The objective of the study was to describe the prevalences of obesity in French adults over a 9-year period. Mailed questionnaire surveys, in 1997, 2000, 2003, and 2006, sampled 20,000 representative French households by the method of quotas. Weight, height, and waist circumference were reported by all members of the selected households ≥18-years. Obesity was defined according to the WHO criteria, BMI >30 kg/m2. The prevalence of adult obesity increased progressively from 8.6% (95% confidence interval: 8.2–8.8) in 1997 to 13.1% (12.7–13.5) in 2006. The increase affected all ages, socioeconomic strata, and regions. Although the prevalence of obesity increased in parallel in men and women from 1997 to 2003, the rate of increase was lower in men between 2003 and 2006. These surveys showed a sharp increase in the prevalence of obesity in France in recent years contrasting with a stable prevalence in the 1980s. The results of the first Obepi surveys prompted the French government to implement a Nutrition and Health National Plan in 2001. Regular monitoring of obesity prevalence in France and neighboring countries is needed to compare future trends.
Between 1980 and 1991 (1), there was only a modest increase in the prevalence of obesity in French adults at the national level, which remained ∼6.5%. This contrasted with the trends documented in neighboring countries such as the United Kingdom (2). A protective effect of the French lifestyle was hypothesized, with the pleasurable and social aspects of eating and the importance of food quality (3). Four surveys with the same methodology, in 1997, 2000, 2003, 2006, document the prevalences of obesity in France, on national representative samples.
Methods and Procedures
At the four 3-yearly surveys, mailed questionnaires were sent by a polling institute, TNS Sofres, to a sample of 20,000 households. This sample is a permanent sample constituted on a voluntary basis to answer various questions. It is selected to be representative of the French population by quotas, based on the distributions of the closest national Institut National de la Statistique et des Études Économiques reference for age class, sex, occupational class, region, and size of city of residence. Five hundred and fifty new households are incorporated each month into this permanent sample replacing 550 households withdrawn because of repeated nonanswers (75%), willingness of the participants (17%), or exceeding the 10-year lifetime limit in the sample (8%).
Measures and definitions
All members of the households aged ≥15 years were asked to measure their weight, height, and waist circumference with the help of another household member whenever possible, before writing the measures on the questionnaire. Detailed illustrated instructions were provided to explain how to measure waist circumference and height. Height was measured against a wall, and the waist circumference, at the level of the umbilicus, both with a tape measure.
Weight and height were used to calculate the BMI. Overweight and obesity were defined according to the WHO definitions (4) as, respectively, 25 ≤ BMI < 30 kg/m2 and BMI ≥30 kg/m2. Obesity was further categorized into moderate (30 ≤ BMI < 35 kg/m2), severe (35 ≤ BMI < 40 kg/m2), and morbid (BMI ≥40 kg/m2). Net monthly household income per capita was computed as the net household income (reported data) divided by the weighted number of subjects in the household with weights of 1 for adults ≥18 years, 0.75 for adolescents (15–17 years), and 0.5 for children (<15 years).
Of the 20,000 questionnaires sent, 73.4, 71.4, 69.6, and 66.6 % were returned in 1997, 2000, 2003, and 2006, respectively. After exclusion of pregnant women, those aged <18 years and people with outlying and missing data, BMI was available for 26,595, 25,659, 24,402, and 22,374 individuals and waist circumference for 24,988, 22,990, 21,083, and 18,896 individuals, respectively.
Comparing subjects included in the analysis (responders with BMI available) with those not included (nonresponders or responders but BMI not available) by χ2 tests showed similar biases for each survey: a significant under-representation of subjects under 55 years, of farmers and of inhabitants of the Paris region and a significant overrepresentation of retired subjects (results not shown). To correct for these differences induced by nonresponse, each sample was weighted using the Raking Adjusted Statistics method (5). Standard deviations took into account the increase in variance resulting from the weighting as proposed by Kish (6). Prevalences of overweight and obesity are given for each survey with 95% confidence intervals.
Prevalences of overweight and obesity
There was a progressive increase in the prevalence of obesity between 1997 and 2006 (Table 1). To analyze whether the changes in the structure of the French population over the 9 years could account for all or part of the observed increased in the prevalence of overweight and obesity, the 2000, 2003, and 2006 data were standardized to the 1997 population age-distribution. The results showed only minor changes, between the original prevalences and those standardized on the 1997 distribution: 10.0% (95% confidence interval: 9.6–10.4), 11.7% (11.2–12.2), 12.9% (12.4–13.4) in 2000, 2003, and 2006, respectively.
Table 1. Prevalences (95% confidence intervals) of overweight (excluding obesity) and obesity by age-class in subjects ≥18-years in the French 1997, 2000, 2003, and 2006 Obepi surveys
Changes in waist circumference
Parallel to the increase in obesity prevalence, there was an increase in the mean waist circumference. In the 1997, 2000, 2003, and 2006 surveys, the weighted means (95% confidence intervals) for men were: 91.2 (91.0–91.4), 92.4 (92.2–92.6), 93.3 (93.1–93.5), and 93.8 (93.5–94.1) cm, and for women 79.7 (79.5–79.9), 81.6 (81.4–81.8), 82.8 (82.6–83.0), and 84.2 (83.1–84.5) cm.
Relationship between sociodemographic factors and obesity
The prevalence of obesity was similar in men and women between 1997 and 2003 and increased between 1997 and 2003 in both sexes (Table 1). Between 2003 and 2006, the prevalence of obesity continued to rise in women from 11.9 to 13.6% (P < 0.05) but the increase was more modest in men from 12.0 to 12.5% which were not significantly different.
From 1997 to 2006, the prevalence of obesity increased in almost all age classes in both men and women (Table 1). The prevalence of overweight remained stable in most age classes. An increase in the prevalence of obesity was observed even in households with high monthly incomes (Figure 1). The same conclusion was reached when the relationship was examined as income per capita (for example in the highest income per capita category, the prevalence increased from 7.6 in 1997 to 11.4% in 2006). An increase in the prevalence of obesity between 1997 and 2006 was also noted in all regions. For example, in the regions with the two extreme prevalences in 1997: in the Paris region it increased from 7.0 to 12.1% and in the north of France from 13.7 to 19.1%.
Previous estimates of the prevalence of obesity in France dated from the 1980s and showed stable prevalence of obesity in French adults aged ≥20 years: 6.4% in men with a slight increase in women, from 6.3 to 7.0% (1). The results of the Obepi surveys suggest a sharp acceleration of obesity prevalences in the 1990s in France. The increase seems to be slowing down, at least in men since 2003, but this trend awaits confirmation.
Some methodological limitations of the Obepi studies have to be acknowledged. The surveys were not based on random samples but on the permanent panel of a poll institute. There was a declining response rate from the first to the fourth survey. However, as the demographic characteristics of nonresponders remained similar over time, it should not affect our results on obesity prevalence. A second limitation arises from the fact that weight, height, and waist circumference were reported and not measured by an investigator. To limit the reporting bias in the Obepi surveys, subjects were asked wherever possible to measure themselves with the aid of another household member, before reporting the anthropometric data. It is likely that this added constraint explains part of the nonresponse. However, the underreporting bias may less affect the interpretation of the changes in the prevalence of obesity over time. Two other estimates of the prevalence of obesity in France come from the three regional French MONICA centers and were based on measured height and weight in 35–64-year-old subjects in 1985–1988 and 1995–1997 (7) and in French men (17–25-year old) drafted for military service between 1985 and 1996 (8). These studies confirm the upward trends found in the Obepi studies in the 1990s. The pitfalls of the methodology of the Obepi surveys have to be weighted against its main advantages. The quota method often achieves actually a better representativity of the national population than random procedure. The low cost and simplicity of the procedure allow the study of large samples and therefore to closely follow rapidly changing situations.
The increase in the prevalence of obesity in France is close to that described recently in other western European countries. In most of these countries (7,9,10,11,12,13), an increase in the prevalence of obesity was documented in the past decade(s) with rates ranging from 0.2 (Netherlands) to 0.9% (United Kingdom) per year in recent years, in comparison to 0.6% in France. The rate in the UK is similar to that in the United States, as estimated from the National Health and Nutrition Examination Survey 1988–1994 and 1999–2000 data (14).
The increase in obesity prevalence found in the Obepi studies was uniform over the country and across ages, regions, and occupational classes. The prevalence of overweight, an intermediate category, was more stable because the percentage of subjects with BMI <25 kg/m2 decreased over time in parallel with the increase in obesity prevalence. In Sweden, the increase in obesity prevalence between 1996/1997 and 2000/2001 was also noted whatever the age, educational level, smoking status, size of the city of residence, country of birth (12). The recent changes in our environment and way of life appear to outweigh the classical cultural and socioeconomic differences in obesity prevalence but do not erase them. The results according to socioeconomic status are however not consistent across countries. In the MONICA studies, between 1979–1989 and 1989–1996, an increase in the difference in mean BMI between subjects with the lowest and the highest educational level was documented in most centers, particularly in women (15). In Switzerland, the increase was higher in men in the high occupational level group and in women in the low occupational level group (16). In the United States, however, the disparity in the prevalence of obesity across socioeconomic classes has decreased over the period 1971–2000 (17).
Obesity prevalences in the elderly are reaching high levels. This may be in part due to a better survival of obese subjects as recently suggested (18). Obesity may no longer be associated with increased mortality rates in elderly obese subjects (19) and there is a debate on the risk and benefit of intentional weight loss in the elderly (18).
France and many other countries in the world have to face the increasing number of obese subjects, and an increasing number of subjects with obesity-related medical complications. The results of the first Obepi surveys were one of the arguments that prompted the French government to implement Nutrition and Health National Plans for the period 2001–2005 and 2006–2008 (http:www.sante.gouv.frhtmpointsurnutrition), with specific goals: to inform the population on healthy food choices and the benefit of physical activity, to develop the nutritional monitoring of the population, to screen and treat obesity and its related complications, and to encourage research in human nutrition.
In that context, we hope to confirm a slowing down in the increase of obesity prevalence in the next Obepi survey in 2009.
The Obepi surveys were funded by Institut Roche, France, with the scientific collaboration of INSERM (National Institute of Health and Medical Research), contracts INSERM-Produits Roche N 97062 and 03114A10. We thank JM Joubert, S Fages, and C Moisan from Produits Roche, G Bonnélye, J Hovart, and P Périé from TNS Sofres, and B Balkau from INSERM unit 780 for their contributions.
M.-A.C. received honorarium for presenting the results of the Obepi surveys in symposia sponsored by Roche. The other authors declared no conflict of interest.