METHODS

We have carried out a search for information in the birth record for all those of South Asian origin who gave birth in Southampton from 1957 to 1996. All the data for the mother and baby in the birth record were extracted and computer coded. The birth records for babies born in Southampton from 1957 to 1996 were checked to identify all babies of South Asian origin. The medical records of patients from the maternity unit of the Southampton General Hospital (now the Princess Anne Hospital) are sent to the District Inactive Library (DIL) after three years. After a further three years the records were stored on microfilms (or microfiche) or more recently the records have been entered on a computerised database. Birth records for the 1990s that have not been microfilmed or entered into the database are kept in their clinic folders and stored in boxes at the DIL. All information recorded in the birth record for the mother and baby were extracted and computer coded.

South Asian people were defined as that group who were resident in Britain and who originally came from the Indian subcontinent, or were the descendants of people originally from the Indian subcontinent (Bangladesh, India, and Pakistan). South Asian names were identified from the birth records using the approach of Henley ¹¹ and with guidance from the local community. By using both first and second names it is possible to correctly classify 98.5% of the South Asian people, compared with a reference judgement.¹² The later records contained information on the mothers’ place of birth, which reduced uncertainties about ethnicity. The records of any students, or the wives of students, from South Asia were not included because it was considered unlikely that they would be resident in the UK in the long term. All measurements were converted to metric units. From the 1970s onwards, the dates of the last menstrual period (LMP) and the expected date of delivery were recorded. The length of gestation was calculated as the difference between the date of the LMP and the actual date of delivery. In cases where the date of the LMP was not known, the midwife or the doctor made an estimate of the duration of gestation based upon the height of the uterine fundus. In later records ultrasound scans were used to estimate fetal size and therefore its “age”. An estimate of maternal weight at 15 weeks gestation was derived by linear regression analysis, using weights taken before 13 weeks, between 13 to 15 weeks, and between 17 to 20 weeks.

To determine the completeness with which people from South Asia had been identified, a check was carried out against the information collected during the 1991 census for Southampton. The Southampton and South West Hampshire Local Research Ethics Committee gave approval for the study to be carried out.

RESULTS

The analyses in this paper have been restricted to singleton births delivered at term (greater than 37 weeks gestation). The total number of singleton births was 2683 and 46 twin births were excluded. Of the singleton births, there were 210 pre-term births; 77 babies without data on gestational age or birth weight, and three babies without a recorded gender, all of whom were excluded. The number of births available for subsequent analysis was 2395.

Table 1 presents the overall mean birth weight for all subjects in the study, as well as mean levels broken down by place of birth, religion, and gestational age. Birth weight tended to be greater in babies born to mothers from Pakistan and Fiji. The lightest babies were likely to be born from mothers from East Africa or the UK. Babies of Muslim mothers tended to be heavier than babies of either Sikh or Hindu mothers. There was an increase in birth weight with gestational age, even though all of the babies included in the study were born at a gestation longer than 37 weeks.

The rest of the analyses presented in this paper will be restricted to mothers either born in the Indian subcontinent (first generation, 1435) or the UK (second generation, 283).

Table 2 presents mean data on booking age, parity, maternal height and weight, as well as birth weight and head circumference broken down by whether the baby was born to a first or second generation mother. First generation mothers were statistically significantly older at age of booking and had more children than second generation mothers. There was no statistically significant difference in maternal height between generations, but first generation mothers tended to be heavier at booking than second generation mothers. Unadjusted birth weights were greater in babies born to first, rather than second, generation mothers; however, the difference between generations was only present between girl babies (who were also lighter than boy babies of either generation). Of babies born to first generation mothers from the Indian subcontinent, 7.5% were classified as low birth weight (less than 2500 g) compared with 11.7% in babies born to second generation mothers. After adjusting for gender, mother’s ethnic group, age, religion, parity, and weight at booking; the statistically significant difference between generations disappeared. Adjusted head circumference did not differ between generations.

Table 3 presents adjusted birth weights for babies born to first and second generation babies broken down by five year periods. The first time period in which second generation mothers gave birth to babies was 1971–75. The number of births is small for 1971–75 and 1976–80 and therefore mean estimates may be unreliable (wide confidence intervals). After adjusting for other factors, babies born to second generation mothers were statistically significantly lighter than babies born to first generation mothers for babies born since 1991 (3042 g compared with 3220 g), but for no other years. For babies born to either first or second generation mothers, after adjusting for factors known to influence birth weight, there was no clear trend for increasing birth weight over time, although babies born to first generation mothers since 1991 were on average the heaviest, before and after adjusting for other factors. Although women born in Pakistan gave birth to heavier babies than women born in India, in neither group was there a trend for increasing birth weight over time (data not presented).

To determine whether it was possible to identify a unit change in birth weight by year of birth, a multiple regression analysis was used, with adjustment for other factors. It was not possible to demonstrate an effect of year of birth on birth weight. Within the total population, there was a difference in birth weight for the different ethnic groups, but for none of these groups was there a consistent pattern of change in birth weight with time. Adjusting for ethnic group did not change the overall pattern.

Similar explorations were carried out to determine any changes with time for head circumference, placental weight, and the ratio of placental weight to birth weight. There was no identifiable change with time for any of the variables (data not presented).

DISCUSSION

The main finding from this study is that, after adjusting for factors known to influence birth weight, there has not been any clear increase in birth weight for either first or second generation babies born in Southampton over the past 40 years. There was no difference in birth weight for babies born to second generation mothers compared with babies born to first generation mothers who were born in the Indian subcontinent. Furthermore, in neither group was there a trend with time to suggest any increase in birth weight. Although women from Pakistan gave birth to heavier babies than women from India, in neither groups was there a trend for increasing birth weight over time.

We have used the data contained in the 1991 census to determine that the data presented here represent a fairly complete record of all South Asian babies born in Southampton. Using the age specific census estimates, we conclude that the observed number of births was very close to that expected, suggesting that the sample in this study is a reasonable reflection of the population. The clinic data used in the study were collected routinely and therefore it is likely that some errors will have occurred. The data were thoroughly checked to eliminate obvious coding errors, but it is not possible when using retrospective material to check the accuracy of the data. We have had to assume that any errors that have occurred are randomly distributed, and that errors were not systematically related to year of study such that an underlying trend was obscured by error. The staff in the clinics where the maternal anthropometry was carried out were not the same as those who made the measurements in the newborn babies, making it unlikely that any errors would be correlated. Any uncertainty around the dates of the LMP, and hence in the estimation of gestational age, might have occurred in the earlier records before ultrasound scanning became a routine procedure, but is unlikely to have been consistent in nature. It was assumed that any estimate of gestational age of more than 44 weeks was probably attributable to an incorrect LMP, and therefore these cases were excluded from the main analysis.

There are two other studies that have looked at intergenerational effects on birth weight, for people from the Indian subcontinent, in the UK. One study, based upon 111 second generation births, showed an increase in birth weight from first to second generation babies.⁹ By contrast, another study based upon 778 second generation births, showed no intergenerational increase in birth weight.¹⁰ Our results, based on 283 second generation births, conform with the findings in the latter study and we were not able to demonstrate any change in birth weight across the generations. Furthermore, we were not able to identify any pattern of change towards increased birth weight in time for the population as a whole or any subgroup within the population. It is possible that either selection and information bias may be present in all three studies, but we are not able to explain how any bias that might have occurred is likely to lead to the patterns present in our study. We estimate that random error could account for a variation in birth weight of about 100 g in our sample. Thus, given the size of our study we would expect to be able to detect differences between year groups of about 100 g. Therefore, it seems unlikely that if a true increase in birth weight exists, either over time or between generations, it has been masked by error.

The birth weight of South Asian babies born in Southampton, is not different to that for South Asian babies born elsewhere in the UK.^{2, 13} The birth weight for infants born in the UK whose mother comes from the Indian subcontinent is higher than the average birth weight in the Indian subcontinent (by about 300 g),^14–16 but when compared with the general population in the UK is some 300 g lighter.¹⁷

It may be that much of the obvious difference in birth weight between South Asian babies born in the Indian subcontinent and the UK is most readily explained by environmental factors that lead to a higher infectious load and a poorer quality of diet during pregnancy. Hence, an immediate consequence of an improved environment, allows birth weight to increase by about 300 g. Whereas a change of this sort might be adequate to explain an initial improvement in birth weight, it is not sufficient to explain the continued difference in birth outcome between people from South Asia and the general population. Nor does it adequately address the apparent lack of intergenerational improvement. Indeed, in the study conducted by Draper the suggestion was that by the third generation the situation was, if anything, worse.¹⁰ In our analysis we sought to explain the difference by adjusting for other factors that are known to affect birth weight, but we were not able to influence the lack of increase in birth weight over time.

There are two important underlying assumptions in this work; firstly that there is no biological reason why babies born to mothers of South Asian origin should not have the same potential as the general UK population. If this assumption is correct, the birth weights reported here represent a marker of constrained growth. Secondly, that babies who are born smaller than they should be, carry both a short and long term risk to their health, and that it is therefore important to do something about it.^{18, 19} Balarajan and Raleigh ²⁰ have shown that perinatal mortality is higher in babies born to mothers from the Indian subcontinent, which they attribute, at least partly, to low birth weight associated with poor nutrition.

Heart disease and diabetes are much more common in immigrants from the Indian subcontinent than the general UK population.^{21, 22} Is this in some way related to their lower birth weight? Research in India, UK, and elsewhere, suggests that there is a close link between size, shape, and body composition at birth and subsequent risk of diabetes and heart disease, which can be moderated by changes in lifestyle, but not completely removed.²³ Based on our findings, the predicted decline in diabetes between first and second generation immigrants, based on improved early nutrition and environment may not materialise, because in fact the environment has not improved or at least has not yet affected fetal growth. Our data would suggest that it would not be appropriate to adopt a complacent approach that assumes simply that the problem will resolve itself in time as social circumstances improve. The persistence of low birth weight within this group of the population requires a focused approach for further research, to determine the specific biological factors that lead to constrained fetal growth. Identifying specific factors that may be amenable to intervention, and that hasten the process of achieving a more desirable, or optimal, birth weight should be an important component of the research agenda.