General obstetrics: Preterm birth and multiple pregnancy in European countries participating in the PERISTAT project


B Blondel, INSERM, U149, 16 Avenue Paul Vaillant-Couturier, 94807 Villejuif Cedex, France. Email


Objective  To compare rates of preterm birth among multiple births in European countries, to estimate their contribution to overall preterm birth rates and to explore factors which could explain differences between preterm birth rates.

Design  Analyses of data from vital statistics, birth registers or national samples of births.

Setting  Eleven member states of the European Union.

Population  All live births or representative samples of births at national or regional level for the year 2000 or most recent year.

Methods  Description of rates of preterm birth before 37 and 32 weeks, estimation of population attributable risks (PAR), study of associations between preterm birth rates in multiples and singletons and nonspontaneous labour using Spearman's rank correlation coefficient.

Main outcome measures  Preterm birth rates, PAR, proportions of deliveries with nonspontaneous onset (caesarean sections before labour or induction of labour).

Results  The proportion of multiple births before 37 weeks varied from 68.4% in Austria to 42.2% in the Republic of Ireland. In half of the countries, over 20% of all preterm births were attributable to multiple births. A strong association was found between the proportions of births before 37 weeks among multiple and singleton births (r= 0.81; P < 0.001). An association was observed between the rates of preterm birth and the proportions of deliveries with nonspontaneous onset among twins.

Conclusions  Wide variations in rates of preterm births and deliveries with nonspontaneous onset were found between countries, suggesting marked differences in clinical practice which could have long-term implications for the health of children from multiple births.


Patterns of multiple births and the context in which they occur have changed considerably since the 1970s. In most developed countries, multiple birth rates have risen mainly because of the increasing maternal age and the availability of fertility treatments.1 However, across Europe the timing and extent of the increase has varied. For example, between 1975 and 2002 the twinning rate increased by 50% in England and Wales and by 90% in the Netherlands.2

In addition to rising multiple birth rates, increase in preterm birth rates has been observed in countries such as Canada, the USA and France.3 As with multiple birth rates, the size of this increase has varied between countries. It would appear that a substantial proportion of the increase may be attributed to a greater tendency for clinicians to intervene actively by inducing labour or undertaking a caesarean before labour, with the intention of preventing maternal or perinatal morbidity and mortality.4,5

To investigate this, we compared preterm birth rates among multiple births in the member states of the European Union, using aggregated data compiled for the PERISTAT project.6 Further, we wished to assess the risk of preterm birth attributable to multiple births in each country, and within the constraints of the available data, to attempt to explain differences in rates of preterm birth. For the last of these objectives, comparisons were also made between multiple and singleton births to test whether differences between member states were specifically related to multiple births.


The PERISTAT project was set up to develop a set of indicators for monitoring and describing perinatal health in Europe, forming part of the European Commission's Health Monitoring Programme.6 After using a Delphi method to compile a list of indicators, a feasibility study was undertaken to assess the extent to which the participating countries were able to provide data to construct these indicators. Standard tables of aggregated data were derived from routine population-based data available in each country for the year 2000, if available, or the data for the most recent year were used.7

All states which were members of the European Union at the time of the project participated and provided data to a varying extent. In most countries, data were derived from civil registration or medical birth registers (Appendix). National data were available for seven countries (Austria, Denmark, Finland, Ireland, Italy, the Netherlands and Sweden). In France, rates of multiple births were available from birth registration but the other data were derived from the national perinatal survey, which included all births in 1 week in 1998. In Germany, data were available for all births in 9 of its 16 states (Bundesländer), accounting for about 70% of all births in Germany. In the UK, data on gestational age were not available for England and Wales but data were provided separately for Scotland and Northern Ireland. For Belgium, only data from Flanders were used. Data for the French-speaking area of Belgium and Greece were omitted because there were inconsistencies in the multiple births data. Portugal and Spain could not provide the relevant data for multiple births. In Luxembourg, the number of multiple births was too small to provide precise estimates of preterm birth rates.

For the PERISTAT project, we obtained specific data on multiple births which were included in three of the tables requested from each country:

  • 1The number of women delivering live or stillborn babies by number of fetuses.
  • 2The number of live births subdivided into singleton births and multiples and gestational age at birth, in weeks.
  • 3The number of live births by method of onset of labour (spontaneous, induced and caesarean before labour) and whether they were singleton or twin preterm or term deliveries.

Multiple maternities in each country were described in terms of maternities which were defined as pregnancies leading to one, two or more registrable births (live births or stillbirths). Rates of births before 32 weeks and between 33 and 36 weeks and the overall preterm birth rates before 37 weeks were described in terms of proportions of multiple live births. Relative risks (RR) and population attributable risks (PAR) for births before 37 weeks, with their confidence intervals,8 were calculated for multiples, with singletons as the reference group. The exact numbers of multiple live births were unknown for some countries. We therefore derived percentages of multiple live births from data about numbers of singleton and multiple maternities in the relevant country.

Finally, we examined the associations between preterm births in multiples and two other indicators, preterm births in singleton births and deliveries with nonspontaneous onset. The latter were defined as caesarean sections before labour or inductions of labour. For the countries of the UK, the data related to elective caesareans, were defined as caesareans where the decision to intervene was taken before the onset of labour. These include a small but unknown number of caesareans, where a decision was taken before labour but the woman was already in labour when the caesarean section was performed. These deliveries with nonspontaneous onset were used as an indicator of the attitude to intervention at the end of pregnancy in each country, despite the fact that we had no information about whether the intervention was undertaken for clinical reasons, such as suspected fetal distress, or for nonmedical reasons. The ecological associations were tested using Spearman's rank correlation coefficient.


Multiple maternity rates varied across countries, ranging from 12.2 per 1000 maternities in Italy to 19.4 per 1000 in the Netherlands (Table 1). Triplet maternity rates were much lower than twin maternity rates and the patterns differed from those observed for twins.

Table 1.  Multiple maternity rates in European countries in 2000*
 Rate per 1000 total maternities**
CountryNumber of maternitiesTwin birthsTriplet and higher order birthsAll multiple births
  • *

    2001 in Austria, 1999 in Republic of Ireland and the Netherlands and 1998 in Italy.

  • **

    Pregnancies leading to one, two or more registrable births.

Austria74 55914.950.4215.37
Belgium (Flanders)60 98718.040.3018.33
Denmark65 99618.320.4418.75
Finland55 85215.880.1616.04
France766 42114.980.2815.26
Germany (9 Bundesländer)549 44915.820.6216.44
Ireland53 54913.000.5213.52
Italy527 21611.730.5212.25
Netherlands197 72618.980.3819.37
Sweden88 33115.990.2016.19
UK (England and Wales)598 58014.240.4414.68
UK (Northern Ireland)21 28114.750.2314.99
UK (Scotland)52 59814.010.3614.37

There were also wide variations in the proportions of multiple births which were preterm (Table 2). The proportion of births before 37 weeks ranged from 68.4% in Austria to 42.2% in Republic of Ireland. The proportion in Republic of Ireland was 62% lower than in Austria and 32% lower than in Flanders. The proportion of births before 32 weeks was also very high, 12.7%, in Austria but there was less variation elsewhere. Variations between countries were fairly similar to the overall proportions of preterm births.

Table 2.  Number (%) of multiple live births before 32 weeks, 32–36 weeks and before 37 weeks in 2000*
Country**n<32 weeks, % (95% CI)32–36 weeks, % (95% CI)<37 weeks, % (95% CI)
  • *

    2001 in Austria, 1999 in Republic of Ireland and the Netherlands and 1998 in France and Italy.

  • **

    Coverage if not entire population of country.

Austria231112.7 (11.4–14.1)55.7 (53.7–57.7)68.4 (66.5–70.3)
Belgium (Flanders)22208.8 (7.6–10.0)47.1 (45.0–49.1)55.9 (53.8–57.9)
Denmark26048.6 (7.6–9.7)36.2 (34.4–38.1)44.9 (42.9–46.8)
Finland17868.4 (7.1–9.7)40.9 (38.6–43.2)49.3 (47.0–51.6)
France (Perinatal survey)4659.0 (6.4–11.6)39.1 (34.7–43.6)48.2 (43.6–52.7)
Germany (9 Bundesländer)17 4599.8 (9.4–10.3)43.5 (42.8–44.2)53.3 (52.6–54.1)
Ireland14638.1 (6.7–9.5)34.1 (31.7–36.5)42.2 (39.6–44.7)
Italy13 1888.3 (7.8–8.8)40.3 (39.4–41.1)48.6 (47.7–49.4)
Netherlands74929.0 (8.4–9.7)39.1 (38.0–40.3)48.2 (47.0–49.3)
Sweden27818.1 (7.1–9.1)35.4 (33.6–37.2)43.5 (41.7–45.4)
UK (Northern Ireland)6548.7 (6.6–10.9)44.3 (40.5–48.2)53.1 (49.2–56.9)
UK (Scotland)14329.9 (8.4–11.5)41.3 (38.7–43.8)51.2 (48.6–53.8)

Figure 1 compares the distributions of multiple live births by gestational age in six countries in order to see whether observed differences in the preterm birth rates between countries are similar at all gestational ages. For this analysis, two contrasting groups of countries were selected: Austria, Flanders and Germany had the highest preterm birth rates, while Denmark, the Republic of Ireland and Sweden had the lowest rates. Between 33 and 36 weeks, the differences between the two groups of countries gradually widen. After 36 weeks, the gestational age distributions become more variable.

Figure 1.

Distribution of gestational age of multiple live births in selected countries.

The minimum RR of preterm live birth for multiples compared with singletons was 8.2, in Austria (Table 3). The highest RR was 10.5, in Finland. At least 17.6% of preterm live births were attributable to multiples in the countries studied. The PAR was above 20% in half of the countries, reaching 24.8 in Denmark.

Table 3.  Relative risk (RR) and population attributable risk (PAR) for preterm live births (<37 weeks) of multiple births compared with singleton births in 2000*
Country**Percent of births <37 weeksRR (95% CI)PAR*** (95% CI)
Multiple birthsSingleton births
  • *

    2001 in Austria, 1999 in Republic of Ireland and the Netherlands and 1998 in France and Italy.

  • **

    Coverage if not entire population of country.

  • ***

    Using the rate of multiple births derived from the number of maternities as detailed in Table 1.

Austria68.48.48.2 (7.9–8.5)18.0 (17.1–18.8)
Belgium (Flanders) (8.9–9.8)23.3 (22.1–24.5)
Denmark44.94.89.4 (8.9–9.9)24.8 (23.5–26.1)
Finland49.34.710.5 (9.8–11.1)23.1 (21.6–24.5)
France (Perinatal survey)48.24.710.3 (9.1–11.6)21.9 (19.0–24.8)
Germany (9 Bundeslander) (8.5–8.8)20.2 (19.8–20.6)
Ireland42.24.49.5 (8.9–10.2)18.8 (17.4–20.3)
Italy48.65.19.6 (9.4–9.8)17.6 (17.1–18.0)
Netherlands48.25.88.3 (8.1–8.6)21.9 (21.3–22.6)
Sweden43.55.28.4 (7.9–8.8)17.7 (18.1–20.2)
UK (Northern Ireland) (8.5–10.2)19.9 (17.9–21.9)
UK (Scotland) (8.0–9.0)17.7 (16.4–18.9)

There was a significant correlation between the proportion of preterm births (<37 weeks) in multiple pregnancies and the proportion of preterm births in singleton pregnancies (rank correlation coefficient r= 0.77, p= 0.003).

Data on the onset of labour were available for eight member states or countries within member states. For twin pregnancies, the rate of births with nonspontaneous onset varied between 43.6 and 62.2%. For singleton births, the rates varied between 15.9 and 46.0% (Table 4). There was an association between the proportion of twin births with nonspontaneous onset and that of singleton births (rank correlation coefficient r= 0.74, p= 0.04). There was also an association between the proportion of twin births with nonspontaneous onset and the proportion of multiple live births before 37 weeks of gestation in each country (rank correlation coefficient r= 0.76, p= 0.03).

Table 4.  Births with nonspontaneous onset* in 2000**
Country***Total n (%)Twins n (%)Singleton births n (%)
  • *

    Induction of labour or caesarean before labour or elective caesarean section in the UK.

  • **

    1998 in France.

  • ***

    Coverage if not entire population of country.

Belgium (Flanders)61 794 (40.0%)2170 (53.7%)59 624 (39.5%)
Denmark66 868 (17.2%)2600 (49.8%)64 268 (15.9%)
Finland56 446 (21.8%)1766 (43.6%)54 680 (21.1%)
France (Perinatal Survey)13 505 (29.5%)465 (48.4%)13 040 (28.9%)
Germany (9 Bundeslander)556 076 (25.8%)17 250 (51.3%)538 823 (25.0%)
Sweden87 767 (17.3%)2526 (44.0%)85 241 (16.5%)
UK (Northern Ireland)21 619 (46.5%)651 (62.2%)20 968 (46.0%)
UK (Scotland)52 042 (34.4%)1375 (50.4%)50 667 (34.0%)


In many European Union states, about half of the children from multiple births were born preterm and they accounted for between 18 and 25% of preterm births in each country. Countries which had high rates of preterm birth among multiples also tended to have high rates of preterm births among singleton babies. On the other hand, there was an association between the proportions of multiple births which were preterm and the proportions of births with nonspontaneous onset.

Our study illustrated the problems inherent in acquiring detailed data for all the participating European countries. The gaps in availability of data on gestational age and the onset of labour meant that the ecological correlations were based on data for a relatively small number of countries. Added to this, many European countries have small populations and thus a small number of multiple births. The rates are therefore subject to fairly sizeable year-to-year fluctuations, as can be seen from the wide confidence intervals. More generally, even when there is adequate statistical power to detect correlations between indicators, these must be interpreted with care since the use of aggregated data limits our ability to draw causal inferences.

Because of the limitations of the data available, we estimated the proportions of preterm births which could be attributed to all multiple births, without distinguishing between twins and triplets or higher order births. PAR are dominated by twins, however, as triplets and higher order births are very much less common. Thus, for example, in Canada over the 3 years (1995–1997 combined), 10.3% of preterm live births were attributable to twins and 1.1% were attributable to triplets or higher order pregnancies.3

In our study, the proportion of preterm births attributable to multiples varied from 18 to 25%. These variations result not only from differences in multiple birth rates but also from differences within countries between preterm birth rates of multiple and singleton births. Thus, high levels of PAR can be largely explained in the Netherlands by a high rate of multiple birth, in Finland and France by a relatively low rate of preterm birth among singleton births compared with multiple births and in Flanders by a combination of both these factors.

Among providers of services for subfertility, there is increasing concern about the need to avoid multiple pregnancies. This is likely to have contributed to the recent decreases in triplet and higher order birth rates which have been observed in many Western countries.2,9 This attitude could also have an impact on rates of twinning. Decreases in twinning rates observed in Finland and Sweden9 could be explained by a substantial rise in the use of single-embryo transfer in in vitro fertilisation (IVF)10 which has taken place without any reduction in success rates.11 Such changes in practice will not have an immediate impact on the overall level of preterm birth in most countries, however. Single-embryo transfer is not yet widely used in assisted conception elsewhere in Europe12 and it is more difficult to avoid a multiple pregnancy when ovarian stimulation is used on its own.13

Over and above the use of medical management of subfertility, the rising age at childbirth is another factor which has contributed to the rise in multiple birth rates. Before these techniques were developed, data for many countries showed that multiple birth rates increased according to mothers’ ages up to the age of 40 years.14,15 From the 1960s up to the end of the twentieth century, mothers’ ages at childbirth increased in most European countries, without any sign of a decline.16 It would therefore seem that the trends in multiple births attributable to this factor are unlikely to be reversed in the near future.

Our analyses show that rates of preterm birth among multiple births vary widely between countries. Differences in the proportions of multiple births which were triplets contribute only very minimally to this because of the small numbers involved. For example, in Germany, where the rate of triplet and higher order births was very high, a halving of the observed triplet rate of 0.62 per 1000 maternities to 0.31 per 1000 would lead to only a modest decrease in the rate of preterm birth among multiple births, from 53.3 to 51.9%.

Other factors could explain the differences between countries in their rates of preterm birth. First, the way in which gestational age is measured can have an impact on rates of preterm birth in both singleton and multiple births. Studies based on very large populations have shown that basing measures of gestational age on the date of the last menstrual period can lead to underestimates of gestational age. Rates of preterm birth based on gestational ages measured only by ultrasound can be 20% higher than when gestational age is based on dates of last menstrual period alone.17,18 It was not known to what extent ultrasound had been used for determination of gestational age in participating countries. Preferentially choosing ultrasound as the method of measurement will have had the effect of shifting the gestational age distribution to the left.19 It was difficult to assess this effect, however, because rates of deliveries with nonspontaneous onset are high among twins and this could affect the whole gestational age distribution. The same applies if some countries rely solely on ultrasound, while others use only the date of last menstrual period. This could contribute to a limited extent to the differences in rates of preterm births, for example between the rates for Austria and Ireland.

Second, rates of preterm birth could be affected by factors which are more common in some countries than in others. While the contribution of subfertility treatments to multiple birth rates is high and varies between countries,20,21 recent studies suggest that the unadjusted risk of delivery before 37 weeks may be slightly higher in twins resulting from IVF or ovarian stimulation compared with other twins.22,23 Moreover, the strength of the association between some socio-demographic characteristics and the risk of preterm birth is much lower for multiple births than for singleton births.24,25 As a result, the risk factors are unlikely to explain entirely the differences in preterm birth rates between countries.

A third factor which could explain variations in the rates of preterm birth is differences between obstetric practices in the participating countries. Clinicians in some countries may be more inclined than those in others to intervene to prevent preterm birth in women with multiple pregnancies. Nevertheless, even if they exist, such differences would have a limited influence, as the effectiveness of the measures which are commonly proposed for preventing preterm birth in multiple pregnancies have not been clearly demonstrated.26–28 In addition, the increasing focus on fetal wellbeing and maternal complications could reduce the impact of measures aimed at reducing preterm birth, if it increasingly leads to decisions to end the pregnancy before term.

To explore possible associations with obstetric practice, we used the proportion of deliveries which were of nonspontaneous onset as an indicator of practice. The ecological association found between preterm births and births with nonspontaneous onset suggests that differences in approaches to obstetric management at the end of pregnancy are likely to contribute to the variations seen in preterm birth rates.

The variations in the rates of births with nonspontaneous onset in twins between countries are consistent with the corresponding rates in singleton births. This shows that policies for managing the end of pregnancy vary considerably from one country to another, even between countries such as Northern Ireland and Scotland which are close to each other and use common definitions. These wide differences raise questions about the implications of these practices for the health of children.

In some countries, it has been suggested that the rise in preterm birth rates in twins has been accompanied by a decrease in stillbirth and neonatal death rates and in neonatal morbidity.4,29 This association would support the hypothesis that the rise in preterm birth is a consequence of changes in the approaches of obstetricians and neonatologists to managing complications of pregnancy, in the light of increasing survival rates among preterm babies. It is not possible to compare variations in mortality in relation to preterm birth rates in our study as the numbers of children from multiple births are too small in most countries. Measures of longer term outcomes, such as rates of cerebral palsy, are also essential. Although it is difficult to compare cerebral palsy rates between countries because of differences in case definition, comparisons between trends in cerebral palsy and trends in preterm birth within countries with cerebral palsy registers can yield some insights. A recent collaborative study using data from long-standing European registers showed that rates of cerebral palsy had not increased for twins born in the 1980s.30

The high rate of births with nonspontaneous onset in some countries raises questions about the optimal gestational age for delivery and the most appropriate method of delivery for multiple births. Observational studies seem to show that mortality and neonatal morbidity are minimal if twins are born at 37 or 38 weeks.31,32 On the other hand, there is currently no evidence in favour of recommending a planned caesarean delivery for twins at term33 or after 32 weeks of pregnancy, except possibly if the first twin is in a breech presentation.34 A randomised trial of routine elective caesarean section for twins is under way and should provide an answer to this question.


Wide differences were found in rates of preterm birth and in births with nonspontaneous onset. This shows that the circumstances in which twins are born varies considerably from one western European country to another. The data suggest that clinicians in the participating European countries have very different approaches to the management of twin pregnancies. It is not currently known whether the very active approaches adopted in countries such as Northern Ireland or Flanders lead to better outcomes than the different policies pursued in countries such as Sweden and Finland.


The PERISTAT project was cofinanced by the European Commission Health Monitoring Programme.

The PERISTAT study group: Steering committee

J. Zeitlin, K. Wildman, G. Bréart, France (project coordinators); S. Alexander, Belgium; H. Barros, Portugal; B. Blondel, France; S. Buitendijk, the Netherlands; M. Gissler, Finland; A. Macfarlane, UK.

The PERISTAT study group: Scientific advisory committee

C. Bakoula, Greece; F. Bolumar, Spain; J. Bottu, Luxembourg; S. Cnattingius, Sweden; M. Cuttini, Italy; P. Defoort, Belgium; J.-B. Gouyon, France; L. Krebs, Denmark; W. Künzel, Germany; N. Lack, Germany; M. Langer, Austria; J. Langhoff-Roos, Denmark; G. Lindmark, Sweden; S. Marchant, UK; N. Montenegro, Portugal; F. Morcillo, Spain; M. Newburn, UK; JG. Nijhuis, the Netherlands; S. Prati, Italy; A. Staines, Ireland; M. Virtanen, Finland; N. Vitoratos, Greece; C. Vutuc, Austria; Y. Wagener, Luxembourg.


Table Appendix..  Data sources used for constructing Tables 2–4
CountryCoverage, if not nationalData source*Year
  • *

    For complete information on data sources see the PERISTAT report.7

  • **

    Bayern, Baden-Württemberg, Berlin, Hessen (data from 2001), Niedersachsen and Bremen, Nordrhein, Sachsen, Thüringen, Westfallen-Lippe, representing 72.6% of all births.

Austria Statistics Austria2001
BelgiumFlandersSPE (Studiecentrum voor Perinatale Epidemiologie)2000
Denmark Danish perinatal database2000
Finland Medical birth registry—STAKES2000
FranceRepresentative sampleNational Perinatal Survey1998
GermanyNine Bündeslander**BAQ—perinatal survey2000*
Ireland National Perinatal Reporting System1999
Italy ISTAT, Civil birth and death registration. Discontinued in 19981998
Netherlands Merged database from professional registers. LVR: data on course of pregnancy and delivery. LNR: diagnoses of the child, duration of hospital stay, treatments1999
Sweden Medical Birth Register2000
UKScotlandInformation and Statistics Division, SMR2 Maternity Discharge Sheet2000
UKNorthern IrelandPerinatal Information, Northern Ireland, aggregated data from child health systems2000