Influence of maternal nutrition on outcome of pregnancy: prospective cohort studyBMJ 1999; 319 doi: https://doi.org/10.1136/bmj.319.7206.339 (Published 07 August 1999) Cite this as: BMJ 1999;319:339
- Fiona Mathews, university research officer (, )
- Patricia Yudkin, university lecturer,
- Andrew Neil, university lecturer
- Division of Public Health and Primary Health Care, Institute of Health Sciences, University of Oxford, PO Box 777, Oxford OX3 7LF
- Correspondence to: F Mathews Department of Zoology, University of Oxford, Oxford OX1 3PS
- Accepted 25 May 1999
Objective: To investigate the relations of maternal diet and smoking during pregnancy to placental and birth weights at term.
Design: Prospective cohort study.
Setting: District general hospital in the south of England.
Participants: 693 pregnant nulliparous white women with singleton pregnancies who were selected from antenatal booking clinics with stratified random sampling.
Main outcome measures: Birth and placental weights at term.
Results: Placental and birth weights were unrelated to the intake of any macronutrient. Early in pregnancy, vitamin C was the only micronutrient independently associated with birth weight after adjustment for maternal height and smoking. Each ln mg increase in vitamin C was associated with a 50.8 g (95% confidence interval 4.6 g to 97.0 g) increase in birth weight. Vitamin C, vitamin E, and folate were each associated with placental weight after adjustment for maternal characteristics. In simultaneous regression, however, vitamin C was the only nutrient predictive of placental weight: each ln mg increase in vitamin C was associated with a 3.2% (0.4 to 6.1) rise in placental weight. No nutrient late in pregnancy was associated with either placental or birth weight.
Conclusions: Concern over the impact of maternal nutrition on the health of the infant has been premature Maternal nutrition, at least in industrialised populations, seems to have only a small effect on placental and birth weights. Other possible determinants of fetal and placental growth should be investigated.
Placental and infant birth weights were not associated with the intake of any macronutrient early or later in pregnancy
After adjustment for the effects of maternal height and smoking, only vitamin C independently predicted birth weight. The expected mean difference in birth weight for infants with mothers in the upper and lower thirds of intake was about 70 g
Vitamin C was the only nutrient that independently predicted placental weight, but again this relation was of doubtful clinical significance
Among relatively well nourished women in industrialised countries, maternal nutrition seems to have only a marginal impact on infant and placental size. Other causes of variation in the size of clinically normal infants should now be investigated
Barker and colleagues have shown strong associations between infant and placental size and the risk of later chronic disease, such as cardiovascular disease and diabetes.1 2 On the basis of data from animal studies and cross country comparisons, poor maternal nutrition has been implicated as one of the key “adverse environmental influences in utero,” which could lead to compromised fetal and placental growth and adverse long term consequences.1 Their observational studies of British women also suggest that maternal diet is an important determinant of infant and placental size.3–5 The “Barker hypothesis” has led to calls for improvements in maternal diet, which have generated some concern among obstetricians.6
Although there is widespread recognition of the importance of adequate maternal nutrition during pregnancy in developing countries, there is considerable uncertainty about its role in industrialised countries, where profound malnutrition is uncommon. Even near starvation, such as occurred during the “Dutch hunger winter,” reduced mean birth weight by only 300 g.7 The evidence from adequately nourished populations is conflicting Observational studies have found only weak and inconsistent associations between intake of macronutrients and infant size,3 5 8 9 and few data are available for micronutrients.9–11 In supplementation trials, micronutrients have not been shown to have important impact on mean birth or placental weights.10–13 Protein and energy supplementation have produced both increases and decreases in birth weight, with high density protein supplements seeming to reduce birth weight.14 15 The results of such trials are difficult to generalise, however, as intakes are often increased well beyond normal levels. In addition, trials are usually designed to detect changes in adverse outcomes of pregnancy rather than in birth weights within the normal range. We therefore conducted a large scale observational study in an attempt to elucidate the role of maternal nutrition in pregnancy. We set out to detect differences in birth weight between women with high and low intakes of nutrients early in pregnancy.2
Participants and methods
Full details of the survey methods are reported elsewhere.16 Briefly, healthy white nulliparous women attending antenatal clinics in Portsmouth between May 1994 and February 1996 were stratified by smoking status, and simple random selection was carried out within each stratum (details on request). Of the 1002 women invited to participate, 963 were recruited. The sampling procedure resulted in the prevalence of smoking among respondents being similar to nationally representative samples of pregnant women.17 18
To detect a 150 g difference in mean birth weight (SD 500 g) between the two extreme thirds of nutrient intake (two sided α=0.05, power=90%) required a sample of 750 women. The expected difference was based on the results of a pilot study. To allow for non-completion of the food diary we aimed to recruit 1000 women.
Structured interviews were conducted by trained researchers during visits to antenatal clinics. Social class was based on the woman's most recent occupation.19 Height, weight, blood pressure, and blood count were as measured routinely. Women were classified as smokers if they reported smoking or if their serum cotinine concentration was greater than 14 ng/ml.20 21
The main method of dietary assessment was a 7 day semiquantitative food diary completed the week after booking.16 Usual diet since booking was reported in a food frequency questionnaire mailed at 28 weeks' gestation. The questionnaire was based closely on that used by the European prospective investigation of cancer (EPIC).22 Reliability tests indicate that diary data correspond better to weighed records than do estimates based on a food frequency questionnaire 23; indeed, the relatively low accuracy and reproducibility of food frequency questionnaires has been widely reported.24 25 This method was nevertheless used in late pregnancy as a second food diary would have compromised participation rates.
Babies were weighed at delivery to the nearest 5 g. Placentas were weighed to the nearest 1 g after the amnion was stripped to the cord, the chorion cut at the edge of the placenta, and the cord removed flush with the placenta.
Over the ranges studied, placental and birth weights showed linear relations with gestational age and were also associated with sex. For clarity, and for comparison with other studies,3–5 individual measurements were adjusted to the mean gestational age and sex of the cohort. All subsequent analyses used these adjusted values.
The relations of placental and birth weights to maternal factors, sex of the baby, and gestational age were examined with analyses prespecified in the protocol. Means were compared by using z tests and proportions by using χ2 tests. Tests of significance were two tailed. The sociodemographic variables considered were smoking status; cigarettes smoked in day before interview (0 and approximately equal groups: 1-8; 9-16; ≥17); maternal age at booking (days); reported weight before conception (kg); maternal weight (kg) at booking; maternal height (m); body mass index (kg/m2) before conception; body mass index at booking; diastolic blood pressure at booking; haemoglobin concentration at booking (g/dl); social class in three groups (I and II; III non-manual and III manual; IV and V); and education in three groups (higher than O level (>GCSE); O level (GCSE grades A-C); and less than O level (GCSE grades D and E)).
The fit of multiple linear regression models was ascertained by examination of residuals. Placental weights were ln transformed to satisfy the assumptions of normality; total intakes of all micronutrients were ln transformed to reduce the leverage of outlying values. Transformation did not materially affect the results. Maternal variables were considered for inclusion if they were significantly (P<0.05) associated with nutrient intake or pregnancy outcome in univariate analysis or if this association had been reported elsewhere. Each model was built by using a combination of forced entry and forward stepwise procedures (criterion for entry was P<0.05 and for removal P>0.10).
Of the 963 women recruited, 917 had live singleton deliveries in Portsmouth. Food diaries were completed by 739 (80.6%) of these women (“respondents”). To permit comparisons with the results of Godfrey et al 4 5 the 46 respondents who delivered before 259 days' gestation were excluded. The results obtained from the entire cohort did not differ from those presented for term deliveries, and no nutrient was associated with preterm delivery.21
Table 1 shows the characteristics of the 693 mothers and babies. The nulliparous women in our cohort were sociodemographically similar to those studied by Godfrey et al 4 5 and their social class distribution was comparable with a nationally representative sample of mothers.26
Dietary intakes —Table 2 shows nutrient intakes (from food and supplements) in early and later pregnancy Nutrient intakes from food alone as well as totals including supplements are shown for iron and folate as supplements made a substantial contribution to intakes for these nutrients.
Maternal characteristics and placental and birth weights —In univariate analyses maternal height, weight before conception, weight at booking, self reported smoking status, and smoking status validated by cotinine measurements were all predictive of birth weight (P<0.001). In multiple regression, however, the only independent predictors of birth weight were smoking status validated by cotinine measurement (P<0.001) and maternal height (P<0.001). Smoking predicted a 104 g (95% confidence interval 47 g to 161 g) decrease in birth weight, and each 10 cm of additional height predicted a 172 g (129 g to 215 g) increase in birth weight. In univariate analyses placental weight was associated with maternal height (P=0.033), weight before pregnancy (P=0.002) and weight at booking (P=0.001) but with no other maternal characteristics. In multiple regression maternal height was the only independent predictor of placental weight (P<0.001), with each 10 cm of additional height predicting a 5% (2% to 8%) increase in placental weight.
Nutrient intakes in early pregnancy and birth weights —Birth weight was positively associated with intakes of vitamin C, vitamin E, and total folate but with no other nutrient (table 3). After adjustment for maternal smoking and height, birth weight remained associated only with vitamin C (P=0.031). There was no interaction between smoking and vitamin C intake. These findings were unaltered by simultaneous adjustment for energy intake.
Nutrient intakes in early pregnancy and placental weights —Placental weight was positively associated with vitamin C, vitamin E, and total folate (table 4) and remained so after adjustment for maternal height. Simultaneous adjustment for energy intake did not alter these relations. As intakes of vitamin C, vitamin E, and folate were correlated, the independent effect of each nutrient was investigated. After adjustment for vitamin C intake no other nutrient independently predicted placental weight.
Nutrient intakes later in pregnancy, birth weight, and placental weight —The 624 respondents to the food frequency questionnaire were similar to the whole cohort in their nutritional intakes early in pregnancy, placental and infant weights, and associations between early pregnancy nutrition and outcome. No nutrient in late pregnancy was significantly associated with any outcome, and this remained true after adjustment for other maternal factors.
In this large and detailed study we found no clinically important effects of maternal nutrition on the placental or birth weight of infants born at term. Dietary intake and smoking status were measured as accurately as possible, and account was taken of maternal characteristics that might confound associations between dietary intake and infant size. The inclusion of only nulliparous women also removed the possible confounding effect of parity.
Vitamin C, but no other nutrient, was positively related to birth weight, with about a 100 g difference between the lowest and highest thirds of intake. The significance of this relation, however, was considerably reduced after adjustment for smoking and maternal height. Vitamin C showed some association with placental weight, but again this relation was weaker after adjustment for maternal height As in previous research, there was no association between any nutrient in later pregnancy and placental or birth weights.3 4 9
Comparison with Barker's data
Our findings differ from those of Barker and colleagues, who studied a cohort with similar social class and age distributions. In early pregnancy, Godfrey et al found significant negative relations between energy intake and placental and birth weights.4 These relations were largely due to strong associations between the outcomes and carbohydrate intakes. It is notable that the median values and the variability of all nutrient intakes, particularly carbohydrate and energy, were much higher than in our study. For example, the median intakes of energy and carbohydrate were 9.8 MJ (2346 kcals) and 303 g, respectively4 compared with 8.5 MJ (2044 kcals) and 256 g in our cohort. The interquartile ranges for carbohydrate and energy consumption were about two thirds greater (125 g v 74.5 g and 3.8 MJ v 2.3 MJ, respectively). This increased variability resulted mainly from an extended right hand tail in the distribution of intakes.
It would be surprising if these differences reflected a greater intrinsic variability in the diets of women from Southampton compared with those in Portsmouth. Not only are the two coastal cities extremely close geographically but they are of similar size and sociodemographic structure. Indeed, there is considerable population exchange between the two conurbations. We are also confident that our lack of positive findings was not the result of greater error in measurement. There is considerable evidence that our main method of dietary assessment—a 7 day food diary—suffers less measurement error than food frequency methods.23–25 It is notable that both the median values and variability in nutrient intakes in our study with food diaries were similar to those obtained by Barker's group when their subjects completed a 4 day food diary; their data, however, were not analysed in relation to outcome measures.27 Further, the variability in our data was almost the same as that for women in the national dietary and nutritional survey of British adults (7 day weighed diary).28
In later pregnancy we, like Godfrey et al, used a food frequency questionnaire. In neither study was any individual nutrient found to predict pregnancy outcome. Godfrey et al, however, found that low intake of meat protein later in pregnancy, in conjunction with high carbohydrate early in pregnancy, was associated with reduced birth and placental weights.4 This observation contrasts with their earlier report that diets with a high proportion of energy from animal protein were associated with reduced birth weight.3 In our study, combinations of nutrients were not investigated, given the lack of main effects for carbohydrate, protein, and total energy; the high correlation between carbohydrate and protein intakes (0.62 in diary and 0.72 in food frequency questionnaire); and the likelihood of “positive” findings due to multiple significance testing.
Maternal malnutrition may be an important determinant of fetal growth in developing countries. Our work suggests that among the reasonably well nourished women of industrialised countries, however, maternal diet in pregnancy has, at most, a small impact on placental and birth weights. We are currently analysing stored serum samples from our cohort to help to clarify the relation of maternal nutrition to the outcome of pregnancy.
We thank Dr R Smith for performing the cotinine analyses; Mr J Bevan and colleagues for permission to include their patients in the study; the midwifery and support staff at St Mary's Hospital Portsmouth for their assistance; Mrs L Willis and Mrs L McRoberts for help in data collection and management; Dr T Key and colleagues for permission to base the food diary and the food frequency questionnaire on those used in the EPIC study; Mr R Fraser for helpful comments; and all the women who participated in the research.
Contributors: FM was the principal investigator. She designed the study, obtained funding, supervised research staff, collected data from pregnant women, measured infants, coded and entered food diaries, performed statistical analyses, and prepared the paper for publication. PY was involved in design of the study and had a major input to the statistical analysis, interpretation of results, and writing of the paper. AN participated in designing the study, obtaining funding, and supervision of the project and contributed to interpretation of the results and preparation of the paper. FM is the guarantor.
Funding The Sir Jules Thorn Charitable Trust and Oxford University Medical Research Fund.
Competing interests None declared.