In view of its increasing prevalence over the past two decades, childhood obesity is now considered a serious public health concern. In the USA, the prevalence of overweight in children and adolescents increased from 11 % in 1988–94(Reference Ogden, Flegal, Carroll and Johnson1) to 17 % in 2003–4(Reference Ogden, Carroll, Curtin, McDowell, Tabak and Flegal2). Similar trends are observed in Europe(Reference Bundred, Kitchiner and Buchan3, Reference Romon, Duhamel, Collinet and Weill4) and in other parts of the world. Furthermore, the risk of overweight or obesity is unequally distributed between social classes(Reference Lamerz, Kuepper-Nybelen, Wehle, Bruning, Trost-Brinkhues, Brenner, Hebebrand and Herpertz-Dahlmann5–Reference Stamatakis, Primatesta, Chinn, Rona and Falascheti7).
The prevention of obesity in general (and childhood obesity in particular) in an ‘obesogenic’ environment has become a true challenge. There are few controlled studies on the prevention of overweight and obesity in children and adolescents. Furthermore, these studies differ in terms of their strategy, setting and duration, and thus do not allow general conclusions to be drawn concerning the value of preventive measures(Reference Summerbell, Waters, Edmunds, Kelly, Brown and Campbell8). A recent review of the literature evaluated twenty-four studies involving a total of 25 896 children; eight of the studies reported that intervention had positive effects on the prevalence of obesity(Reference Flodmark, Marcus and Britton9). However, the authors of the review were not able to determine any correlations between the type of intervention and its subsequent success (or failure). They suggested that a combination of long-term actions (promoting healthy dietary habits and physical activity throughout society) should have more powerful effects than measures limited to the school environment alone. Currently, most studies are school-based and there is a lack of reports on long-term programmes.
For over 12 years, we have had the opportunity to monitor a population exposed to a nutritional and physical activity information programme, the Fleurbaix–Laventie Ville Santé (FLVS) study. Fleurbaix and Laventie (FL) are two small towns in northern France. Three different periods can be distinguished: the first period (1992–7) focused on nutrition education in schools; the second period (1997–2002) was a longitudinal observational study on the determinants of weight changes; and the third period (2002–7) targeted physical activity and nutrition in the FL population as a whole via a series of community-based actions. The aim of the present paper is to describe BMI trends and changes in the prevalence of childhood overweight in this setting. The analysis takes into account the influence of parental social class. The prevalence of childhood overweight in the 2004–5 school year in FL was compared with that in two comparison towns (CT) with similar socio-economic backgrounds but no prevention programme in operation.
Methods
Overview of the FLVS studies
The two towns (FL) were the setting of three consecutive studies (FLVS I to III).
FLVS I (1992–7)
The aim of the FLVS I study(Reference Lafay, Vray, Boute and Basdevant10, Reference Lafay, Mennen, Basdevant, Charles, Borys, Eschwege and Romon11) was to evaluate the effects of a school-based nutrition education programme on the eating habits of the whole family. The programme consisted of both knowledge-oriented and practical approaches to nutrition. This stage was a school-based intervention and addressed children only.
The knowledge-oriented approach aimed at improving children’s knowledge of the characteristics of foods and nutrients, healthy eating habits, and food processing and labelling, and was implemented throughout the teaching syllabus. This education programme was set up by the school teachers themselves (n 44). Teachers had been trained in the basics of nutrition by nutritionists and dietitians and had developed their pedagogical methodology under the control of the Regional Board of Education (a senior counsellor was in charge of the project). During the first period there was no special focus on physical activity.
Through a range of practical, cross-cutting initiatives, the practical approach promoted pleasant, affordable and diversified food, discovery meals in school cafeterias, cooking classes, visits to farms and various food shops, and family breakfasts organized in schools with monitoring from dietitians.
The activities were originally intended to last for 5 years but are still running in all FL schools.
FLVS II (1997–2002)
FLVS II(Reference de Lauzon-Guillain, Basdevant, Romon, Karlsson, Borys and Charles12–Reference Romon, Lafay, Bresson, Oppert, Borys, Kettaneh and Charles15) was a longitudinal epidemiological study to investigate the determinants of weight and fat mass changes in the FL population. Every two years, 294 families living in FL had a clinical examination and were administered questionnaires about food habits, eating behaviour and physical activity.
This stage was supposed to be an observational study without intervention; however, the school-based interventions of the first stage were maintained and there was a steadily increasing commitment of the community at large so that progressively the intervention addressed both children and adults.
FLVS III (2002–7)
FLVS III comprised two parts. First, in 2003, a health check-up was offered at home to 3000 volunteers from the FL population. It included a fasting blood sample, a clinical examination and a questionnaire aiming at screening unhealthy habits (smoking, physical inactivity, alcohol consumption). Second, from 2004 onwards, subjects identified on the basis of this initial check-up (individuals with overweight, high blood pressure, high level of sedentary behaviour, unhealthy eating habits, hyperlipidaemia and smokers) were offered family-oriented advice on healthy eating and physical activity provided by a dietitian, who referred to the general practitioner in cases of identified health problems (including childhood obesity).
From 1999 on, the community at large became increasingly committed to the programme. Two dietitians were employed to perform interventions in schools and associations and at various meetings in the town for both children and adults. The town councils supported actions in favour of physical activity, new sporting facilities were built and sport educators were employed to promote physical activity in primary schools, walking-to-school days were organized, and family activities were also organized. Various local stakeholders (general practitioners, pharmacists, shopkeepers, sporting and cultural associations) set up family activities focused on a ‘healthy lifestyle’.
Over the whole period covered by the present report (1992–2007), regular communication including newsletters (three per annum) and press releases (one per annum) supported the project. Media interest about the programme was important during the period: 212 articles appeared in local newspapers, 190 in the medical press and 208 in the national press, and there were twenty-five television reports and seventy-five radio reports.
The comparison towns
Two CT (Bois-Grenier and Violaines, located in the same general area as FL and with similar socio-economic characteristics) were selected in 1992, when the first nutritional intervention programme began. In the same year, volunteer families in the CT participated in a dietary survey(Reference Lafay, Vray, Boute and Basdevant10). No information programmes or interventions related to nutrition or physical activity were performed in the CT over the period in question (1992–2005).
Subjects
In 1992, 2000(Reference Heude, Lafay, Borys, Thibult, Lommez, Romon, Ducimetiere and Charles16) and then annually from 2002 to 2004, body weight and height were measured in the entire population of 5- to 12-year-old children attending schools in FL. These measurements were performed during the school year, which runs from early September to early July.
Data collected in 1992 and 2000 have been described previously(Reference Heude, Lafay, Borys, Thibult, Lommez, Romon, Ducimetiere and Charles16). In 2002, 2003 and 2004, we measured 515, 592 and 633 children, respectively (corresponding to a participation rate of 95–98 % of all eligible individuals).
In the CT, a total of 349 children were measured in the 2004–5 school year (98 % of the towns’ school population).
Anthropometric methods and definition of overweight
The same group of research nurses measured all children and collected all other relevant data.
Weight was determined in light clothes (without shoes) to the nearest 0·1 kg, using either a calibrated scale for 5- and 6-year-olds or a bipodal bioimpedance device (model TBF 310; Tanita, Tokyo, Japan) for 7- to 12-year-olds. Standing height (without shoes) was measured to the nearest 0·5 cm. BMI was calculated as weight divided by the square of height (kg/m2). Overweight prevalence was determined using the gender- and age-specific cut-offs according to the International Obesity Taskforce (6 month categories) published by Cole et al.(Reference Cole, Bellizzi, Flegal and Dietz17).
Sociodemographic data
Data on parental occupation and year of arrival in FL and CT were collected in June 2005, via a telephone interview. Parental occupation was defined by the occupation of the father in a two-parent family or by that of the single parent in a single-parent family(Reference Rose and Pevalin18). The data were categorized into four classes: class I, professional and managerial occupations; class II, intermediate occupations (employees, own-account workers); class III, lower occupations (lower technical occupations, semi-routine and routine occupations); and class IV, unemployed.
Statistical analysis
Comparisons between 1992 and 2000 have been published previously(Reference Heude, Lafay, Borys, Thibult, Lommez, Romon, Ducimetiere and Charles16). Overall prevalences of overweight in 1992, 2000 and 2004 are given with 95 % confidence intervals.
Our statistical analysis focused on the three school years beginning in 2002, 2003 and 2004. BMI was compared between the different years by using a multivariate mixed model that took account of repeated observations (the SAS MIXED procedure, with an unstructured variance–covariance matrix), since some of the children were represented in the survey at several successive annual examinations. We adjusted the model for age and for height, since BMI was correlated with height in these children (Spearman correlation = 0·40 for boys and girls, P < 0·0001). The adjustment for height is discussed elsewhere(Reference Heude, Lafay, Borys, Thibult, Lommez, Romon, Ducimetiere and Charles16). Prior to these analyses, BMI values were log-transformed in order to normalize their distribution.
The odds ratios for the age-adjusted prevalence of overweight were calculated using generalized estimation equations (the SAS GENMOD procedure) which take repeated observations into account.
We also modelled the children’s BMI and overweight status in 2004 as a function of the following explanatory variables in a multivariate linear regression model: parental social status, time spent living in the intervention towns, age, gender and height.
Comparisons between FL and the CT were performed with multivariate linear regression models adjusted for height and age for weight and BMI, and with Pearson χ 2 tests for overweight prevalence (by gender and social class).
The SAS statistical software package version 9·1 (SAS Institute Inc., Cary, NC, USA) was used for all analyses. The significance threshold was set to P < 0·05.
Results
Trends in overweight prevalence in FLVS children (1992–2004)
Figure 1 shows the changes over time in the prevalence of childhood overweight (including obesity). Between 1992 and 2000, there was an increase in BMI and overweight prevalence(Reference Heude, Lafay, Borys, Thibult, Lommez, Romon, Ducimetiere and Charles16). From 2000 to 2004, there was a decrease in the prevalence of overweight which was more pronounced in girls. For girls and boys taken together, the overweight prevalence was 13·2 % (sixty-eight children), 10·5 % (sixty-two children) and 8·8 % (fifty-six children) in 2002, 2003 and 2004, respectively. For the same years, the prevalence was 9·5 %, 7·7 % and 7·4 % for the boys and 17·1 %, 13·6 % and 10·4 % for the girls. When the years 1992 and 2004 are compared, there was no significant difference in the prevalence of overweight for either boys or girls.
Changes over time in overweight (including obesity) prevalence in boys (▪) and girls (●) since the start of the Fleurbaix–Laventie Ville Santé (FLVS) study. Values are means with their 95 % confidence intervals represented by vertical bars
The adjusted anthropometric characteristics and odds for the prevalence of overweight (including obesity) are shown in Table 1. The adjusted BMI decreased significantly from 2002 to 2004, for both girls and boys. However, when compared with 2002, the age-adjusted OR for overweight in 2003 and 2004 was significantly lower only for girls.
Age and anthropometric characteristics of children in 2002, 2003 and 2004 by gender, and odds ratio for overweight in 2003 and 2004 compared with 2002: Fleurbaix–Laventie Ville Santé (FLVS) study
ref, referent category
†No adjustment.
‡Adjusted for age and taking into account repeated observations.
§Including obesity, according to gender- and-age-specific BMI cut-offs of the International Obesity TaskforceReference Cole, Bellizzi, Flegal and Dietz(17).
Multivariate analysis of the 2004 BMI data
Our analysis revealed that only parental social class (P = 0·01) and height (P = 0·001) were significantly related to BMI. Multivariate logistic regression analysis of overweight status yielded the same results (data not shown).
Comparison between intervention and comparison towns
BMI and prevalence of overweight
Mean age, weight, BMI and overweight prevalence in 2004 for boys and girls in the CT are shown in Table 2, in comparison with FL. For boys, mean age was significantly higher in the CT. When adjusted for age and height, BMI was significantly higher in the CT than in FL for both genders. The crude overweight prevalence was significantly higher in the CT than in FL for both genders. For girls and boys taken together, the prevalence was 8·8 % in FL and 17·8 % in the CT (P < 0·0001).
Age, weight, height, BMI and overweight prevalence comparisons between schoolchildren in the intervention (FL) and comparison towns (CT) in 2004: Fleurbaix–Laventie Ville Santé (FLVS) study
†Adjusted for age.
‡Including obesity, according to gender- and-age-specific BMI cut-offs of the International Obesity TaskforceReference Cole, Bellizzi, Flegal and Dietz(17).
Influence of social class
The two sets of towns did not differ in terms of social class distribution. Classes I to IV represented respectively 15·7 % (n 53), 53·2 % (n 180), 27·5 % (n 93) and 3·5 % (n 12) for the CT and 17·1 % (n 107), 52·3 % (n 327), 26·2 % (n 164) and 4·3 % (n 27) for FL (P = 0·8).
Figure 2 shows the prevalence of overweight in CT and FL as a function of parental social class. Class IV was not included in the figure because of the small sample size. In both sets of towns, there was a clear between-class difference (P = 0·003 for FL, P = 0·005 for CT), with a higher prevalence for class III in FL and for classes II and III in CT. Indeed, the disparity between the two sets of towns in terms of overweight prevalence was explained by the prevalence differences in classes II (P = 0·0004) and III (P = 0·02).
Prevalence of overweight (including obesity) in 2004 in Fleurbaix and Laventie (intervention towns; ▪) and nearby comparison towns (□) as a function of parental social class: Fleurbaix–Laventie Ville Santé (FLVS) study. The social classes were defined as follows: class I, professional and managerial occupations; class II, intermediate occupations (employees, own-account workers); class III, lower occupations (lower technical occupations, semi-routine and routine occupations). Significantly different from the comparison towns (χ 2 test): *P < 0·05, ***P < 0·001
Discussion
The aim of the present paper was to analyse recent trends in the prevalence of childhood overweight in a community which has been exposed to nutritional interventions for 12 years. In the intervention towns, after an initial rise, there was a significant decrease in mean BMI and the prevalence of overweight tended to decrease. Moreover, in 2004, the overweight prevalence was significantly lower when compared with nearby comparison towns.
There is now a growing body of data on the prevention of childhood obesity. In general, all school-based interventions aimed at preventing overweight and obesity produce some improvement in health knowledge and health-related behaviour but, to date, most have not had a significant impact on the weight status of the participating children(Reference Caballero, Clay and Davis19–Reference Veugelers and Fitzgerald23). Studies using targeted prevention strategies also improved health-related behaviour and were most successful (in terms of weight status) when parents were involved in the intervention(Reference Epstein, Valoski, Vara, McCurley, Wisniewski, Kalarchian, Klein and Shrager24). The follow-up times vary considerably from study to study. Furthermore, in most cases, the length of time over which interventions were being conducted was too short to produce an identifiable modification of weight status. More recent studies have shown encouraging results. The Kiel Obesity Prevention Study (KOPS) combined school-based interventions for overweight children and family-based interventions for their families. The preliminary results of the 4-year follow-up showed a small decrease in the prevalence of overweight in 9- to 11-year-old girls. However, the prevalence of overweight was very high in the non-intervention group (60·0 %), relative to the intervention group (46·2 %)(Reference Danielzik, Pust, Landsberg and Muller25). A recent study has published the 4-year follow-up results after a 6-year health and nutrition programme in primary schools; the mean BMI of the control group was higher than in the intervention group. However, this was a longitudinal study and the follow-up assessment was based on only 60 % of the initial study population(Reference Manios and Kafatos26). In contrast, the results of our cross-sectional study suggest that weight trends can indeed be modified at the population level.
Our results span over a 12-year follow-up period. The first interventional study started in 1992. At the time, childhood obesity was not considered to be a public health problem; in fact, the programme focused on improving nutritional knowledge and the objective was to see whether school-based information would improve dietary habits at home. Furthermore, this first study did not target increased physical activity. Over the next five years, no interventions (other than provision of nutritional information in schools) took place – only the observational study of the families was pursued. However, given that 60 % of the families were involved in this follow-up, there was ongoing involvement of schools and growing community interest. Even though the initial focus was not on obesity and physical activity, support from the town councils led to increased promotion of physical activity. The first evaluation (conducted at the end of the first intervention) showed a non-significant trend towards an increased prevalence of overweight between 1992 and 2000(Reference Heude, Lafay, Borys, Thibult, Lommez, Romon, Ducimetiere and Charles16). Taken in isolation, these initial results could have been interpreted as indicating the school-based programme’s lack of efficacy. In fact, this was the first step in an awareness-raising process for FL families – a process which has gradually intensified over time and prompted the initiation of a whole-community intervention programme in 2000. It is noteworthy that children at school in 2002 were born between 1991 and 1997, i.e. after the beginning of the first nutritional intervention in FL (1992). It is tempting to suggest that the nutrition-oriented context in FL was more effective in preventing overweight in children born after 1992 than in reversing overweight in their older counterparts. Since the two towns are now associated in the media with a healthy lifestyle, our data could also have been biased by the arrival of families who place greater importance on good health. However, our results do not reveal any differences between families living in FL at the start of the intervention and those who arrived afterwards.
Our results confirm the strong influence of parental social class on the risk of overweight and obesity. For children from high social classes, there was no difference between the CT and FL – mainly because there was no increase in overweight prevalence in this population. These results are in accordance with a previous study performed in the same area, which showed that between 1989 and 1999, there was no change in either mean BMI or overweight prevalence in 5-year-old children from the highest social class(Reference Romon, Duhamel, Collinet and Weill4). Recently, Stamatakis et al. have reported similar results in 5- to 10-year-old English children: those from higher-income households had lower odds (OR = 0·74) of being obese than their peers from lower social classes(Reference Stamatakis, Primatesta, Chinn, Rona and Falascheti7). Many earlier studies have examined the relationship with parental socio-economic status (SES); most reported an inverse correlation between SES and obesity(Reference Danielzik, Czerwinski-Mast, Langnase, Dilba and Muller27–Reference Janssen, Boyce, Simpson and Pickett29). Some studies found that the lowest overweight prevalence was observed in the two most extreme socio-economic groups(Reference Moreno, Sarria, Fleta, Rodriguez, Gonzalez and Bueno30). This lack of inter-study consistency may be explained by differences in the study populations(Reference Livingstone, McCaffrey and Rennie31) or SES indicators examined. The indicator used in the present study has been shown to be a good measure of life chances(Reference Rose and Pevalin18).
A limitation of the present work is that we did not have the data concerning the schools attended by the children, and therefore we were not able to adjust for a school clustering effect. A second limitation is that we lack the corresponding data for the CT. At the beginning of the FLVS study, an analysis of a sub-sample of families (in both FL and the CT) enrolled in a dietary habits study in 1992 did not show any differences in terms of the prevalence of childhood overweight (12·6 % in the CT and 11·2 % in FL; NS). However these results were based on reported and not measured weight and height (L Lafay, JM Borys, MA Charles and M Romon, unpublished results). It has been suggested that comparison of the impact of an intervention with a control group undergoing the same height and weight assessments could have the same sort of impact on the control children’s behaviour as the intervention itself, and may thus lead to an underestimation of the latter’s effect(Reference Summerbell, Waters, Edmunds, Kelly, Brown and Campbell8). The prevalence that we now observe in the CT is very close to the prevalence observed around the same time in other French studies. In 1999, the overweight prevalence was 16·9 % in 5-year-olds(Reference Romon, Duhamel, Collinet and Weill4) and in 2000 it was 17·9 % in 7- to 9-year-old children(Reference Rolland-Cachera, Castetbon, Arnault, Bellisle, Romano, Lehingue, Frelut and Hercberg32). In 2000, the French Ministry of Health launched PNNS (Programme National Nutrition Santé – National Nutrition and Health Programme); we cannot exclude an effect of this France-wide intervention promoting healthy eating and physical activity at the national level (mostly through nationwide media campaigns). Results from the 2006 cross-sectional French survey showed that 14·1 % of children aged 3–10 years were overweight(Reference Lioret, Maire, Touvier, Dubuisson, Dufour, Calamassi-Tran, Volatier and Lafay33). These data, together with the 2005 comparison with the CT, suggest that the decreasing prevalence observed in FL does not reflect a decreasing prevalence of overweight in the country as a whole due to the PNNS campaign.
The present study provided a good opportunity to perform long-term follow-up of a population with repeated, cross-sectional studies. Its main finding is that, at the community level, a reversal of the trend towards increasing childhood overweight can be observed. These data are in line with two recent studies(Reference Sanigorski, Bell, Kremer, Cuttler and Swinburn34, Reference Foster, Sherman, Borradaile, Grundy, Vander Veur, Nachmani, Karpyn, Kumanyika and Shults35); these studies are short-term studies and long-term maintenance of their results remains to be demonstrated. In all cases, the changes occurred in the context of a number of interventions which were not merely limited to the school environment. Furthermore, the FLVS study shows that the social classes at greater risk of overweight may indeed benefit from nutritional and health-related interventions. The absence of measurement of mediating variables prevents us from defining which aspects of the interventions were actually effective. If there is a causal link between the interventions and the decrease in overweight prevalence, it is worth emphasizing that a follow-up period of more than 10 years was needed to yield these results. The first years of the intervention were aimed at improving nutrition knowledge and did not target behaviour as such. Finally, the present study suggests that long-term results may be achieved through ongoing support from society as a whole, including parents, schools and government agencies(Reference Ockene, Edgerton, Teutsch, Marion, Miller, Genevro, Loveland-Cherry, Fielding and Briss36).
Acknowledgements
The Fleurbaix–Laventie Ville Santé Study was supported by grants from the CEDUS (Centre for Sugar Research and Information), Fournier Pharma, Lesieur, Nestlé France, Produits Roche, Go-Sports companies, and the Conseil Regional Nord Pas de Calais. The authors declare that they have no conflict of interest. M.R. conceptualized and wrote the manuscript; A.L. was involved in the design of the study and data collection; M.T. performed the statistical analyses; A.B. participated in the conceptualization and design of the study and reviewed the paper; J.M.O. participated in the conceptualization and design of the study, data analysis and contributed to writing the manuscript; J.L.B. participated in data analysis and reviewed the paper; P.D. participated in data analysis and reviewed the paper; M.A.C. participated in data analysis and contributed to writing the manuscript; J.M.B. participated in the conceptualization and design of the study and contributed to writing the manuscript. The help of D. Fraser in editorial revision of the manuscript was greatly appreciated.
