点击显示 收起
1 From the Department of Pediatrics, Children's Hospital, CHU-Toulouse, France (BJ, MD, and MT); INSERM U558, Epidemiology and Public Health, Toulouse, France (BJ, NA, CC, and HG); LN Pharma, Toulouse, France (LN-P); and the Department of Clinical Pharmacology, CHU-Toulouse, France (LN-P)
2 Supported by grants from the French Social Security System. 3 Reprints not available. Address correspondence to N Ahluwalia, INSERM U558, Hôpital Paule De Viguier, 330 Avenue de Grande Bretagne, TSA 70034, 31059 TOULOUSE CEDEX 9, France. E-mail: ahluwalia.n{at}chu-toulouse.fr.
ABSTRACT
Background: Pediatric overweight and obesity are becoming an epidemic worldwide, which indicates the need for formulating preventive programs and policies during a child's early years.
Objective: We identified factors associated with overweight in young children in southwestern France.
Design: Children [n = 1780;
Conclusions: A family history of overweight or diabetes, overweight in the first 2 y of life, and television viewing are associated with overweight at 4 y. These factors should be considered in developing programs for the prevention of overweight in early childhood.
Key Words: Children • preschoolers • overweight • obesity • risk factors • school-based study
INTRODUCTION
The epidemic of obesity and associated diseases poses a serious public health challenge worldwide (1). A major concern is the dramatic and continuing increase in the prevalence of overweight and obesity in children in most developed countries. For instance, between 1963 and 2002, the prevalence of overweight increased markedly, from 4.2% to 15.8%, in 6–11-y-olds in the United States (2). Overweight and obesity in childhood are associated with substantial psychosocial consequences as well as with increased cardiovascular disease risk (ie, hypertension, dyslipidemia, and hyperinsulinemia) (3). Furthermore pediatric overweight and obesity can have long-term health consequences in adulthood such as metabolic syndrome, obesity, and associated complications (4-8). Therefore, early prevention of overweight is emerging as an important strategy to reduce associated short- and long-term morbidity.
Excess body weight usually results from a complex interaction of genetic and lifestyle factors. Few studies have been conducted, however, on factors associated with overweight in young children (aged <6 y). Well-designed studies with sufficient statistical power, considering several risk factors simultaneously, and including potential confounding factors in the design are even more limited in this age group. Furthermore, few studies have examined the influence of birth weight and growth characteristics in infancy on subsequent overweight (9-11). The identification of risk factors is critical to develop effective intervention programs for the prevention of childhood overweight and obesity. Therefore, our interest was to determine the factors associated with overweight in young children in southwestern France. Specifically, we examined the association of overweight with sociodemographic factors, parental and grandparental characteristics concerning obesity and chronic disease, and the child's nutrient intake, television-viewing practices, and participation in sports. Furthermore, the association of overweight during early childhood (at birth, 9 mo, and 24 mo) with overweight at 4 y was also examined.
SUBJECTS AND METHODS
Subjects
The present study was conducted after obtaining permission from the Inspector of Academy, Ministry of Education, responsible for the Haute-Garonne department. The study was carried out in kindergarten because most children >3 y of age attend these schools in France. One hundred and fifty-six kindergartens in the Haute-Garonne department were contacted to participate in the study; 79 agreed to participate. Flyers and information packets were then sent to the parents or legal guardians of all 3–4-y-old children (n = 3736) via their teachers.
The project was submitted to the Ethical Committee for Biomedical Research ("Comité Consultatif de Protection des Personnes dans la Recherche Biomédicale"), Toulouse, France, and was forwarded to the National Committee for Information and Liberty ("Commission Nationale de l'Informatique et des Libertés") for approval. The study protocol and informed consent documents were then approved by this committee. Fifty-one percent (n = 1910) of the parents or legal guardians indicated an interest for their child to participate in the study and provided a signed informed consent. One hundred and thirty children did not come to the medical visit; therefore, the final cohort consisted of 1780 subjects (Figure 1).
FIGURE 1.. Flow diagram of subject enrollment and available data.
Study design
A study packet was mailed to the parents of participating children (n = 1910); it contained several questionnaires concerning 1) medical history for themselves, their parents, and the participating child; 2) the child's participation in organized sports and television-viewing practices; and 3) the child's 3-d dietary record. Parents were asked to fill out these questionnaires and were requested to send them along with the child's official medical record (Carnet de Santé) on the day of the medical examination at school. In France, as part of the well-child practice, physicians are required to record medical information concerning the child at birth, 9 mo, and 24 mo on the child's official medical record. This information includes gestational age, weight, and height at birth as well as BMI at subsequent visits.
On the day of the medical visit, the study physician copied the data from the child's official medical record and determined weight and height. Each child's weight was recorded in minimal clothing without shoes to the nearest 0.5 kg by using a Seca scale (Seca, Hamburg, Germany) and height was recorded to the nearest 0.5 cm with a stadiometer (Seca) fixed to the wall by using standardized techniques. Children were classified overweight or obese by the study physician with the use of French reference curves for BMI (12). Those with a BMI 90th percentile (z score 1.28) were considered overweight and those with a BMI 97th percentile (z score 1.90) were considered obese. Parents of children considered overweight or obese were given a letter to receive follow-up care from their family practitioner and were offered to be included in the health professionals network for the prevention and treatment of pediatric obesity [Réseau ville-hôpital pour la prévention et la prise en charge de l'obésité pédiatrique RéPOP]. Considering the gestational age, height, and weight at birth recorded on the official medical record, children were classified into 3 categories: small for gestational age (SGA), appropriate for gestational age (AGA), and large for gestational age (LGA) on the basis of the French reference Audipog standards (13).
Data concerning family medical history, television-viewing habits, and participation in organized sports practices
Parents were asked to report their weight and height as well as the existence of obesity (yes or no) in the child's maternal and paternal grandparents. They also provided information regarding the presence of diabetes, hypercholesterolemia, and hypertension for themselves and for the child's paternal and maternal grandparents. Parents also reported whether the child participated in organized physical activity or sports (yes or no), the kind of sport or activity practiced, the time spent (h/wk) on the activities, and the number of hours per day their child watched television.
Dietary assessment
Parents were asked to keep a dietary record for 3 d, including a weekend day, in the study booklet. Detailed written instructions were given in the booklet, which included an example of a child's intake for one day, covering all food groups, methods of preparation, and amounts consumed at various meals (breakfast, lunch, dinner, midmorning and -afternoon snacks, as well as snacks consumed at any other hour). Estimates of portion sizes were also provided as pictures in the booklet. Details concerning the meals consumed at school were obtained from the cafeteria staff by the study dietitians. In total, dietary records for 1252 children were received; 58 were illegible. All dietary records from overweight children (n = 98) and a randomly selected subset of dietary records on 50% of the nonoverweight children (n = 548) were analyzed by using 4-D Client PROFIL IV software Profil dossier Version VI (Société Audit et Conseil en Informatique Médicale, Saint Doulchard, France). This sample size allowed detecting a difference of at least one-third SD between the overweight and nonoverweight children for each sex (with the use of two-tailed tests, = 0.05; statistical power = 0.95).
Sociodemographic factors
The school location provided 2 variables that served as proxies for sociodemographic status: region (urban or nonurban) and schooling in underprivileged area (yes or no). Schools in Toulouse and its suburbs were considered as urban and those in villages as nonurban. The sample consisted of primarily white children (95%). We used the French Ministry of National Education's definition of "zones for prioritizing education" to identify schools located in underprivileged areas.
Statistical analysis
Statistical analyses were conducted by using SAS version 8.0 (SAS Institute, Cary, NC). All continuous variables examined in the study were normally distributed. Multivariate logistic regression analysis was conducted to examine the relation of child's sex, sociodemographic factors, and medical history variables of parents and grandparents with the risk of overweight at 4 y. Univariate associations were examined by using chi-square tests; when P < 0.20, variables were entered into a stepwise logistic regression model by using the backward option (n = 1026). Results are reported as odds ratios (ORs) and 95% CIs. Similar approaches were taken to examine the association of overweight at 4 y with early growth indicators, ie, growth status at birth and BMI at 9 and 24 mo (n = 1339), and with television viewing and participation in organized physical activity (n = 593). The interaction between sex and other main effects were tested and, when significant, the models were rerun for each sex separately.
Analysis of variance (ANOVA) was carried out to examine the differences in nutrient intake as well as in energy intake from various meals. The model tested the main effects of weight status (overweight compared with not overweight) and sex (boys compared with girls) as well as the interaction between weight and sex. Because the interaction effect (weight status x sex) was significant for several variables examined, we conducted these analyses separately for each sex by using Student's t test. Statistical significance was set at P < 0.05.
RESULTS
The study cohort consisted of 1780 subjects (899 boys and 881 girls) with a mean (±SD) age of 3.9 ± 0.4 y. Prevalence of overweight was 9.1%, of which 3.8% were considered obese, by using French reference standards. By using the international cutoffs (14) recommended by the International Obesity Task Force (IOTF), the prevalence estimates for overweight and obesity in the current study were 6.9% and 1.8%, respectively. A significant association between sex and overweight was observed with chi-square analysis: 7% of boys and 11.2% of girls were overweight (P < 0.01). Most (68.7%) of the children lived in urban areas (Toulouse and its suburbs), and only16.5% attended kindergarten in areas considered as underprivileged.
Among the demographic and family history variables examined, sex, region (urban or rural), school in underprivileged area (yes or no), as well as the presence of overweight in parents and the number of diabetic or obese grandparents were associated with overweight at 4 y (P < 0.20) and are therefore presented in Table 1. No significant association between overweight at 4 y and the presence of diabetes, hypertension, or hypercholesterolemia in parents was observed; the frequency of these conditions was 0.9%, 1.8%, and 7.1%, respectively, for fathers and 0.8%, 1.5%, and 4.7%, respectively, for mothers. In the multivariate analyses, being female, having an overweight mother or both parents overweight, and having 1 diabetic grandparent were independently associated with overweight at 4 y (Table 1). In multivariate analyses, overweight at 4 y was not significantly associated with weight status at birth; however, it was significantly associated with overweight at later ages (9 and 24 mo) (Table 2).
View this table:
TABLE 1. Sociodemographic factors and family medical history–related factors associated with overweight at 4 y
View this table:
TABLE 2. Association of neonatal growth status and BMI at 9 and 24 mo with overweight at 4 y1
For the most part, nutrient intakes did not differ significantly between overweight and nonoverweight children for either sex; importantly, however, overweight boys had significantly greater energy and lipid intake than did nonoverweight boys (Table 3). This was substantiated by higher energy intakes at most main meals of the day, namely lunch and dinner. Interestingly, no significant differences in energy intakes from snacks were noted for either boys or girls (Table 4).
View this table:
TABLE 3. Nutrient intakes by weight status for boys and girls1
View this table:
TABLE 4. Energy intake at various meals by weight status for boys and girls1
Only a subsample of parents (n = 593) responded to questions on television viewing and participation in organized sports or physical activities for their children. Only 2.4% of children never watched television and 27.5% watched >1 h/d. No significant sex differences were noted for television viewing (1 or >1 h). About 25% children participated in some kind of organized sport activity such as gymnastics, exercise, swimming, or dancing. Girls tended to participate in organized sports more than did boys (32.2% compared with 18.1%, respectively; P < 0.05). In a multivariate analysis, there was a significant interaction between sex and participation in organized sports; therefore, models were run separately for boys and girls (Table 5). Viewing >1 h television per day was associated with overweight in both sexes, but participation in organized sport activities was associated with overweight in girls only, without any relation with the time spent on these activities.
View this table:
TABLE 5. Association of television viewing and participation in organized physical activity with overweight at 4 y1
DISCUSSION
Overweight and obesity in children can be affected by several factors, such as parental and grandparental characteristics, birth and growth related variables, and environmental factors. Most studies have addressed the association of overweight with these factors individually; however, to our knowledge, few studies have simultaneously determined several risk factors for overweight including diet in young children <6 y of age (9, 15). Thus, the current study yields important information on factors associated with overweight at the young age of 4 y.
This cross-sectional study of 4-y-olds was based on voluntary participation of kindergarteners. Although the sample was not representative, it included children from both rural and urban areas and covered schools in both underprivileged and other areas spanning most socioeconomic situations. The prevalence of obesity in the current study was lower than that reported in 4-y-old British children (16) and in 2–5-y-old American children (17). It is recognized that the use of different reference standards renders the comparison of prevalence estimates across countries difficult. By using the IOTF definition of overweight, the prevalence of overweight in the current study remained lower than that reported in studies conducted in 4–6-y-olds in other countries that used the same international cutoffs (18, 19). To our knowledge, the only comparable data in France are based on a national survey in 1994, which reported the prevalence of overweight and obesity of 14.2% and 2.4%, respectively, in 3–14-y-old children with the use of the IOTF definition. These higher estimates may be related to the wider age-range of children examined in that study (20) and perhaps to the fact that our study was conducted on a convenience sample in southwestern France.
The chief objective of the present study was to identify factors associated with overweight in 4 y olds considering several factors simultaneously. Overweight was associated with female sex in the current study; this finding has been reported in some (10, 21, 22), but not all (9, 17), studies. The association of parental overweight with overweight in children has been reported consistently (9, 10, 23). In the current study, the association of maternal overweight remained significant even after control for other risk factors in the multivariate analysis, which is consistent with the findings of Padez et al (11) and Stettler et al (22). However, the association with paternal overweight was no longer significant.
Few studies have examined the trends in intrafamilial overweight over 3 generations. In the current study, although parent-declared obesity in one or more grandparent was associated with overweight in children in the univariate analysis, as reported also by Polley et al (24), this association did not remain significant in the multivariate analysis. Interestingly, however, having at least one diabetic grandparent remained significantly associated with overweight even after accounting for other risk factors. It is plausible that diabetes in grandparents acts here as a proxy for obesity and represents a measure less subject to reporting error. To our knowledge, this finding is novel and underlines the importance of genetics and shared lifestyle practices across generations in affecting body weight in children (24, 25).
The association of childhood obesity with growth indicators at birth and in early childhood is of growing interest (9, 11, 22, 26). Our finding that neither SGA nor LGA was significantly associated with overweight at a later age is consistent with the finding by Neville and Walker (26). In separate analyses, we also examined the association of birth weight with overweight at 4 y and found a positive association; OR (95% CI) for birth weight, per 1000 g, was 1.6 (1.2–2.3). This association was no longer significant after control for the family characteristics listed in Table 1 and for overweight at 9 and 24 mo (data not shown). Our data did not enable us to examine the relation of weight gain before 2 y of age with overweight at 4 y. However, an association between weight gain from birth to 1 or 2 y and subsequent overweight has been reported (9, 27). Our finding that children who were overweight at either 9 or 24 mo were more likely to remain overweight at 4 y is noteworthy; those children who were overweight at both 9 and 24 mo had the greatest risk of remaining overweight. A similar association between weight at 8 mo and at 18 mo with obesity at a later age (7 y) was recently reported (9). Taken together, these findings suggest that efforts for obesity prevention in childhood need to be initiated at very young ages.
Lifestyle factors, such as dietary intake practices, television viewing, and physical activity, have been recognized as important predictors of body weight. Interestingly, we did not find major differences in either macro- or micronutrient intakes between overweight and nonoverweight children; this was particularly true for girls. Note that overweight boys, however, had significantly greater daily energy intake (by 100 kcal/d) than did their normal-weight counterparts, and this difference was observed for the main meals of the day (lunch and dinner) but not for snacks. Dietary factors have not been associated with measures of body fat, BMI, and overweight in the few studies that have been conducted in preschoolers (9, 15, 28, 29) even though such associations have been reported in older children. Atkin and Davies (28) suggested that the relation between macronutrients and adiposity may develop over years; thus, the age group examined in the current study (4 y) may have been too young to show such an effect. The association of energy intake with overweight in boys but not girls observed in the current study may be related to cultural expectations and differences in attitudes toward overweight in girls and boys.
Our finding of a positive association between overweight and television viewing is consistent with the findings of other studies conducted in young and older children (9, 11, 30, 31). However, it is in contrast to the findings of Jago et al (15) who reported an association between television-watching and BMI only when children were 6–7-y-old and not at earlier ages. In the current study, children who watched television >1 h/d tended to participate less often in organized sport activities than did those who watched less television (chi-square: 3.6; P = 0.06); however, television viewing was not associated significantly with total energy intake or with energy intake from snacks.
Physical activity has generally been negatively associated with body fat and BMI in older children (28, 31, 32). We did not examine physical activity patterns per se, but qualitatively assessed child's participation in organized sports as reported by parents and found that it was positively associated with overweight in girls only. This is in contrast to the lack of association reported in a small study (n = 41) of 3–7-y-olds (33) and a negative association reported in older children (34, 35) but consistent with the findings of Jago et al (15) in children who were 3–4 y of age. The surprisingly positive association between practicing organized sports and overweight at 4 y could be partly explained by the fact that parents of overweight children are more prone to enroll their children in organized sports activities to prevent their child from becoming obese. In our study, this parental attitude seems to affect girls only, who tended to be overweight more often than boys. We considered the possibility of low response rate on the physical activity questionnaire leading to bias; however, the responders and nonresponders did not differ significantly in any of the sociodemographic and family characteristics, birth weight–related variables, or lifestyle factors (diet and television viewing), with the exception of region (nonresponse rate was 55% in rural areas compared with 73% in urban areas)—a variable which was not significantly associated with overweight (Table 1).
Some limitations of the study must be considered when interpreting the findings. Data on several variables were based on the declaration of parents. Such data are less precise and can lead to misclassification. Statistically, this increases the probability of missing a true association but is unlikely to induce a false significant relation. The cross-sectional nature of the study did not enable us to draw inferences concerning causes and effects. Future longitudinal studies involving these factors are therefore warranted to establish the temporal nature and causality of these associations. Furthermore, the study population was composed primarily of white subjects, which did not allow us to examine the trends across different ethnic backgrounds.
In conclusion, the current cross-sectional study showed the importance of hereditary and lifestyle factors (parental overweight; diabetic grandparents; overweight at 9 mo, 24 mo, or both; and television viewing) as risk factors for overweight at 4 y. The association of energy consumption with overweight was significant for boys but not for girls. Prevention of overweight and obesity in children is generally focused on change in lifestyle during later childhood or adolescence. Our results suggest that intervention may be indicated earlier in infancy and during the toddler years to tackle the increasing prevalence of obesity.
ACKNOWLEDGMENTS
We thank Pierre Machicot and Richard Cosaert from the Caisse Primaire d'Assurance Maladie (social security) of Haute Garonne region for their assistance with subject recruitment. We thank Francoise Collet, Danielle Augier, Danielle Leautier, and Claudine Guitard from the Protection Maternelle et Infantile (Well-child program), Haute Garonne, for their assistance in obtaining information from children's medical records. We thank Nadia Lounis, Johanne Galino, and Corinne Fleury for their assistance in data collection.
The authors' contributions were as follows—MT: design of study, funding for the study, overseeing study, and contributing to the manuscript development; HG: design of the study, supervision of data analysis and contributing to the manuscript development; BJ: overseeing study and writing of the manuscript; LN-P: overseeing the study and contributing to manuscript development; MD: contributing to data collection and analysis, and contributing to the manuscript development; CC: conducted data analysis under the supervision of HG and NA and contributed to the manuscript development; NA: study design, supervision of data analysis, and writing of the manuscript. None of the authors had any conflict of interest.
REFERENCES