Maternal birthweight and diet in pregnancy in relation to the infant's thinness at birth
Correspondence: Dr K. Godfrey, Medical Research Council Environmental Epidemiology Unit (University of Southampton), Southampton General Hospital, Southampton SO16 6YD, UK.
Objective To examine how maternal diet in pregnancy and parental body size and birthweight influence an infant's thinness at birth measured by a low ponderal index.
Design An observational study of newborn infants and their parents.
Setting Southampton, England.
Population Five hundred and thirty-eight infants born at term.
Main outcome measure Ponderal index at birth.
Results Women who had a high intake of carbohydrate in early pregnancy and a low intake of dairy protein in late pregnancy tended to have infants that were thin at birth (P= 0.01 and P= 0.03, respectively, in a simultaneous analysis). Women who themselves had a low birthweight also tended to have thin infants, ponderal index falling from 28.3 kg/m3 to 26.2 kg/m3 as the women's birthweights decreased from more than 4.0 kg to 2.5 kg or less (P < 0.0001). Tall fathers had thin infants, but ponderal index was not related to the women's heights or the fathers’ birthweights.
Conclusion These associations may reflect constraints on placental development imposed by a woman's nutrition in pregnancy and during her own intrauterine life. Effects of the father's height may be mediated through genetic influences on skeletal growth.
Many newborn infants are thin, lacking skeletal muscle as well as subcutaneous tissue1. People who were thin at birth tend to have reduced glucose tolerance as children and to be insulin resistant in adult life2–4. Studies in Britain, recently replicated in Sweden, have shown that such infants tend to develop non-insulin dependent diabetes as adults3,5. They also tend to have other components of the insulin resistance syndrome, which include raised blood pressure and dyslipidaemia3,6. They are at increased risk of death from coronary heart disease7. These long term associations of thinness at birth are thought to result from persisting consequences of fetal adaptations to undernutrition, which include restriction in growth of skeletal muscle and alteration of its metabolism4,8.
A considerable body of evidence in humans and in animals, including the results of embryo transfer studies, shows that size at birth is primarily determined by the mother, whose influence acts more through the intrauterine environment than through the genes she passes to her infant9–12. Thinness at birth, conventionally measured by a low ponderal index (birthweight/length3), has been attributed to undernutrition of the fetus13. Little is known about the maternal influences that lead to thinness at birth. Low energy intake and low weight gain in pregnancy have been implicated14. Infants born after the Dutch ‘hunger winter’ of 1944 to 1945 were thin15. Following on from our initial report of the effects of maternal nutrition on placental and birthweights16, we present here an analysis of the relations between maternal diet in pregnancy, parental body size and birth-weight and ponderal index in the offspring at birth.
We have described the study methods in our previous report on placental and birthweights16. Caucasian women (n= 655) aged 16 years or older with singleton pregnancies at < 17 weeks of gestation attended the midwives’ antenatal booking clinic at the Princess Anne Maternity Hospital, Southampton. Of these women, twelve miscarried or had a termination of pregnancy, and seven gave birth outside the district. Five hundred and ninety-six (94%) of the remaining 636 agreed to participate in the study, and of these 557 were delivered at term ( 259 days gestation, 37 weeks).
The women were visited at home by a trained research nurse in early and late pregnancy (median duration of gestation 15.3 and 32.7 weeks, respectively) and asked about their menstrual and obstetric history and current smoking habits. The women and the infants’ fathers were requested to ask their parents about their own birthweights which were obtained for 506 women and 446 fathers. The woman's height was measured in the antenatal clinic and her first recorded weight in pregnancy documented. The fathers were asked their height at the home visit. For all except 15 women and 13 fathers social class could be allocated according to their current or last occupation18,19.
In early and late pregnancy a food-frequency questionnaire was administered to the women. It assessed the average frequency of consumption of 100 foods or food groups in the three months preceding the visit. The nutrient content of a standard portion of each food was multiplied by its reported frequency of use to calculate average daily nutrient intake20. The early pregnancy estimates of nutrient intake have been validated against those determined from food diaries kept over four days20.
The gestation at delivery was estimated using an algorithm that took account of the woman's menstrual history and ultrasound scan data17. The infant was weighed at birth to the nearest 5 g using digital scales. The placenta was weighed on digital scales after trimming by stripping the amnion to the cord, cutting the chorion at the edge of the placenta and removing the cord flush with the placenta. Crown-heel length was measured three times (to the nearest millimetre) using a neonatal stadiometer, and the mean used in the analysis. Before the study fieldworkers were trained, and during the study inter-observer tests of repeatability were performed at monthly intervals. Discrepancies in measurements of crown-heel length between the four fieldworkers were small compared with the overall ‘between subject’ standard deviation (SD), the maximum difference between any two observers being 4 mm compared with an SD of 20 mm. The women gave informed consent and the study was approved by the local ethics committee.
The size of the sample studied was determined by considerations relating to a parallel study in the women examining maternal iron stores and placental volume in early pregnancy17. Statistical analysis was by tabulation of means. Multiple linear regression was used to take account of the independent effects of potential confounding variables. Levels of significance refer to regression analysis of continuous variables. Nutrient intakes were log transformed where necessary to satisfy assumptions of normality.
Having excluded three women who were not visited in late pregnancy and 16 for whom the infant's length at birth was not recorded because of hip instability, admission to the special care baby unit or early discharge from hospital, there remained 538 women who gave birth at term (85% of the sample of 636). Their characteristics are shown in Table 1, together with those of the infants and the infants’ fathers. Half of the women were primiparous; a quarter smoked regularly in early pregnancy and half were from social class III16. The male infants were larger than the female infants but had lower ponderal indices. As expected, placental weight was strongly associated with ponderal index, a 1 SD decrease in placental weight being associated with a 0.38 SD decrease in ponderal index. All anthropometric measurements increased with gestation, ponderal index rising by 0.031 kg/m3 per day. We therefore adjusted the infants’ measurements for gender and gestation.
Table 1. Characteristics of the infants bom at term and their parents.
|Women|| || || || |
| Height (m)|| ||1.63||(0.06)||535|
| Early pregnancy weight (kg)|| ||64.5||(12.8)||538|
| Own birthweight (g)|| ||3318||(537)||492|
| Age (years)|| ||26.3||(4.9)||538|
|Fathers|| || || || |
| Height (m)|| ||1.78||(0.07)||528|
| Own birthweight (g)|| ||3431||(591)||435|
| Gestation (days)||281.3||281.3||(9.3)||538|
| Birthweight (g)||3529||3351||(486)||538|
| Placental weight (g)||543||524||(121)||524|
| Head circumference (cm)||35.5||34.8||(1.3)||538|
|Ponderal index (kg/m3)||26.9||27.3||(2.1)||538|
We have shown previously that women who had a high intake of carbohydrate in early pregnancy, followed by a low intake of protein, specifically dairy protein, in late pregnancy had infants with reduced placental weights16. Table 2 shows that carbohydrate and dairy protein intakes also had independent and additive associations with ponderal index. In a simultaneous analysis, ponderal index fell with higher carbohydrate intakes in early pregnancy (P= 0.01) and with lower dairy protein intakes in late pregnancy (P= 0.03). After taking account of carbohydrate and dairy protein intakes, the variance in ponderal index explained was not increased by taking account of fat or meat protein intakes. Ponderal index was unrelated to the reported severity of nausea experienced in early pregnancy.
Table 2. Mean ponderal index (kg/m3) at birth adjusted for the infant's gender and duration of gestation according to the woman's daily intake of carbohydrate in early pregnancy and dairy protein in late pregnancy. Overall SD = 2.1; values in parentheses are numbers of subjects.
|18.5||27.0 (72)||27.0 (49)||26.5 (53)||26.8 (174)|
|>18.5–26.5||27.2 (74)||27.2 (63)||26.9 (50)||27.1 (187)|
|>26.5||27.6 (34)||27.4 (61)||27.2 (82)||27.3 (177)|
|All||27.2 (180)||27.2 (173)||26.9 (185)||27.1 (538)|
Table 3 shows ponderal index and placental weight in relation to parental height and birthweight. The infants of short women were lighter and shorter but maternal height was unrelated to the infants’ ponderal indices. Women who themselves had low birthweights had infants that were lighter and shorter; they were also thinner, ponderal index falling from 28-3 to 26-2 kg/m3 as maternal birthweight fell from more than 4-0 kg to 2-5 kg or less (P < 0-0001). Maternal height and birth- weight were both associated with placental weight and in a simultaneous regression each association remained statistically significant (P= 0-01 for height and 0-002 for birthweight).
Table 3. Mean birthweight, length, ponderal index and placental weight adjusted for the infant's gender and duration of gestation according to parental height and birthweight. P values are for linear regression of continuous variables.
|Women|| || || || || |
|Height* (cm)|| || || || || |
| P|| ||< 0.0001||< 0-0001||0.7||<0.0001|
|Own birthweight (g)|| || || || || |
| P|| ||<0.0001||0.002||<0.0001||0.0001|
|Fathers|| || || || || |
|Height* (cm)|| || || || || |
|Own birthweight (g)|| || || || || |
The infants of short fathers were lighter and shorter (Table 3). There was, however, a proportionately greater decrease in length than birthweight, and the ponderal indices of infants of shorter fathers was therefore higher than that of infants of taller fathers. The infants of fathers who had low birthweight were lighter and shorter, but paternal birthweight was unrelated to the infants’ ponderal indices. While paternal height and birthweight were both associated with placental weight in univariate analyses, only birthweight was statistically significant (P= 0.001) in a simultaneous regression.
Women who smoked had infants with lower birthweights and shorter lengths than infants of nonsmoking women, but smoking was unrelated to ponderal index (27.1 kg/m3 in both groups). Primiparous women also had infants with lower birthweights and shorter lengths than the infants of multiparous women, and their ponderal indices were lower, 26.9 compared with 27.3 kg/m3 (P= 0.02). After taking account of maternal height, smoking and parity, maternal weight in early pregnancy was only weakly related (P= 0.06) to the infants’ birthweights and was unrelated to ponderal index. Although women of lower social class had smaller infants, this effect was explained by their shorter stature, and neither the women's or the fathers’ social class was related to the infants’ ponderal indices.
When analysed by simultaneous regression, taking account of gestational age, gender and parity, maternal diet (Table 2), maternal birthweight and paternal height (Table 3) were independently associated with ponderal index at birth. These associations were also independent of paternal social class and maternal social class, height and smoking.
We have shown that thinness at birth, measured by a low ponderal index, is related to a woman's diet in pregnancy, to her own birthweight and to the father's height. The women in our study were an unselected group who delivered at term and had a social class distribution that was similar to all women in England and Wales19.
Maternal dietary intakes in pregnancy were assessed using a food frequency questionnaire. Although such questionnaires can be subject to bias21, validation of the questionnaire used has shown that it gives an assessment of diet that can be used to rank the nutrient intakes of individuals20. We found that women who had a high intake of carbohydrate in early pregnancy and a low intake of dairy protein in late pregnancy tended to have infants that were thin at birth (Table 2). The effect of dairy, but not meat protein, could reflect differences in the amino acid composition of the proteins or parallel differences in the micronutrient contents of these two sources of dietary protein. Our study design did not allow assessment of the effects of dietary intakes in the weeks just prior to delivery. Studies in farm animals indicate, however, that outside the setting of extreme famine thinness at birth is commonly a consequence of poor placental development resulting from influences acting much earlier in gestation22,23. We have previously shown that the combination of high carbohydrate intakes in early pregnancy and low dairy protein intakes in late pregnancy is associated with small placental size16, perhaps a result of nutritional effects on maternal hormones and cytokines involved in the regulation of placental growth24. Our current findings suggest that, as in farm animals, nutritional constraint of human placental growth can lead to thinness at birth.
Whereas the heights of women in our study were measured in the antenatal clinic, fathers’ heights were reported by them. Resulting inaccuracies would tend to diminish associations between the father's height and birth size. Birthweights were obtained for 91% of the women and 80% of the fathers. Since our comparisons were made within the study sample, bias due to incomplete ascertainment of birthweight would be introduced if associations with ponderal index differed in infants with known and unknown parental birthweights. There is no reason to suspect this. In keeping with other studies, validation of these recalled weights against a sample of those recorded at the time showed a high level of agreement16.
While the woman's height was unrelated to the infant's ponderal index, tall fathers had thinner infants. This association does not seem to be mediated by impaired placental development since father's height is not related to placental weight independently of father's birthweight. Perhaps genetic influences associated with paternal height promote high rates of skeletal growth in the offspring which outstrip the supply of nutrients for soft tissue deposition.
In contrast to the associations with parental height it was the woman's birthweight rather than father's which was associated with ponderal index; women who had low birthweights had thin infants. This is consistent with previous studies showing that a woman's birthweight influences the weight of her infant25,26, and provides the first evidence that women whose own fetal growth was retarded tend to have thin infants. This effect could explain why some women tend to have thin infants in successive pregnancies27.
The finding that maternal but not paternal birthweight is associated with thinness is further evidence against the hypothesis that a genetic abnormality underlies both thinness at birth and later insulin resistance3,5. Our results suggest that the association between maternal birthweight and thinness at birth may be mediated through impaired placental development; low maternal birthweight was strongly associated with low placental weight, independently of maternal current height or weight. An interpretation of this is that low growth rates in utero among female fetuses may lead to impaired placentation when they reproduce, perhaps as a result of alterations in the uterine vasculature determined during fetal life. Impaired placentation may then in turn result in failure of the fetus to achieve its full growth potential or lead to actual fetal wasting28.
In conclusion, thinness at birth is known to be associated with insulin resistance and with the development of non-insulin dependent diabetes and coronary heart disease in adult life. Our results suggest that causes of a low ponderal index include impaired placental growth, a consequence of either a woman's low growth rate in utero, or imbalances in her carbohydrate and protein intakes during pregnancy. Paternally-derived genes controlling skeletal growth may also play a role. Though these effects could be of long term importance for the development of cardiovascular disease, they are not currently a basis for changing dietary recommendations to pregnant women. A strategy to increase the ponderal index of infants must include promotion of intrauterine growth in the present generation of infants for the benefit of the next one. As Mellanby29 wrote sixty years ago, “It is certain that the significance of correct nutrition in child-bearing does not begin in pregnancy itself or even in the adult female before pregnancy. It looms large as soon as a female child is bom and indeed in its intrauterine life”.
The authors would like to thank the women who gave their time; the antenatal clinic, labour ward and postnatal ward staff for their considerable assistance in the study; Mr T. Wheeler and Professor E. J. Thomas for their guidance and for allowing us to include their patients; Mr D. Howe for advice and for performing ultrasound scans. The fieldwork was carried out by S. Crofts, V. Davill, J. Hammond, L. Greenaway, S. Mitcham and S. White. The study was funded by the Dunhill Trust and the Medical Research Council. Dr Godfrey was in receipt of a Medical Research Council Training Fellowship.