Intensity of perinatal care, extreme prematurity and sensorimotor outcome at 2 years corrected age: evidence from the EPIPAGE-2 cohort study

This study examined the impact of the intensity of perinatal care on outcomes at 2 years of age for extremely preterm babies born in French level 3 hospitals in 2011. Among survivors, there was no evidence of a difference in the adverse primary outcome of sensorimotor disability. There were, however, important differences in the beneficial outcome of survival without sensorimotor morbidity for babies born in hospitals of high-intensity perinatal care compared with those born in hospitals of low intensity among the populations of foetuses alive at admission to hospital and in live births, but no differences when considering babies admitted to neonatal intensive care. Results remained consistent in sensitivity analyses restricted to 24 to 26 weeks gestation.

Strengths of our study include the fact it is a large, prospectively collected national cohort with standardised definitions of outcomes following international recommendations. Data collection was comprehensive [2, 4] with an overall follow-up rate of 83.3% for the sensorimotor outcome for the children included in this study. This is better than or similar to other population cohorts at a similar age [26, 27]. We used multiple imputation [24] to help mitigate this issue, but ideally, follow-up would be higher as the impact of missing data is difficult to ascertain: it is not possible to know whether children lost to follow-up are more or less likely to be impaired [28].

The measure of perinatal care intensity we used makes no assumptions about appropriate or ‘best’ practices. Instead, we assumed ‘active’ management would use any and all appropriate available techniques to ensure the baby survives. In the scenario we examined, this translates to hospitals providing treatment such that babies are subsequently admitted into neonatal intensive care. Following maternal admission to hospital, foetal deaths prior to delivery (stillbirths) or in the delivery room were therefore interpreted as a sign of inactive management. We restricted our measure of activity to include only births at 24 to 25 weeks gestational age as these are the ages when there was the greatest variability in practice in France in 2011: babies born before 24 weeks gestation almost uniformly did not receive active care, whereas at 26 weeks gestation, the inverse was true [15, 17].

That there were very few admissions at 24 to 25 weeks gestation in some hospitals, meaning that the indicator was partly based on small numbers, could be a weakness. We addressed the possible imprecision of the intensity ratio for these hospitals by weighting the mean around which the groups were constructed. A counter point is that we included almost all births at 22–26 weeks gestation occurring in level 3 hospitals in France. Only three of 65 units did not admit any women who would subsequently deliver at 24–25 weeks gestation—meaning we were unable to calculate perinatal intensity ratios. However, between them, there were only nine deliveries at 22–26 weeks gestation in these three hospitals.

The overall size of the cohort may also be perceived as a problem. In fact, the numbers in EPIPAGE-2 do not differ greatly from other cohorts. The first EPICure in 1995 included 4004 births under 26 weeks gestation with 308 survivors to 2.5 years of age in Great Britain and Ireland over 10 months [29, 30]; the second EPICure in 2006 included 2326 births below 27 weeks gestation with 1031 survivors to 3 years of age in England during 1 year [26, 31]; EXPRESS reported 774 births below 27 weeks gestation with 491 survivors at 2 years of age in Sweden during a 4-year period [27, 32]; and EPIPAGE-2 reported 2205 births below 27 weeks gestation with 545 survivors at 2 years in France in 8 months of 2011 [2, 4]. Compared to England in 2006, there was a higher proportion of babies born in a level 3 unit in France in 2011 (79.0% [2] compared to 56.4% in the EPICure 2 study [33]). For comparison, there are observational studies with larger numbers of subjects, but these tend to have been conducted by research networks without defined geographical coverage over a number of years.

We also note that the distribution of gestational age at delivery appears to differ between the intensity groups. This may be due to the fact that women admitted to high-intensity hospitals with threatened preterm labour are more likely than those admitted to low-intensity hospitals to have treatments aimed at prolonging the pregnancy—and of delivering a liveborn baby in good condition such that the baby is subsequently admitted to neonatal intensive care. This hypothesis is supported by the associations seen with active management treatments (e.g. tocolysis and antenatal steroids) which were more frequently used in women who delivered in high-intensity hospitals. However, there was no difference in overall length of maternal admission to hospital prior to delivery. An alternative hypothesis is, therefore, that the skewed distribution is linked to the construction of the indicator (as it uses babies born at both 24 and 25 weeks gestation). This does not explain differences in the numbers of babies born at 22, 23 or 26 weeks, though, so there may be a different reason or combination of reasons for the differences seen. We accounted for this potential bias by including gestational age in all adjusted analyses. Similarly, while there were no differences in the distribution of multiple pregnancies between intensity groups, we felt it was important to account for this in our models as there was potential for this to influence the ratios used to classify hospitals.

Another criticism might be that the indicator used as the exposure is contained within one of the outcomes (survival without sensorimotor disability), and thus, a positive association is to be expected. While this may be partly true—an ‘active’ hospital is one that has improved ‘perinatal’ survival (defined here as survival prior to admission to neonatal intensive care)—our definition of intensity only included foetuses subsequently delivered at 24 to 25 weeks gestation whereas the morbidity-free survival outcome also included deliveries occurring at 22, 23 and 26 weeks gestation. Consequently, the results were not certain, particularly given the high degree of variability in the numbers of babies born in each hospital at each week of gestation. Furthermore, intensity was measured at the hospital level but assigned to individual subjects. This is the same method and the same population as previously used by investigators in the EXPRESS study [14]. In that study, they used rates of obstetric treatments such as administration of antenatal steroids, tocolysis and caesarean section—all of which are strongly linked to improved survival prior to neonatal unit admission [34]—as contributory factors towards their ‘regional activity score’. The EXPRESS investigators additionally performed analyses in the live-born population, as did we. However, by investigating the population who were admitted to neonatal intensive care, our results indicate that survival differences only occur prior to intensive care unit admission without subsequent impact on morbidity status. In turn, this suggests that it is not possible to identify in the delivery room those babies who will have a better—or worse—long-term outcome. Finally, we note that in our study, survival was but a secondary outcome: there is no obvious link between our measure of perinatal intensity and the primary outcome of sensorimotor disability at 2 years of age.

Previous studies have investigated the relationship between the volume of admissions and outcome [33] or between measures of activity based on individual treatments such as administration of antenatal steroids, tocolysis, magnesium sulphate, caesarean section or postnatal attempts at resuscitation (for example, intubation or the administration of surfactant) [9, 14, 35, 36]. Such indicators are dependent upon the presenting circumstances of each patient and on local management protocols. They therefore may not be good indicators of differences in the intensity of active perinatal management between hospitals or regions due to differences in population make-up or varying protocols. For example, tocolysis may be of use in women who have pPROM but is not indicated in the absence of active labour. Antenatal steroids are recommended for threatened preterm delivery [37], but there may be insufficient time prior to delivery for administration. Both of these indicators consequently vary according to the presenting population. There are also differences postnatally. Some hospitals routinely intubate babies born at extremely preterm gestations. Others seek to avoid intubation altogether during the first 72 h of life, preferring to utilise techniques such as continuous positive airway pressure along with less invasive surfactant administration via a narrow-bore endotracheal catheter during spontaneous respiration [38]. The validity of our indicator was additionally highlighted by showing strong relationships with such perinatal management factors.

Although there was only one survivor born below 24 weeks GA, attitudes in France towards extreme preterm birth are not dissimilar to some other European countries [7]. Recent studies from the Netherlands demonstrate most variation in attitudes among obstetricians and neonatologists occurring at 24 and 25 weeks gestation [39], with significant mortality at these gestations [40]. A study comparing five different European regions showed the largest differences in survival for babies born at 500 g or more at 24 weeks gestation; results for births at 23 and 25 weeks and below 500 g were much more similar [41]. More importantly, perhaps, there have been changes over time in attitudes to extremely preterm birth across Europe, and hospitals that have actively changed their policies seem to have experienced larger survival gains [42]. We also note that our findings are consistent with those from studies in Sweden [14] and the USA [9].

This study described a novel measure of the intensity of perinatal care comprised of common vital measures which could be applied broadly elsewhere. In this population of women delivering at extremely preterm gestations, we found no evidence of increased sensorimotor impairment at 2 years of age for babies born in hospitals with a higher intensity of active perinatal management. We demonstrated an important improvement in survival without sensorimotor disability in populations of foetuses alive at maternal admission to hospital and in live births. Together, these results indicate that hospitals with higher perinatal intensity levels improve survival without increasing sensorimotor morbidity at 2 years of age. They support findings from other populations that outcomes are better following higher intensity treatment at birth and strengthen the case that more aggressive treatment does not lead to increased levels of morbidity.