Dr EJWM Troe, The Generation R Study Group (Ae-029), Erasmus Medical Centre, PO Box 2040, 3000 CA Rotterdam, The Netherlands. Email firstname.lastname@example.org
Objective To examine whether differences in birthweight of various ethnic groups residing in the Netherlands can be explained by determinants of birthweight.
Design Population-based birth cohort study.
Setting Data of pregnant women and their partners in Rotterdam, the Netherlands.
Population We examined data of 6044 pregnant women with a Dutch, Moroccan, Turkish, Capeverdean, Antillean, Surinamese-Creole, Surinamese-Hindustani and Surinamese-other ethnic background.
Methods Regression analyses were used to assess the impact of biomedical, socio-demographic and lifestyle-related determinants on birthweight differences.
Main outcome measure Birthweight was established immediately after delivery in grams.
Results Compared with mean birthweight of offspring of Dutch women (3485 g, SD 555), the mean birthweight was lower in all non-Dutch populations, except in Moroccans. Differences ranged from an 88-g lower birthweight in offspring of the Turkish women to a 424-g lower birthweight in offspring of Surinamese-Hindustani women. Differences in gestational age, maternal and paternal height largely explained the lower birthweight in the Turkish, Antillean, Surinamese-Creole and Surinamese-other populations. Differences in birthweight between the Dutch and the Capeverdean and Surinamese-Hindustani populations could only partly be explained by the studied determinants.
Conclusions These results confirm significant differences in birthweight between ethnic populations that can only partly be understood from established determinants of birthweight. The part that is understood points to the importance of determinants that cannot easily be modified, such as parental height. Further study is necessary to obtain a fuller understanding.
Differences in birthweight between populations of different races or ethnic groups are well documented.1–3 In the USA, several studies have shown that the proportion of low birthweight (<2500 g) is elevated in Black population.4–7 Similar findings are seen in the UK in infants of South-Asian, Black-Caribbean and Black-African descent compared with UK-born white mothers.8 Also, in the Netherlands, studies have shown differences in birthweight and the proportion of low birthweight between the non-Dutch and Dutch populations.9,10 Birthweight is strongly related to perinatal and infant mortality.11,12 The risk of adverse outcomes is seen not only in those with a low birthweight but also in the broad spectrum of birthweight. Additionally, birthweight is also associated with morbidity and mortality in later life; for example, associations of low birthweight with diseases during adulthood, such as diabetes mellitus and cardiovascular disease, have been observed.13,14
The lower mean birthweight and higher proportion of low birthweight among minority populations could be the result of various determinants.15,16 Several determinants have been identified that are associated with birthweight, such as gestational age,17 parity,18 socio-economic status,19 marital status,20 maternal age,21,22 maternal height, body mass index (BMI), pre-pregnancy weight,23,24 smoking25 and alcohol use.26
These determinants of birthweight vary across ethnic populations. It is still unclear to what extent the lower birthweight of ethnic minority populations can be explained by these determinants. It might be that a lower birthweight in ethnic populations for a large part is related to unchangeable genetic determinants. By the study of determinants of birthweight in a variety of ethnic groups, we aim to help understand which genetic, physiological, socio-demographic and lifestyle determinants explain ethnic differences in birthweight.
The Generation R study is multi-ethnic, population-based birth cohort study, with a large sample size, in which detailed information about a large number of potential determinants is available. Within the Generation R Study, we studied if biomedical, socio-demographic and lifestyle-related determinants explain differences in birthweight between the Dutch and the non-Dutch populations.
This study is embedded in the Generation R Study, a prospective population-based cohort study from fetal life until young adulthood. The Generation R study is designed to identify early environmental and genetic determinants of growth, development and health in fetal life, childhood and adulthood and has been described previously in detail.27,28 Briefly, all pregnant women and their partners in a previously defined area in Rotterdam, the Netherlands, were approached at their first antenatal visit. Most women spoke Dutch; if not, the study was explained and questionnaires were available in their own language.28 In total, 8880 pregnant women with a delivery date between April 2002 and January 2006 enrolled in the prenatal part of the study. Data in pregnancy were collected from physical examinations, fetal ultrasounds and questionnaires. The partners of pregnant women had one physical examination at enrolment and received one questionnaire. Pregnant women were usually seen for the first time before the 18th week of the pregnancy and in total three times during pregnancy, in early pregnancy (gestational age < 18 weeks), mid-pregnancy (gestational age 18–25 weeks) and late pregnancy (gestational age ≥ 25 weeks) in a research setting. The individual time scheme of these assessments depended on the specific gestational age at enrolment. The children form a prenatal recruited birth cohort, thus the overall response of the study can be calculated at birth. Of all eligible children in the study area, 61% participated at birth in the study.28 The Medical Ethics Committee of the Erasmus Medical Centre approved the Generation R Study. Written informed consent was obtained from all participants.
Ethnic background of the pregnant women was assessed by country of birth of herself and her parents. Information about countries of birth was obtained by questionnaire. The participating pregnant woman was of non-Dutch ethnic origin if one of her parents was born abroad (definition Statistics Netherlands29). If both parents were born abroad, the country of birth of the participant’s mother decided on the ethnic background. Besides women of Dutch ethnic background, a distinction was made among the non-Dutch minority populations in this study: Moroccan, Turkish, Capeverdean, Antillean and Surinamese. Women with an ethnic background other than these were grouped as ‘other-western’ for European, North American, Oceanean, Japanese and Indonesian, and as ‘other non-western’ for African, Asian (except Japanese and Indonesian) and South- and Central American. Women with a Surinamese background are of mixed ethnic origin, mainly consisting of Hindustanis originating from India and Creoles from Africa. These women were asked about their ethnic origin and further classified as: Surinamese-Hindustani, Surinamese-Creole or Surinamese-other.
Birthweight of the infant was obtained from community midwifery and hospital registries. Birthweight was established directly postpartum and expressed in grams.
Gestational age was established by fetal ultrasound examination during the first ultrasound visit.30 Maternal and paternal height were measured at time of enrolment.
Information about maternal age, marital status, educational level and parity was obtained by questionnaires. Maternal age was assessed at enrolment in the study and was a continuous variable. Marital status of the pregnant woman was classified into three categories: (1) married, (2) cohabiting and (3) single parent. Educational level of the pregnant woman was assessed by the highest completed education and reclassified into three categories: (1) primary school, (2) secondary school and (3) higher education. Parity was classified in two categories: (1) nulliparous and (2) multiparous.
Maternal weight was measured without shoes and heavy clothing at time of enrolment. BMI at early gestation was calculated from maternal weight and maternal height (weight/height2 [kg/m2]) and adjusted for gestational age at intake. Maternal smoking and alcohol use were assessed by questionnaires by asking pregnant women whether they smoked/consumed alcohol during pregnancy (yes/no).
Population for analysis
For the present analyses, data of all prenatal enrolled women were available (n = 8880). Women with missing data on their ethnic background (n = 1102) were excluded from analyses. Women with twin pregnancies (n = 94), abortion or intrauterine death (n = 95) and perinatal death (n = 39) were excluded from analyses. Also excluded were women who withdrew from the study (n = 2) and were lost to follow up (n = 109). Of the remaining 7439 pregnant women, maternal height and maternal weight were missing in 66 subjects, who were excluded from analyses. Analyses were carried out in pregnant women with a Dutch, Turkish, Moroccan, Surinamese-Creole, Surinamese-Hindustani, Surinamese-other, Antillean and Capeverdean ethnic background (n = 6044). The results of women with an ‘other-western’ and ‘other non-western’ ethnic background were not presented because of the mixed composition of these populations (n = 1329). The analyses with paternal height were restricted to the population of pregnant women with whom the partner participated in the study (n = 4471).
The non-Dutch populations under consideration were compared with the Dutch population (reference). Differences in baseline characteristics of the non-Dutch populations were compared with the Dutch population using the chi-square statistic for categorical variables and analysis of variants for continuous variables. Univariate linear regression analysis was used to study the associations of ethnic background with birthweight. Gender of the infant was controlled for in all analyses as potential confounder. To study the individual effect of each potential determinant, each determinant was separately tested. We used separate categories for missing information on the determinants, which were added to the model. A bootstrap analysis was used to calculate a confidence interval around the change of the inequality when adjusting for the separate determinants.31
In the multivariate analyses, we examined the relationship between ethnic background and birthweight with different models. The different variables were gradually entered into the models by taking into account the hierarchical causal position of the variables.32 Gestational age was considered as confounding determinant and first entered into the model. Next the other biomedical variables, maternal and paternal height were added to the model. To this new model, educational level was added, and a similar procedure was repeated for the demographic and lifestyle-related variables. Differences in birthweight with the 95% CI presented were the differences in mean birthweight in grams of the offspring of non-Dutch women compared with the offspring of Dutch women. All statistical analyses were performed using Statistical Package of Social Sciences version 11.0 for Windows (SPSS Inc., Chicago, IL, USA) and Splus 6.0 Professional Release 1.
Characteristics of the pregnant women according to ethnic background are shown in Table 1. Gestational age was lower in the Capeverdean, Antillean and all Surinamese populations than in the Dutch population. Compared with Dutch women, women of the non-Dutch minority populations and their partners were of lower height and of lower maternal age. Women of the non-Dutch populations were more frequently low educated than those of the Dutch population. Turkish and Moroccan pregnant women were more frequent married and multiparous than Dutch women. Maternal BMI at early gestation was higher in all non-Dutch populations than in the Dutch population, except in the Surinamese-Hindustani population. Turkish women most frequently smoked during pregnancy, while Moroccan women hardly smoked during pregnancy. Compared with the Dutch women, women of the non-Dutch minority populations less frequently consumed alcohol during pregnancy.
Table 1. Subject characteristics
Values are means (SD) or percentages.
P values are result of chi-square test for categorical variables or analysis of variants for continuous variables.
Table 2 shows the mean birthweight of the offspring per ethnic population. The mean birthweight of offspring of Dutch women was 3485 g (SD 556). Offspring of Moroccan women had the highest mean birthweight (3517 g, SD 493), while offspring of Surinamese-Hindustani women had the lowest mean birthweight (3061 g, SD 541). All the studied ethnic populations, except the Moroccan population, had a significant lower mean birthweight than the Dutch population.
Table 2. Mean birthweight per ethnic population and difference in birthweight compared with the Dutch population
Values are means (SD) or differences (95% CI).
Mean birthweight (g)
Difference in birthweight (g)
+32 (−19 to 84)
−88 (−133 to −43)
−258 (−323 to −192)
−274 (−344 to −204)
−292 (−366 to −218)
−424 (−496 to −352)
−203 (−279 to −126)
In additional analyses, we tested the effect of the individual determinants on the observed differences in birthweight between the Dutch and non-Dutch populations. Maternal and parental height were the most important determinants of the lower birthweight in the non-Dutch populations. The lower gestational age was another important determinant, specifically in the Antillean, Surinamese-Creole, Surinamese-Hindustani and Surinamese-other populations. Educational level, maternal age and marital status had a smaller but still significant contribution to the observed differences in birthweight between the Dutch and non-Dutch populations (detailed information available in Appendix S1).
In Figure 1A–G, the results of the multivariate analyses are shown. The differences in birthweight of the Turkish, Surinamese-Creole and Surinamese-other populations compared with the Dutch population became nonsignificant after adjustment for gestational age, maternal and paternal height (model 3, Figure 1B, E, G). After additional adjustment for educational level, the difference in birthweight between the Antillean and Dutch population was not significant any more (model 4, Figure 1D). When the demographic determinants were added to the model (model 5), a further decrease in birthweight differences was observed in the Capeverdean, Antillean and all Surinamese populations. In the last step, lifestyle-related determinants were added to the model (model 6). Differences in birthweight with the Dutch population could only partly be explained in the Capeverdean and Surinamese-Hindustani populations (Figure 1C, F).
Our main finding is that the lower mean birthweight in the non-Dutch populations compared with the Dutch population to a large extent is determined by the shorter gestational age and lower maternal and paternal height in these populations. Educational level, maternal age and marital status had a smaller, but significant, contribution to the differences in birthweight between the Dutch and non-Dutch populations. Differences in birthweight between the Dutch and the Turkish, Antillean, Surinamese-Creole and Surinamese-other populations could largely be explained by the observed determinants. However, the differences in birthweight between the Dutch and the Capeverdean and Surinamese-Hindustani populations could only be partly understood by the established determinants.
The strength of this study is the population-based cohort with a large number of subjects in the studied non-Dutch populations studied from early pregnancy. Questionnaires in different languages were available to allow pregnant non-Dutch women who do not understand the Dutch language well enough to participate in the Generation R Study.28 Detailed information about a large number of potential determinants was available in this study.
There are some limitations that we might have to consider regarding this study. Information on ethnic background of the pregnant woman was missing for 1102 subjects, due to missing data on country of birth of the pregnant woman and her parents. Compared with the offspring of pregnant Dutch women, birthweight in this population (missing ethnic background) was 109 g (95% CI −146 to −72) lower. The response in the Generation R study is 61%. Selective participation of pregnant women could have influenced the observed mean birthweight of different ethnic populations. We compared birthweights in our study with the Netherlands Perinatal Registry. Although a different definition of ethnic background limited an accurate comparison, we observed no differences in mean birthweight for the Surinamese-Creole and Surinamese-Hindustani populations (Generation R: 3192 and 3061 g; Netherlands Perinatal Registry: 3128 and 3043 g, respectively). In contrast, birthweight of offspring of Dutch women differed between the Generation R study and the Netherlands Perinatal Registry (Generation R: 3485 g; Netherlands Perinatal Registry: 3353 g, respectively). Thus, selective participation might have influenced the magnitude of birthweight differences between ethnic populations. However, it is unlikely that this results in a different association between the studied determinants and birthweight in those women in the study and those who are not studied.
We classified the ethnic background of the pregnant women by country of birth of the participating women and their parents. Classification by country of birth has been proposed by Statistics Netherlands as method of choice to classify common ethnic groups in the Netherlands. It is widely applied in national registries and health policy applications. Advantages of this classification are that it is objective and stable over time. However, limitations are that it does not distinguish third generation migrants, does not take into account the heterogeneity of ethnic groups and therefore does not differentiate between ethnic subgroups (such as Hindustani and Creoles). Several other approaches to classify ethnic groups have been proposed in the international literature,33 including self-classification of ethnicity or race and classification of the nationality of a subject. However, accurate classification of race, ethnicity and nationality is limited because of the subjective nature and can change over time. In additional analyses of our data, we compared the magnitude of birthweight differences between ethnic groups based on self-classification with ethnic groups based on the classification by country of birth. Despite these theoretical differences between both classifications, in our data, birthweight differences between ethnic groups were virtually identical.
The associations between ethnic background and birthweight could be affected in the non-Dutch populations because of selection effects. Over 60% of the pregnant non-Dutch women were of first generation. These first generation women will have had relatively good health, and their health potential may have contributed to a higher birthweight of their children. Therefore, the healthy migrant effect might have influenced our findings.
We observed only a small effect of lifestyle determinants on ethnic differences in birthweight. This small effect could be because our study did not optimally capture the lifestyle characteristics. The small effect of smoking and alcohol consumption could be due to the use of binary variables for smoking and alcohol consumption. Possibly, residual effects of smoking and alcohol consumption are present, and the impact of smoking and alcohol consumption might be more expressive if a more refined classification was available. It could also be that unmeasured lifestyle determinants, such as illicit drug use, physical activity and food habits, can contribute to ethnic differences in birthweight.
We were unable to study several determinants that might be of importance in explaining the differences in birthweight. Data on gestational weight gain and energy intake during pregnancy, two important variables to predict birthweight,34,35 were not available for present analysis. Also, data on blood pressure during pregnancy, pregnancy-interval and pregnancy-related diseases were not available for present analysis.
Ethnic differences in birthweight
Our results demonstrated that, compared with the Dutch population, the mean birthweight is lower in all studied non-Dutch populations, except in the Moroccan population. These results are in line with a previous study in the Netherlands.10 Although the mechanisms of these disparities are not well understood, our results suggest that gestational age and parental height most strongly determined the lower birthweight in the non-Dutch populations compared with the Dutch population.
In our study, shorter gestational age was of particular importance in the Antillean and all Surinamese populations, which is in line with previous studies in the Netherlands, indicating that preterm birth is more frequently seen in the black (mainly Surinamese-Creole) and Hindustani populations.36,37 Also, in our study, a higher proportion of births were preterm (<37 weeks) in these populations compared with the Dutch population (preterm birth: Antillean 7.8%, Surinamese-Creoles 7.6%, Surinamese-Hindustani 7.6% and Surinamese-other 6.7% compared with Dutch 5.0%, results not shown). Prevention of preterm births in these populations might therefore reduce birthweight differences. Single motherhood might be associated with several factors that may increase the risk of preterm births. Single mothers might be exposed to more stress during pregnancy than mothers who have a partner. Maternal stress during pregnancy seems to be associated with preterm birth.38 Besides, single mothers might engage in more risky sexual behaviour, which may lead to a higher prevalence of urogenital infections. Several studies have found associations between urogenital infections and preterm birth.39,40 These hypotheses need to be elucidated in future studies to develop prevention strategies in these populations. Another factor that might influence gestational age at birth and/or the proportion of preterm births in the non-Dutch populations is the use of prenatal care. Several studies have suggested that prenatal care use and gestational age at first visit are associated with unfavourable pregnancy outcomes.41,42 However, complete data about prenatal care use were available only in a small sample of our study population. In this subsample, we examined whether ethnic differences in gestational age and preterm birth could be explained by differences in prenatal care use. These separate analyses showed that, although ethnic differences in prenatal care use were found, these differences did not affect the ethnic differences in gestational age and preterm birth.
This study showed that maternal and paternal height are important determinants for the lower birthweight in all non-Dutch populations. Several studies have reported an effect of maternal height on birthweight.23,43 Maternal and paternal height represent a complex conjuncture of genetic and environmental influences, especially long-term dietary intake and nutritional status. The observed shorter height in the non-Dutch populations might reflect the long-term poorer nutritional and living conditions they were exposed to. In a separate analysis, maternal height of second generation women in our study showed to be significantly taller compared with their first generation counterparts (results not shown). These results indicate that, although height is partly genetically determined, environmental influences might affect the mean height of populations. Reducing the socio-economic and environmental inequalities between ethnic populations could in long-term decrease differences in birthweight. Except for the large effect of maternal height on birthweight, short maternal height is also associated with other unfavourable pregnancy outcomes. Previous studies have reported an increased risk for idiopathic preterm labour, prolonged labour and caesarean delivery in women with short height.44–46
The socio-demographic determinants, maternal age, educational level and marital status, had a smaller but significant contribution in explaining differences in birthweight between ethnic populations. Especially, the extremes of these determinants, that is teenage motherhood, low-educated women and single motherhood, could result in an increased risk of lower birthweight.20,22,47 The exact pathways of teenage and single motherhood and low education leading to lower birthweight are unclear. Possible pathways to low birthweight might be less optimal use of prenatal care, more unhealthy behaviours (drug use, risk full sexual behaviour), more stress and unfavourable material conditions. Further study of these determinants could help obtain a fuller understanding of the causal pathways.
In our study lifestyle-related determinants only had a marginal role in explaining ethnic differences in birthweight. Maternal BMI at early gestation, which was higher in women with a non-Dutch ethnic background than in Dutch women, lead to a higher birthweight. However, obesity is also associated with an increased risk of gestational diabetes, pre-eclampsia, caesarean section and neonatal hypoglycaemia.48 Attention to the high BMI at early gestation in the non-Dutch populations is wanted because of these possible harmful complications. Since maternal smoking during pregnancy is quite equally distributed among the ethnic populations, it does not contribute much to the observed differences in birthweight. Prevention of maternal smoking, however, remains needed across all populations because of the detrimental prenatal and postnatal effects.49
Our findings suggest that differences in birthweight between ethnic populations can only partly be understood from established determinants of birthweight and that further study is necessary to obtain a fuller understanding. The part that is understood points to the importance of differences in parental height and gestational age. These risk factors are partly determined by genetic factors and cannot easily be modified. In short run, only small reductions of differences in birthweight between ethnic populations seem feasible.
The Generation R Study is conducted by the Erasmus Medical Centre in close collaboration with the School of Law and Faculty of Social Sciences of the Erasmus University Rotterdam; the Municipal Health Service Rotterdam area, Rotterdam; the Rotterdam Homecare Foundation, Rotterdam and the Stichting Trombosedienst and Artsenlaboratorium Rijnmond (STAR), Rotterdam. We gratefully acknowledge the contribution of GPs, hospitals, midwives and pharmacies in Rotterdam. The first phase of the Generation R Study was made possible by financial support from the Erasmus Medical Centre, Rotterdam; the Erasmus University Rotterdam and the Netherlands Organisation for Health Research and Development (ZonMw).