Main

Between 11 and 22% of prematurely born infants have evidence of intrauterine growth restriction (IUGR) (1,2), yet the influence of fetal growth restriction on respiratory morbidity remains controversial. Some studies have highlighted that small size at birth was associated with a reduction in the incidence of respiratory distress syndrome (RDS) (3,4), but other studies suggested no influence (5,6,7,8,9) or a significant increased risk (10,11). Subsequently, there have been three large studies (12,13,14) that demonstrated an association between being born small for gestational age (SGA) and an increased risk of respiratory morbidity. In a study of 2,487 infants born before 37 wk of gestation, SGA infants had a significantly greater risk of developing chronic lung disease or dying and had a longer hospital stay (12). In another series, which included 2,764 very low birth weight infants born between 24 and 31 wk of gestation, SGA infants had a 3.42-fold increased risk of bronchopulmonary dysplasia (BPD) and a 4.5-fold increased risk of death (13). In the third study, of 19,759 singleton, very low birth weight infants, born between 25 and 30 wk of gestation, IUGR was associated with increased risks of RDS and neonatal death (14). In the latter study (14), prenatal corticosteroid use was associated with decreased risks of both RDS and neonatal death. In none of the previous or the more recently reported studies (12,13,14), however, were infants routinely exposed to postnatal surfactant, and in the majority of studies, antenatal steroids were not routinely used. The aim of this study was to determine whether among infants born very prematurely, that is <29 wk of gestational age, who had been routinely exposed to both antenatal steroids and postnatal surfactant, neonatal and infant respiratory morbidity was increased in those who were born SGA.

Results

Of 797 infants recruited to the United Kingdom Oscillation Study (UKOS), 22% (174/797) were SGA infants and 205 (26%) infants died before hospital discharge. A comparison of maternal factors by SGA status showed significant positive associations between SGA and whether the mother had preexisting hypertension, pregnancy-induced hypertension or thyroid disease, prolonged interval between membrane rupture and delivery, proven chorioamnionitis, use of tocolytic drugs, and delivery by cesarean section ( Table 1 ). The SGA infants as compared with those who were not SGA were born at a later gestational age (P < 0.001). The proportion of SGA infants increased from 0% at 23 wk of gestation to 28% at 28 wk of gestation (Supplementary Table S1 online). None of the maternal factors showed significant associations with neonatal outcomes and so were not included in later multifactorial models in accordance with the a priori analysis plan.

Table 1 Maternal and infant characteristics by SGA status
Full size table

SGA was associated with significantly higher rates of supplementary oxygen dependency at 36 wk postmenstrual age (PMA) (71 vs. 53%, P < 0.001) and at hospital discharge (30 vs. 19%, P = 0.008), pulmonary hemorrhage (18 vs. 11%, P = 0.016), and mortality (34 vs. 23%, P = 0.004). SGA infants stayed in the hospital significantly longer than those who were not SGA (median of 107 vs. 88 d, P < 0.001) ( Tables 2 and 3 ). Those effects remained significant and were greater in size after adjusting for sex and gestational age ( Table 3 ). Regression analysis using birth weight z-score as the exposure variable showed a very similar picture but with stronger effects of birth weight after adjustment for gestational age and sex ( Table 3 ).

Table 2 Neonatal outcome by SGA status
Full size table
Table 3 The relationship between neonatal outcomes and being small for gestational age, adjusted for birth factors
Full size table

Twenty-four month follow-up respiratory data from questionnaires completed in the prespecified window (i.e., when the children were between 22 and 28 mo of age) were available for 367 infants. Children with follow-up data tended to be sicker than those who were lost to follow-up (Supplementary Table S2 online), but the differences were not statistically significantly different. Overall, 50% of parents reported that their children had had cough and 58% had received chest medicines ( Table 4 ). A greater proportion of SGA infants required hospital respiratory readmissions (30 vs. 19%, P = 0.035), but there were no significant associations of SGA and follow-up respiratory outcomes after adjustments ( Table 5 ). In addition, there was no significant correlation between a diagnosis of BPD and longer-term respiratory outcomes. Birth weight z-score was negatively associated with respiratory hospital admissions and use of chest medicines. Such associations remained significant and became more marked after adjusting for sex and gestational age ( Table 5 ). After further adjustment for oxygen dependency at 36 wk PMA, the relationships among birth weight z-score, respiratory hospital admission, and use of chest medicines remained significant. Final adjustment for sex, gestational age, oxygen dependency at 36 wk PMA, living with siblings younger than 5 y of age, having a twin/triplet, and living with a mother or father who smoked showed that five adverse respiratory outcomes (respiratory hospital admission, cough, use of any chest medicine, use of bronchodilators or inhaled steroids, and use of antibiotics) at follow-up were negatively and significantly associated with birth weight z-score.

Table 4 Respiratory outcomes at 24-mo follow-up by SGA status
Full size table
Table 5 Relationships between being small for gestational age and respiratory outcomes at 24 mo
Full size table

We performed further analyses using oxygen dependency at 28 d rather than oxygen dependency at 36 wk PMA to see if this affected the findings; the results were not appreciably different and so are not reported here.

Discussion

We have demonstrated that SGA, very prematurely born infants, despite routine exposure to antenatal steroids and postnatal surfactant, had increased respiratory morbidity in the neonatal period and at follow-up as compared with those who were not SGA. The associations remained significant after adjusting for infant factors and possible risk factors for chronic respiratory morbidity. As compared with using a cutoff to define abnormal size at birth, we showed even stronger relationships between adverse respiratory outcomes and birth weight z-scores, which would indicate that poor growth in utero adversely affects neonatal and infant respiratory outcomes. It has been suggested that neonatal growth standards based on healthy populations could improve the identification of very preterm neonates as SGA and at risk of intraventricular hemorrhage (1). In this study, although we found a number of maternal diseases to be associated with being SGA, none of those factors showed significant associations with neonatal outcomes and hence were not included in later models. Our results are in line with those of another study (15), which showed that among a very preterm cohort stratifying for pregnancy complications yielded similar risk patterns; that is, BPD declined continuously with increasing birth weight. Our finding of stronger relationships using birth weight z-score rather than the binary outcome, above or below the 10th percentile, suggests that the 10th percentile is not a true threshold and that growth should be analyzed as a continuum. A threshold, however, is useful clinically and so it is recommended that analyses include both the binary and continuous outcomes to provide a complete picture.

The mortality rate in the UKOS overall was 26%. Infants born at 25 wk plus 6 d or less had a mortality rate of 42%, which is comparable with the 39% mortality rate of infants in the EPICURE study, who were all born at 25 wk of gestation or less but in 1995 (16). In a prospective cohort study in a geographically defined population in the UK during 2000–2005, the survival rate to discharge of infants born at <25 + 6 wk was 47% (17). By contrast to previous studies, the majority (>90% of infants) were routinely exposed to antenatal steroids and postnatal surfactant. In one recently reported study (2) among infants with a birth weight <1,000 g and exposed to antenatal steroids and postnatal surfactant, those with SGA had a lower incidence of RDS and BPD. Antenatal steroid use, however, was <80%; it is not clear how many infants received surfactant, and infants up to 34 wk were included in the analysis. Our results demonstrate that neither antenatal steroid nor postnatal surfactant exposure mitigated the adverse influence of IUGR on neonatal and infant respiratory morbidity. Those findings are in accordance with the results of randomized controlled trials demonstrating that neither treatment reduces BPD (18,19).

We did not find a significant correlation between a diagnosis of BPD and longer-term, adverse respiratory outcomes. BPD, in this study, was diagnosed in infants who were receiving supplementary oxygen beyond 36 wk postmenstrual age. No oxygen reduction test was undertaken (20). Such a definition has previously been shown to correlate poorly with longer-term respiratory outcomes (21). Diagnosing infants as having BPD because they are supplementary oxygen dependent at 28 d and then assessing BPD severity at 36 wk PMA (22) has been shown to correlate much better with longer-term outcomes (23).

The lack of effect of antenatal glucocorticoids on BPD development has been ascribed to increased survival of very immature infants, but an alternative hypothesis is that antenatal steroids are the first hit taken by the fetal lung, which primes the lung for more ventilation-induced injury (24). Antenatal glucocorticoid treatment is given to women at risk of preterm delivery because it decreases the risk of death, RDS, and intraventricular hemorrhage (25). Preterm rupture of the membranes and histological chorioamnionitis are not contraindications to maternal glucocorticoids (26,27), and women with undiagnosed chorioamnionitis with preterm labor currently receive corticosteroids to enhance fetal lung maturation (26,27). In a subgroup of preterm infants, however, it has been suggested maternal glucorticoids may increase lung inflammation, which could promote the subsequent development of BPD (28).

There are possible limitations to our study; these include that 2-y follow-up data were not available for all survivors. The children who were followed up tended to be sicker as compared with those who were not, but the differences did not reach statistical significance. The choice of growth standards affects the cutoff to define an SGA infant. We used British charts from 1990 (29) as also used in many other studies, including the EPICURE study (16). More recent standards produced from the United States (30) are based on larger numbers, but their ethnic mix is different from that in the UK study and so they were not suitable. The infants included in the current study had been entered into a randomized trial examining the efficacy of high-frequency oscillation and conventional mechanical ventilation as modes of respiratory support for very prematurely born infants (31). We have, however, previously shown that both short-term (31) and 2-y follow-up (32) outcomes were similar in the two groups. Respiratory follow-up data at 22–28 mo was collected during a pediatrician consultation; therefore, it is possible that there may have been faulty parental recall. However, this would have similarly affected parents of children with or without SGA. In addition, the analysis of the results of the contemporaneously collected neonatal data yielded similar findings to the analysis of the follow-up data, that is, increased respiratory morbidity in the children who had been born SGA.

Our results demonstrated among infants born very prematurely that those who were SGA had greater longer-term respiratory morbidity, and these results are predictable from those obtained from studies on animal models. In a series of studies of respiratory development in growth-restricted ovine fetuses, Harding et al. (33) demonstrated that the lungs of growth-restricted fetuses were structurally immature and their tracheal and bronchial development was impaired. Other studies have demonstrated that normal lung development is critically dependent on the presence of appropriate oxygen tensions and nutrition (34), which are reduced in SGA fetuses. In both sheep (35) and rats (36), the reduced number of alveoli associated with IUGR at birth persisted into adulthood with no evidence of catch-up growth. Those data are in line with the finding that lung function at school age was reduced in prematurely born infants who had experienced IUGR as compared with gender- and age-matched children born at term (37).

In conclusion, we have demonstrated that being born SGA, despite routine exposure of antenatal steroids and postnatal surfactant, is associated with increased neonatal and infant respiratory morbidity, which was not explained by other risk factors. Our results emphasize the importance of investigating therapies to prevent fetal growth restriction and improving the postnatal management of affected infants to improve their long-term outcome.

Methods

Data were examined from infants recruited into the UKOS, a multicenter, randomized controlled trial conducted between August 1998 and January 2001 (31). Infants were randomized to either high-frequency oscillation or conventional mechanical ventilation within 1 h of birth. The study was approved by the South Thames Multicentre Research Ethics Committee and the local research ethics committee at each participating center. Parents gave informed written consent for the infant to take part in the randomized trial and the follow-up.

Data were available for the mother’s age, ethnicity, smoking during pregnancy, preexisting or pregnancy-induced hypertension requiring treatment, preexisting or pregnancy-induced diabetes requiring treatment, thyroid disease, use of antenatal corticosteroids taken before birth, whether an antepartum hemorrhage or rupture of membranes had occurred, use of tocolytic drugs, and mode of delivery. Infant factors obtained from the database included birth weight; sex; gestational age; multiple birth; Apgar scores at 1, 5, and 10 min; and the use of surfactant. Infants were diagnosed as having been born SGA if their birth weight was <10th percentile (i.e., birth weight z-score ≤−1.28). The birth weight z-score (i.e., SD score) was calculated for each infant using the British 1990 reference (29).

Neonatal outcomes were BPD, that is, supplementary oxygen dependency at 36 wk PMA and hospital discharge; an oxygen reduction test was not undertaken (20). A secondary outcome (see sensitivity analysis) was oxygen dependency at 28 d. Other outcomes were air leak, pulmonary hemorrhage, death, and use of postnatal corticosteroids. The infants were followed up to a corrected age of 24 mo, when pediatricians completed questionnaires by asking parents about their infant’s respiratory health and risk factors for respiratory morbidity. The pediatricians were asked to complete the questionnaire within a prespecified window, that is, when the children were between 22 and 28 mo of age. The questions asked were as follows: whether the child suffered from cough and/or wheeze in the past 12 mo; and whether the child had had hospital admission(s) for respiratory problems and/or had received any medications (bronchodilators, inhaled steroids, or antibiotics) for respiratory disorders.; Questions were also asked about risk factors for respiratory morbidity, that is whether the child lived with long-haired pets or in a rented accommodation; if there had been multiple births; whether the infant lived with a mother or father who smoked or had contact with smokers for >15 h per wk; or whether the infant lived with siblings who were <5 y of age. Follow-up respiratory outcomes were respiratory hospital admissions and use of chest medicines (bronchodilators, inhaled steroids, and/or antibiotics).

Analysis

The analyses were performed using both the dichotomized SGA (yes/no) and the birth weight z-score because the latter analysis gives greater statistical power (38). A statistical analysis plan was established a priori to avoid ad hoc modeling decisions: we first explored maternal factors as well as which birth factors affected intrauterine growth, and then, after adjusting for those, which neonatal outcomes were influenced by SGA. Analysis was then undertaken to determine any effects of SGA on respiratory morbidity after adjusting, in three stages, for all confounding maternal and infant factors, neonatal outcomes, and sociodemographic risk factors. The choice of factors for adjustment was guided by previous analyses of the UKOS data (39).

Unifactorial analyses were initially undertaken, and variables significant at P < 0.10 were then modeled in multifactorial regressions. Logistic regression or linear regression was used to examine associations for binary or continuous outcomes, respectively. Odds ratios for SGA/not SGA and the odds ratio equivalent to 1 SD decrease in birth weight z-score are reported. Skewed continuous data were log-transformed as appropriate. Robust SEs were used to take into account the clustering among multiple births (40). All analyses were conducted using Stata v 12.1 (2012; Statacorp, College Station, TX).

Statement of Financial Support

J.W.L. was supported by a National Institute for Health Research/Health Technology Assessment grant and the National Institute for Health Research Biomedical Research Centre at Guy’s and St Thomas’ Hospital/King’s College London. A.G. and N.M. are National Institute for Health Research Senior Investigators. The United Kingdom Oscillation Study was funded by the Medical Research Council, and N.M. receives partial funding from the Department of Health’s National Institute for Health Research Biomedical Research Centre’s funding scheme at University College London Hospitals/University College London.