Maternal dietary patterns during pregnancy and body composition of the child at age 6 y: the Generation R Study1,2
3 Department of Epidemiology, 4The Generation R Study Group, 5Department of Pediatrics, 6Department of Obstetrics and Gynecology, 7Department of Internal Medicine, and 8Department of Public Health, Erasmus Medical Center, Rotterdam, Netherlands, and 9Leiden University College, The Hague, Netherlands
ABSTRACT Background: Maternal diet during pregnancy may affect body composition of the offspring later in life, but evidence is still scarce. Objective: We aimed to examine whether maternal dietary patterns during pregnancy are associated with body composition of the child at age 6 y. Methods: This study was performed among 2695 Dutch motherchild pairs from a population-based prospective cohort study from fetal life onward. Maternal diet was assessed in early pregnancy by a 293-item semiquantitative food-frequency questionnaire. Vegetable, fish, and oil; nuts, soy, and high-fiber cereals; and margarine, snacks, and sugar dietary patterns were derived from principal component analysis. We measured weight and height of the child at age 6 y at the research center. Total body fat and regional fat mass percentages of the child were assessed with dual-energy X-ray absorptiometry. Results: In the crude models, statistically significant associations were found for higher adherence to the vegetable, fish, and oil dietary pattern and the nuts, soy, and high-fiber cereals dietary pattern with lower body mass index, lower fat mass index, and lower risk of being overweight, but none of these associations remained significant after adjustment for sociodemographic and lifestyle factors. We found no associations between the margarine, snacks, and sugar dietary pattern and any of the outcomes. Conclusion: Our results suggest that the associations between maternal dietary patterns during pregnancy and body composition of the child at age 6 y are to a large extent explained by sociodemographic and lifestyle factors of mother and child. Am J Clin Nutr 2015;102:873–80. Keywords: child body composition, diet during pregnancy, dietary patterns, epidemiology, fetal programming
INTRODUCTION
Childhood overweight and obesity are a major public health problem, and in 2010, the prevalence was estimated to be 11.7% in developed countries (1). Overweight and obesity during childhood can cause several health complications, including insulin resistance and high blood pressure (2). In addition, overweight and obese children are more likely to become overweight adults (3) and have a greater risk of developing chronic diseases, such as type 2 diabetes mellitus and cardiovascular diseases (4).
Maternal lifestyle during pregnancy has been suggested to influence the risk of obesity in childhood. Professor David Barker was one of the first who proposed this phenomenon of fetal programming, also known as the “Barker hypothesis” (5). Maternal diet during pregnancy is one of the lifestyle factors that may have an effect on fetal programming (6, 7). Besides a direct effect of maternal diet on fetal growth, epigenetic alterations in the fetus may change fetal metabolism, which in turn could alter growth or body composition unfavorably later in life (8). Furthermore, specific groups of mothers might have different nutritional requirements and therefore may respond differently to the effects of diet, which could result in an altered body composition of the children within these subgroups. Some studies indeed have shown an association between nutrition during pregnancy and body composition of the offspring. Higher maternal blood levels of n–6 fatty acids (9) and folate (10) have been associated with higher fat mass in the offspring at age 6 y. However, other studies did not replicate these findings (11, 12). Only one study investigated food group intake during pregnancy, which showed that a higher maternal meat intake was associated with a higher fat mass in adolescents, but no associations with vegetables, fish, fruit, and milk were found (13). All of these studies investigated single nutrients or foods, but people do not eat isolated nutrients or foods but whole diets instead. The approach of analyzing dietary pattern accounts for synergy in whole diets and also considers the degree of interaction among nutrients (14). To our knowledge, only one study investigated the association between maternal diet during pregnancy and infant body composition with dietary pattern analysis, and this study looked at 1
Supported by the Erasmus Medical Center, Rotterdam; the Erasmus University, Rotterdam; the Dutch Ministry of Health, Welfare and Sport; and the Netherlands Organization for Health Research and Development (ZonMw). VWVJ received an additional grant from the Netherlands Organisation for Health Research and Development (ZonMW VIDI: 016.136.361). 2 Supplemental Tables 1–5 are available from the “Supplemental data” link in the online posting of the article and from the same link in the online table of contents at http://ajcn.nutrition.org. 10 These authors equally contributed equally to this study. *To whom correspondence should be addressed. E-mail:
[email protected]. Received November 11, 2014. Accepted for publication July 20, 2015. First published online September 9, 2015; doi: 10.3945/ajcn.114.102905.
Am J Clin Nutr 2015;102:873–80. Printed in USA. Ó 2015 American Society for Nutrition
873
Downloaded from https://academic.oup.com/ajcn/article-abstract/102/4/873/4564532 by guest on 07 January 2019
Marion van den Broek,3,10 Elisabeth TM Leermakers,3,4,5,10 Vincent WV Jaddoe,3,4,5 Eric AP Steegers,6 Fernando Rivadeneira,3,4,7 Hein Raat,8 Albert Hofman,3 Oscar H Franco,3 and Jessica C Kiefte-de Jong3,9*
874
VAN DEN BROEK ET AL.
body composition when the infant was up to 6 mo of age (15). The latter study found no association, raising the question whether the children in the study population were too young to see the effects of fetal programming. Hence, the purpose of this study was to investigate the influence of different maternal dietary patterns during pregnancy on multiple indexes of body composition in the offspring at age 6 y. In addition, we adjusted for several potential confounders to assess whether this association was independent of sociodemographic and lifestyle factors.
Study design The present study was embedded within the Generation R Study, an ongoing population-based prospective cohort study from fetal life onward that has been described in detail (16). In short, the Generation R Study is conducted in Rotterdam, the second largest city in the Netherlands. All mothers with a delivery date between April 2002 and January 2006 were eligible. In total, 9778 mothers were enrolled in the study, of whom 7069 were included during early pregnancy. Consent was provided by 92% of the mothers for postnatal follow-up of their children. Of the live-born children, 81% still participated in the study at age 6 y. The study was conducted in accordance with the World Medical Association Declaration of Helsinki and was approved by the Medical Ethics Committee at Erasmus Medical Center, University Medical Center Rotterdam, Netherlands. Written consent was obtained from all participants. Population for analysis A flowchart of the selection process of the study population is shown in Figure 1. Because cultural differences could influence the definition of dietary patterns and the dietary assessment was designed for a Dutch population, only the subset of mothers of Dutch national origin were eligible for this study (n = 4097), with dietary data being available in 3559 mothers. Mothers who had a multiple pregnancy (n = 102), underwent abortion (n = 8), had a stillbirth (n = 16), or were lost to follow-up (n = 3) were excluded. Dietary patterns were determined in the 3479 mothers with a singleton live birth. At the child’s age of 6 y, 790 mother-child pairs did not visit the research center, leading to a population for analysis of 2689. Data of fat mass and lean mass were available in 94% of these children (n = 2520). Dietary assessment Information about dietary intake in early pregnancy (median 13.4 wk of gestation; 95% range: 9.9, 22.8 wk) was obtained with a self-administered semiquantitative food-frequency questionnaire (FFQ). This FFQ was validated in a Dutch elderly population and was adapted for use during pregnancy (17). The FFQ was validated with three 24-h recalls in a group of Dutch pregnant women in Rotterdam (n = 71), who were visiting the midwifery. The intraclass correlation coefficients for energy-adjusted macronutrient intake were between 0.48 and 0.68 (N Bergen, Erasmus Medical Center, unpublished data, 2015). The FFQ consists of 293 items and
FIGURE 1 Selection process of study population in the Generation R cohort, Rotterdam, Netherlands. FFQ, food-frequency questionnaire; PCA, principal component analysis.
comprises questions about consumption frequency, portion size, preparation method, and additions to the dish. Dutch household measures and photographs were used to make approximations about portion size (18). Mean daily nutrient intake was calculated with the use of the Dutch food composition table from 2006 (19). Principal component analysis (14) was used to determine dietary patterns. First, the 293 individual food items were reduced to 23 food groups (Table 1 and Supplemental Table 1). This division was based on the Voedselconsumptiepeiling classification (20), but some adjustments toward this division have been made to better capture specific nutrients (e.g., dividing cereals into low- and high-fiber cereals). Only factors (i.e., dietary patterns) with an eigenvalue of $1.5 were extracted to reduce bias as a result of multiple testing and to identify the most common dietary patterns in the study population. This decision was based on a combination of the scree plot, the Kaiser criterion (21), and the interpretability of the factors. The Varimax rotation was chosen because this generally leads to a number of components, with each having a small number of large factor loadings and a large number of small factor loadings (22). This enhances the interpretation of the dietary patterns because each dietary pattern represents only a small number of food groups (and thus leads to less heterogeneous dietary patterns). Subsequently, a factor loading was calculated for each food group, which illustrates the extent to which each food group is correlated with the specific dietary pattern. The 3 highest positive factor loadings per dietary pattern were used to label the dietary pattern (Table 1). For each mother, regression-based scores were extracted, which are SD scores that correspond with the similarity of one’s
Downloaded from https://academic.oup.com/ajcn/article-abstract/102/4/873/4564532 by guest on 07 January 2019
METHODS
875
MATERNAL DIET AND CHILD BODY COMPOSITION TABLE 1 Factor loadings of food groups in dietary patterns of the mothers during pregnancy (n = 3479)1 Food group
Nuts, soy, and high-fiber cereals
Margarine, snacks, and sugar
0.05 0.782 0.13 0.26 20.15 0.24 0.23 0.09 0.452 0.27 0.742 20.06 20.11 0.05 0.28 20.14 0.13 0.35 0.05 0.20 0.03 0.00 0.44
20.53 0.17 0.37 20.26 0.29 0.432 20.16 20.54 0.24 0.05 0.08 20.03 0.13 0.08 0.35 20.28 0.28 20.00 20.09 20.02 0.642 0.392 20.02
0.21 20.03 0.02 0.29 0.16 0.36 0.25 0.33 20.11 0.19 20.12 0.622 0.562 0.402 0.10 0.29 20.02 20.04 0.39 0.15 0.30 20.10 0.07
Food groups with a factor loading $0.2 or #20.2 are considered to have a strong association with a dietary pattern. 2 The 3 highest positive factor loadings per dietary pattern, which were used to label the pattern. 1
diet to one of the dietary patterns, and these scores were used as adherence scores for these dietary patterns. Subsequently, the adherence scores of the mothers of the population for analysis (n = 2689) were categorized into quartiles. Child body composition At age 6 y, all children were invited to visit our dedicated research facility in Sophia Children’s Hospital, Rotterdam, Netherlands, where measures of body composition were obtained by the use of standardized procedures by well-trained research staff. Weight was measured with light clothing and without shoes to the nearest gram with an electronic scale (SECA), and height was measured in standing position to the nearest 0.1 cm by a stadiometer (Holtain Limited). BMI was calculated as the weight in kilograms divided by the height in meters squared (kg/m2). With the use of BMI, children were classified as overweight or obese according to the cutoff points from the International Obesity Task Force (23). These cutoff points were age and sex specific. For example for 6-y-old children, a boy would classify as overweight with a BMI .17.55 kg/m2 and a girl would classify as overweight with a BMI .17.34 kg/m2. We used dual-energy X-ray absorptiometry scans (iDXA scanner; GE Healthcare) (24) to assess total body fat and regional fat mass percentages. The iDXA scanner can measure total body fat with an accuracy of ,0.25% CV (24). Children were scanned in a supine position with their feet together in a neutral position and hands flat and pronated by their sides. The fat-free mass index was calculated as [lean mass (kg) + bone mass (kg)]/[height2 (m)]. The fat mass index was calculated as [fat mass (kg)]/[height2 (m)]. Percentage body fat was calculated as
100% 3 [total body fat mass (g)]/[total body mass (fat mass + lean mass + bone mass) (g)]. The android/gynoid fat mass ratio was calculated by dividing the abdominal fat mass by the fat mass around the hips, thighs, and buttocks. For each outcome, age- and sex-specific SD scores were calculated based on the total population. Covariates Several sociodemographic, medical, and behavioral characteristics were considered possible confounders. Information regarding maternal age, prepregnancy BMI (using self-reported prepregnancy weight and height measured at intake), education (low compared with high), family income [,V2200 compared with $V2200 (US$2400)/mo], parity (nulliparous compared with multiparous), maternal smoking (never during pregnancy, until pregnancy was known, or continued throughout pregnancy), maternal continuation of alcohol during pregnancy (never during pregnancy, until pregnancy was known, or continued throughout pregnancy), folic acid supplementation (not used, started during first 10 wk, or started periconceptionally), stress during pregnancy (Global Severity Index), and vomiting and feeling nauseous (once a week or less compared with daily or a few days a week) was obtained from prenatal questionnaires sent in different trimesters. Information on sex of the child, gestational age, and birth weight was available from obstetric records assessed in hospital registries and midwife practices (16). Information regarding breastfeeding of the child (never, partially breastfed in the first 4 mo, or exclusively breastfed for at least 4 mo) was collected by
Downloaded from https://academic.oup.com/ajcn/article-abstract/102/4/873/4564532 by guest on 07 January 2019
Potatoes and other tubers Vegetables Fruits High-fat dairy Low-fat dairy High-fiber cereals Low-fiber cereals Meat Fish and shellfish Eggs Vegetable oils Margarine and butter Sugar and confections Snacks Coffee and tea Sugar-containing beverages Light drinks and water Alcoholic beverages Condiments and sauces Soups and bouillon Nuts, seeds, and olives Soy products Legumes
Vegetables, fish, and oil
876
VAN DEN BROEK ET AL.
a combination of delivery reports and postnatal questionnaires. Other postnatal questionnaires completed by the mother included information on watching television (hours) at 2 y of age and participation in sports (yes compared with no) at 6 y of age. Statistical analyses
able model. Stratified analyses were performed when a statistically significant interaction was found. Also, based on the hypothesis that growth patterns might differ between girls and boys (27), we stratified our analyses for sex of the child. For all the analyses, P , 0.05 was considered statistically significant. The statistical analyses were performed with SPSS Statistics 22.0 (SPSS, Inc.).
Models
Missing data To prevent bias due to missing data, we used multiple imputation (26) (N = 10, for details and results: Supplemental Table 2 and Supplemental Table 3) to replace missing values on covariates. Analyses were performed in each of the 10 imputed data sets separately, and final results were pooled. Additional analyses Some variables were expected to be effect modifiers, because the effects of the different dietary patterns on child body composition might differ within strata of these variables. Therefore, we tested for possible interactions between dietary patterns and maternal prepregnancy BMI, maternal folic acid use, maternal smoking during pregnancy, vomiting during pregnancy, nausea during pregnancy, and maternal energy intake by adding an interaction term (i.e., adherence score 3 stratum) to the multivari-
RESULTS
Study population We identified the following dietary patterns within our study population: vegetable, fish, and oils; nuts, soy, and high-fiber cereals; and margarine, snacks, and sugar. These 3 dietary patterns account for 25.8% of the variance in food consumption. Maternal and child characteristics of the study population are presented in Table 2. Most mothers followed higher education (62.7%) and were also mostly nulliparous (61.9%). In addition, most mothers (75.9%) did not smoke during pregnancy, but approximately half of the mothers in the study population continued TABLE 2 Characteristics of the participants (n = 2689)1 Maternal characteristic Age, y Prepregnancy BMI, kg/m2 Gestational age at enrollment, wk Educational level, n (%) Primary or secondary Higher Missing Household income, V/mo, n (%) ,22004 $2200 Missing Parity, n (%) 0 $1 Missing Smoking, n (%) Never during pregnancy Until pregnancy was known Continued during pregnancy Missing Alcohol, n (%) Never during pregnancy Until pregnancy was known Continued during pregnancy Missing Folic acid supplement use, n (%) No Started first 10 wk Started periconceptional Missing Total energy intake, kcal/d Stress during pregnancy, Global Severity Index score
Value 31.7 6 4.22 23.3 6 3.9 13.6 (9.9, 21.5)3 1019 (37.8) 1640 (61.7) 36 (1.3) 591 (23.8) 1893 (76.2) 211 (7.8) 1665 (61.9) 1026 (38.1) 4 (0.1) 1886 236 363 210
(75.9) (9.5) (14.6) (7.8)
776 413 1278 228
(28.8) (15.3) (51.8) (8.5)
203 (9.2) 734 (33.1) 1281 (57.8) 477 (17.7) 2153 6 503 0.12 (0.00, 0.77)
1 Values are absolute numbers; valid percentages in parentheses. Missing values for continuous variables were 373 (13.8%) for prepregnancy BMI and 284 (10.5%) for stress during pregnancy. 2 Mean 6 SD (all such values). 3 Median; 95% range in parentheses (all such values). 4 V2200 = US$2400.
Downloaded from https://academic.oup.com/ajcn/article-abstract/102/4/873/4564532 by guest on 07 January 2019
We used standardized scores of exposures (i.e., maternal dietary patterns) and outcomes (i.e., child body composition measures), which were both normally distributed. We also assessed the normality of the distribution of the residuals and their variance, and the distributional assumptions of the residuals for linear regression were met. All associations were then assessed in a crude model. In addition, we used multivariable linear regression to determine the association between different dietary patterns and child BMI, fat-free mass index, fat mass index, total body fat percentage, or android/gynoid fat mass ratio. Multivariable logistic regression was used to assess the association between the different dietary patterns and the risk of being overweight. The dietary patterns were analyzed in quartiles of adherence, and trend tests were performed with the adherence score as a continuous variable. Potential confounders were selected on the basis of the association with the exposure and/or the outcome. Each confounder was then entered individually into a linear regression model of dietary patterns and child BMI, total body fat percentage, or android/ gynoid fat mass ratio at age 6 y. Confounders were included in the model when they induced a change in effect estimate of at least 5% for one of the outcomes and included for all outcomes (25). Hence, the same multivariable models were used for all outcomes. A cutoff of 5% change in effect estimate was chosen, because many factors are known to influence both maternal diet and child body composition, and the study population was seen as large enough for this low cutoff. Furthermore, we created a multivariable model with and without maternal total energy intake to assess whether the association between the dietary patterns and child body composition was explained by total energy intake, because an association found between a dietary pattern that consists mainly of high-energy foods and a disease outcome may be an effect of a high-calorie intake instead of being determined by the specific foods of a pattern.
Reference 20.04 (20.13, 0.05) 20.07 (20.16, 0.02) 20.07 (20.16, 0.02) 0.21 Reference 20.03 (20.12, 0.06) 0.03 (20.07, 0.12) 0.07 (20.02, 0.17) 0.03 Reference 0.04 (20.06, 0.13) 20.03 (20.14, 0.09) 20.02 (20.17, 0.13) 0.46
Reference -0.09 (20.18, 20.00)* -0.15 (20.23, 20.06)** -0.15 (20.23, 20.06)** ,0.01 Reference 20.08 (20.17, 0.01) 20.06 (20.20, 0.02) 20.05 (20.14, 0.03) 0.44 Reference 0.04 (20.05, 0.12) 20.01 (20.10, 0.08) 20.02 (20.07, 0.10) 0.98
Multivariable
Reference 0.06 (20.04, 0.16) 20.01 (20.10, 0.10) 0.00 (20.10, 0.10) 0.66
Reference 0.03 (20.08, 0.13) 0.08 (20.03, 0.18) 0.13 (0.03, 0.23)** 0.01
Reference 0.02 (20.08, 0.12) 0.01 (20.09, 0.11) 0.07 (20.03, 0.17) 0.35
Crude
Reference 0.01 (20.10, 0.11) 20.10 (20.23, 0.03) 20.16 (20.33, 0.01) 0.01
Reference 0.01 (20.09, 0.11) 0.06 (20.05, 0.16) 0.12 (20.01, 0.23) 0.01
Reference 20.02 (20.12, 0.08) 20.04 (20.14, 0.07) 0.00 (20.10, 0.11) 0.79
Multivariable
Fat-free mass index (n = 2520)
Reference 20.06 (20.14, 0.03) 0.01 (20.08, 0.09) 0.04 (20.05, 0.13) 0.25
Reference -0.15 (20.24, 20.06)* -0.14 (20.23, 20.05)* -0.16 (20.25, 20.07)* 0.01
Reference 0.02 (20.06, 0.11) 0.03 (20.08, 0.13) 0.03 (20.11, 0.17) 0.33
Reference 20.07 (20.16, 0.01) -0.09 (20.17, 20.00)* 20.09 (20.18, 0.00) 0.30
Reference -0.16 (20.25, 20.08)* -0.22 (20.30, 20.13)* -0.24 (20.32, 20.15)* 0.01
Reference 20.01 (20.09, 0.08) 20.01 (20.08, 0.08) 20.02 (20.10, 0.07) 0.90
Multivariable
Fat mass index (n = 2520) Crude
Downloaded from https://academic.oup.com/ajcn/article-abstract/102/4/873/4564532 by guest on 07 January 2019
Values (regression coefficients with 95% CIs in parentheses) reflect differences in outcomes (age- and sex-specific SD scores) for quartiles 2–4 compared with the lowest quartile. Trend tests were performed with an adherence score (SD scores) as a continuous variable in the model. The multivariable model was also adjusted for maternal age, gestational age at dietary assessment, smoking, folic acid supplement use, continuation of alcohol during pregnancy, educational level, family income, parity, prepregnancy BMI, stress during pregnancy, child sex, breastfeeding, television watching at 2 y, and participation in sports at 6 y. *P , 0.05, **P , 0.01. Q, quartile.
1
Vegetable, fish, and oil dietary pattern Q1 low Q2 Q3 Q4 high P value, per SD Nuts, soy, and high-fiber cereals dietary pattern Q1 low Q2 Q3 Q4 high P value, per SD Margarine, snacks, and sugar dietary pattern Q1 low Q2 Q3 Q4 high P value, per SD
Crude
BMI (n = 2689)
TABLE 3 Associations between maternal dietary patterns and offspring body composition at age 6 y1
MATERNAL DIET AND CHILD BODY COMPOSITION
877
878
VAN DEN BROEK ET AL.
Associations of dietary patterns with offspring body composition
Associations of dietary patterns with child overweight/obesity Compared with the lowest quartile, quartiles 3 and 4 of the vegetable, fish, and oil dietary pattern and quartiles 2, 3, and 4 of the nuts, soy, and high-fiber cereals dietary pattern were significantly associated with a lower risk of the child being overweight in the crude model (Figure 2). After adjustment for sociodemographic and lifestyle variables, these associations were no longer significant. We observed no significant association between the margarine, snacks, and sugar dietary pattern and the child’s risk of being overweight. Additional adjustment for total energy intake did not alter these results (data not shown). Additional analyses
FIGURE 2 Associations between maternal dietary patterns and a child’s risk of being overweight at age 6 y (n = 2689). ORs (with 95% CIs) reflect the risk of being overweight or obese for quartiles 2–4 compared with the lowest quartile. The multivariable model was adjusted for maternal age at intake, gestational age at dietary assessment, smoking, folic acid supplement use, continuation of alcohol during pregnancy, educational level, family income, parity, prepregnancy BMI, stress during pregnancy, sex of the child, breastfeeding, watching television at 2 y, and playing sports at 6 y. *P , 0.05, **P , 0.01. Q, quartile.
drinking alcohol while they were pregnant (51.8%). Characteristics of the mothers and children who were excluded during the selection of the study population are shown in Supplemental Table 4.
We stratified for sex of the child. Results did not differ within strata of sex of the child (data not shown). We observed a significant interaction between maternal folic acid supplement use and the nuts, soy, and high-fiber cereals dietary pattern on BMI of the child (P , 0.01). Stratification by maternal folic acid use showed a significant association (P-trend , 0.01) between the nuts, soy, and high-fiber cereals dietary pattern and BMI of the child only in the group of mothers who started folic acid periconceptionally (P-trend = 0.01) and not in those who started in the first 10 weeks (P-trend = 0.82), or those who never used folic acid supplements (P-trend = 0.32) (data not shown). No significant interactions were found for maternal prepregnancy BMI (all P . 0.27), maternal smoking during pregnancy (all P . 0.14), maternal energy intake (all P . 0.28), vomiting during pregnancy (all P . 0.09), or feeling nauseous during pregnancy (all P . 0.32).
Downloaded from https://academic.oup.com/ajcn/article-abstract/102/4/873/4564532 by guest on 07 January 2019
Higher adherence to the vegetable, fish, and oil dietary pattern was associated with lower BMI in the offspring in the crude model (Table 3). After adjustment for lifestyle and sociodemographic factors, results were no longer significant. We found no significant associations between the nuts, soy, and high-fiber cereals or the margarine, snacks, and sugar dietary pattern and child BMI. Analyses of fat-free mass index and fat mass index revealed that the crude association between the vegetable, fish, and oil dietary pattern and lower BMI was driven by a lower fat mass index. Also, higher adherence to the nuts, soy, and high-fiber cereals dietary pattern was associated with a lower fat mass index. The linear trend (per SD of adherence score) was significant for both patterns (P , 0.01), although the results from the quartiles suggested that the difference between quartiles 1 and 2 was largest. Nevertheless, for both dietary patterns, no consistent significant associations remained after adjustment for sociodemographic and lifestyle variables (Table 3). Additional adjustment for total energy intake did not alter the results (data not shown). Supplemental Table 5 shows the associations of the dietary patterns with offspring body fat percentage and android/gynoid fat ratio. Higher adherence to the vegetable, fish, and oil dietary pattern and the nuts, soy, and high-fiber cereals dietary pattern was significantly associated with lower body fat percentage and lower android/gynoid fat ratio in the offspring in the crude model. However, the associations did not remain significant after adjustment for confounders.
MATERNAL DIET AND CHILD BODY COMPOSITION DISCUSSION
sociodemographic factors were collected at the time of birth of these children and are likely to have changed during the follow-up. To our knowledge, only one previous study investigated the association between maternal diet during pregnancy and child body composition with the use of dietary pattern analysis (15). This study did not find a significant association between maternal dietary patterns during the third trimester of pregnancy and infant weight at 4–6 mo. They did not use principal component analysis to derive a posteriori dietary patterns; instead, dietary patterns were evaluated by a priori diet scores (i.e., Alternative Healthy Eating Index for Pregnancy and Mediterranean diet) and therefore are likely to reflect optimal patterns. Thus, the focus of that study was more on a healthy diet, whereas an unhealthy diet during pregnancy might be more likely to unfavorably influence child body composition. Our study has several strengths, including a large sample size, prospective data collection, and information on a large number of potential confounders. Also, the comprehensive measures of the body composition of the children enabled us to investigate different aspects of body composition. In addition, the use of dietary patterns has the advantage that it reflects the entire diet of the mothers, thereby considering the interactions among nutrients and synergy in diets (14). We used an FFQ to assess maternal diet. An FFQ measures habitual diet and is considered a valid approach to assess dietary patterns (34). This study also has some limitations to consider. The definition of the food groups for identifying the dietary patterns was restricted by choices made in the design of the FFQ. For example, in one answer in the FFQ, both vegetables with a high and a low folic acid content were combined, but it might be interesting to make a division between them based on an earlier study that reported that a higher maternal folate concentration during pregnancy was associated with a higher adiposity in the offspring at age 6 y (10). Nevertheless, for other food groups, it was possible to make a division, such as the dairy products food group, which we divided into high fat and low fat. Another limitation of the study is that a self-reported diet is prone to measurement error, because it relies on memory and people are likely to underestimate their nutritional intake (35). Usually, this leads to an underestimation of any true association. We corrected for this by adjusting for total energy intake in an additional model, which reduces potential systematic measurement error (36), and this did not alter our results. As in many prospective studies, we had missing data on our covariates, which may lead to attrition bias. However, we used multiple imputation to impute missing values, which has shown to be a reliable method to minimize attrition bias (37). Furthermore, during the selection of the study population, 34.2% of the participants were excluded. The 1402 mothers who were excluded were on average lower educated (50.7% had only primary or secondary education compared with 37.8% of the 2695 mothers who were included), were more likely to continue smoking during pregnancy (24.1% compared with 14.6% in the study population), and were less likely to start folic acid supplements periconceptionally (49.2% compared with 57.8% in the study population). Thus, our study population has a selection toward a more healthy population, and results in other populations might not be similar. In a population where unhealthy dietary patterns are more clearly present, an association between maternal diet and child’s body composition might be easier to detect. Nevertheless, our novel findings suggest that prevention of obesity in children should not focus solely on maternal diet during pregnancy but that other factors, such as maternal BMI, smoking,
Downloaded from https://academic.oup.com/ajcn/article-abstract/102/4/873/4564532 by guest on 07 January 2019
In a population-based prospective birth cohort, we observed that adherence to a vegetable, fish, and oil; nuts, soy, and highfiber cereals; or margarine, snacks, and sugar dietary pattern during pregnancy was not independently associated with child body composition at age 6 y. All significant associations in the crude models of the vegetable, fish, and oil and the nuts, soy, and high-fiber cereals dietary patterns were no longer significant after adjustment for sociodemographic and lifestyle factors. Maternal diet is one component of the intrauterine environment that is critical to fetal development (8). In previous studies, a maternal Western diet (28) and wheat products diet (29) during pregnancy were shown to be associated with a lower birth weight, which in turn could alter later body composition (30, 31). On the basis of the phenomenon of fetal programming, which emphasizes prenatal nutrition as a key determinant for the increased risk of diseases later in life (5), we hypothesized that maternal diet during pregnancy could have a substantial influence on a child’s body composition. There might be several explanations why the present findings do not correspond with the hypothesis. Dietary patterns derived from principal component analysis are data driven and thus differ across populations. In our population, the most pronounced dietary patterns were characterized by high intake of healthy components. A more unhealthy diet, such as a Western diet, which is characterized by an excessive amount of saturated fat and carbohydrates, may have a larger impact on a child’s body composition beyond other lifestyle and sociodemographic factors. An animal study demonstrated that 10-wk-old rats of mothers fed a junk-food diet during gestation and lactation had an increased body weight (32). In addition, previous studies on the influence of maternal nutrient intake on child body composition showed that food components that are seen as more unhealthy, such as high intake of n–6 fatty acids or glucose, are associated with a higher BMI at age 3 y (33) or fat mass at age 6 y (9), respectively. The margarine, snacks, and sugar dietary pattern identified in our study has some similarities to a Western diet, but the pattern is also characterized by some relative “healthier” foods, such as a higher intake of nuts, seeds, olives, and high-fiber cereals, which may explain the discrepancy between our results and those from other studies. Another explanation that our findings do not correspond with the hypothesis could be that previously reported associations between maternal nutrient intake during pregnancy and child body composition may not be independent. Observational studies on diet and body composition are very susceptible to bias, because sedentary behavior, physical activity, and other factors that are related to child body composition are often clustered with specific dietary patterns. It is thus likely that these previously observed associations can at least partly be explained by other characteristics. In our study, all significant associations in the crude models were no longer significant after adjustment for sociodemographic and lifestyle factors. Other studies that did find associations between maternal diet during pregnancy and child body composition did not always adjust for maternal covariates, such as maternal folic acid use (9, 13) or prepregnancy BMI (10, 13), which were important confounders in our analyses. In addition, the study that reported the association between a high meat intake and a high fat mass in 16-yold adolescents (13) may have the benefits of the longer follow-up period, but this also increases the risk of residual confounding. More confounding factors could be involved, and some of these factors may need to be measured more frequently. For example,
879
880
VAN DEN BROEK ET AL.
The authors’ responsibilities were as follows—MvdB, ETML, and JCK-dJ: designed and conducted the research; VWVJ, EAPS, FR, HR, AH, and OHF: provided essential materials; MvdB and ETML: performed statistical analyses; MvdB, ETML, and JCK-dJ: wrote the manuscript: VWVJ, EAPS, FR, HR, AH, and OHF: critically revised the manuscript; JCK-dJ: had primary responsibility for the final content; and all authors: read and approved the final manuscript. ETML, OHF, and JCKdJ work in ErasmusAGE, a center for aging research across the life course funded by Nestlé Nutrition (Nestec Ltd.), Metagenics, and AXA. Nestlé Nutrition (Nestec Ltd.), Metagenics, and AXA had no role in design and conduct of the study; collection, management, analysis, and interpretation of the data; and preparation, review or approval of the manuscript. The rest of the authors declared no conflicts of interest related to this study.
REFERENCES 1. de Onis M, Blo¨ssner M, Borghi E. Global prevalence and trends of overweight and obesity among preschool children. Am J Clin Nutr 2010;92:1257–64. 2. Lobstein T, Baur L, Uauy R. Obesity in children and young people: a crisis in public health. Obes Rev 2004;5:4–104. 3. Singh AS, Mulder C, Twisk JWR, Van Mechelen W, Chinapaw MJM. Tracking of childhood overweight into adulthood: a systematic review of the literature. Obes Rev 2008;9:474–88. 4. Reilly JJ, Kelly J. Long-term impact of overweight and obesity in childhood and adolescence on morbidity and premature mortality in adulthood: systematic review. Int J Obes (Lond) 2011;35:891–8. 5. Barker DJP, Osmond C. Infant mortality, childhood nutrition, and ischaemic heart disease in England and Wales. Lancet 1986;1:1077–81. 6. Cetin I, Berti C, Calabrese S. Role of micronutrients in the periconceptional period. Hum Reprod Update 2010;16:80–95. 7. Harding JE. The nutritional basis of the fetal origins of adult disease. Int J Epidemiol 2001;30:15–23. 8. Wu G, Bazer FW, Cudd TA, Meininger CJ, Spencer TE. Maternal nutrition and fetal development. J Nutr 2004;134:2169–72. 9. Moon RJ, Harvey NC, Robinson SM, Ntani G, Davies JH, Inskip HM, Godfrey KM, Dennison EM, Calder PC, Cooper C. Maternal plasma polyunsaturated fatty acid status in late pregnancy is associated with offspring body composition in childhood. J Clin Endocrinol Metab 2013;98:299–307. 10. Yajnik CS, Deshpande SS, Jackson AA, Refsum H, Rao S, Fisher DJ, Bhat DS, Naik SS, Coyaji KJ, Joglekar CV. Vitamin B12 and folate concentrations during pregnancy and insulin resistance in the offspring: the Pune Maternal Nutrition Study. Diabetologia 2008;51:29–38. 11. Lewis SJ, Leary S, Davey Smith G, Ness A. Body composition at age 9 years, maternal folate intake during pregnancy and methyltetrahydrofolate reductase (MTHFR) C677T genotype. Br J Nutr 2009;102:493. 12. Hauner H, Much D, Vollhardt C, Brunner S, Schmid D, Sedlmeier EM, Heimberg E, Schuster T, Zimmermann A, Schneider KT, et al. Effect of reducing the n–6:n–3 long-chain PUFA ratio during pregnancy and lactation on infant adipose tissue growth within the first year of life: an open-label randomized controlled trial. Am J Clin Nutr 2012;95:383–94.
13. Yin J, Quinn S, Dwyer T, Ponsonby AL, Jones G. Maternal diet, breastfeeding and adolescent body composition: a 16-year prospective study. Eur J Clin Nutr 2012;66:1329–34. 14. Hu FB. Dietary pattern analysis: a new direction in nutritional epidemiology. Curr Opin Lipidol 2002;13:3–9. 15. Poon AK, Yeung E, Boghossian N, Albert PS, Zhang C. Maternal dietary patterns during third trimester in association with birthweight characteristics and early infant growth. Scientifica 2013;2013:786409. 16. Jaddoe VW, Duijn C, Franco O, Heijden A, Iizendoorn M, Jongste J, Lugt A, Mackenbach J, Moll H, Raat H, et al. The Generation R Study: design and cohort update 2012. Eur J Epidemiol 2012;27:739–56. 17. Klipstein-Grobusch K, den Breeijen JH, Goldbohm RA, Geleijnse JM, Hofman A, Grobbee DE, Witteman JCM. Dietary assessment in the elderly: validation of a semiquantitative food frequency questionnaire. Eur J Clin Nutr 1998;52:588–96. 18. Donders-Engelen M, Heijden van der L, Hulshof KF. Maten, gewichten en codenummers. Wageningen (Netherlands): Human Nutrition of TNO and Wageningen University; 2003. 19. Netherlands Nutrition Centre. Nevo: Dutch food composition database 2006. The Hague: Netherlands Nutrition Centre; 2006. 20. Ocké MC, Hulshof KFAM, Buurma-Rethans EJM, Brants HAM, Drijvers JJMM, ter Doest D. Voedselconsumptiepeiling 2003. Samenvatting werkwijze en evaluatie. Bilthoven (Netherlands): RIVM; 2004. 21. Hatcher L. A step-by-step approach to using the SAS system for factor analysis and structural equation modeling. Cary (NC): SAS Institute; 1994. 22. Kaiser H. The varimax criterion for analytic rotation in factor analysis. Psychometrika 1958;23:187–200. 23. Cole TJ, Bellizzi MC, Flegal KM, Dietz WH. Establishing a standard definition for child overweight and obesity worldwide: international survey. BMJ 2000;320:1240–3. 24. Kaul S, Rothney MP, Peters DM, Wacker WK, Davis CE, Shapiro MD, Ergun DL. Dual-energy X-ray absorptiometry for quantification of visceral fat. Obesity (Silver Spring) 2012;20:1313–8. 25. Mickey RM, Greenland S. The impact of confounder selection criteria on effect estimation. Am J Epidemiol 1989;129:125–37. 26. Rubin DB, Schenker N. Multiple imputation in health-care databases: an overview and some applications. Stat Med 1991;10:585–98. 27. Komiya S, Eto C, Otoki K, Teramoto K, Shimizu F, Shimamoto H. Gender differences in body fat of low-and high-body-mass children: relationship with body mass index. Eur J Appl Physiol 2000;82:16–23. 28. Knudsen VK, Orozova-Bekkevold IM, Mikkelsen TB, Wolff S, Olsen SF. Major dietary patterns in pregnancy and fetal growth. Eur J Clin Nutr 2008;62:463–70. 29. Okubo H, Miyake Y, Sasaki S, Tanaka K, Murakami K, Hirota Y. Maternal dietary patterns in pregnancy and fetal growth in Japan: the Osaka Maternal and Child Health Study. Br J Nutr 2012;107: 1526–33. 30. Loos RJ, Beunen G, Fagard R, Derom C, Vlietinck R. Birth weight and body composition in young adult men—a prospective twin study. Int J Obes Relat Metab Disord 2001;25:1537–45. 31. Barker M, Robinson S, Osmond C, Barker DJP. Birth weight and body fat distribution in adolescent girls. Arch Dis Child 1997;77:381–3. 32. Bayol SA, Farrington SJ, Stickland NC. A maternal ‘junk food’ diet in pregnancy and lactation promotes an exacerbated taste for ’junk food’ and a greater propensity for obesity in rat offspring. Br J Nutr 2007;98: 843–51. 33. Deierlein AL, Siega-Riz AM, Chantala K, Herring AH. The association between maternal glucose concentration and child BMI at age 3 years. Diabetes Care 2011;34:480–4. 34. Hu FB, Rimm E, Smith-Warner SA, Feskanich D, Stampfer MJ, Ascherio A, Sampson L, Willett WC. Reproducibility and validity of dietary patterns assessed with a food-frequency questionnaire. Am J Clin Nutr 1999;69:243–9. 35. Lutomski JE, van den Broeck J, Harrington J, Shiely F, Perry IJ. Sociodemographic, lifestyle, mental health and dietary factors associated with direction of misreporting of energy intake. Public Health Nutr 2011;14:532–41. 36. Freedman LS, Schatzkin A, Midthune D, Kipnis V. Dealing with dietary measurement error in nutritional cohort studies. J Natl Cancer Inst 2011;103:1086–92. 37. Sterne JAC, White IR, Carlin JB, Spratt M, Royston P, Kenward MG, Wood AM, Carpenter JR. Multiple imputation for missing data in epidemiological and clinical research: potential and pitfalls. BMJ 2009;338:b2393.
Downloaded from https://academic.oup.com/ajcn/article-abstract/102/4/873/4564532 by guest on 07 January 2019
educational level, and continuation of alcohol consumption during pregnancy, may be more important targets for obesity prevention. In addition, our findings raise the question regarding external validity of the Barker hypothesis in relatively healthy populations and highlight the importance for future researchers in this area to be aware of confounding bias. In conclusion, our results do not support the hypotheses that maternal dietary patterns during pregnancy are independently associated with body composition of the offspring at age 6 y when taking into account sociodemographic and lifestyle factors of both mothers and their children. It might be worthwhile to further investigate the influence of maternal dietary patterns during pregnancy on child body composition in other populations or subgroups because dietary patterns may differ between populations and considering the limited number of studies on this topic.