Impact of Intensive Care Practices on Short-Term and Long-term Outcomes for Extremely Preterm Infants: Comparison Between the British Isles and France

METHODS

The EPICure study included all births from 20 to 25 completed weeks of gestation between March and December 1995 inclusive from all of the maternity units in the United Kingdom and the Republic of Ireland (N = 4004).⁹ The EPIPAGE study included all births between 22 and 32 weeks of gestation in 9 regions of France between January and December 1997 (N = 3673).¹⁰ In the EPIPAGE study, only 24 children were born alive before 23 weeks of gestation; 2 were transferred to a NICU, and none survived to discharge. Thus, for this study, the analysis was conducted only for children who were born between 23 and 25 completed weeks of gestation in both countries.

In the EPICure study, all births were recorded on a delivery room log. For infants who were admitted to NICU, gestational age was reevaluated: it was based on the last menstrual period or on the results of a routine ultrasound scan performed before 20 weeks of gestation. When both were available and the expected date of delivery derived from scans differed from that based on the last menstrual period by >14 days, the scan gestation was used. Seven additional children were included because pediatricians assessed gestational age as <26 weeks in the absence of dating information. Details of neonatal admissions were collected from NICU. We included all 1892 infants who were born between 23 and 25 weeks, 6 days of gestation, 786 of whom were transferred to NICU. At discharge from hospital, 312 children were alive. At 30 months (corrected for gestational age), detailed outcome was known for 286 (92%) children of the 312 children alive at discharge: 6 died between discharge and 30 months, and 280 were examined in a hospital outpatient clinic or at home.¹¹ At 6 years, 239 (78% survivors) underwent psychological testing.¹²

In the EPIPAGE study, gestational age was taken as the best obstetric estimate on the basis of the date of the last menstrual period and a routine early prenatal ultrasound. Infants were recruited in maternity units, and the initial data were collected in maternity and NICU during 1997. A complementary recruitment in NICU was performed for infants who were born before 27 weeks in 1998 in 7 of the 9 regions but only for those who were admitted to a NICU. There was no statistical difference between infants who were born in 1997 and 1998 and transferred to NICU as far as gestational age, Clinical Risk Index for Babies (CRIB) score,¹³ cerebral lesions, gender, and pregnancy complications were concerned. Thus, analysis of the mortality rate before transfer to NICU was performed only for births in 1997 (456 births 23–25 weeks of gestation), whereas analysis for NICU admissions included infants who were born in 1997 and 1998, 127 infants in 1997, and 106 in 1998 (Table 1). Of these 233 infants, 131 children were alive at discharge and 2 died between discharge and 5 years. Detailed questionnaires were completed for 113 children (87% of survivors) by the doctor who provided ongoing care at 2 years of age (uncorrected for prematurity). At 5 years, 74 children (57% of survivors) underwent psychological evaluation.

The following outcomes were compared between the 2 data sets: perinatal deaths (stillbirths [including deaths before and during labor] and deaths in the delivery room and NICU), brain lesions in the neonatal period, and neurodevelopmental outcome. Brain lesions were reported on the basis of the worst ultrasound scan in the neonatal period. Hemorrhages were classified as subependymal, intraventricular, or parenchymal. Ventriculomegaly was defined as a ventricular index >4 mm above the 97th percentile in the EPICure study and was clinically estimated in the EPIPAGE study. Severe cerebral lesions included the combination of ventriculomegaly or cystic leukomalacia or intraparenchymal hemorrhage. In both studies, the definition of cerebral palsy (CP) proposed by the European Surveillance of Cerebral Palsy Network was used,¹⁴ and the Mental Processing Composite (MPC) of the Kaufman Assessment Battery for Children (K-ABC)¹⁵ was measured. This is a global score, similar to IQ, which is age-standardized to a mean of 100; cognitive impairment is defined as scores <70 (−2 SD). MPC was evaluated for 201 children at 6 years (EPICure) and for 74 children at 5 years (EPIPAGE). In the EPICure study, 38 additional children were not evaluated by the K-ABC usually because of severe cognitive impairment (below the “floor” of the test) or severe associated disability; for these, the NEPSY¹⁶ or the Griffiths¹⁷ scales were used. In the EPIPAGE study, 2 children were not assessed because of motor and/or sensorineural disability. Only children who were evaluated by using the K-ABC were included in the comparisons of the 2 populations.

Population characteristics compared were gestational age, gender, birth weight, singleton/multiple pregnancy, pregnancy complications, and CRIB score.¹³ Preterm premature rupture of membranes (PPROM) was defined as birth that occurred at least 24 hours after membrane rupture; spontaneous labor as preterm labor with intact membranes; preeclampsia as hypertension appearing in pregnancy and with proteinuria; and hemorrhage as vaginal bleeding after 20 weeks of gestation, regardless of cause after exclusion of local hemorrhage from the genital tract in the EPICure study and as antepartum hemorrhage after the first trimester, including placental abruption, placenta previa, and other hemorrhagic syndromes, in the EPIPAGE study.

The following practices were compared: administration of antenatal steroids, tocolysis, mode of delivery, and the proportion of deaths after discontinuation of intensive care. The type of tocolytic agent used was not recorded in the EPICure study. Only intravenous betamimetics (84% of tocolysis) were considered in the EPIPAGE study. Death after discontinuation of intensive care was defined as a death that occurred after withholding and/or withdrawal of treatment in the EPIPAGE study. In the EPICure study, doctors were asked whether “intensive care was electively discontinued” after a “formal decision to withdraw care after appropriate discussion” and excluded situations in which death followed “an acute deterioration and unsuccessful resuscitation.”⁹

Comparisons between the 2 populations were performed for mortality rates, neonatal morbidity, and neurologic outcome in total and by week of gestation. The Pearson χ² statistic was used to compare proportions. Gestational age was treated as a continuous variable. For understanding differences in outcomes, comparisons of characteristics of infants; complications of pregnancy; discontinued treatment rates; and prenatal, perinatal, and postnatal practices were made between children who were transferred to a NICU. Logistic regression was used to study the relationships between outcome (mortality and morbidity) and the country of birth. Linear regression was used to compare continuous variables. We adjusted for the characteristics of the population and treatments that are known to be associated with each outcome. The level of statistical significance retained was P < .05. SAS software was used (SAS Institute, Inc, Cary, NC). Data from the EPICure study had all identifiers removed before be combined with the EPIPAGE data set. The EPIPAGE study was approved by the French Data Protection Authority (Commission Nationale de l'Informatique et des Libertés). The follow-up phase of the EPICure study was approved by the Trent Multicenter Research Ethics Committee, and signed informed consent was obtained from participating parents.

RESULTS

In the EPICure study, there were similar numbers of births at each gestational week; in contrast, the proportions of total births and births recorded as live born in the EPIPAGE study at 23 and 24 weeks was significantly lower than those recorded in the EPICure study (Table 2). Hence, a smaller proportion of infants in the EPIPAGE study were admitted for neonatal care at these earlier gestations, and the EPIPAGE survivor population comprised a greater proportion of infants who were born at 25 weeks. There was no difference in birth weight between the 2 studies after adjustment for gestational age among live births, infants who were admitted to NICU, and survivors. Birth weight was missing for 68% of stillbirths in the EPIPAGE study; therefore, we could not compare birth weight for total births.

The proportion of stillbirths and deaths in delivery suites was significantly higher in the EPIPAGE than in the EPICure study (Table 3). The reverse was true for deaths in NICU, overall and at each gestational week; consequently, the proportion of children who were alive at discharge did not differ significantly between the 2 populations after adjustment for gestational age.

Having adjusted for gestational age, there were no significant differences between the studies in the rate of hemorrhage or ventricular enlargement (Table 4). Cystic leukomalacia was more frequent in the EPIPAGE study compared with the EPICure study, but there were similar frequencies of severe brain lesions. Intensive care was withdrawn more frequently in the EPICure study than in the EPIPAGE study (55% vs 47%), but the difference was not significant. Among infants who died after discontinued treatment, the rate of severe brain lesion was lower in the EPICure study than in the EPIPAGE study (41% vs 67%; odds ratio after adjustment for gestational age: 0.4 [95% confidence interval: 0.2–0.7]). The risk for severe brain lesions was 17% in the EPICure study and 11% in the EPIPAGE study among survivors at discharge. In the EPICure study, 55 (19.6%) children were described with CP compared with 18 (15.9%) in the 1997–1998 EPIPAGE cohort. At 5 to 6 years of age, mean (SD) MPC scores were 87.9 (13.4) and 86.9 (16.1), respectively; and 9.5% and 13.5% of children, respectively, had an MPC score <70. In the EPICure study, when other estimates of IQ were included, mean scores fell to 82.2 (19.2).

Multiple birth was significantly more prevalent in the EPIPAGE study than in EPICure NICU admissions (Table 5). PPROM was more common in the EPIPAGE study, but the proportion of deliveries after spontaneous labor was lower and the combined frequency of PPROM, and spontaneous labor was similar in the 2 studies. Tocolytic agents were used more frequently in the EPIPAGE study, and cesarean section before labor was more common in the EPICure study. Antenatal steroids were administered with similar frequencies. CRIB scores and birth weight, after adjustment for gestation, did not differ between the 2 studies (Table 5). In the EPICure study, surfactant was administered more frequently.

The higher NICU mortality rate in the EPICure study remained essentially unchanged after adjustment for the population characteristics and practices (Table 6). There was no significant difference in severe brain lesions and overall cognition between the 2 studies. Although there was a higher risk for CP in the EPICure study compared with the EPIPAGE study, this difference was not statistically significant.

The rates of severe cerebral lesions during the neonatal period in children who were not assessed at 2 years compared with the children assessed were 13% vs 18% in the EPICure study and 24% vs 9% in the EPIPAGE study, but these differences were not significant. There were no statistically significant differences in the gestational age of children who were not assessed and children who were assessed at 2 years. The rates of neonatal severe cerebral lesions in children who were not assessed at 5 to 6 years compared with the children who were assessed were significantly higher in both studies: 28% vs 11% in the EPICure study and 18% vs 5% in the EPIPAGE study, respectively. Similar differences were observed for CP at 2 years: 45% of children who were not assessed at 5 to 6 years had CP at 2 years versus 9% in those who were assessed in the EPICure study and 27% vs 10% in the EPIPAGE study.

For the EPICure study, the denominator (N = 781 358) was calculated on the basis of the number of live births provided by the Office for National Statistics for England and Wales in 1995, the Northern Ireland Statistics and Research Agency in 1996, the General Register Office for Scotland in 1995, and the number of births for the Central Statistics Office for the Republic of Ireland in 1995. In the EPICure study, births were enrolled from March to December. Although the percentage of preterm births has been shown to be higher in winter in London, the discrepancy seems very low for births between 24 and 32 weeks¹⁸; therefore, the denominator was recalculated for 10 months on the basis of the assumption of a stable percentage of extremely preterm births throughout the year (N = 651 132). The number of live births for the EPIPAGE study (N = 274 324) was provided by the National Institute for Statistics and Economic Studies. The rates (including stillbirth and live births) per 1000 live births were 0.96 at 23 weeks, 0.98 at 24 weeks, and 0.97 at 25 weeks in the EPICure study compared with 0.50, 0.42, and 0.74 in the EPIPAGE study, respectively.

DISCUSSION

Despite clear differences in the immediate neonatal decision-making between the 2 studies, the outcome in terms of survival at discharge, CP, and cognitive impairment was not different between the EPICure and EPIPAGE studies. Compared with previous comparative studies in extreme preterm births,^7,19 the main strengths of this comparison are, first, to detail all types of neonatal mortality (deaths in delivery room, deaths in NICU, and withdrawing/withholding intensive care) and long-term outcomes; second, to have included a relatively large population of extremely preterm infants; and, last, to be able to take into account characteristics of the populations and treatments.

In both studies, births were identified in the delivery suite with only minimal differences in the way in which gestation was calculated. In the EPICure study gestational age was formally checked again against an algorithm after admission to NICU and modified in a small number of cases. Differences in classification might have occurred between terminations of pregnancy, stillbirths, and live births in the EPICure and/or EPIPAGE study. In the mid-1990s, stillbirths were registered after 23 weeks in the United Kingdom and only after 27 weeks in France; therefore, we might have anticipated a higher proportion of stillbirths at 24 to 25 weeks in the EPICure study than in the EPIPAGE study. The higher level of legal age for registering in France, however, may have led to greater numbers of infants’ being classified as dead when they might have been assigned as alive in the United Kingdom. Some terminations of pregnancy were included as stillbirths in the EPICure study, whereas, in France, some stillbirths after withholding of intensive care might have been classified as terminations of pregnancy. It seems likely that such misclassification was limited because we observed a lower rate of stillbirths among all births in the EPICure study compared with the EPIPAGE study. Although no estimation is available, we believe that it should not affect the conclusions of the study. Most obvious, the distribution of all births in the EPIPAGE study was toward higher gestational ages compared with the EPICure study. A higher use in France of tocolytic agents that are known to delay the delivery²⁰ may also be contributory. Furthermore, there was no difference in birth weight for gestation among live births in the EPIPAGE study and the EPICure study despite there being more multiple births in the EPIPAGE study, possibly as a result of a higher rate of multiple births after assisted reproduction in France.²¹

We estimated that the extremely preterm birth rate per 1000 births was close to 1 in the EPICure study, similar to values between 0.94 and 0.92 in Denmark,²² whereas the rate was between 0.40 and 0.50 at 23 and 24 weeks and approximately 0.75 at 25 weeks in the EPIPAGE study. It is interesting to speculate that infants who were born at 25 weeks in the EPIPAGE study might represent infants who were delivered at earlier gestational ages in the EPICure study as a result of less active attempts to prolong pregnancy. Furthermore, the distribution of gestational age in 2 other contemporary cohorts, in Belgium²³ and Norway,²⁴ was intermediate between those of the EPICure and EPIPAGE studies, and, in each study, the number of births rose between 23 through 25 weeks of gestation.

Socioeconomic status was not recorded in the same way in both studies and therefore not included. Although the impact of socioeconomic status should be limited in terms of CP,²⁵ it is an important moderator of cognitive outcome²⁶; however, extremely preterm births occur more frequently in deprived populations, and we have no reason to think that there would have been an important difference in the pattern of socioeconomic status between these births in the 2 countries.

The proportion of children who were not assessed at approximately 2 years was 10% in the EPICure study and 14% in the EPIPAGE study, and 36% and 44%, respectively, were not assessed by using K-ABC at 5 years. For children who were not assessed at 2 years, more had severe cerebral lesions during the neonatal period in the EPIPAGE study and a lower rate of severe cerebral lesions in the EPICure study compared with the children who were assessed, but these differences were not significant. At 5 to 6 years, for children who were not assessed, the frequency of cerebral lesions in the neonatal period and CP at 2 years were significantly higher in both studies, particularly in the EPICure study. Thus, the proportion with CP might be overestimated in the EPICure study and underestimated in the EPIPAGE study and the prevalence of cognitive impairment underestimated in both studies but particularly in the EPICure study, because we used only MPC scores and ignored those with scores below the test floor; hence, developmental attainment is likely to be slightly overestimated and the rate of impairment is likely to be underestimated. In both studies, approximately 1% of children had severe impairment (score <55).

Population data on extremely preterm births have been rare until recently. To our knowledge, the EPICure and EPIPAGE studies are the only recent gestation- and population-based cohorts of extremely preterm infants with long-term follow-up. In Denmark, a national study followed extremely preterm children who were born in 1994–1995 up to 5 years, but the number of survivors who were born at 23 to 25 weeks was low (N = 54).²⁷ This study seems to indicate very similar death rates in maternity wards compared with the EPIPAGE study at 23 to 24 weeks.²² The Norwegian population-based cohort included births in 1999 and 2000 between 22 and 27 weeks or a birth weight of 500 to 999 g.²⁴ Among infants who were born at 23 to 25 weeks, they observed a higher NICU admission rate among all births than either the EPIPAGE or EPICure study and a higher survival rate at discharge. This is in contrast to the findings of EPIBEL,²³ a population-based study in Belgium conducted in 1999–2000 in which a higher proportion of live-born infants at 23 to 25 weeks’ gestation were admitted for intensive care than in our 2 studies, but the death rate during intensive care lies between the mortality rates in the EPIPAGE and EPICure studies.

The only previous study that compared childhood outcome between 2 population-based cohorts of extremely preterm children found a rate of disabling CP of 17.2% in New Jersey in the United States among children who were born in 1984–1987 (N = 146) versus 3.4% in the Netherlands among children who were born in 1983 (N = 143).⁷ We did not expect such a large difference in our study, because the differences in practice between the 2 cohorts is not as extreme as those in the previous study.

In a European study conducted in 1996–997 to describe resuscitation practices in NICU, although most neonatologists had been involved in end-of-life limitation of treatments, decision-making experiences varied between countries. For example, 73% of French neonatologists compared with only 4% of British neonatologists declared having administered drugs with the purpose of ending life in NICU at least once.²⁸ Conversely, a greater proportion of British than French neonatologists declared having withheld intensive care (91% vs 73%) and withdrawn mechanical ventilation (89% vs 76%). Our study is a rare opportunity to compare between countries the actual practice in terms of withholding and withdrawal of treatment. Our results suggest that withholding intensive care was more frequent in France than in the British Isles, whereas the rate of treatment withdrawal was lower. In keeping with this difference in strategies, cesarean section was more often used in the EPICure study, and tocolysis was more common in France.²⁹

The rate of severe brain lesions was 24% in both studies among infants who were admitted to NICU. Despite a relatively high number of missing data and a low number of infants by week of gestational age, practices of treatment withdrawal seemed less correlated to severe brain lesions in the EPICure study than in the EPIPAGE study. Consequently, among survivors, the rate of severe brain lesions was higher in the EPICure study (17%) than in the EPIPAGE study (11%). Furthermore, because severe brain lesion might better predict CP than severe cognitive delay,³⁰ it may explain the slightly higher rate of CP but not of MPC score <70 in the EPICure study.

Our study confirms a high risk for mortality and long-term impairment in children who are born at extremely low gestational ages. Significant gaps in our understanding of the effects of clinical management have been identified,³¹ and ethical practice varies between countries and individual physicians.²⁸ Our failure to find one strategy that was superior to the other leaves room for the expression of different collective and individual values between families and doctors and in the ethical debate.

CONCLUSIONS

The comparison of the “wait until certain” strategy and the “statistical prognostic” strategy have been highlighted: the first approach results in a lower mortality rate, whereas a “statistical” one results in less impairment.⁷ In our study, the approach in the EPICure population seems closer to the “individual” strategy,⁸ whereas in the EPIPAGE population the strategy seems closer to the “statistical” one. Our results suggest that these 2 strategies do not produce significant differences in terms of short-term or of long-term outcomes. We look forward to long-term outcomes from other ongoing studies to explore this finding further.