点击显示 收起
1 From the Department of Zoology, University of Oxford, Oxford, United Kingdom (FM); the Center for Veterinary Medicine, Food and Drug Administration, Laurel, MD (LY); and the Division of Public Health and Primary Health Care, University of Oxford, Institute of Health Sciences, Oxford, United Kingdom (AN).
2 Supported by the UK Department of Health and The Sir Jules Thorn Charitable Trust. FM is a Royal Society Research Fellow. 3 Address reprint requests to F Mathews, Department of Zoology, South Parks Road, Oxford OX1 3PS, United Kingdom. E-mail: fiona.mathews{at}zoology.ox.ac.uk..
ABSTRACT
Background: Compromised fetal growth may program chronic diseases of adulthood, and it has been suggested that maternal nutrition is a major determinant of fetal growth. We previously found no clinically significant associations between maternal diet and the size of the infant and placenta at birth in a large cohort of white women living in the United Kingdom.
Objective: The objective was to examine the relations between indexes of maternal nutritional status in pregnancy and the birth and placental weights of infants born at term.
Design: We conducted a prospective cohort study of 798 white nulliparous women with singleton pregnancies. Blood samples were obtained at 16 and 28 wk of gestation.
Results: The concentration of most nutrients was not associated with pregnancy outcome. High retinol and hemoglobin concentrations in late, but not in early, pregnancy were strongly and independently associated with lower birth weight and smaller placental size at birth. Each 0.1-µmol increase in retinol predicted a 20.8-g (95% CI: 9.2, 32.5 g) decrease in birth weight (P < 0.001), and each 0.1-g/L increase in hemoglobin predicted a 61.5-g (95% CI: 28.5, 94.4 g) decrease in birth weight (P < 0.001).
Conclusions: We found negative associations between birth and placental weights and maternal retinol and hemoglobin concentrations. These relations may be causal or may reflect an underlying metabolic dysfunction, such as failure of plasma volume expansion. Our results provide no evidence that having high circulating nutrient concentrations, for example, through the use of supplements, would improve infant and placental growth.
Key Words: Pregnancy • nutrition • birth weight • placenta • smoking • human
INTRODUCTION
The early origins hypothesis suggests that compromised fetal development, even in clinically healthy infants, is a major determinant of the risk of chronic diseases in adulthood (1). It has been argued that maternal nutrition in pregnancy plays a pivotal role in regulating fetal growth and, hence, the infant's future risk of disease (2). We previously reported our investigation of the relation between maternal dietary intake and infant and placental size (3). In contrast with the finding of Barker and colleagues, who generated the early origins hypothesis and studied a similar cohort (4), we found no association between placental and infant birth weights and the intake of any macronutrient. After adjustment for other explanatory variables, vitamin C was the only micronutrient associated with either placental or birth weights. Even these associations were of doubtful clinical significance. However, nutrient availability may not be closely associated with dietary intake because factors such as smoking status and physiologic variations in absorption may influence circulating concentrations. Circulating nutrient concentrations are a more functionally relevant exposure. We therefore now present the results of biochemical assays made on serum samples from our cohort.
SUBJECTS AND METHODS
Study design and population
The full details of the survey methods are reported elsewhere (3, 5). The sampling frame for the study was all pregnant nulliparous women in the geographic catchment area of St Mary's Hospital, Portsmouth, United Kingdom. Regardless of their medical history, all pregnant women were referred by their family physician to hospital antenatal clinics at the time of the study. White nulliparous women attending their first hospital antenatal clinic between May 1994 and February 1996 were stratified by self-reported smoking status (smoker or nonsmoker). Simple random selection was carried out within each stratum. Of the 1002 women invited to participate, 963 were recruited; sample size calculations were reported elsewhere (3). The selection procedure meant that the prevalence of smoking among respondents was similar to nationally representative samples of pregnant women (6, 7). At their first antenatal clinic visit, 90% of subjects were between 14 and 17 wk gestation (range: 9–20 wk). Information on subject age, education, socioeconomic status, smoking habits, and use of nutritional supplements was obtained at this clinic by trained researchers using a standard questionnaire. Social class was based on the woman's most recent occupation (8). The study was approved by the hospital's local ethics committee and by the Ethics Committee of Oxford University Medical School. All subjects gave their informed consent to participate.
Blood samples for nutritional analyses and hematology were obtained from subjects at 16 wk of gestation, or "early pregnancy." The samples were taken during routine venipuncture for "triple test" genetic abnormality screening (the triple test measures -fetoprotein, human chorionic gonadotropin, and estriol concentrations), which screens for Down Syndrome and neural tube defects. A second blood sample was obtained at 28 wk of gestation, or "later pregnancy." Samples were collected into plain evacuated tubes and refrigerated immediately. The serum was stored at -70°C within 3 h of collection for samples taken at booking clinics. Blood samples taken later in gestation were processed within 24 h of collection; the delay was due to the need for blood to arrive from outlying midwives' clinics. Blood samples for routine hematology tests were collected into evacuated tubes containing EDTA. In addition to blood indexes of nutritional status, dietary intake was assessed as described elsewhere (3, 5). The main method of dietary assessment was a 7-d semiquantitative food diary kept in the week after booking, with detailed instruction and support being provided by trained personnel. The usual diet thereafter was reported in a food-frequency questionnaire mailed at 28 wk of gestation.
Infants were weighed at delivery to the nearest 5 g on digital scales (Seca 757; Seca Ltd, Birmingham, United Kingdom). Placentas were weighed to the nearest 1 g on digital scales (Soehnle Quanta; Soehnle-Waagen GmbH & Co, Murrhardt, Germany). All scales were checked weekly against standard metal weights and were calibrated if necessary. A standardized method was used to prepare the placentas for weighing: the amnion was stripped to the cord, the chorion was cut at the edge of the placenta, and the cord was removed flush with the placenta. Data on obstetric history and maternal anthropometric measurements were abstracted from the mother's hospital records after delivery. Gestational age at delivery (in days) was based on the date of the last menstrual period (LMP). However, ultrasound scan dates were used if the LMP date was unknown (n = 19) or if a scan taken at < 20 wk gave a gestational age that differed by > 14 d from that given by the LMP date (n = 70). Pediatric assessment of maturity at delivery was accepted if it differed by > 28 d from that calculated from the LMP date or ultrasound (n = 3). No gestational age estimates were missing.
Laboratory investigations
Serum ferritin and vitamin B-12 were measured with the use of a microparticle enzyme immunoassay (Abbott Laboratories, Chicago) and folate was measured with an ion capture assay (Abbott Laboratories) at the Reference Hematology Laboratory, John Radcliffe Hospital, Oxford, United Kingdom. The laboratory subscribes to the UK National External Quality Assessment Scheme. Selenium was measured in duplicate (mean used in analysis) by inductively coupled serum mass spectrometry at the SAS Trace Element Unit, Southampton General Hospital, United Kingdom (9, 10). The laboratory participates in 2 external quality-assessment schemes and performed the selenium assays for the National Diet and Nutrition Surveys (11). Lipid-soluble antioxidant assays were conducted with the use of HPLC at the Imperial Cancer Research Fund Clinical Trials Service Unit, Oxford, United Kingdom. Samples taken at different time points from the same individual were analyzed simultaneously to reduce analytic variability. Total serum cholesterol was measured with an enzymatic colorimetric method (MPR2 CHOD-PAP kit with Preciset Cholesterol Standard; Boehringer Mannheim, Lewes, United Kingdom) at the Diabetes Research Laboratories, Radcliffe Infirmary, Oxford, United Kingdom. The intraassay CVs for the assays were as follows: folate, 2.2–8.3%; ferritin, 1.8–4.9%; vitamin B-12, 3.9–4.5%; selenium, < 10%; cholesterol, < 2.0%; and lipid-soluble vitamins, 4.6–6.7%. The corresponding interassay CVs were as follows: folate, 2.2–9.3%; ferritin, 3.3–6.2%; vitamin B-12, 6.4–8.5%; selenium, 3.1–7.8%; cholesterol, < 2%; and lipid-soluble vitamins, 6.7–11.8%. Routine hematologic measurements were performed as part of the subject's standard antenatal care at St Mary's Hospital Portsmouth, which subscribes to the UK National External Quality Assessment Scheme. Blood counts were made with the use of a Coulter counter (Coulter Electronics Ltd, Luton, United Kingdom). Serum cotinine concentrations were measured with a radioimmunoassay, as described elsewhere (12, 13).
It was not possible in all cases to obtain sufficient blood for the full range of analyses, as indicated in the results. In addition, serum ferritin and vitamin B-12 were measured only in those women whose routine hematologic data were available. Some samples were not analyzed for technical reasons (principally sample hemolysis).
Data analysis
We used SPSS for WINDOWS (version 10.0; Chicago) to analyze the data. Tests of significance were two-tailed. P < 0.05 was considered statistically significant, but because of the large number of tests conducted, we interpreted P values cautiously throughout and considered values < 0.05 but > 0.01 as marginal. After a preliminary univariate analyses, the data were examined by using multiple linear regression. The fit of models was ascertained by an examination of residuals. Each model was built by using a combination of forced entry and forward stepwise procedures: where the latter was used, the criterion for entry was P < 0.05 and for removal P > 0.10. To avoid instability in the models, only cases with valid data for all of the variables in that particular analysis (no missing values) were included.
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, the residuals between the observed measurements and those predicted by linear regression were computed. These residuals were then added to the mean for the cohort. The measures presented are therefore individually adjusted to the mean gestational age and sex of the cohort. Placental weights were loge transformed to satisfy the assumptions of normality. The relations of placental and birth weights to maternal factors (including nutritional status), sex of the infant, and gestational age were examined with analyses prespecified in the protocol as described below. To reduce the number of possible exposure variables tested, the only dietary variable included in the current analyses was vitamin C, which we had previously found to be the only nutritional variable associated with either placental or birth weight (3).
The following covariates were examined as previously described (3): sex of infant (male, female), smoking status (0 = nonsmoker, 1 = smoker), number of cigarettes smoked on the day before the interview (0 and approximately equal groups: 1–8, 9–16, and =" BORDER="0">17), maternal age at first hospital visit (days), reported preconceptional weight (kg), maternal weight (kg) at first hospital visit, maternal height (m), body mass index (BMI; in kg/m2) before conception, BMI at first hospital visit (16 wk of gestation), diastolic blood pressure at first hospital visit, hemoglobin concentration at first hospital visit (g/L), social class in 3 groups (I and II, IIINM and IIIM, and IV and V, where I is the highest class), and education in 3 groups (higher than ordinary-level school-leaving examination or equivalent, ordinary level, and lower than ordinary level). Women were classified as smokers if they reported smoking or if their serum cotinine concentration was > 14 µg/L (14). Full details are given elsewhere (12, 13).
RESULTS
Nine hundred sixty-three women were recruited, and 917 of them had live singleton deliveries in Portsmouth. Sufficient blood for the nutritional analysis was obtained from 849 (93%) of these women at 16 wk of gestation, and this group did not differ significantly by age or social class from the rest of the cohort. For comparison with other studies, the 51 subjects with preterm deliveries (< 259 d gestation) were excluded, which gave a sample size of 798 for our primary analyses. Food diaries had been completed by 650 (81%) of these women (data presented previously; 3). A third-trimester blood sample was available for 641 (80%) of the women. For clarity, data were excluded from the 39 women with term deliveries who had provided a blood sample in the third, but not in the second, trimester; the failure to obtain a sample was due to an oversight by the midwife or to a collapsed vein during sampling. A flow chart of subject participation in the study is shown in Figure 1. The characteristics of the mothers and infants included in the final analysis are shown in Table 1. The median serum nutrient concentrations and hematologic indexes of mothers in early and later pregnancy are shown in Table 2. Selenium, ferritin, folate, and vitamin B-12 concentrations were measured only in early pregnancy, because of financial constraints.
FIGURE 1.. Flow diagram of subject participation throughout the study. The n values shown are maxima. The precise values depend on the sample volume of blood available for each subject and assay.
View this table:
TABLE 1. Characteristics of 798 mothers and infants
View this table:
TABLE 2. Serum nutrient concentrations and hematologic indexes of mothers in early and later pregnancy
Birth weight
Each nutrient was initially examined separately as a predictor in univariate analyses. In early pregnancy, only lutein (P = 0.015), cholesterol (P = 0.032), ferritin (P = 0.014), and folate (P = 0.002) were associated with birth weight. As reported previously, birth weight was also associated with maternal height and smoking status (3). After adjustment for smoking and height, cholesterol, ferritin, and folate were marginally significant predictors of infant weight (Table 3).
View this table:
TABLE 3. Relation of nutritional and hematologic profiles to birth weight (g) in multiple linear regression after adjustment for maternal height and smoking1
In later pregnancy, there were negative associations between birth weight and maternal lutein (P = 0.017), retinol (P = 0.003), hemoglobin (P = 0.009), and packed cell volume (PCV) (P = 0.014) concentrations. Retinol, hemoglobin, and PCV remained strongly associated with birth weight after adjustment for maternal smoking and height (Table 3). Because of the high degree of colinearity between hemoglobin and PCV, comparisons of sequential and adjusted regression models were made to identify the stronger predictor of birth weight. We found that after adjustment for hemoglobin, PCV was no longer a significant predictor (P = 0.675), and this variable was therefore not included in further analyses. Similarly, after adjustment for hemoglobin concentrations in later pregnancy, ferritin concentration in early pregnancy (used as another index of iron status) was no longer a significant predictor of birth weight (P = 0.237).
The changes in hemoglobin and retinol concentrations during pregnancy were also predictors of birth weight (P < 0.001 and P = 0.002, respectively, after adjustment for smoking and height): the larger the decrease as pregnancy progressed, the bigger the infant. No other change in nutrient concentration was significantly associated with birth weight. In neither early nor later pregnancy were there any significant interactions between maternal smoking and nutrient status.
Data on both retinol and hemoglobin were available for 558 women. In multiple linear regression analysis, which also included maternal smoking and height, each of these variables was independently predictive of birth weight after adjustment for the other (retinol: partial R2 = 0.016, P = 0.002; hemoglobin: partial R2 = 0.020, P = 0.001). Intake of vitamin C was the only dietary factor found in our previous analysis to be an independent predictor of birth weight (3). It remained a significant predictor, and the results for the other variables were unchanged, when it was entered into a simultaneous regression analysis (Table 4). Overall, the model explained 37% (adjusted R2) of the variability in raw birth weights.
View this table:
TABLE 4. Raw and standardized regression coefficients for predictors of birth weight (g) in simultaneous multiple linear regression1
Placental weight
Maternal height was the only nonnutritional variable associated with placental weight (P < 0.001). Selenium concentrations in early pregnancy were negatively associated with placental weight. After adjustment for height, each unit increase in serum selenium concentration predicted a decrease in loge placental weight of 11% (Table 5). -Carotene, cholesterol, and ferritin in early pregnancy were all also associated with placental weight, but these relations were marginal (P > = 0.02 in each case) and did not remain significant after adjustment for selenium concentrations. As for birth weight, vitamin C intake had previously been found to be the only dietary variable associated with placental size (3). The relations of both selenium and vitamin C to placental weight were essentially unchanged when they were regressed simultaneously (P = 0.029 and 0.016, respectively; n = 493).
View this table:
TABLE 5. Relation of nutritional and hematologic profiles to loge placental weight (g) in multiple linear regression after adjustment for maternal height1
Later in pregnancy, strong negative associations were found with placental weight for retinol, hemoglobin, and PCV, and there was a marginal association for cholesterol. As with birth weight, there were highly significant associations between the change in retinol, hemoglobin, and PCV during pregnancy and the size of the placenta (P < 0.001 in each case). Greater reductions were associated with larger placentas. For the other nutrients we measured at 2 stages in pregnancy, none of the differences over time were associated with placental size (P > 0.2 in each case). Because of the high correlation between hemoglobin and PCV, we conducted sequential regression analyses (having adjusted for maternal height). Hemoglobin was the stronger predictor of placental weight, with PCV no longer being significant (P = 0.662). The data for PCV were therefore not explored further.
Data on both retinol and hemoglobin in later pregnancy, together with a trimmed placental weight, were available from 516 women. In simultaneous regression analysis, retinol, hemoglobin, and maternal height were all independently predictive of loge placental weight. These relations were unchanged by the addition of dietary vitamin C into the model (Table 6). However, it is important to note that overall, maternal height, retinol, hemoglobin, dietary vitamin C, infant gestational age, and sex accounted for only 11% of the variability in placental weights.
View this table:
TABLE 6. Raw and standardized regression coefficients for predictors of placental weight (g) in multiple linear regression1
Similar results were obtained from a multiple linear regression that simultaneously adjusted for cholesterol concentration in later pregnancy rather than dietary vitamin C (n = 429: retinol, P = 0.001; hemoglobin, P = 0.001; height, P = 0.011; and cholesterol, P = 0.016). Because of the restricted number of persons with data for all variables, it was not possible to build a robust model containing both cholesterol and vitamin C.
DISCUSSION
We reported the nutritional status of a large group of pregnant white women from the United Kingdom. Most indexes of nutritional status were not associated with either the birth or placental weights of term infants. These findings are supported by our earlier work on the dietary intake of this cohort (3). However, high retinol and hemoglobin concentrations in later pregnancy were associated with lower birth and placental weights. On average, women with the lowest retinol concentrations (bottom 5%) had infants who were 160 g heavier than were those with the highest retinal concentrations (top 5%). There was a 180-g difference in birth weight for women with the lowest and highest concentrations of hemoglobin. Nevertheless, the mean weights of infants born to women with the top 5% of circulating retinol or hemoglobin concentrations were still clinically normal (3242 and 3284 g, respectively). There was no evidence that the effect of maternal smoking on birth weight was modulated by the women's nutrient status, as might be expected were the effects of smoking mainly due to oxidant exposure.
The study had a large sample size and a prospective design. One of the strengths of the cohort is that it is similar, in terms of both smoking status and socioeconomic class, to nationally representative samples of pregnant women in the United Kingdom (6, 7, 15). The study is unusual in that it collected data on a wide range of nutrients and considered both dietary intake and circulating nutrient concentrations as indexes of exposure. Care was taken to assess potential confounding variables accurately, and smoking status was validated on the basis of serum cotinine. The inclusion of only nulliparous women avoids the major confounding effect of parity, which contributes to both pregnancy outcome and health behaviors.
The main weakness of the study, as with any observational design, is the possibility that unknown confounding factors were not adjusted for. It was also not possible to investigate every nutrient: recent work suggests that n-3 fatty acids prolong gestation and increase birth weight, for example (16, 17). Finally, the relations observed between nutrition and pregnancy outcome may differ in extreme circumstances (such as in starvation or in animal experiments). Nevertheless, the issue of whether maternal nutritional status does affect pregnancy outcome in women from industrialized countries is of great clinical and public health importance and must be distinguished from that of whether such effects can be produced (18).
Debate surrounds the use of both retinol and iron supplements during pregnancy. High doses of retinol are teratogenic, and in some countries pregnant women are advised to avoid retinol-containing supplements and liver (19). However, this advice may lead to vitamin A deficiency (20). Retinol is important in cell differentiation and growth (21). We had therefore hypothesized that low serum retinol is associated with reduced fetal size, but our data suggest the opposite relation. In the only other study of this nutrient, no association was found between plasma retinol and birth weight in a sample of 423 American women (22).
Serum retinol is a relatively insensitive indicator of body vitamin A status. Only 1% of the body's reserves circulate in the serum, and homeostatic mechanisms control concentrations via retinol binding protein concentrations. It is therefore not surprising that we found no associations between dietary vitamin A and birth and placental weights (3). In contrast, fetuses have limited control over their retinol concentrations. They have low liver stores and do not synthesize retinol binding protein until late in gestation (23). Small differences in maternal serum retinol concentrations may therefore be important for the fetus.
Alternatively, the negative relation between serum retinol in later pregnancy and fetal growth may not be causal but reflects some underlying metabolic disturbance. Placental blood flow is critically important to fetal growth (24). The delivery of many hormones and metabolites to the fetus depends both on maternal concentrations and placental perfusion rates. Thus, maternal nutrient concentrations could be high, yet delivery to the fetus low, if there is poor placental blood flow. Plasma volume expansion is one of the prime mechanisms for maintaining blood flow in pregnancy. Low plasma volume expansion is associated with poor fetal growth (25), and it could also curb the normal decrease in plasma nutrient concentrations during pregnancy. Our finding that the mothers of lighter infants had smaller reductions in retinol concentrations during pregnancy supports this hypothesis. Alternatively, retinol concentrations may directly influence placental blood flow and, hence, indirectly affect placental and fetal growth. This pathway has been noted in rats: protein-energy malnutrition lowered placental and fetal growth via reduced placental blood flow (26).
A mechanism involving blood flow may also explain the results for hemoglobin. As did retinol, hemoglobin had inverse relations with infant and placental size that were apparent in later, but not in early, pregnancy. Inverse relations of hemoglobin to placental and infant size, as well as a U-shaped relation, have been reported (27-30). Hemoglobin concentrations only weakly reflect nutritional status in developed countries but are strongly influenced by plasma volume, especially in later pregnancy. At high hemoglobin concentrations, blood becomes more viscous and the efficiency of placental perfusion is reduced. Additionally, low flow rates themselves reinforce increased blood viscosity (31). One small study found that after accounting for the effects of plasma volume, the relation with birth weight disappears (31). PCV, like hemoglobin, is also influenced by plasma volume. The highly significant negative relation we found between PCV and birth and placental weights concords with several other cohort studies (31-35) and supports our blood flow hypothesis. We used ferritin as an index of iron stores: the lack of convincing associations with pregnancy outcome again suggests that the associations for hemoglobin were not primarily nutritional in origin.
Most of the other nutrients we investigated have a much stronger relation to short-term dietary intake than do retinol, hemoglobin, or ferritin (36). We therefore expect them to have a weaker relation with plasma volume, because of day-to-day variation. The imprecision of snapshot measurements may explain why there were no significant negative associations between most other nutrients and birth or placental weights, although many showed negative trends.
It is notable that retinol and hemoglobin concentrations are associated with placental weight and with birth weight. The finding that placental and birth weights are causally related strengthens our interpretation that the associations are real. Our results do not suggest that higher maternal circulating concentrations of nutrients improve fetal and placental growth among the relatively well nourished women of industrialized countries. Rather, the opposite relation was observed. Although this association may not be causal, caution should be exercised in the use of supplements containing retinol until further evidence is available. Similarly, a routine reversal of normal declines in hemoglobin in well-nourished women seems unwise, particularly because high hematocrit and hemoglobin values (35) and routine supplementation (37) are associated with low birth weight and preterm delivery. The vast majority of the variance in placental and infant weights remains unexplained, and resources should be directed toward investigating other determinants. We conclude that the maternal intake of the nutrients measured is unlikely to be an important determinant of the long-term health of infants.
ACKNOWLEDGMENTS
We thank the staff of St Mary's Hospital, Portsmouth, United Kingdom, for their support; the women who participated in the project; Bob Wenlock for helpful advice; and John Darley, Trevor Delves, Sue Manley, David Manning, Lin McRoberts, and Linda Willis for technical assistance with the data collection and nutrient assays.
FM had the original idea for the study. FM and AN devised the research questions, formulated the study design, and obtained funding. FM collected the data, performed the statistical analysis, wrote the first draft of the paper, and is the manuscript's guarantor. LY provided advice on the biochemical assays of nutritional status and supervised the assays of the lipid-soluble vitamins when she was based at the Imperial Cancer Research Fund Clinical Trials Service Unit, Oxford, United Kingdom. All 3 authors contributed to the preparation of the manuscript. None of the authors had any financial or personal interest in any company or organization sponsoring the research and had no advisory board affiliations.
REFERENCES