Determinants of overweight in a cohort of Dutch children

¹ From the Departments of Human Biology (NV, DLAP, ECMM, FB, PR, GH, and MSW-P) and Methodology and Statistics (ADMK), Maastricht University, Maastricht, Netherlands

² Supported by Maastricht University.

³ Reprints not available. Address correspondence to N Vogels, Maastricht University, Human Biology, PO Box 616, 6200 MD Maastricht, Netherlands. E-mail: n.vogels{at}hb.unimaas.nl.

ABSTRACT  
Background: To improve the effective prevention and treatment of obesity, it is important to focus on body weight (BW) development and its determinants during childhood.

Objective: The aim of the present study was to investigate the effects of early development, parental and genetic variables, and behavioral determinants on overweight at 12 y.

Design: In a Dutch cohort of 105 children, anthropometric measurements were conducted from birth until age 7 y. At age 12 y, anthropometric measurements were obtained again, as were measurements of body composition, leptin concentration, 3 polymorphisms, and physical activity, and the Three-Factor Eating Questionnaire was conducted. In addition, parental body mass indexes (BMIs, in kg/m²) and Three-Factor Eating Questionnaire scores were determined.

Results: The children’s mean (±SD) BMI at 12 y was 19.0 ± 2.6, and 15.2% were classified as overweight. From the first year of life, BMI tracked significantly with BMI at age 12 y (r = 0.24, P < 0.05). Linear regression analyses showed that a rapid increase in BW during the first year of life, a high BMI of the father, and a high dietary restraint score of the mother were significantly associated with overweight at age 12 y (P < 0.05). No significant genetic relation was observed. In addition, overweight was positively associated with dietary restraint of the child, and percentage body fat was negatively associated with the child’s activity score (P < 0.05).

Conclusions: In this homogeneous cohort of normal-weight to moderately overweight children, tracking of BMI during childhood took place from the first year of life. Overweight at age 12 y was predicted by an early rapid increase in BW and parental influences. Overweight during childhood may be maintained or even promoted by a high dietary restraint score and low physical activity.

Key Words: Childhood obesity • dietary restraint • physical activity • body composition

INTRODUCTION  
Childhood obesity is emerging as a major health problem (1, 2). It is associated with several risk factors for heart disease and other chronic diseases in later life (3). Most obese children remain obese as adults (4, 5), a progression that is referred to as "tracking" of overweight. Tracking is defined as the maintenance of a relative position in the population over time (6). Body weight (BW) development during childhood may be determined by early development and parental and genetic influences; however, behavioral factors, such as attitudes toward eating and physical activity, may also play an important role. With respect to early development and parental influences, catch-up growth, birth weight, breastfeeding, and parental obesity may affect carryover of obesity from childhood into adulthood (4, 5, 7-10). With respect to genetic determinants, 3 possible polymorphisms may be involved. The nuclear fatty acid receptor peroxisome proliferator-activated receptor (PPAR-) represents a direct link between adiposity, food response, and control of appetite. The glucocorticoid receptor (GRL) gene has an important role in the metabolism of adipose tissue and in the regulation of abdominal fat distribution (11). Both genes have been found to be associated with obesity (12-15). The neurocytokine ciliary neurotrophic factor (CNTF) exerts its multiple effects through a receptor complex that shares remarkable similarities with the receptor for leptin (16), yet no clear associations with obesity have been found (12, 16, 17).

BW during childhood may also be affected by dietary restrained eating behavior. Overweight may develop because dietary restraint leads to a reaction of overeating and, as a result, may cause an additional increase in dietary restraint. The parents’ restrained eating behavior may also affect this phenomenon. Dietary restraint refers to conscious restriction of food intake to achieve or maintain a preferred BW (18-20) and is reflected in a relatively high score on the cognitive restraint factor of the Three-Factor Eating Questionnaire (TFEQ) (18). Birch et al (21) found that the dietary restraint of mothers and their perceptions about their daughters’ risk of overweight predicted maternal-child feeding practices. This in turn predicted the daughters’ eating and relative weight. Finally, development of overweight during childhood may be associated with low physical activity through a reduction of energy expenditure (22, 23).

A multiple linear regression model was used to predict the combined effect of the different early development, parental, and genetic variables. Behavioral factors were tested in a second model. We hypothesized that a low birth weight, lack of breastfeeding, a high catch-up growth, various parental characteristics, and certain genotypes were related to overweight in later years. Also, behavioral characteristics of the child, eg, the activity or dietary restraint score, can affect or may be the result of overweight status. In addition, tracking of body mass index (BMI; in kg/m²) during a 12-y period in a cohort of Dutch children, of which a small percentage (15%) were overweight, was investigated.

SUBJECTS AND METHODS  
Subjects
Subjects were recruited from a Dutch cohort of white children born between 1990 and 1993 (24). As infants, these children and their mothers participated in a study of essential fatty acids during pregnancy and pregnancy outcome (24) and in a study performed between 1997 and 2000 on the long-term effects of fetal essential fatty acid availability (25). Anthropometric data, as well as information about breastfeeding, were available for the children. No interventions were provided. To evaluate the development of obesity and related variables, a follow-up study was performed in 2004. All children who participated in the previous 2 studies (n = 259) received an information letter about a new study on childhood obesity. When no response was obtained, phone calls were made in an attempt to contact the parents. We were able to contact the parents of 169 children. Of these, 60 refused and 4 retracted their initial consent. In total, 105 children participated (60 boys and 45 girls; Table 1). The BWs and BMIs of the 105 children who participated in the study did not differ significantly at birth, 1 y, or 7 y of age from those of the 154 children who did not participate. Each child and either the father or mother gave written informed consent to participate in the study, which was approved by the Central Committee Human Research in The Hague and by the Medical Ethical Committee of the Maastricht University Hospital.

View this table:
TABLE 1. Neonatal and childhood characteristics of the subjects by status of overweight at the mean age of 12 y¹

 
Study design
The children’s BW, height, BMI, body composition, leptin concentration, 3 relevant polymorphisms, TFEQ scores, and physical activity were measured by the end of 2004 when the mean age of the children was 12.4 y (age range: 11–14 y). BW, height, BMI, and TFEQ scores of both parents were determined at this time. These data and the results from the studies of Rump et al (24, 25) were used to find predictors of and associations with overweight at the mean age of 12 y.

Determinants
Anthropometric measurements
At the mean age of 12 y, the children’s height was measured with the use of a wall-mounted stadiometer (Seca model 220; Hamburg, Germany) and BW was measured with the use of a digital balance, which was accurate to 0.1 kg (Sauter D7470, Ebingen, Germany). Measurements were done while the children wore only underwear, after an overnight fast, and after the children voided their bladders. BMI was calculated as BW (in kg)/height² (in m). BMI in childhood changes substantially with age. To define normal weight, overweight, and obesity in children, we used the specific cutoffs described by Cole et al (3). They developed age- and sex-specific cutoffs for BMI for overweight and obesity in children using dataset-specific centiles linked to adult cutoffs. Waist circumferences were measured at the site of the smallest circumference between the rib cage and the ileac crest while the subjects were standing.

Body composition
Body composition was measured by using the deuterium dilution technique. ²H₂O dilution was used to measure total body water (TBW). The subjects were asked to collect a urine sample in the evening just before drinking the deuterium-enriched water solution. After ingestion of this solution, the subjects went to bed and no additional consumption was allowed for this period. Ten hours after drinking the water solution, another urine sample was collected. The dilution of the deuterium isotope is a measure of the TBW of the subject. Deuterium was measured in the urine samples with an isotope ratio mass spectrometer (VG-Isogas Aqua Sira; VG Isogas, Middlewich, England). TBW was obtained by dividing the measured deuterium dilution space by 1.04. Fat-free mass (FFM) was calculated by dividing TBW by the hydration factor 0.73 (26-28). Fat mass (FM) was determined as BW – FFM.

Leptin
Serum leptin concentrations were measured with a double antibody, sandwich-type enzyme-linked immunosorbent assay that used a monoclonal antibody specific for human leptin. The lower limit of detection is 0.5 µg/L and the upper limit is 50 µg/L. The intra- and interassay CVs were 9% and 12%, respectively. The leptin concentrations of normal-weight subjects range from 2 to 12 µg/L.

Early development determinants
BW, height, and BMI were measured at birth, 0.5 y, 1 y, 2 y, 3 y, 4 y, and 7 y of age. In addition, duration of breastfeeding or formula feeding was determined (24, 25) (Table 1). The increase in BW during the first year of life was calculated by subtracting BW at birth from BW at the age of 1 y.

Parental characteristics
Both parents reported their actual BW, which was measured at home according to our standard instructions (as described above). Height, which was originally measured with the use of a wall-mounted stadiometer, was obtained from the parents’ passports. BMI was calculated as BW (in kg)/height² (in m). In addition, the parents’ attitude toward eating was measured with the TFEQ (Table 2).

View this table:
TABLE 2. Parental characteristics¹

 
Determination of genotypes
The genomic DNAs of 119 subjects were isolated from peripheral blood leukocytes with a QIAamp kit (Qiagen, Hilden, Germany).

PPAR-2 genotyping
A 270-base pair (bp) fragment of the PPAR-2 gene was generated from genomic DNA by PCR with a forward primer, 5-GCCAATTCAAGCCCAGTC-3, and a mutagenic reverse primer, 5-GATATGTTTGCAGACAGTGTATCAGTGAAGGAATCGCTTTCCG-3, the latter of which introduces a BstU-I restriction site only when the C->G substitution at nucleotide 34 is present in relation to the Pro12Ala polymorphism (29). The PCR products were digested with BstU-I at 60 °C for 60 min, subjected to electrophoresis on a 2.5% agarose gel, and stained with ethidium bromide. The expected products after digestion with BstU-I are 270 bp for PP homozygotes, 227 and 43 bp for AA homozygotes, and 270, 227, and 43 bp for PA heterozygotes.

GRL genotyping
A 87-bp fragment of the GRL gene was generated from genomic DNA by PCR by using the forward primer 5-GCTCACAGGGTTCTTGCCATA-3 and the reverse primer 5-TTGCACCATGTTGACACCAAT-3. The latter primer includes a CG polymorphism in intron 2, 646 nucleotides downstream from exon 2 (30). The PCR products were digested with BclI at 50 °C for 60 min, subjected to electrophoresis on a 3% agarose gel, and stained with ethidium bromide. The expected products after digestion with BclI are 87 bp for GG homozygotes, 47 and 40 bp for CC homozygotes, and 87, 47, and 40 bp for GC heterozygotes.

CNTF genotyping
A 134-bp fragment encompassing the null mutation at position –6 before the second exon of the CNTF gene was generated from genomic DNA by PCR with the use of the forward primer 5-CCAGAGAGATGAGTGAGATTTTGT-3 and the reverse primer 5-CAGGTTGATGT-TCTTGTTCATGCC-3 (31). The PCR products were digested with HaeIII at 37 °C for 60 min, subjected to electrophoresis on a 2.5% agarose gel, and stained with ethidium bromide. The expected products after digestion with HaeIII are 94 and 40 bp for GG homozygotes, 134 bp for AA homozygotes, and 134, 94, and 40 bp for GA heterozygotes.

Attitude toward eating
Eating behavior was analyzed by using a validated Dutch translation of the TFEQ (18, 32); some questions were translated into an easier, understandable Dutch language for the children. The TFEQ consists of 3 factors that measure a person’s attitude toward eating. Dietary restraint (factor 1) reflects the extent to which persons attempt to cognitively control their food intake (18). Inhibition of restraint (factor 2 or disinhibition) reflects individual differences in the extent to which release from the cognitive suppression of eating occurs in response to the presence of palatable food or other disinhibiting stimuli, such as emotional distress (18). Factor 3 refers to the subjective feeling of hunger (18).

Physical activity
Physical activity was estimated with the Baecke Questionnaire (33). This questionnaire consists of 3 components: work activity, sports activity, and leisure activity. It has been validated with doubly labeled water (34). For the children, the work index was replaced by a school index with exactly the same questions.

Statistical analysis
Student t tests (for the continuous variables) and chi-square tests (for the nominal variables) were performed to determine differences in single variables between the groups (ie, lean compared with overweight). The relation of the dependent variable BMI or percentage body fat (%BF) at the mean age of 12 y was analyzed with correction for possible confounder variables by using stepwise (backward) multiple linear regression analyses (SPSS for WINDOWS, version 11.5; SPSS Inc, Chicago, IL). First, the independent variables that are not changeable by willpower (ie, age, sex, birth weight, increase in BW during the first year of life, breastfeeding or formula feeding, BMI, TFEQ outcomes of the parents, and genetic background of the child) were tested. Second, the contribution of the changeable variables (ie, activity scores and TFEQ outcomes) was determined. To test the significance of the contribution of the changeable variables, an F test was used. Correlations between BMI from birth until the age of 12 y were evaluated as Pearson correlation coefficients. All tests were two-sided, and differences were considered significant at P < 0.05. Values are expressed as means (±SDs).

RESULTS  
Anthropometric measurements
Data on BW, BMI, body composition, and leptin concentrations were collected from 60 boys and 45 girls of whom, at the mean age of 12 y, 15.2% were overweight and none were obese. Overweight status was measured by using 2 measures: BMI and %BF (35). All overweight children were overweight based on both measures. Overweight status measured with BMI was defined on the basis of age- and sex-appropriate international standards (3). Moreover, BMI was strongly positively correlated with %BF (r = 0.67, P < 0.001). Leptin concentrations were also strongly correlated with %BF (r = 0.76, P < 0.0001; data not shown), which has been shown before (36, 37).

The characteristics of the subjects (n = 105), including the differences between the children’s lean and overweight status at age 12 y, are shown in Table 1. Student’s t test analyses showed that the 16 children who were characterized as overweight (3) at age 12 y already had significantly higher BWs and BMIs at the age of 7 y than did the 89 lean children. At the mean age of 12 y, the BW, BMI, waist circumferences, %BF, FM, FFM, and leptin concentrations of the overweight children were significantly higher than those of the lean children. Their dietary restraint and inhibition of restraint scores were also significantly higher than those of the lean children.

The effects of differential early development and parental and genetic variables, corrected for age and sex, on the children’s overweight status at the age of 12 y were tested in 2 linear regression models with BMI and %BF as the dependent variables. Because there were only 16 overweight children in this Dutch cohort, logistic regression analysis with lean versus overweight as the dependent variable was not appropriate. Because the main limitation of BMI is that it does not distinguish FM from lean mass (38), and because it is excessive body fat that causes the most health consequences (35, 38), we used %BF as well as BMI as indicators of the children’s overweight status.

Early development and parental and genetic variables
The final linear regression model is shown in Model 1 (Table 3 and Table 4), with BMI or %BF as the dependent variables and the available possible early development and parental and genetic determinants as initial independent variables. From the model with BMI as the dependent variable (Table 3), we concluded that 4 variables, namely sex (P < 0.05), BMI of the father (P < 0.05), dietary restraint score of the mother (P < 0.01), and rapid increase in BW during the first year of life (P < 0.01), together predicted 23.9% of the variance in BMI at the age of 12 y. From the model with %BF as the dependent variable (Table 4), we concluded that sex (P < 0.01), BMI of the father (P < 0.01), inhibition of restraint score of the mother (P < 0.01), general feeling of hunger of the mother (P < 0.05), and rapid increase in BW during the first year of life (P < 0.05) were the variables that together predicted 29.7% of the variance in %BF at the age of 12 y. In both models, all variables, except for the general feelings of hunger of the mother, were positively associated with the dependent variable. Female sex was associated with both a higher BMI and a higher %BF. The difference between the 2 models was that the 3 different TFEQ scores of the mother were associated differently with the BMI and the %BF of the child. Neither of the 2 models showed a significant association between childhood overweight and the 3 genotypes of the PPAR-2, GRL, and CNTF genes.

View this table:
TABLE 3. Multiple linear regression analysis with BMI as the dependent variable (n = 105)¹

 

View this table:
TABLE 4. Multiple linear regression analysis with percentage body fat as the dependent variable (n = 105)¹

 
Behavioral variables
It was not clear whether behavior variables were predictors or effects of the weight status of the child. The additional contribution of behavioral variables (variables that are changeable by willpower) above that of the different early development and parental and genetic variables (variables that are not changeable by willpower) are shown in Model 2 (Table 3 and 4), again with BMI or %BF as the dependent variables. Only the final models are presented. The model with BMI as the dependent variable showed that sex (P < 0.05), dietary restraint score of the mother (P < 0.01), rapid increase in BW during the first year of life (P < 0.05), dietary restraint score of the child (P < 0.05), and inhibition of restraint score of the child (P = 0.05) were significantly associated with BMI of the child. These variables accounted for 36.5% of the variance in BMI. The increase in R² was 0.126, which implies that there was an increase in the explained variance of 12.6% when behavioral variables were added to the model. The model with %BF as the dependent variable (Table 4) showed that sex (P < 0.01), BMI of the father (P < 0.05), inhibition of restraint of the mother (P < 0.01), rapid increase in BW during the first year of life (P < 0.05), physical activity score of the child (P < 0.01), and dietary restraint score of the child (P < 0.05) were significantly associated with %BF at the age of 12 y, with an explained variance of 39.4%. This R² is 0.098 higher than in the first model for %BF; thus, when behavioral variables were added to the model, the explained variance increased by 9.8%. All variables mentioned above were positively correlated except for the activity score, which was negatively correlated with %BF. The main difference between the 2 models was the influence of BMI of the father and the activity score of the child, which both appeared to be associated only with %BF.

BMI tracking
BMI at age 12 y was significantly associated with BMI at age 1 y (r = 0.24, P < 0.05) as well as with BMIs at all other ages tested. Pearson correlation coefficients for childhood BMIs by age are shown in Table 5. The table shows that the Pearson correlation coefficient for BMI tracking from age 7 y to age 12 y was 0.76, which indicates that the BMI at age 7 y explained 58% of the variance in BMI at the age of 12 y. BMI development is presented in Figure 1, where the mean BMIs of the lean and overweight children at different time points are shown retrospectively. The division of the children into the lean or overweight group was based on their weight status at the age of 12 y, as previously described (3).

View this table:
TABLE 5. Pearson correlation coefficients (r) of BMI in children from birth to the mean age of 12 y¹

 

View larger version (16K):
FIGURE 1.. Development of body mass index (BMI; in kg/m²) in the lean (n = 89) and overweight (n = 16) children. *Significantly different from lean children, P < 0.05 (t test).

 

DISCUSSION  
The study describes the effects of differential early development, parental and genetic influences, and behavioral influences on overweight at the mean age of 12 y in a Dutch cohort of children who were born between 1990 and 1993. In addition, BMI tracking from birth until the age of 12 y was assessed.

In this cohort, 15.2% of the children were overweight (BMI: 23.8 kg/m²; %BF: 30.6%) and 84.8% were lean (BMI: 18.1 kg/m²; %BF: 17.9%). In 2003, the prevalence of childhood obesity in the Netherlands was 12%, which was the lowest in Europe (39). In the United States and the United Kingdom, for example, the prevalence is nearly double that figure (40). In the present study, girls had a significantly higher %BF at age 12 y than did boys. At this age, most girls had already reached puberty and, therefore, their body composition had changed.

With respect to the early development variables, a rapid increase in BW during the first year of life (from birth until the age of 1 y) appeared to be a significant predictor of a high BMI and a high %BF at age 12 y. A simple linear regression analysis showed that a rapid increase in BW during the first year of life was not significantly associated with birth weight (data not shown). This indicates that in this cohort of Dutch children there was no catch-up growth. Catch-up growth was earlier described by Ong et al (7), who found that children that showed catch-up growth between ages 0 and 2 y (ie, a low birth weight in combination with a rapid increase in BW thereafter) were heavier, taller, and fatter at age 5 y than were the other children. From the present study, we concluded that only a rapid increase in BW during the first year of life predicted overweight at the age of 12 y independent of whether catch-up growth occurred.

Birth weight was not significantly associated with BMI or %BF. In the literature, results are conflicting; some studies report a J-shaped or U-shaped relation, with a higher prevalence of obesity seen for the lowest and highest birth weights, which suggests a more complex association (41). Furthermore, we found no significant association between overweight and breastfeeding. Recently, other researchers reported that breastfeeding for 2 mo was protective against childhood obesity (10, 42, 43). A reason why we found no significant association in this cohort may be a sample size that is too small to uncover the effect or that there was only a relatively small percentage of overweight children. Both of these are limitations of the present study, especially the sample size, which is low compared with epidemiologic studies.

With respect to parental influences, children of overweight parents are more likely to be overweight themselves (9, 44). In the present study, the BMI of the father was positively correlated with both BMI and %BF of the child; having an overweight father in this cohort of Dutch children is therefore an important predictor of the child becoming overweight. The overall high restraint score of the mothers (data not shown) and the significant correlation between the mother’s dietary restraint score and the BMI of the child may be a possible explanation for the fact that we found no significant association between the BMI of the mother and the weight status of the child. Overweight mothers pass on their genetic background to their children. When these mothers "successfully dietary restraint" (45, 46)—ie, when their high restraint scores help them to control their weight successfully—they will have a normal weight despite their heredity. This may hide the possible relation between the BMI of the mother and the weight status of the child. The dietary restraint score of the mother was a significant predictor of the BMI of the child, and the disinhibition score of the mother was a significant predictor of the %BF of the child. Cutting et al (47) found that both the dietary restraint and inhibition of restraint scores of the mother predicted overweight of the daughter. They speculated that mother’s dietary restraint and disinhibition may serve as a source of information for daughters regarding what cues should trigger eating and how much to eat (47). Children learn a lot about food and eating from the family environment. Importantly, mothers, as the primary caretakers, provide children with meals consisting of foods that they themselves prefer or foods they believe are healthy. They also teach the child what and how much to eat (47). These maternal controls are influenced in part by the mother’s own dieting and weight-control beliefs and attitudes (21).

No significant associations between overweight and the different genotypes of the PPAR-2, GRL, and CNTF genes were observed in the children. In adults, we recently found that the GRL gene had a direct influence on weight maintenance after weight loss. For the PPAR-2 gene, there was an indirect relation (12). Both genes play an important biological role in fat cell differentiation. A possible reason why we did not observe such relations for these particular genotypes in children could be that there was no stimulated experimental situation, such as exposure to a diet (and being a successful or unsuccessful dieter). Other possibilities are the rather small sample size of the cohort and the small percentage of overweight in this Dutch cohort.

With respect to the child’s eating behavior, a high dietary restraint score in children significantly predicted overweight (both BMI and %BF). Also, inhibition of restraint was positively associated with BMI. This confirms previous observations (48, 49). Rather than a preventive method for becoming overweight (50), a high dietary restraint score in children may be interpreted as a consequence of being overweight. Lean children showed no dietary restrained eating behavior. In children, a high score on dietary restraint and inhibition of restraint, and consequently a high focus on diet, may therefore be a risk factor to become even more overweight (48).

Percentage body fat was negatively associated with physical activity, as scored on the Baecke Questionnaire. A high physical activity score, and thus energy expenditure, was associated with a low %BF. The finding of no significant correlation with BMI may be explained by the fact that BMI does not distinguish FM from lean mass (38) and therefore does not take body composition into account. Moreover, the accumulation of FM may cause one to become less active; it is very hard to define a cause or effect here. Future research should combine the use of Baecke Questionnaires with triaxial accelerometers to provide more detailed information about physical activity (51).

From the first year of life, BMI tracked significantly with BMI at the age of 12 y. This implies that BMI at a very young age may already be an important predictor of the BMI in later life, thus stressing the importance of overweight prevention in these young age categories. BMI tracking during childhood and from childhood into adulthood has been reported before (5, 8, 52, 53). The present study confirms BMI tracking, even with the low prevalence of overweight in the Netherlands. This, and the fact that we found important associations with childhood overweight in this relatively small cohort, was a strength of the present study. Another strength was the accurate method used to measure body composition.

In conclusion, BMI measured in the first year of life was a predictor of BMI at the age of 12 y. Predictors of overweight at age 12 y were an early rapid increase in BW, a high BMI of the father, and restrained eating behavior of the mother. No significant genetic relation was observed. Moreover, overweight children appeared to be more concerned about diet and had a lower activity score than did lean children, factors that may maintain or even promote childhood overweight.

ACKNOWLEDGMENTS  
We thank our subjects for their thoughtful participation in this study. We thank Loek Wouters and Wendy Sluijsmans for their assistance and Kathleen Melanson (University of Rhode Island) for editing the English text.

NV carried out the study, collected and analyzed the data, and wrote most of the manuscript. DLAP and FB (supervised by ECMM) were involved in the DNA analysis, anthropometric and body composition measurements, and the questionnaire scoring. PR (supervised by GH) collected the variables of the children from birth until 7 y. ADMK supervised the statistical analyses. The planning of the study, processing of the results, and writing of the manuscript were performed under the general supervision of MSW-P. The authors had no conflict of interest.

REFERENCES