Obstetric interventions for babies born before 28 weeks of gestation in Europe: results of the MOSAIC study


LAA Kollée, Professor of Paediatrics, Children’s Hospital, Radboud University Nijmegen Medical Centre, PO Box 9101, 6500 HB Nijmegen, the Netherlands. E-mail l.kollee@cukz.umcn.nl


Objective  To describe obstetric intervention for extremely preterm births in ten European regions and assess its impact on mortality and short term morbidity.

Design  Prospective observational cohort study.

Setting  Ten regions from nine countries participating in the ‘Models of Organising Access to Intensive Care for Very Preterm Babies in Europe’ (MOSAIC) project.

Population  All births from 22 to 29 weeks of gestation (n = 4146) in 2003, excluding terminations of pregnancy.

Methods  Comparison of three obstetric interventions (antenatal corticosteroids, antenatal transfer and caesarean section for fetal indication) rates at 22–23, 24–25 and 26–27 weeks to that at 28–29 weeks and the association of the level of intervention with pregnancy outcome.

Main outcome measures  Use of antenatal corticosteroids, antenatal transfer and caesarean section by two-week gestational age groups as well as a composite score of these three interventions. Outcomes included stillbirth, in-hospital mortality and intraventricular haemorrhage (IVH) grades III and IV and/or periventricular leucomalacia (PVL) and bronchopulmonary dysplasia (BPD).

Results  There were large differences between regions in interventions for births at 22–23 and 24–25 weeks. Differences were most pronounced at 24–25 weeks; in some regions these babies received the same care as babies of 28–29 weeks, whereas elsewhere levels of intervention were distinctly lower. Before 26 weeks and especially at 24–25 weeks, there was an association between the composite intervention score and mortality. No association was observed at 26–27 weeks. For survivors at 24–25 weeks, the intervention score was associated with higher rates of BPD, but not with IVH or PVL.

Conclusions  There are large differences between European regions in obstetric practices at the lower limit of viability and these are related to outcome, especially at 24–25 weeks.


Neonatal mortality because of preterm delivery has decreased significantly during the last decades.1–4 This decrease is usually attributed to advances in perinatal care. Important examples are regionalisation of high-risk pregnancies, antenatal corticosteroids, improved artificial ventilation techniques, surfactant administration, more sophisticated diagnostic tools and advanced monitoring of vital parameters.5,6 However, survival at the lowest range of gestational age is still low, particularly when geographically defined populations are considered.7 Therefore, the appropriateness of full intensive care for these infants continues to be questioned.8,9 As stated by William Silverman, ‘the dramatic increase of technical power in neonatal medicine has made possible an unprecedented form of extremism’ and ‘there is, to put it mildly, a pervasive suspicion that overtreatment of seriously compromised neonates is now widespread’.10

Large differences have been documented between European neonatal intensive care units with respect to ethical policies related to the treatment of extremely preterm babies.11 However, little comparative data exist on the variation of obstetrical practices and their impact on outcomes.

The objective of this study, which was part of the ‘Models of Organising Access to Intensive Care for Very Preterm Babies in Europe’ (MOSAIC) project, was to describe the differences in the use of three selected obstetric interventions in very preterm births and to assess whether the extent of intervention was correlated with mortality and short-term morbidity.


The MOSAIC study was a prospective cohort study of all births from 22+0 to 31+6 weeks of gestation in ten study regions in nine European countries in 2003. Participating regions were: Flanders in Belgium (BE), the Eastern region of Denmark (DK), the Ile-de-France region of France (FR), the State of Hesse in Germany (GE), Lazio in Italy (IT), the Central and Eastern regions of the Netherlands (NL), Wielkopolska and Lubuskie in Poland (PL), the Northern region of Portugal (PO) and the Northern and the former Trent region of the UK.12 These regions reflect the catchment areas of perinatal centres, although they do not necessarily correspond to administrative regions. Between 30 000 and 65 000 live births took place in the MOSAIC regions in 2003, except for Ile-de-France (135 000 births). The study period was from 1 January through 31 December 2003 with the exception, again, of the Ile-de-France region where births were included from 1 February to 31 August 2003.

Gestational age was based on the best estimate of the obstetrical team caring for the pregnant woman using information on last menstrual period and ultrasound investigation. Ultrasound scans for dating pregnancies are part of routine antenatal care in the MOSAIC regions.

The MOSAIC consortium selected data items with common definitions to be collected on a structured data sheet for each baby. In some regions with pre-existing data collection systems, MOSAIC items were added to existing protocols. Data were abstracted from records in the neonatal units for babies admitted to neonatal care. For stillbirths and babies dying before admission to a neonatal unit, the data sheets were filled in from obstetric records. In the regions in France, the Netherlands and the UK, the data sheets were filled in by external investigators, whereas in the other countries, they were completed by the units’ medical personnel. In all regions, however, the MOSAIC researchers reviewed the completed forms for missing and inconsistent information. In Belgium, Poland, Italy and France, the forms were translated into local languages in accordance with standard procedures. The MOSAIC instrument was pre-tested in all regions to verify that the required information was available in medical records and that the proposed definitions were applicable. Clarifications were made to the instrument after the pre-test.

The MOSAIC protocol required the cross-checking of inclusions in the study with birth registers of the maternity units to verify completeness of recruitment. However, some live births of borderline viability may have escaped inclusion if they occurred outside the obstetric ward or been misclassified as stillbirths rather than delivery room deaths if they died soon after birth. In the MOSAIC cohort stillbirth rates at <24 weeks, 24–27 weeks and 28–31 weeks groups varied between regions and some of these variations may be caused by misreporting.13

This article studied a sub-sample of the total cohort and included all births below 30 completed weeks of gestation, excluding terminations of pregnancy (TOP).

Three key obstetric interventions were selected: administration of corticosteroids to the mother, antenatal transfer to a level III perinatal centre and caesarean section for fetal reasons before or during labour. Caesarean sections for haemorrhage, (pre)eclampsia, HELLP syndrome and maternal disease were not considered. For antenatal corticosteroids, the definition used in MOSAIC was administration of any corticosteroids. In the Hesse region of Germany, however, this definition was not implemented consistently, as their routine data collection uses a stricter definition––a full course of antenatal steroids 24 hours before delivery. For 20% of the sample in Hesse, records were abstracted to verify corticosteroid administration using the MOSAIC definition.

Rates of missing data for these interventions were low; when >15%, they are flagged in the tables. Missing data were most frequent for administration of corticosteroids at very early gestations. Because of the differences in the definition of antenatal corticosteroid administration in Hesse, actual values are not reported in the table, but the ratios to the rate at 28–29 weeks are presented.

For each region, the rates of the three selected obstetric interventions carried out at each of the lower gestational age categories (22–23, 24–25 and 26–27 weeks) were computed and divided by those in the 28–29 weeks group. The latter was regarded as the reference group, under the assumption that active obstetric intervention at this stage of pregnancy is the standard and performed routinely in all participating regions. The resulting ratios may be viewed as measure of the frequency of application of each intervention at the earlier gestational ages standardised for practices in the more mature neonates in the same region. A ratio of one would indicate the same rate of intervention in the index gestation band as at 28–29 weeks. In contrast, a ratio below one would indicate less-active management in the index gestation band compared with the reference group.

Finally, a composite obstetric intervention score was computed for each region and index gestational age category by averaging the ratios for the three key interventions. For Hesse, the intervention score was computed with and without the ratio for antenatal steroids and also using the ratio computed on the 20% of the sample.

Three outcome measures were identified: 1) stillbirth rate, computed as the number of fetal deaths (terminations of pregnancy excluded), on the total number of births; 2) in-hospital mortality, that is the sum of intra-partum, labour ward and neonatal ward deaths during hospital stay, on the population of fetuses alive at onset of labour; 3) short-term morbidity, defined as a diagnosis of intraventricular haemorrhage grade III or IV (IVH) and/or periventricular leucomalacia (PVL) and bronchopulmonary dysplasia (BPD) among survivors to discharge from hospital. IVH was classified according to Papile.14 Bronchopulmonary dysplasia was defined as oxygen dependency or mechanical ventilation at 36 weeks.

The associations between the composite intervention score and regional outcomes were assessed through the Spearman’s rank correlation coefficients. Analyses were conducted with STATA 8.0 SE (STATA statistical software: Release 8.0.; Stata Corp., College Station, TX, USA, 2003).


Table 1 reports the total number of births (excluding terminations of pregnancy), the number of fetuses alive at the onset of labour and the number of stillbirths in the four gestational age groups in the participating regions. There were 4146 deliveries from 22+0 to 29+6 weeks of gestation, 1569 (30.5%) of them in the reference group (28–29 weeks).

Table 1.   Study population: very preterm births in the European regions participating in the MOSAIC project
Very preterm births except terminations of pregnancy by gestation groups
 22–23 weeks24–25 weeks26–27 weeks28–29 weeks
 All birthsStillbirthsAlive at onset of laborAll birthsStillbirthsAlive at onset of laborAll birthsStillbirthsAlive at onset of laborAll birthsStillbirthsAlive at onset of labor
BE: Flanders64453010229761323010517628152
DK: Eastern23161339336585531031490
FR: Ile-de-France15412775143551061944015630347258
GE: Hesse6040279619781191310718218164
IT: Lazio7358247723618915751097102
NL: Eastern/Central59324152282869205112321102
PL; Wielko/Lubuskie8137488521659823751142590
PO: Northern221314502133671257991883
UK: Trent76513911840861613213523729210
UK: Northern49361974215793207412310113

Table 2 shows the numbers and proportions of fetuses alive at the onset of labour that received the selected interventions by gestational age and region, as well as the ratios comparing the rates in the lower gestational age groups with those of the reference group. In most regions, around 80% of total births at 28–29 weeks received antenatal steroids. Rates were not significantly different at 26–27 weeks apart from the Polish region where the rate was lower. At 24–25 weeks, antenatal steroid use ranged from 18% in the Netherlands to 100% in Denmark, with rates for most regions close to 60%. Except for Denmark and UK North, steroid use at this gestational age was significantly lower than in the older babies (28–29 weeks). In all regions, antenatal steroids were only used in a minority of cases at 22–23 weeks and the ratios with respect to practices at 28–29 weeks were always significantly lower than one.

Table 2.   Selected obstetrical interventions by gestational age group and ratios comparing with 28–29 of GA, fetuses alive at onset of labor
 22–2324–2526–2728–29Ratio of rate at 28–29 weeks (95% CI)*
  1. *Bold italics to indicate that CI does not include 1.

  2. **More than 15 missing data.

  3. ***Different definition, percentages not presented (see methods).

  4. ****excluding maternal reasons.

Antenatal steroids
BE: Flanders2928766810478150810.34 (0.19–0.62)0.84 (0.71–0.99)0.96 (0.84–1.09)
DK: Eastern122531 100**41  93**81810.31 (0.11–0.82)1.23 (1.11–1.36)1.14 (0.99–1.30)
FR:Ile-de-France743965814983248880.03 (0.01–0.12)0.67 (0.56–0.79)0.95 (0.87–1.04)
DE: Hesse***2577--106--160--0.41 (0.18–0.90)0.97 (0.74–1.29)1.01 (0.79–1.30)
IT: Lazio22185761727294820.22 (0.09–0.54)0.75 (0.60–0.94)0.88 (0.74–1.05)
NL: Eastern/Central41528185186102870.06 (0.01–0.22)0.20 (0.09–0.45)0.99 (0.87–1.13)
PL: Wielko/Lubiskie45336450756089790.42 (0.28–0.65)0.64 (0.48–0.84)0.76 (0.62–0.95)
PO: Northern14293361578183780.36 (0.16–0.84)0.77 (0.57–1.04)1.03 (0.87–1.22)
UK: Trent3520856413387209890.22 (0.12–0.44)0.71 (0.60–0.84)0.98 (0.90–1.06)
UK: Northern1833509266 95**109830.40 (0.21–0.78)1.11 (0.99–1.25)1.16 (1.04–1.28)
Antenatal transfer
BE: Flanders3033762810543152360.94 (0.54–1.62)0.78 (0.51–1.19)1.21 (0.89–1.64)
DK: Eastern110361447988150.000.94 (0.36–2.44)0.58 (0.20–1.67)
FR: Ile-de-France7511063315647258520.03 (0.00–1.18)0.63 (0.70–0.85)0.91 (0.74–1.11)
DE: Hesse2627753710327158251.09 (0.55–2.17)1.51 (1.01–2.26)1.10 (0.73–1.67)
IT: Lazio2486015731099110.75 (0.18–3.16)1.35 (0.59–3.07)0.86 (0.35–2.12)
NL: Eastern/Central41528215165102580.08 (0.22–0.33)0.37 (0.18–0.76)1.12 (0.86–1.45)
PL: Wielko/Lubiskie470649731487180.000.51 (0.21–1.23)0.74 (0.36–1.54)
PO: Northern14213336574483340.64 (0.22–1.81)1.08 (0.63–1.85)1.30 (0.85–1.98)
UK: Trent3910862713526210180.57 (0.21–1.50)1.48 (0.94–2.32)1.43 (0.95–2.15)
UK: Northern181755457340112400.41 (0.14–1.19)1.13 (0.78–1.63)0.99 (0.69–1.42)
Caesarean section****
BE: Flanders2706978548124580.000.12 (0.05–0.29)0.83 (0.64–1.08)
DK: Eastern1283155468072600.14 (0.02–0.92)0.92 (0.63–1.33)1.35 (1.06–1.71)
FR: Ile-de-France75196713749216650.02 (0.00–0.14)0.11 (0.05–0.23)0.75 (0.62–0.92)
DE: Hesse21  19**54788483130880.22 (0.09–0.53)0.89 (0.76–1.03)0.95 (0.85–1.07)
IT: Lazio2255531675179780.06 (0.01–0.39)0.39 (0.26–0.59)0.65 (0.50–0.84)
NL: Eastern/Central410280463785380.000.000.98 (0.62–1.56)
PL: Wielko/Lubiskie4625817584766550.04 (0.01–0.28)0.32 (.17–0.58)0.85 (0.60–1.21)
PO: Northern13232532486767720.32 (0.12–0.88)0.45 (0.25–0.81)0.93 (0.72–1.20)
UK: Trent370721010134177500.000.20 (0.10–0.40)0.68 (0.50–0.92)
UK: Northern1905315632990470.000.32 (0.16–0.64)0.61 (0.39–0.96)

The percentages of preterm deliveries at 28–29 weeks of gestation that were preceded by antenatal transfer to a level III centre ranged from 11% in Italy to 58% in the Dutch region. Differences with earlier gestational age groups were less marked than for antenatal corticosteroid use; however, in most regions, the cohort of babies at 22–23 weeks was less likely to be transferred in utero and the difference with older babies was significant in the Netherlands . At 24–25 weeks, in utero transfer practices were closer to those in the reference group, except in the French and Dutch regions. In Hesse, transfer at 24–25 weeks was significantly more likely than in the 28–29 group.

There was considerable variation in the rates of caesarean section for fetal reasons in the reference group (from 38% in the Netherlands to 88% in Hesse). This intervention was used rarely in births at 22–23 weeks, except in the German and Polish regions. From 24 weeks on, practices varied more markedly between regions. In some (Denmark and Germany), caesarean sections were performed as frequently as in the reference group, while in others (French, Italian and UK regions), even at 26–27 weeks, caesarean sections were less common than at 28–29 weeks.

Figure 1 shows the composite intervention scores in the European regions for gestational age, calculated as the average of the three individual interventions ratios shown in Table 2. They differed between regions, but, in general, the composite scores at 22–23 weeks were low and increased with gestational age. At 22–23 weeks gestation, the intervention score was lowest in the Netherlands and in France, indicating a non aggressive management of these births, while the opposite was found in Germany. At 24–25 weeks, the scores varied from quite low in the Netherlands to very high in Germany and Denmark. At 26–27 weeks of gestation, interventions were performed at a similar degree as in the reference group and scores were above 0.80 in all regions except in Italy and Poland.

Figure 1.

 Composite intervention scores by region and gestational age group, calculated as average ratio of the three interventions at 22–23, 24–25 and 27–27 weeks compared with that at 28–29 weeks.

Figure 2 illustrates the association between the composite intervention scores and mortality outcomes. At 22–23 and 24–25 weeks, there was a significant association between the intervention scores and decreasing in-hospital mortality (rho −0.91, P < 0.001; rho −0.721, P = 0.019 respectively). At 24–25 weeks only, there was also an association between the score and the stillbirth rate (rho −0.697, P = 0.025). No association was found between the score and mortality outcomes at 26–27 weeks, where variability in both score values and mortality was low.

Figure 2.

 Association between the composite intervention scores at 22–23, 24–25 and 26–27 weeks and mortality outcome in ten European regions.

Figures 3 and 4 illustrate the association between mortality and the individual intervention ratios. The general patterns were similar to those observed with the composite score. At 22–23 weeks, the in-hospital mortality rate was correlated with antenatal transfer and, marginally, with antenatal steroid administration (Figure 3). No association was evident with regard to the stillbirth rate. The strongest associations were seen at 24–25 weeks, where the administration of antenatal corticosteroids ratio showed a significant association with decreasing stillbirths and in-hospital mortality (Figure 4). At this gestational age, antenatal transfer was only marginally related to in-hospital mortality and caesarean section was related to stillbirths. None of the associations was significant for the 26 to 27 week group, in line with results observed using the composite score (data not shown in figures).

Figure 3.

 Association between the individual intervention scores at 22–23 weeks and mortality outcome in ten European regions.

Figure 4.

 Association between the individual intervention scores at 24–25 weeks and mortality outcome in ten European regions.

For the 24–25 weeks group only, we explored also the correlation between intervention score and morbidity, defined as severe IVH/PVL and BPD among survivors (Figure 5). As in the Netherlands there were no survivors at this gestational age, only nine regions were included in this graph. No association between the scores and IVH or PVL was present, but there was a statistically significant increased incidence of BPD with higher intervention scores. At 22–23 weeks, the number of survivors was too low (n = 11) for this analysis. At 26–27 weeks, where no association with mortality was detected, rank correlations associating intervention scores with morbidity were also not significant.

Figure 5.

 Association between the composite intervention scores at 24–25 weeks and IVH/PVL and BPD in nine European regions. There were no survivors in the Netherlands.


The results of this observational study clearly demonstrate that extremely preterm births are treated differently in the different European regions of the MOSAIC study, with largest variability at 24–25 weeks. In some regions (Denmark, Germany and Italy), babies at that gestational age routinely received active treatment similary to those at 28–29 weeks, while in the Dutch region treatment at 24–25 weeks was conservative. From 26 weeks onwards, babies were treated similarly to those of the reference group in most of the study regions, although in France, Italy and UK, the decision to perform a caesarean section was taken less often than for the more mature group. The different intervention policies were associated with stillbirth and in-hospital mortality, especially at 24–25 weeks. At 26–27 weeks, both the rate of interventions and the rate of survival were more uniform in every region, preventing the possibility to explore correlations. In the most immature fetuses (22–23 weeks), we found that active obstetrical management was associated with increased in-hospital survival. Nevertheless, the number of survivors was so low—only 11 children overall– that we could not explore the effects on morbidity. At 24–25 weeks, where the impact of obstetrical interventions on survival appears to be the highest, no association was found with severe neurological morbidity among survivors, which is a reassuring finding, although balanced by the increasing frequency of BPD.

Variability of attitudes and self-reported practices towards management of severely diseased neonates has already been reported in Europe 15,16 However, this is the first study comparing actual practices in a large area-based sample of births across ten European countries. Additional strengths are the large sample size, prospective design and homogeneous data collection strategy. A potential weakness is our choice to use treatment at 28–29 weeks as the reference, for it is based on consensus rather than objective data. Nonetheless, our assumption that active obstetric management at that gestational age was the accepted standard, regardless of country or region, is supported by the fact that the frequency of antenatal steroid administration in this group of patients was around 70% or higher in all regions. The variation in the frequency of caesarean section from 30 to 81% at this gestational age might be explained by other differences in policies than intervention attitude. Furthermore, we did not want to choose a more mature group of very preterm babies (i.e., 30–31 weeks), as these babies are sometimes cared for in less-specialised centres in the participating regions.17

Although our study recorded information at the individual level, we chose to carry out an ecological analysis looking at the association between different treatment policies, as measured by the composite intervention score, and outcomes at regional level. This analysis allowed us to standardise intervention rates using a region-specific standard, namely the rate of intervention at 28–29 weeks. Rather than ‘proving’ the effectiveness of the selected obstetric interventions, either by themselves or in association, our findings show that where births at 24–25 weeks gestational age are treated as or almost as intensively as births at more mature gestational ages, mortality outcomes improve without increasing neurological short-term morbidity. However, increased survival is linked to increased frequency of BPD in this group.

The paucity of literature concerning the role of the obstetrician in the chain of perinatal decision making at the margin of viability is striking, since this profession has the first contact with pregnancies that result in extremely preterm delivery. Obstetricians have to make crucial decisions concerning maternal treatment, including the use of tocolytic drugs, administration of corticosteroids, referral to a level III centre and, the most far-reaching, performance of a caesarean section.

The effect of tocolysis is limited.18 However, the administration of corticosteroids for fetal lung maturation in cases of threatened preterm delivery has been shown to be highly effective in improving the outcome of babies born before 32 weeks. It results in less-severe lung disease and fewer related complications, including fewer cases of intraventricular haemorrhage and chronic lung disease.19–21 Transfer of the pregnant woman to a tertiary level centre for very preterm delivery is also associated with higher levels of survival.22–24 The use of caesarean section for very preterm deliveries in the absence of other obstetric indications is controversial and there is an ongoing debate about its impact on very preterm survival and morbidity.25–27 Observational studies on this subject have produced contradictory results, although in most cases, they found that caesarean delivery did not reduce the risk of mortality overall.26,28–31 For some selected subgroups, such as fetuses in breech presentation, multiple births or those with growth restriction, caesarean section was associated with improved survival,26,28,30–33 although not consistently.34 Because of the indication bias associated with the decision to carry out a caesarean section, however, observational studies cannot distinguish with certainty between the effect of the caesarean section and the underlying complications associated with the decision to perform one. Furthermore, for extremely preterm babies, a caesarean delivery is often a marker of a more general decision to actively intervene and may thus be linked to other medical interventions, which are responsible for greater survival rates.35–37 Overall, the appropriateness of intervening with operative delivery for extremely preterm infants remains uncertain.38

Our results indicate that obstetricians generally regard the administration of corticosteroids as good clinical practice at 26–27 and, to a lesser extent, at 24–25 weeks of gestation. Dutch obstetricians clearly differed from other European colleagues in their attitude towards antenatal steroid administration, seldom prescribing antenatal steroids before 26 weeks of gestation. However, this may change in the near future as Dutch professional bodies have recently advocated more active intervention at 25 weeks.39

The variability between the European regions in rates of antenatal transfer to tertiary centres for older preterm infants may be explained by differences in organisation of perinatal care between countries and regions.40 At 22–23 weeks, rates of antenatal transfer were negligible in most regions. In the Netherlands, the low rate of antenatal transfer below 26 weeks confirms the conservative approach to the treatment of extremely preterm neonates there, despite the high level of regionalisation of high-risk perinatal care. In France, babies at 24–25 weeks were also transferred in utero only half as often as babies of 28 and 29 weeks. In the other regions, however, in utero transfer rates at 24–25 weeks of gestational age did not differ significantly from the reference group.

The decision to perform a caesarean section for fetal distress at an extremely preterm gestational age may save the life of the baby, at least on the short-term, but the procedure is invasive for the mother and the most costly of the three selected for this study in terms of possible negative side effects. Particularly at a very early gestational age, the immediate benefits to the fetus must be weighted against possible adverse effects for current and future pregnancies, as well as with long-term infant outcomes. Within the reference group of babies of 28–29 weeks, practices related to caesarean section clearly varied between regions, but the generally high rates indicate that this was the standard of care at this gestation in most of them. The frequency of caesarean section before 26 weeks was low almost everywhere, however, except in Hesse. At 26–27 weeks, rates were lowest in the French, Italian and UK regions.

The variability across regions persisted when the three interventions were considered together as a composite intervention score. The high rate of intervention at 22–23 weeks in the Polish, Portuguese and German regions indicates a preference by physicians for active management, even in case of extremely preterm birth. This may point to a general cultural attitude in favour of maintaining life under all circumstances. For Germany, the findings are in agreement with previous studies in Europe.15,41 In contrast, in the Dutch region, there was a clear threshold for active intervention from 26 weeks onwards. In the other regions, this threshold appeared to be placed 2 weeks earlier, but in a significant number of cases, a decision was made not to actively intervene.

As the effectiveness of clinical interventions are difficult to assess in observational studies such as this one, we cannot determine with certainty whether the lower mortality results from the application of the interventions or, more generally, from the whole perinatal treatment package that goes together with the obstetrician’s decision to intervene. When we correlated the ratios for the individual interventions with mortality, we had results similar to the analyses of the composite score. Associations were strongest for the 24–25 week group and not significant at 26–27 weeks. Some of the interventions appeared to have a slightly different effect by type of mortality outcome. Antenatal transfer was more strongly associated with lower in-hospital mortality, while caesarean section ratios were significantly related to the stillbirth rate, as might be expected. Use of antenatal steroids was associated with outcomes at 24–25 weeks, but not at 22–23 weeks. As these data show, however, discriminating between the impact of these individual interventions is complex. Their use indicates an overall intention to treat the fetus and positive expectations for survival. Other studies have also found that willingness to intervene is a major determinant of survival. An active obstetric policy in Germany was demonstrated to result in higher survival rates up to 24 hours for babies at the margin of viability.36 The different ‘attitudes’ between New Jersey, USA and the Netherlands to preterm deliveries from 23 to 26 weeks in the 1980s affected outcome substantially: the death rate in New Jersey, USA was 10% versus 45% in Netherlands.37

In our study, we could not explore the relationship between active obstetric management and long-term neurological outcome among survivors. However, short-term indices of neonatal neurological morbidity in survivors were not related with the level of intervention in this cohort of very preterm infants, while respiratory morbidity was associated with higher composite score values in infants that were born at 24–25 weeks of gestation. That the occurrence of IVH and PVL as short-term morbidity indicators for cerebral damage was not clearly influenced by the interventions would be in line with the findings of a Swedish study, in which pro-active management of extremely preterm births in the North of Sweden resulted in higher 1-year survival without major morbidity compared with more selective management in the South.35 It is not yet established whether the increased survival of preterm infants recorded over the last decades has changed rates of long-term disability of perinatal origin. However, in the previously mentioned study comparing outcomes in New Jersey and the Netherlands, differences in attitudes affected not only mortality but also longer term outcome. For example, the rate of cerebral palsy among survivors at age two was 17% in New Jersey, while it was only 3.4% at age five in the Netherlands.37 More recently, a cost-effectiveness analysis about caesarean section for fetal indications at 24 weeks showed that both intact survival (16.8% versus 12.9%) and survival with major morbidity (39.2% versus 19.4%) were higher with willingness to perform surgery25 Some data suggest that increased survival might be linked to increased short- and long-term morbidities, particularly for infants at the border of viability.42–44

The decision to intervene in extremely preterm births is complex and observational studies such as MOSAIC cannot answer the question of whether more active intervention is justified.

Long-term follow-up studies are necessary to understand fully the impact of the decisions made by health professionals and outcomes at the limits of viability.

The large differences in intervention approaches and related mortality figures at gestational ages below 26 weeks may influence the results of comparative perinatal mortality studies, particularly, when they are carried out at the international level. This should be taken into account in the interpretation of the results, especially when mortality figures are used as indicators of quality of care.


We conclude that at the margin of viability, particularly around 24–25 weeks of gestation, differences in short-term perinatal outcome reflect in part differences in intervention policies. Different mortality rates between regions or countries may partly be determined by differences in ethical attitude and resulting treatment policies. Such differences do not necessarily reflect quality of care as these are two different notions. Appropriate ethical decision making at the margin of viability requires strong cooperation between obstetricians and neonatologists and provision of good information for parents to balance perinatal treatment aiming at optimal outcomes with a reasonable chance of quality of life in surviving infants.

Disclosure of interests

The authors have nothing to disclose with regard to potential conflicts of interest.

Contribution to authorship

All authors were involved in the design of the study and writing of the manuscript.

L. Kollée, M. Cuttini, R. Agostino, K. Boerch, J.-L. Chabernaud, E. Draper, L. Gortner, W. Künzel, R.Maier, J. Mazela, D. Milligan P. Van Reempts and T. Weber were involved in the data collection. J. Zeitlin had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis. J. Zeitlin, D. Delmas, L. Kollée, E. Papiernik, A. den Ouden and G. Bréart were involved in the data analysis. All authors have approved the final version of the manuscript.

Details of ethics approval

No ethics approval was necessary for this specific study. Ethics approval was sought for the collection of the data as required in each of the participating regions.


This study was financed by a grant from the European Commission Research Directorate (contract nr. QLG4-2001-01907).


The authors would like to thank Dr Brad Manktelow for his statistical help and the personnel in the maternity and neonatal units in the regions participating in the MOSAIC project for their cooperation.


The MOSAIC Research Group

Belgium, Flanders (E. Martens, G. Martens, A. Bekaert, P. Van Reempts); Denmark, Eastern Denmark (K. Boerch, T. Weber, B. Peitersen); France, Ile-de-France (G. Bréart, JL. Chabernaud, D. Delmas, P-H Jarreau, E. Papiernik); Germany, Hesse (L. Gortner, W. Künzel, R. Maier, B. Misselwitz, S. Schmidt); Italy, Lazio (R. Agostino, D. Di Lallo, R. Paesano); Netherlands, Eastern & Central (L. den Ouden, L. Kollée, G. Visser, J. Gerrits, R. de Heus); Poland, Northern Wielkopolska & Lubuskie (G. Breborowicz, J. Gadzinowski, J. Mazela); Portugal, Northern Region (H. Barros, I. Campos, M. Carrapato, UK, Trent Region (E. Draper, D. Field, J. Konje); UK, Northern Region (A. Fenton, D. Milligan, S. Sturgiss); INSERM U149, Paris (G. Bréart, B. Blondel, H.Pilkington, J. Zeitlin); External contributors (M. Cuttini, S. Petrou). Steering committee (E. Papiernik, project leader, J. Zeitlin, research coordinator, G. Bréart, E. Draper, L. Kollée).