, P. Vittu (Lille St. Antoine); I Hamon, J.M. Hascoet (Nancy); L. Egreteau, P. Morville (Reims); J. Matis, U. Siméoni, J. Messer (Strasbourg).
- Received March 23, 2000.
- Accepted March 28, 2001.
Reprint requests to (L.E.) American Memorial Hospital, 49 Rue Cognacq Jay, 51092 Reims Cedex, France. E-mail:
- BPD =
- bronchopulmonary dysplasia •
- COD =
- chronic oxygen dependency •
- GA =
- gestational age •
- CLD =
- chronic lung disease •
- PCA =
- postconceptional age •
- CRIB =
- critical risk index for infants •
- IUGR =
- intrauterine growth restriction •
- RDS =
- respiratory distress syndrome
- Halliday HL
- Cooke RW
- Shennan AT,
- Dunn MS,
- Ohlsson A,
- Lennox K
- ↵Truffert P, Rakza T, Liska A, Lequien P. Prématuritéextrême: les données de la littérature permettent elles de définir un pronostic? In: Treisser A, ed. 27e Journées Nationales de la Société Française de Médecine Périnatale. Paris, France: Arnette Blackwell; 1997:97–111
- Avery ME,
- Tooley WH,
- Keller JB,
- et al.
- Lenoir S,
- Grandjean H,
- Leloup M,
- Claris O,
- Hascoet JM,
- le Groupe Osiris-France
- Fowlie PW,
- Gould CR,
- Parry GJ,
- Philips G,
- Tarnow-Mordi WO,
- Strang D
- Mc Colley SA
- Carlisle SC,
- Moxley MA,
- Longmore WJ
- ↵Soll RF, McQueen MC. Respiratory distress syndrome. In: Sinclair JC, Bracken MB, eds. Effective Care of the Newborn Infant. New York, NY: Oxford University Press; 1992:325–358
- ↵Groneck P, Gote-Speer B, Oppermann M, Eiffert H, Speer CP. Association of pulmonary inflammation and increased microvascular permeability during the development of bronchopulmonary dysplasia: a sequential analysis of inflammatory mediators in respiratory fluid of high risk preterm infants. Pediatrics. 1994;93:712–718
- Copyright © 2001 American Academy of Pediatrics