Article Text
Abstract
Background The mean birth weight of offspring of Bangladeshi, Indian and Pakistani women tends to be among the lowest of any ethnic groups regardless of country of residence. However, it is unclear whether the mean birth weight of South Asian offspring born in England and Wales is higher among those whose mothers were themselves born in England and Wales compared to those whose mothers were born in the Indian sub-continent.
Methods We used cross-sectional data from a unique linkage of routine records for the whole of England and Wales (2005–2006, n=861 654) to estimate mean birth weights of the live singleton offspring of Bangladeshi, Indian, Pakistani or White British ethnicity according to whether maternal place of birth was England and Wales or the Indian sub-continent.
Results Offspring of women born in the Indian sub-continent were slightly heavier at birth than offspring of South Asian women born in England and Wales even after adjustment for gestational age, maternal age and parity (Bangladeshi 28 g, 95% CI 10 to 46; Indian 31 g, 95% CI 20 to 42; Pakistani 21 g, 95% CI 12 to 29).
Conclusions There is no indication that the mean birth weight of South Asian offspring of women born in England and Wales is higher than the mean birth weight of those whose mothers were born in the Indian sub-continent. This suggests a shared physiological tendency for down-regulation of fetal growth transmissible across generations. Within the UK, there is unlikely to be any appreciable increase in mean birth weight of South Asian babies over the next few decades.
- Fetal growth
- ethnicity
- migrant studies
- birthweight
- births si
- ethnic minorities si
- fetal
- migration & health
Statistics from Altmetric.com
Introduction
Investigating health status changes among people migrating from their country of origin to one with a different environment (including exposure to pathogens, pollutants, diet and other health-related behaviours) can be informative about underlying aetiology. This approach has been applied to cancer and cardiovascular disease as in the classic investigations of mortality among Japanese migrants to the USA.1 Less commonly it has also been used in studies of perinatal outcomes.2 3 It may be a particularly fruitful approach when there are large differences in outcomes between the country of origin and destination as seen for birth weight between the Indian sub-continent and Britain. Mean birth weights in the countries of the Indian sub-continent are low compared to those in many other parts of the world.4 5 This is mirrored in birth weight differences in Britain among South Asians compared to the White British population. Studies of UK populations have typically found offspring of South Asian women to be around 250 g lighter at birth than those of White British women with a corresponding higher prevalence of low birth weight6–12 as well as higher rates of perinatal mortality.13 In 2008, the first national birthweight data by ethnicity for England and Wales (for 2005) confirmed these observations14: while only 5.6% of White British live singleton births were of low birth weight, around 10% of Indian, Pakistani and Bangladeshi infants were so classified. Indeed, of all ethnic groups examined (including black African and Caribbean), those who were Indian, Pakistani or Bangladeshi were the lightest at birth. Among those born at 40 weeks gestation, Indian and Bangladeshi babies were 300 g lighter than White British babies indicating a reduced fetal growth rate. This tendency of women of South Asian origin to produce lighter babies than other ethnic groups has been observed in other countries.3 15–20 However, the reasons for these differences are not well-understood.
It is possible that fixed genetic effects may account for some of this effect in which case one would not expect fetal growth rates to change in response to migration to a different environment. However, if the primary drivers of these ethnic differences are to be found in environmental differences in the maternal life-course (from conception onwards), it might be expected that they would attenuate with acculturation over generations—that is, convergence in behaviour and lifestyle with that of the British population as a whole. Specifically, one might expect to see increases in the birth weight of South Asian babies according to the number of generations since their family originally migrated. However, it should be noted that changes in birth outcome between generations, as well as those resulting from positive improvements in the environment, may also be affected by the adoption of negative behaviours, including smoking and a less healthy diet. This generational hypothesis has been examined in four studies conducted in the UK but their findings have been inconclusive due to limitations of the data they analysed,11 21–23 particularly their relatively small size.
Obtaining a more definitive answer to whether birth weight is increasing across generations of migrants from the Indian sub-continent is important not least because South Asians form a sizeable part of the British population. At the 2001 Census, 2.1 million gave their ethnicity as being either Indian, Pakistani or Bangladeshi making up 3.6% of the total population of Great Britain.24 From a public health perspective it would also give a crucial indication of whether we may expect any diminution of these ethnic differences in birth weight over the next few decades. A more definitive test of the generational hypothesis would also provide important evidence about how responsive these particular ethnic differences in fetal growth are to apparently substantial changes in the maternal environment.
By linking two routine national data sources that, for the first time, make it possible to classify births according to ethnicity and maternal country of birth, we investigate whether the low birth weight seen in babies of South Asian ethnicity born in England and Wales has increased between offspring of women born in the Indian sub-continent and offspring of women born in England and Wales.
Methods
Data sources
Since 2002, National Health Service (NHS) numbers have been allocated at birth for all babies born in England, Wales and the Isle of Man. Each birth is notified to the Central Issuing System which allocates NHS numbers and through this system key birth details are collected centrally on what is known as the NHS Numbers for Babies (NN4B) dataset.25 For all NHS numbers issued since the beginning of 2005, the Office for National Statistics (ONS) has received a subset of the variables contained in this dataset, which it has been possible to link with other datasets containing NHS number.25 The quality and completeness of the NN4B dataset has been shown to be generally good.26 We analysed a dataset produced from the linkage of routine birth registration records for babies born alive in England and Wales in 2005 and 2006 to NN4B records.14 26–29
Data items
The following information was available from the birth registration data: mother's country of birth, birth weight, whether a birth is live or stillborn, singleton or multiple, and information on maternal age, marital status/registration type, parity and National Statistics Socio-Economic Classification (NS-SEC). Marital status/registration type refers to whether the birth was registered inside or outside marriage and, if outside marriage, who registered the birth. The NS-SEC is based on fathers' occupation and is, therefore, only available for births occurring in marriage or jointly registered by both parents; it is only coded for every tenth live birth registered. Information on parity is only collected for mothers of babies born within marriage who are asked the number of previous live and stillborn children by their present or any former husband.
The NN4B record requests information on the ‘ethnic category (baby)’ as defined by the mother using specified categories that match those used in the 2001 Census. The three Asian or Asian British groups, Bangladeshi, Indian and Pakistani, and the White British group are included in this analysis. Throughout this paper the Asian/Asian British groups are referred to individually as Bangladeshi, Indian and Pakistani, and collectively as South Asian. While ethnicity refers to the baby, in the case of the four ethnic groups of interest here it seems safe to assume that the mother's ethnic group will be the same. NN4B also collects information on gestational age at delivery. As information on ethnic group and gestational age comes from NN4B data these data items are not available for birth registration records that have not linked to an NN4B record. The NN4B data specification asks for gestation length in weeks ‘calculated from relevant menstrual data held within the maternity system’.26 However, it is now widely accepted that ultrasound gives a better estimate of gestational age and it is very likely that a appreciable proportion of the gestational lengths recorded will have been based on ultrasound.26
Information on maternal country of birth was used to distinguish between South Asian babies according to whether the mother was a first or second generation migrant. In line with established terminology, second generation migrants were those born in England and Wales although strictly speaking people born in Britain are not ‘migrants’. First generation migrants were those born in the Indian sub-continent defined specifically for each ethnic group: Bangladeshi mothers born in Bangladesh, Indian mothers born in India and Pakistani mothers born in Pakistan. It should be noted that the second generation mothers will include some women who are in fact third or higher generation migrants, although the proportion of these will be small.
Information was not available on a number of key potential confounders, including maternal weight, height, smoking in pregnancy, interval between births and maternal diet. In the Discussion we consider the potential impact of these factors in providing an explanation for the differences we observe where possible by citing evidence from other data sources from the UK.
Analytic strategy
Our primary interest was in estimating mean birth weight of Bangladeshi, Indian and Pakistani babies and (1) seeing whether within each South Asian ethnic group mean birth weight differed between the offspring of first and second generation migrants, and (2) contrasting these means with those of White British babies. Since birth weight is known to vary according to factors including gestational age, parity and maternal age, we undertook analyses in which adjustment was made for these. Since data on parity were only available for births registered in marriage, the final analyses were restricted to babies born in marriage in order to take account of this important potential confounder. As shown in the Results section, this restriction had little impact on Bangladeshi, Indian or Pakistani births as the majority of them are born in marriage.
We used cross-tabulations and linear regression to assess the association of ethnicity cross-classified by maternal country of birth on mean birth weight adjusting for a range of factors (sex, gestational age, parity and maternal age). We repeated these analyses for the 10% subset with NS-SEC coded to see whether additional adjustment for this socioeconomic measure affected our findings. All statistical analyses were conducted using STATA statistical software (version 9).30 The final comparisons of birth weights within ethnic group according to maternal country of birth adjusted for various factors were estimated from models based on all births in marriage with complete data by using the post-estimation command lincom that provides point estimates and 95% CIs for specified linear combinations of coefficients.
Results
There were 1 315 352 babies born alive in England and Wales in 2005 and 2006, 1 276 198 of which were singletons. For both years, 99.9% of these birth registration records were successfully linked to a NN4B record. Our findings were very similar for 2005 and 2006; thus, we present results for the 2 years combined. We have restricted our attention to the 861 654 live singleton births of White British, Bangladeshi, Indian or Pakistani ethnicity where either the mother was born in England and Wales, or in Bangladesh where births were of Bangladeshi ethnicity, in India if Indian, or in Pakistan if Pakistani. As shown in table 1, South Asian births whose mothers were born elsewhere were excluded, as were live singleton births in ethnic groups not relevant to our analysis and those whose ethnicity was not known or the birth registration record was not linked to a NN4B record.
The frequency distribution and mean birth weights classified by key social and biological characteristics are shown in table 2 for all live singleton births and separately for the subset born in marriage. Mean birth weight was higher in babies of White British ethnicity than those of Indian, Pakistani or Bangladeshi ethnicities; the difference was largest for Bangladeshi and smallest for Pakistani births. For White British births, those in marriage had a higher mean birth weight than all births, while no such difference was seen by marital status for South Asian births.
In each South Asian ethnic group, mean birth weight was lower, and the percentage of low birth weight (defined as <2500 g) higher, among offspring of second compared to first generation migrants (table 3). The gestational age, maternal age and NS-SEC indicators shown in table 3 differ across ethnic groups. Consistent with other studies, there is evidence of a slightly higher risk of South Asian births being preterm.31 Within each ethnic group there is also some evidence of variation by maternal country of birth for other factors: (1) the percentage with maternal age under 25 years was higher in Bangladeshi and Pakistani offspring of second compared to first generation mothers (p<0.001); this pattern was not seen for Indians; (2) more Bangladeshi babies of second compared to first generation migrant mothers were classified to managerial and professional occupations (p<0.001), while the opposite pattern was seen for Pakistanis (p<0.001). Indians had a very much higher proportion of women in this NS-SEC category and had the smallest difference according to migrant generation. Among South Asians these patterns were the same regardless of whether the analysis was based on all live singleton births or restricted to those born in marriage. This is not surprising as the vast majority of South Asian births were to mothers who were married, although as expected this percentage was slightly lower among second generation mothers. However, as is well-known, among the White British, mean birth weight differed considerably according to whether the birth was registered in marriage or not.
Large differences in birth weight between South Asian and White British babies were evident having adjusted for sex, gestational age and maternal age, with the mean birth weights of South Asian babies being 200–300 g lower (figure 1). Further restricting these analyses to the 10% subset with known NS-SEC produced a very similar picture with additional adjustment for NS-SEC having very little effect (data not shown). However, what is particularly striking in figure 1 is that the inter-generational differences in mean birth weight within each ethnic group are very small.
Turning to this central issue of differences in birth weight between offspring of first and second generation migrants, table 4 shows these differences adjusted for potential confounders. Within each of the South Asian ethnic groups there was a consistent tendency for offspring of first generation migrants (ie, mothers born in Bangladesh, India or Pakistan) to be very slightly heavier than offspring of mothers born in England and Wales. Because the vast majority of births to women in these ethnic groups occur in marriage, these estimates (restricted to births in marriage) are almost the same as those based on all live singleton births (data not shown). Full adjustment for sex, gestational age, maternal age and number of previous live and stillbirths (Model 4) reduces the difference according to maternal country of birth to between 20–30 g in all three ethnic groups. A very similar picture is evident when further restricted to the 10% subset with known NS-SEC with additional adjustment for NS-SEC having almost no effect on these differences (data not shown).
Discussion
These analyses do not support the generational hypothesis that predicts increased birth weight across successive generations of South Asian migrants to the UK. In fact, the contrary is found with Bangladeshi, Indian and Pakistani babies with mothers born in England and Wales being on average very slightly lighter at birth than those with mothers born in the Indian sub-continent (first generation migrants to Britain). Thus, the differences in birth weight between South Asian and White British babies are in fact a little larger for the offspring of second generation than for first generation migrants from the Indian sub-continent.
Our analyses, based on national data for England and Wales, provide the most precise and representative estimates to date of the simultaneous impact of South Asian ethnicity and maternal country of birth on birth weight. However, these data do have some weaknesses. There is uncertainty surrounding the recording of ethnic group in the NN4B dataset which is meant to be as reported by the mother but in practice is likely to be a mixture of that reported by the mother and that reported by healthcare professionals (without asking the mother).14 However, in general, such uncertainties do not seem to have biased our estimates of birth weight by ethnicity per se as they are similar to those estimated from other large datasets in which ethnicity was obtained using different approaches.11 32
Four other studies have tested the generational hypothesis in the UK. Three were based on local populations21–23 and one on a 1% sample of the national population.11 Compared to the nearly 90 000 South Asian births analysed in this paper, these were much smaller investigations based on between 332 and 4321 South Asian births. The smallest found weak evidence that birth weights of offspring of second generation Asian mothers were higher than those of first generation migrants.22 A Southampton-based study found no such difference after adjustment for maternal height and weight at booking, gestational age, maternal age and parity.23 The other two studies found very slightly greater birth weights among first compared to second generation migrants (differences of 33 g11 and 22 g21), which are very similar to those we found although the differences in these two studies did not reach statistical significance at the 5% level.
Ethnic differences in birth weight by maternal country of birth have been the subject of investigation in the USA, although the focus has been mainly on differences between the black and white populations. These have found that the offspring of black mothers who were themselves born in the USA have lower rates of fetal growth than the offspring of black mothers born outside the USA with both groups having lower rates than whites.33 34 However, the parallels between this and the observations concerning South Asian births in the UK are tenuous. First, the mean birth weight of offspring of non-USA born black mothers is closer to those of white mothers than they are to offspring of black USA-born mothers. In contrast, our analysis for England and Wales shows the birth weight of South Asian babies remains much lower than White British babies regardless of maternal country of birth. Second, unlike the South Asian population in Britain, the majority of the black population born in the USA has lived there for many generations.
Smoking and other negative influences
Acculturation among migrants to the UK appears to be associated with increased prevalence of health-related behaviours such as smoking that may have an adverse effect on fetal growth.35 These may entirely or partly obscure any positive influences on fetal growth associated with being a second generation migrant. We looked at this by analysing data from the Health Survey for England 200436 (HSE) for women of childbearing age (16–49 years). We found the age-adjusted prevalence of current cigarette smoking in Bangladeshi women was 2% in both first and second generation migrants. However, it was higher among second compared to first generation Indian (9% vs 2%) and Pakistani (8% vs 3%) women (p<0.05).
Some women give up smoking when they know they are pregnant. Despite this, in the Millenium Cohort Study the prevalence of smoking in pregnancy in England was reported to be 4% in the combined group of Pakistanis and Bangladeshis, and 6% among Indians.35 However, if we assume that the HSE prevalence figures do apply to pregnancy, as smoking is associated with a 200 g birthweight deficit,37 the reported differences in smoking prevalence within each ethnic group would result in offspring of second generation Bangladeshi, Indian and Pakistani migrants weighing 1 g, 14 g and 10 g less than first generation migrants of the same ethnicity. Thus, smoking alone cannot be masking any important tendency for offspring of second generation mothers to be bigger than those of first generation mothers, although it may explain some of the small birth weight deficit in second compared to first generation migrant offspring.
Nutritional status and height
A key assumption of the generational hypothesis is that over their life-course, second generation migrants are likely to have experienced better nutrition than first generation migrants. This would be expected to be reflected in increases in maternal adult height, which in turn has been shown to be associated with birth weight.38 We looked at this question by analysing data on heights of women aged 16–49 years in two contemporary British cohorts. In the HSE 200436 we found second generation Bangladeshi, Indian and Pakistani women to be about 3 cm taller than first generation migrants. A similar difference was observed in the Millenium Cohort Study for Bangladeshi but not for Indian or Pakistani women.
While it is unclear whether there have been net improvements in nutritional status as proxied by height, these results suggest that maternal height cannot explain the slightly lower birth weight of second compared to first generation migrant offspring. In this context it should be noted that the substantial increases in attained adult height over the twentieth century in high income countries have occurred alongside minimal increases in birth weight.39 To this extent, temporal changes in maternal height within populations do not appear to be as strongly related to size at birth as maternal height and birth weight are cross-sectionally.38
Finally, it is important to mention the potential mediating role of maternal diet in pregnancy, although evidence for this having an impact on fetal growth is limited. Unfortunately, very little systematic work has been done on the diets of South Asians in the UK40 41 and even less on changes between generations.42 However, it is known that there are culturally specific beliefs about the types of food that are appropriate to eat in pregnancy, although how far these are adhered to among migrants from the Indian sub-continent to the UK is unclear.43
It could be argued that second generation migrants may adopt a less healthy westernised diet. This has been suggested as a potential explanation for worse pregnancy outcomes among second compared to first generation Latino migrants to the USA.44 However, this sort of phenomenon is unlikely to provide an explanation for our observation that the birth weights of offspring of second generation migrants were slightly smaller than that of first generation migrants.
Heterogeneity of genetic and socio-cultural ancestry
It could be argued that our comparisons across generations are flawed as there may be essential differences in genetic and cultural ancestry of the different waves of migrants. Recent analyses of the population structure of India underline the great genetic diversity of subpopulations of the Indian sub-continent.45 Certainly, the reasons for migration to the UK from the Indian sub-continent have varied over time with the pressures for migration out as well as the incentives and constraints on entry to the UK changing with local and global political and economic forces. These factors are not the same for migrants from each of the three countries we have studied.46 Further aspects of cultural and behavioural differences between Bangladeshis, Indians and Pakistanis are manifest across such diverse domains as religious affiliation and cardiovascular risk.47 This heterogeneity is also reflected in our own data, as has already been noted in discussion of table 3, which shows appreciable differences in the socio-demographic characteristics both between and within the South Asian groups. However, it is precisely this heterogeneity that makes it all the more remarkable that the birth weight differences (relative to White British) between first and second generation migrant offspring are almost the same for each ethnic group. One potential explanation for this is that there is a common, underlying (physiological) trait shared by Bangladeshis, Indians and Pakistanis that downregulates fetal growth, and which is insensitive to changes in the maternal environment, at least over two generations.
Physiological downregulation
What could be the nature of such a physiological (as distinct from pathological) tendency for downregulation of fetal growth that persists across generations? Such a mechanism could operate both through effects on maternal size as well as through placental regulation of fetal growth. However, it would not have to be a genetic, fixed effect. Instead it could be the result of environmentally induced epigenetic changes. Kuzawa has suggested that fetal growth may take some generations to respond to improvements in maternal nutrition and environment.48 He postulates that an intergenerational (epigenetic) mechanism of ‘phenotypic inertia’ has evolved to ensure that only improvements in nutrition sustained over several generations will lead to increases in fetal growth. This is supported by work on macaque monkeys which suggests that constraints on fetal growth took five generations to ‘wash out’.49 Studies of third and even fourth generation migrants are going to be required before this specific hypothesis can be tested in humans.
In conclusion, these analyses provide strong evidence that the birth weight of Bangladeshi, Indian or Pakistani offspring of women born in Britain has not increased compared to the offspring of women who migrated to Britain from the Indian sub-continent. In fact, offspring of second generation migrants are slightly lighter than those of first generation migrants. Some of this difference between generations of migrants may be explained by the adoption of adverse health-related behaviours such as smoking. However, these behaviours are still far too uncommon to explain differences in size at birth relative to the White British population, which remain substantial. The persistence of low rates of fetal growth among the offspring of second generation South Asian mothers may be due to minimal changes in the pertinent aspects of the behaviour or life-course between first and second generation migrants. However, it may reflect a common physiological constraint on fetal growth that could have an epigenetic basis, and that takes more than two generations to be modified, or to genetic differences in fetal growth potential. Whatever, the explanation, within the UK, it is unlikely that we will see substantial increases in mean birth weights of South Asians over the next few decades. In the meantime, it would be highly desirable for further research to be undertaken on this issue, in particular looking at how birth weight of offspring changes across generations within families, rather than in aggregate as we have presented.
What is already known on this subject
Babies born in the Indian sub-continent have low birth weight compared to those born elsewhere and similar differences are seen in Britain between babies of South Asian and White British ethnicities. The reasons behind these differences are not well-understood.
If they include environmental differences in the maternal life-course, one would expect birth weight differences to diminish over successive generations of migrants. Studies of this issue to date have been inconclusive.
What this study adds
Using contemporary data for all births in England and Wales it is clear for the first time that the birth weight of South Asian offspring of women born in Britain has not increased compared to the offspring of women who migrated to Britain from the Indian sub-continent. In fact, the babies with mothers born in Britain are slightly lighter at birth than those with mothers born in the Indian sub-continent. Within the UK, there is unlikely to be any appreciable increase in mean birth weight of South Asians over the next few decades.
Acknowledgments
We thank Pat Doyle, Shah Ebrahim, Michael Kramer, Debbie Lawlor and Liam Smeeth for comments on an earlier draft.
References
Footnotes
Competing interests None declared.
Ethics approval Approval for the use of these data sources for linkage and the production of statistical data was given by the North East London Ethics Committee and also the Patient Information Advisory Group under Section 60 of the Health and Social Care Act 2001.
Provenance and peer review Not commissioned; externally peer reviewed.