- Top of page
Objective To test whether being small for gestational age, defined as having a birthweight less than the 10th centile of intrauterine growth references, is a risk factor for preterm delivery for singleton live births.
Design A case-control study.
Setting Maternity hospitals in 16 European countries.
Sample Four thousand and seven hundred preterm infants between 22 and 36 completed weeks of gestation and 6460 control infants between 37 and 40 weeks of gestation.
Methods Newborn babies are identified as being small for gestational age using customised reference standards derived from models of fetal growth. The impact of being small for gestational age on preterm delivery is estimated using logistic regression.
Main outcome measure Spontaneous or induced preterm delivery.
Results Being small for gestational age is significantly associated with preterm birth, although the magnitude of this association differs greatly by type of delivery and gestational age. Over 40% of induced preterm births for reasons other than the premature rupture of membranes are small for gestational age compared with 10.7% of control infants (OR 6.41). For spontaneous or premature rupture of membranes related preterm births, the association is also significant, but weaker (OR 1.51). The relationship between growth restriction and preterm delivery is strongest for preterm births before 34 weeks of gestation.
Conclusions These findings highlight the phenomenon of abnormal fetal growth in all premature infants and, in particular, infants delivered by medical decision for reasons other than premature rupture of membranes. The observed association between being small for gestational age and preterm delivery among spontaneous preterm births merits further attention because the causal mechanisms are not well understood.
- Top of page
The previous use of birthweight thresholds to identify premature newborns discouraged reflection on the possible links between intrauterine growth restriction and preterm delivery. As the use of ultrasound increased confidence in gestational age estimates, empirical studies distinguished the risks associated with being small for gestational age from those associated with low gestational age1–5, but did not consider the possibility that intrauterine growth restriction was related to preterm birth itself. Commonly used definitions of intrauterine growth restriction by gestational age are derived from samples of live born infants and do not include information on the growth of non-preterm infants who are still in utero.
Nonetheless, the supposition that preterm newborns deviate from normal fetal growth is common among researchers developing fetal growth standards and comparing fetal growth by ultrasound to observed birth-weights6–8. Several studies have compared actual birthweights and measurements, such as biparietal diameter and abdominal circumference, among preterm infants to references for intrauterine growth and have found that preterm infants have lower values than expected in normal fetuses8–12.
The aim of this paper is to analyse the relationship between intrauterine growth restriction and preterm birth using data from the European Program of Occupational Risks and Pregnancy Outcome (EUROPOP) case-control study of the determinants of preterm birth, undertaken between 1994 and 1997 in 17 European countries13. Models have been developed to estimate birthweight from ultrasound measurements of normal fetuses in utero and they can be used to describe the pattern of fetal growth6,8,12,14. In this paper we use methods developed by Gardosi et al.14 based on Hadlock's fetal growth formula15 to identify small for gestational age fetuses. The method of Gardosi et al. also incorporates biological factors known to affect variations in normal birthweight such as the sex of the infant and the height, weight and parity of the mother in order to improve the accuracy of thresholds for identifying growth restriction16–18. Once we identify small for gestational age fetuses, we test whether or not they have an increased risk of being delivered preterm, either spontaneously or by medical decision.
- Top of page
Between 1994 and 1997, the EUROPOP study was carried out in 17 European countries using the same case-control design and questionnaire. The principal aim of the study was to explore the relationship between occupational risk factors and preterm birth. The study and its methods have been described elsewhere13,19,20. Participating countries were: the Czech Republic; Finland; France; Germany; Greece; Hungary; Ireland; Italy; the Netherlands; Poland; Romania; Russia; Scotland (UK); Slovenia; Spain; Sweden; and Turkey. Data from Turkey are not used in this analysis because the control population was incomplete. Between one and thirteen maternity hospitals from each country participated in the study.
Cases included all preterm singleton live and stillbirths, defined as births between 22 and 36 completed weeks of amenorrhoea, born in participating maternity hospitals during the study period. Controls were randomly selected from singleton births. Every 10th singleton live birth or stillbirth at 37 or more completed weeks of gestation in each maternity unit was included in the sample. In The Netherlands, the interviews took place at home within the first month due to the high rates of home delivery in this country. Information about the mother, her pregnancy and the newborn were collected using the same questionnaire in all countries. Interviewers administered the questionnaire to the mother and abstracted information from her medical records.
Gestational age is based on obstetrical estimates, which are derived from the last menstrual period and ultrasound measures. Of approximately 90% of the sample with data on the timing of the first ultrasound scan (89.5% controls; 87.2% of cases), 64% of cases and 65% of controls had an ultrasound before 15 weeks and 87% of cases and 86% of controls had at least one ultrasound before 21 weeks.
The analyses presented below only use data on live born infants, since the time of death of the fetus was not recorded. Moreover, it is commonly thought that the dead fetus begins a process of weight loss in the hours following death. Malformed infants are also excluded from the analyses (16 controls and 32 cases). Fifty-three cases and 44 controls were excluded because gestational age estimates (obstetric, ultrasound and last menstrual period) were discordant and there was a doubt about their case or control status. Birthweight outliers were identified by detecting extremes of each birthweight distribution by gestational age. Most outliers were clearly erroneous (i.e. over 3.5 standard deviations from the mean). In some cases, when there was a gap of more than 20% between an extreme value and the previous value, the case was excluded even if it was within 3 standard deviations of the mean. This was done for the earlier gestational ages with fewer observations and thus less accurate estimates of the distribution and the mean. The data cleaning process resulted in 35 exclusions for cases and 22 for controls. Data on birthweight were already missing for 13 controls and 39 cases. Maternal height, weight, and parity, which are necessary for developing the growth reference standards, were missing for 5% of cases and 5% of controls. The final sample of live born infants includes 4700 cases and 7827 controls.
Cases are divided into two groups: 1. spontaneous preterm births and preterm births after premature rupture of membranes (n= 3526), and 2. induced preterm births by medical decisions for reasons other than premature rupture of membranes (n= 1174, described as ‘othIND’). This division is necessary because it is not possible to evaluate if intrauterine growth restriction is part of a biological process associated with preterm delivery in the second group. These pregnancies could have ended in a preterm live birth, a preterm stillbirth, a term live birth, or a term stillbirth in the absence of a decision to induce delivery. The control population for comparative analyses of growth restriction excludes term births after 40 weeks because of the difficulty of using fetal growth curves to define small for gestational age infants at 41 and 42 weeks, as described below (control population between 37 and 40 weeks, n= 6460).
For the comparison between the proportions of small for gestational age newborns between cases and controls χ2 tests are used. Unadjusted and adjusted odds ratios of the impact of being small for gestational age on preterm birth are obtained using logistic regression. Covariates are selected in accordance with previous analyses of the impact of obstetrical, medical and socio-economic factors on preterm delivery in this sample19,20.
Defining small for gestational age newborns
The adjustable fetal weight standards developed by Gardosi et al.14 are used to identify small for gestational age newborns. Small for gestational age newborns are defined as having a birthweight below the 10th centile of these reference standards. This method incorporates two distinct principles. The first is to model the growth trajectory of the fetus as a proportion of average birth-weight at 40 weeks. Gardosi et al. base this model on Hadlock's formula15 for fetal growth established from ultrasound measures from 10 to 40 weeks of gestation. Observed variation at term in their population of ultrasound dated singleton births in Nottingham makes it possible to define a coefficient of variation which is used to derive 10th and 90th centiles. Using this method, Gardosi et al. derive 10th, 50th and 90th centiles for fetal growth at all gestational ages. The references for fetuses born at 41 and 42 weeks are not based on observed fetal weights, however, since Hadlock's formula was only developed on fetuses before 41 weeks of gestation. The references for 41 and 42 week fetuses are derived by extrapolating the model beyond 40 weeks.
The second component of their method is to adjust these standards to reflect biological characteristics that are known to affect birthweight. They estimate the impact of maternal parity, weight, height, ethnic group, smoking and sex of the newborn on average birthweight at term using a linear regression model in the Nottingham population of 38,114 singleton births between 259 and 294 days of gestation. The coefficients from this regression make it possible to predict the average weight of a fetus at 40 weeks of gestation given its sex and its mother's characteristics. The mother's smoking status is not taken into consideration for generating predicted weights which are based on the model for nonsmokers. The proportional growth curves can be applied to this predicted weight to derive customised reference centiles at all gestational ages.
Observed birthweights by gestational age groups from the EUROPOP study are presented in Table 1. The birthweight of 3490 g at 40 weeks in this sample is lower than in Gardosi's Nottingham sample (3532 g), and the standard deviation is higher, 435 g versus 389 g. This latter result is not surprising as the variation in the EUROPOP sample, which includes women from 16 European countries, would be expected to be greater.
Table 1. Birthweight (g) in the EUROPOP sample by gestational age (GA), live births only.
| || ||Birthweight (g)|
|GA group (weeks)||n||10th centile||Median||90th centile||Mean(SD)|
Table 2 presents the results of Gardosi's model of birthweight on maternal characteristics for singleton births between 37 and 42 weeks of gestation using the EUROPOP data; the coefficients derived by Gardosi on his Nottingham sample are also shown. Most of the coefficients derived from these two samples are not significantly different, with the exception of maternal weight and the impact of smoking for the category 10–19 cigarettes/day. These differences in the coefficients have little impact on the classification of small for gestational age infants: 99.3% of the EUROPOP sample are similarly classified whether we use Gardosi's coefficients from his Nottingham sample or those derived from the EUROPOP sample (99.2% for the controls and 99.5% for cases).
Table 2. Models of the impact of maternal and infant characteristics on birthweight at term (Gardosi's model specification adjusted for gestational age at 280 days).
|Variable||EUROPOP data ß(95% CI)||Gardosi's Nottingham data ß*|
|Maternal height (cm)||8.8 (7.1, 10.4)||7.8|
|Maternal weight (g)||7.2 (6.0, 8.4)||8.7|
|Maternal weight (g) squared||−0.135 (422, −0.05)||−0.117|
|Maternal weight (g) cubed||0.0014 (−0.0003, −0.0032)||0.00072|
|Parity (ref = 0)|| || |
| 1||87 (66, 108)||108|
| 3||128 (80, 176)||150|
| 4||140 (88, 191)||150|
|Sex of child (+ for a boy, – for a girl)||66 (56, 74.4)||58.4|
|Smoking (ref = 0)|| || |
| < 10 cigarettedday||−157 (−188, −125)||−153|
| 10–19 cigaretted/day||−128(−167, −89)||−215|
| ≥ 20 cigaretteds/day||−182(−252, −113)||−246|
Although Gardosi's model adjusts birthweight references for ethnic origin, we decided not to include country of residence for the EUROPOP analyses because of the relatively small sample sizes in each country and the conceptual difficulties of defining ethnic groups. The variable ‘country of residence’ is included as a control in logistic regressions.
- Top of page
Table 3 presents the percentage of small for gestational age live born infants in the EUROPOP sample by type of delivery and gestational age. In the sample of infants at term, slightly more than 10% are identified as small for gestational age.
Table 3. Percent of live born infants c 10 centile of growth reference standards* (small for gestational age) by type of delivery. Values are given as n (%). GA = gestational age; PROM = premature rupture of membranes; NS = not significant.
|GA groups (weeks)||All births||Spontaneous deliveries and PROM||Induced deliveries not associated with PROM|
|22–24||80 (15.0)||73 (13.7)||7 (28.6)|
|25–27||212 (27.4)||164 (16.5)||48 (64.6)|
|28–30||432 (29.9)||306 (19.3)||126 (55.6)|
|31–33||941 (30.0)||652 (19.6)||289 (53.3)|
|34–36||3035 (19.6)||2331 (15.2)||704 (34.2)|
|39–40||4771 (11.0)||3878 (10.7)||888 (12.2)|
Preterm infants in this sample are more likely to be small for gestational age than term infants. Twenty-three percent of preterm infants between 22 and 36 weeks of gestation are below the 10th centile of the reference standards. However, this proportion differs significantly by both mode of delivery and gestational age. The highest proportion of small for gestational age infants are found between 28 and 33 weeks of gestation when approximately one-third of all preterm infants are classified as small for gestational age. The proportion of small for gestational age infants is lowest among infants in the 22 to 24 week group (15%) and in the 34 to 36 week group (19.6%)
Across all gestational ages, a greater proportion of infants delivered preterm by medical decision not associated with premature rupture of membranes are classified as small for gestational age than preterm infants born after spontaneous labour or premature rupture of membranes. Between 25 and 33 weeks, over half of induced preterm births not associated with premature rupture of membranes are small for gestational age. This proportion is lower, but still high for the group of later preterm births between 34 and 36 weeks, which is 34%. The prevalence of small for gestational age in induced and spontaneous term deliveries is similar for infants born at 39 to 40 weeks. There is a small, yet statistically significant (P= 0.044), difference of 3% in the prevalence of small for gestational age by mode of delivery for the group of term infants born between 37 and 38 weeks.
Small for gestational age is more prevalent among spontaneous preterm births and those following premature rupture of membranes when compared with term infants, but this difference is less marked than that observed for induced preterm births not associated with premature rupture of membranes. Nonetheless, between 28 and 33 weeks about 20% of the preterm infants in this group are small for gestational age.
Table 4 presents odds ratios for preterm birth when the fetus is small for gestational age for all births and by mode of delivery. Preterm births are also subdivided by gestational ages. The control population for this analysis is live births between 37 and 40 weeks of gestation. The models in column 1 displays odds ratio adjusted for country of residence. The models in column 2 present odds ratio adjusted for the set of variables previously identified as significant correlates of preterm birth in this sample. Variables included in the model are: maternal age; obstetrical history; age at completion of schooling; marital status; smoking during pregnancy; body mass index; and country of residence.
Table 4. Odds ratios for preterm birth for fetuses with a birthweight < loth centile of reference standards* by type of delivery and gestational age (GA) group. Control population are all live births between 37 and 40 weeks of gestation. Values are given as n or OR (95% CI). OR = odds ratio; CI = confidence interval; PROM = premature rupture of membranes.
|GA (weeks)||OR adjusted for country of residence (n= 6460 controls)||OR adjusted for country of residence and socio-demographic risk factors † (n= 6396 controls)‡|
|All preterm births||4700||4658|
| 22–36||2.44 (2.20–2.72)||2.33 (2.09–2.60)|
| By GA group|| || |
| 22–27||2.69 (2.32–3.58)||2.44 (1.81–3.30)|
| 28–33||3.60 (3.12–4.16)||3.57 (3.07–4.14)|
| 34–36||2.05(1.82–2.31)||1.93 (1.70–2.18)|
|Spontaneous preterm + PROM||3526||3493|
| 22–36||1.61 (1.43–1.82)||1.51 (1.33–1.71)|
| By GA group|| || |
| 22–27||1.53 (1.06–2.21)||1.36 (0.91–1.99)|
| 28–33||2.03 (1.70–2.45)||1.98 (1.64–2.39)|
|Other induced preterm births||1174||1165|
| 22–36||6.41 (5.53–7.43)||6.38 (5.47–7.45)|
| By GA group|| || |
| 22–27||13.83(7.94–24.08)||14.88 (8.25–26.86)|
| 28–33||10.25 (8.24–12.75)||10.48(8.35–13.15)|
| 34–36||4.58 (3.82–5.50)||4.56(3.77–5.51)|
Adjusting for covariates does not have a significant impact on the odds ratios. For spontaneous preterm births, the odds ratios decrease very slightly when other variables are included in the model. This would be expected if being small for gestational age was associated with other variables which were more directly related to preterm delivery. The adjusted odds ratio for induced preterm deliveries not associated with premature rupture of membranes stay at the same level.
The overall adjusted risk of preterm delivery for a small for gestational age fetus is 2–3, although the odds ratios differ greatly by type of delivery and gestational age. The adjusted risk of spontaneous preterm birth and premature rupture of membrane preterm delivery for a small for gestational age fetus is estimated at 1.51. The risk is highest for spontaneous or premature rupture of membrane-associated preterm births between 28 and 33 weeks of gestation, when the odds ratio for being small for gestational age is approximately 2. Extremely early preterm births (22 to 27 weeks) and late preterm births (34 to 36 weeks) have an odds ratio of 1.4. The magnitude of the odds ratio for induced births for reasons other than premature rupture of membranes is much higher across all gestational ages. For all induced births, the odds ratio associated with being small for gestational age is 6.4, with a high of 14.9 in the earliest gestational age group and 4.6 for induced deliveries which occur between 34 and 36 weeks of gestation.
- Top of page
Our comparisons of preterm and term infants support the hypothesis that there is an excess of small for gestational age infants among live born preterm infants. Twenty-three percent of preterm infants are below the 10th centile of fetal growth standards.
The appropriate threshold for defining growth restriction is a subject of debate. Most analyses rely on establishing a weight cut off by gestational age, with the knowledge that not all fetuses identified as small for gestational age are growth restricted. The 10th centile is the conventional cut off for defining small for gestational age, although relative increases in fetal mortality have been noted in infants under the 15th centile of growth reference standards21, and other analyses use stricter definitions. A further difficulty is the potential variation in the critical threshold with duration of gestation. One analysis found that the risk of fetal death is associated with increasingly smaller deviations from normal growth as fetuses approach term22. Although these issues are important for the use of growth norms, they could not be addressed using the present data which do not include morbidity or mortality outcomes. We adopted the 10th centile for defining small for gestational age infants because this convention was integrated into our analytic model and because it provided sufficient statistical power to test for the presence of an association with preterm birth. In addition, by taking into consideration biological factors, such as maternal weight, height and the sex of the infant, the methods proposed by Gardosi improve the identification of infants that are small for their genetically determined growth potential.
The proportions of small for gestational age preterm infants differ greatly by type of delivery, and there are clearly two distinct, but related, phenomena that underlie the observed relationship between small for gestational age and preterm birth. The proportion of small for gestational age infants is highest for induced deliveries not preceded by premature rupture of membranes. The odds of being delivered preterm by medical decision is multiplied by 6 for growth restricted infants. These findings do not come as a surprise since intrauterine growth restriction is an indication of the failure of the fetus to thrive in utero and is one indication for the medical decision to induce delivery before term. Moreover, many pathological maternal conditions, such as hypertension, which lead to induction of labour, are associated with intrauterine growth restriction23.
The magnitude of the relationship between fetal growth restriction and induced preterm births is noteworthy, however. Induced preterm births are an important and increasing proportion of the preterm population. In France, data from representative sample surveys of births in 10 regions show that induced deliveries rose from 7.3% of preterm births in 1972 to 26.8% in 1988/1989. Induced preterm deliveries not associated with premature rupture of membranes increased from 3.8% to 29.4% in singleton live births from two birth cohorts in Finland in 1966 and 198625. A recent review of other population based studies on indicated preterm births from the United States and Europe found these births ranged from 21% to 37% of all preterm births23. The growth patterns of this sizeable and increasing group of preterm births are distinct from those resulting from spontaneous preterm labour or premature rupture of membranes and place these newborns at a higher risk of adverse outcomes.
This analysis also found significantly more small for gestational age infants among preterm births after spontaneous labour or premature rupture of membranes than among term births. The effect on these preterm births is strongest between 28 and 33 weeks of gestation when small for gestational age is associated with an almost twofold increase in preterm delivery. This result provides evidence to support an association between fetal growth restriction and spontaneous preterm delivery.
There are several reasons to expect a relationship between intrauterine growth restriction and spontaneous preterm delivery. The first reason centres on hormonal pathways. One mechanism, for which support is accumulating, is the role of corticotrophin-releasing hormone as a regulator of parturition both at term and preterm26. Maternal peripheral corticotrophin-releasing hormone has been found to be predictive of preterm labour in the second trimester27. Maternal plasma corticotrophin-releasing hormone concentrations are elevated in pregnancy induced hypertension and intrauterine growth restriction, and a relationship between maternal and fetal stress and increased levels of placental corticotrophin-releasing hormone has been postulated26,28–30. Preterm labour has also been associated with decreased inactivation of prostaglandins through a reduction of prostaglandin dehydrogenase activity31. The downward regulation of this enzyme could be related to increased corticotrophin-releasing hormone or the rise in fetal cortisol secretion26. These hormonal processes may be associated with other adaptive strategies to reduced nutrient availability observed among growth retarded infants, such as advanced neurological, pulmonary and placental maturation32,33.
A more speculative hypothesis involves in utero infections34. Certain infections are related to both preterm labour and intrauterine growth restriction35. Placental insufficiency may also generate fetal immune abnormalities which increase vulnerability to infection36. A final hypothesis is that the tensile strength of the amnion and chorion is diminished by the same conditions that retard fetal growth, thereby reducing the strength of the fetal membranes which contributes to premature rupture of membranes and preterm delivery37.
The true magnitude of the association between growth restriction and spontaneous preterm delivery is probably underestimated in our analyses for two reasons. The first is selection bias introduced by the medical practice of induction. When induction is used frequently in situations of growth restriction, it is no longer possible to evaluate with confidence the strength of the association between intrauterine growth restriction and spontaneous preterm delivery since the outcome for induced small for gestational age infants in the absence of medical intervention is unknown.
The second source of bias is the exclusion of stillbirths from this analysis. We excluded stillbirths because of the technical difficulties of establishing the exact time of death. Moreover, for deaths preceding the onset of labour, delivery before term can be viewed as a consequence of the fetal demise whether it occurs by natural processes or by induction. However, the exclusion of fetuses who were alive when labour started constitutes a potentially important source of bias for the assessment of the relationship between fetal growth restriction and spontaneous preterm birth, especially at earlier gestational ages. In the EUROPOP sample, stillbirths represented 37.1% of all preterm births before 28 weeks, 9.1% between 28 and 33 weeks, and 1.9% between 34 and 36 weeks. Intrapartum stillbirths constitute a much higher proportion of total stillbirths at earlier gestational ages; one study from the French district of Seine-Saint-Denis found that perpartum deaths represented 58.1% of all stillbirths at less than 28 weeks of gestation versus 16.2% between 28 and 36 weeks of gestation38. Most studies of preterm stillbirths have found a high proportion of small for gestational age fetuses5,22,39. Thus, the inclusion of intrapartum deaths in this analysis would most likely reinforce the association between growth restriction and spontaneous preterm birth, especially in the lower gestational age groups. In fact, selection bias due to fetal death may explain the relatively small prevalence of small for gestational age among spontaneous preterm live births between 22 and 27 weeks of gestation.
The validity of these results depends on the accuracy of this model of fetal growth. The conclusion that there is a significant proportion of growth restricted infants among indicated preterm births would seem robust to the choice of models, given the magnitude of the odds ratios for this group. However, for the findings concerning spontaneous preterm births where odds ratios, although significant, range between 1.4 and 2.0, we must be more cautious.
The model of Gardosi et al.14 of fetal growth is based on three assumptions:
different fetuses will follow a similar pattern but different growth velocities to reach their respective birthweights at the end of a normal pregnancy;
the effects of covariates (maternal characteristics) are consistent at different gestations and apply similarly for different subgroups; and
the coefficient of variation remains constant at all gestations.”
Mongelli and Gardosi40 have published evidence to support the first hypothesis. Support for the third comes from ultrasound studies of fetal growth6,12. There is little empirical evidence to back the second hypothesis. In fact, one study of a high risk population found that in utero weight differences associated with maternal stature were apparent from 30 weeks onward, but that associations with fetal sex and maternal parity appeared later41.
Other studies, however, have obtained similar results to ours using different models. In the study by Secher et al.11 birthweight distributions derived from preterm births were significantly different from ultrasound measurement of term births at the same gestational ages, especially for early preterm births. At 32 weeks, they concluded that the 10th centile of fetal growth norms corresponds to the 25th centile of norms derived from the weight for gestational age chart based on live births. This finding is similar to ours: the proportions of small for gestational age infants are higher in early preterm births (28–32 weeks), when about one-third are classified as small for gestational age. In Ott's study8, the proportion of infants below the 10th centile of his intrauterine growth curves varies between 33% and 27% between 31 and 33 weeks, respectively, and ranges between 23% and 12% between 34 and 36 weeks, respectively. Finally, the Swedish intrauterine growth curves developed from serial ultrasound fetometry published by Marsal et al.12 found that 32% of all preterm infants < 30 weeks are less than 2 standard deviations below mean birthweight compared with 11% for infants between 30 and 36 weeks and 2.6% of infants at term. A comparison between the fetal growth model developed by Gardosi et al. and the Swedish references found that the 10th centile projected by the Swedish model was between 4% and 8% higher for gestational ages < 37 weeks, than the 10th centile derived using the model of Gardosi et al. adjusted to reflect the mean birthweight in the Swedish population. The 10th centiles between 37 and 40 weeks were almost identical. While this discrepancy is a reminder that the actual value for the 10th centile fluctuates from model to model, this direct comparison with the Swedish references does not suggest that the model used here overestimates growth restriction.
Our findings support the development and use of intrauterine fetal growth curves to monitor fetal growth in the preterm period instead of postnatal growth curves derived from the population of preterm newborns. They also highlight the phenomenon of abnormal fetal growth among both induced and spontaneous preterm births. The association between small for gestational age and spontaneous preterm birth is of interest because a better understanding of its biological explanations could shed light on the aetiological mechanisms of preterm birth. Greater attention to the subgroup of growth restricted preterm infants in epidemiological studies is also necessary to determine if they have distinct risk factors or differential responses to prevention efforts42.
The authors wish to thank the members of the EUROPOP group listed below and express their appreciation to the interviewers and technical teams in each participating country. We are also grateful for the computer and analysis assistance provided by Ms N. Lelong. The EUROPOP study was financed by the European Union BIOMED project BMH1-CT94-1041.
Members of the EUROPOP Steering Committee
G. C. Di Renzo (Project Leader; Perugia, Italy); G. Bréart (Paris, France); E. Papiernik (Paris, France); Lord N. Patel (Dundee, UK); M. J. Saurel-Cubizolles (Villejuif, France); D. Taylor (Leicester, UK); S. Todini (Perugia, Italy).
Members of the EUROPOP National Staff
M. Kudela & M. Vetr (Olomouc, Czech Republic); A. Heikkilä, R. Erkkola & J. Forström (Turku, Finland); E. Papiernik & P. Lucidarme (Paris, France); J. Tafforeau (Brussels, Belgium); W. Künzel & J. Herrero-Garcia (Giessen, Germany); J. Dudenhausen & W. Henrich (Berlin, Germany); A. Antsaklis & G. Haritatos (Athens, Greece); L. Kovacs, T. Nyari & G. Bartfai (Szeged, Hungary); C. O'Herlihy, J. Murphy & H. Stewart (Dublin, Ireland); G. C. Di Renzo, P. L. Bruschettini & P. Moscioni (Perugia, Italy); E. Cosmi; A. Spinelli & D. Serena (Rome, Italy); G. H. Breborowicz & A. Anholcer (Poznan, Poland); F. Stamatian (Cluj, Romania); A. V. Mikhailov (St. Petersburg, Russia); M. Pajntar, M. Pirc & I. Verdenik (Ljubljana, Slovenia); V. Escribà-Aguir (Valencia, Spain); J. M. Carrera (Barcelona, Spain); K. Marsal & H. Stale (Malmö, Sweden); S. Buitendijk, K. van der Pal (Leiden, The Netherlands); H. van Geijn (Amsterdam, The Netherlands); O. Gökmen, C. Güler & T. Caglar (Ankara, Turkey); P. Owen (Dundee, UK).