Article Text


Interpreting data on the health outcomes of extremely preterm babies
  1. Jennifer Zeitlin,
  2. Pierre-Yves Ancel
  1. U953 INSERM, Epidemiological Research Unit on Perinatal and Women's and Children's Health, Paris, France
  1. Correspondence to Jennifer Zeitlin, U953 INSERM, Epidemiological Research Unit on Perinatal and Women's and Children's Health, Hôpital Saint-Vincent de Paul, 82 avenue Denfert-Rochereau, Paris 75104, France; jennifer.zeitlin{at}

Statistics from

The data presented by Rattihalli et al1 on the evolution of health outcomes for babies born before 26 weeks gestation in the Trent region show rising survival rates (27–46%) between 1991–1993 and 2001–2003 accompanied by a slightly increased probability of severe disability among survivors (35–39%). These results raise multiple questions: Why were there such large gains in survival but no improvement in impairment rates? Can these results be used to draw conclusions about trends in outcome in other regions of Europe or in the USA? How can these data be used to guide clinical decisions for these infants whose probability of being severely disabled is over 1 in 3?

Trying to contextualise these findings using the available scientific literature is daunting. There is convergence among studies showing increases in survival for very preterm infants throughout the 1980s and early 1990s. Some recent data, however, suggest that survival increases slowed or stopped starting in the late 1990s.2,,4 Survival among babies of less than 26 weeks gestation from the state of Victoria, Australia, increased between 1991/1992 and 1997 from 33% to 54%, but decreased to 43% in 2005.2 The National Institute of Child Health and Human Development (NICHD) Neonatal Research Network reported stable mortality between 2003 and 2007 among babies of less than 28 weeks gestation.4 In contrast, survival improved for infants of less than 28 weeks gestation in the Parisian region between 1997 and 2003, in the southern region of Sweden between 1995 and 2004 and in Japan between 1990 and 2005.5,,7

Assessing the trends in longer term health outcomes after preterm birth is even more challenging because studies use different definitions of disability and evaluate children at varying ages. In this context, this study from the Trent region is immensely valuable because it uses the same measures in the same population over time. It gives pause for reflection that (to our knowledge) this study is one of only three population-based comparisons of longer term developmental outcomes in very preterm infants over time published in the scientific literature.2 8 Studies such as these provide robust and truly comparable measures of the evolution in health over time and can be used to compare these trends between health systems.

Editorial on the paper by Rattihalli et al (see page F329)

Data from the other population-based study covering a comparable time period in the state of Victoria show that severe disability among survivors born at less than 28 weeks gestation rose from 1991–1992 (7.8%) to 1997 (15.4%), but decreased in 2005 (3.7%).2 Rates of cerebral palsy followed a similar, but less pronounced trend (11.0%, 12.1% and 9.8%) and were not statistically significantly. These data suggest that there may be a lag effect, whereby initial increases in survival are not immediately translated into better longer term outcomes and may even be associated with increases in rates of impairment. Support for this hypothesised pathway also comes from the literature on cerebral palsy prevalence. While rising cerebral palsy rates were documented in the very preterm population initially, many studies have begun to note downward trends for these infants, although mainly among those over 1000 g or 28 weeks gestation.9 For infants less than 28 weeks gestation, the Swedish population-based register documented declines between 1995–1998 and 1999–2002,10 as did a study from northern Alberta.11 This latter study finds a clear inverted U curve, with cerebral palsy prevalence rising sharply as survival improves throughout the 1980s and early 1990s and declining after 1995. Another single centre study in Cleveland also reported rising rates of impairment followed by reductions.12

Other studies have not reported an inverted U-shaped pattern and have found both increases and decreases in impairment over time. In the NICHD Neonatal Research Network increases in survival for infants of 22–26 weeks gestation over the period 1993–1994 to 1997–1998 were concomitant with improvements in the Bayley Mental Development Index (MDI)and neurodevelopmental impairment, but not rates of cerebral palsy.13 For infants of less than 25 weeks gestation, however, Bayley MDI scores decreased in the 1990s.14 In Nova Scotia, cerebral palsy increased significantly between 1993–1997 and 1998–2002 for infants of 24–30 weeks gestation.15 Studies from Finland and the Netherlands reported stagnating or increasing rates of neurodevelopmental impairment for low birthweight infants (<750 or 1000 g).3 16 The study from Trent is consistent with these latter studies, in which no improvement in the longer term outcomes of survivors is found.

Reviewing this literature is unsettling because it is not possible to identify consistent and common trends over time17 or the reasons for the differences between studies. Part of the obstacle is methodological; studies report on different subgroups of very preterm infants (<25 weeks,14 <26 weeks,1 <27 weeks,13 <28 weeks,2 6 24–27 weeks,5 24–30 weeks,15 <1000 g,3 13 500–999 g,12 ≥24 weeks and ≤750 g,16 <29 weeks and <1500 g,4 20–27 weeks and 500–1250 g11) and do not systematically present results by weeks of gestational age. Studies are also population based, networks of units or individual units raising questions about differential evolutions in recruitment over time. The need in the future is to promote population-based studies which routinely collect data on long term outcome for preterm infants. Presenting data by gestational age (despite small sample sizes, and perhaps in online tables only) would be extremely helpful for researchers and clinicians seeking to reconstitute external comparisons for their populations of interest. Despite these methodological differences, however, it seems safe to conclude that trends over time have differed for geographically and temporally distinct populations.

The good news is that more and better data are forthcoming in Europe. Many regions are implementing routine longer term follow-up of very preterm infants, as in the Trent region, or establishing successive population-based cohort studies of very preterm infants (such as EPICure 1 and 2, EPIPAGE 1 (1997) and 2 (to start in 2011), or the European EPICE cohort (2011)) or developing longer term follow-up protocols within neonatal networks (such as the EURONEONET project). Collaboration between European cerebral palsy registers continues to expand through the Surveillance of Cerebral Palsy in Europe (SCPE) Network. In addition to the possibility of producing more comparable data on long term outcomes for very preterm infants, these initiatives provide an opportunity to get more clarity in three critical areas which may explain the observed differences between populations of very preterm infants.

First, ethical decision-making at the limit of viability can be extremely variable between units and countries and this affects the types of care babies receive and their survival probabilities.18,,20 These decisions are made in both the obstetric and neonatal wards. It is of interest that the number of extremely preterm live births did not increase in Trent over the period under study in contrast with other population-based studies of very preterm births over time.5 7 11 21 More active medical intervention has been associated with higher rates of live birth among very preterm infants.22 Caesarean section rates for infants born between 24 and 25 weeks, excluding caesareans for maternal indications, varied from 0% to 78% between European regions in 2003.23 While differences in survival have been associated with more active intervention,23 more research is needed on associations with longer term disability among survivors.

Second, health systems and healthcare practices may also affect probabilities of longer term impairment. Comparisons across health systems have revealed significant differences in mortality and short term morbidity.24 25 For instance, the MOSAIC study of infants born below 32 weeks gestation in Europe found twofold differences in rates of grade III/IV intraventricular haemorrhage and periventricular leukomalacia (2–10%) and bronchopulmonary dysplasia (10–21%) between regions. This study, and others, have revealed significant differences in the organisation of care and use of medical interventions between regions of Europe.25,,29 The extent of regionalisation, the volume of neonatal nurseries and the use of evidence-based practices and protocols could lead to varying patterns in outcome.

Third, infant characteristics and their evolution over time can explain differences in longer term outcomes. One of the reasons why gains in survival may not be reflected in reduced morbidity is that higher risk infants survive. Thus, risk adjusted morbidity improves, but is obscured by an increased pool of high risk infants. While this hypothesis has been supported by many studies which find a decrease in birth weight and gestational age in later cohorts, it is also plausible that the increased use of antenatal corticosteroids and prophylactic surfactant positively affects the risk profile of the cohort. Other factors to consider are the types of pregnancy complications which have been associated with risk of cerebral palsy and neurodevelopmental impairment.30 31

Providing prognostic data adjusted for population and clinical risk factors makes data on longer term outcome more useful. In the extremely preterm population where risks of poor outcome are alarmingly high, on average, and staggeringly high in the presence of multiple risk factors, using other information in addition to gestational age to refine the prognosis is particularly important for guiding the decisions of clinicians and parents.32 A recent study estimated that the mortality risk of an infant born at 24 weeks gestation with a birth weight under the 25th percentile who had not received antenatal corticosteroids was 88% in contrast with 34% for a 24-week infant with a birth weight over the 75th percentile who had received steroids.33 Models could be refined further using information from the delivery room and the first days of life.34 Risk adjusted data may also be more generalisable, although these prognostic models require validation in varying geographic and temporal contexts.

A final point concerns the population that should be studied in the future. Despite understandable concern about extremely preterm infants at the limit of viability, maintaining a focus on all very preterm infants – up to at least 31 weeks gestation – is important. In the French EPIPAGE study of very preterm infants, 83% of infants with cerebral palsy and 88% of those with a developmental score less than 70 were born between 27 and 31 weeks gestation.35 While rates of disability among extremely preterm infants are starkly higher than those at later gestations, infants with gestational ages of less than 27 weeks make up a minority of infants with cerebral palsy and other developmental problems linked to very preterm birth.


  1. 1.
  2. 2.
  3. 3.
  4. 4.
  5. 5.
  6. 6.
  7. 7.
  8. 8.
  9. 9.
  10. 10.
  11. 11.
  12. 12.
  13. 13.
  14. 14.
  15. 15.
  16. 16.
  17. 17.
  18. 18.
  19. 19.
  20. 20.
  21. 21.
  22. 22.
  23. 23.
  24. 24.
  25. 25.
  26. 26.
  27. 27.
  28. 28.
  29. 29.
  30. 30.
  31. 31.
  32. 32.
  33. 33.
  34. 34.
  35. 34.
View Abstract


  • Competing interests None.

  • Provenance and peer review Commissioned; not externally peer reviewed.

Request Permissions

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.