Obesity and depressive disorders are central public health issues and the relationship between depression and weight remains unclear [1, 2].
Our study shows that irrespective of gender, the prevalence of overweight and obesity at 17 years of age is respectively 10.8% and 1.9%. Figures in France at age 14 are 7% for overweight and 4% for obesity [30, 31]. According to gender, the present results show a prevalence of overweight (12.9% for boys and 8.7% for girls) that is comparable to that found in the international literature, ranging from 5.2% to 28.9% for boys and from 8.1% to 31.0% for girls . As in the literature, we found differences in prevalence according to parental professional status and parental marital status [6, 30, 31].
The prevalence of depression in the present study on the basis of the ADRS measure specific to adolescents was 4.5% among boys and 10.4% among girls. In the literature, the data on prevalence of depression varies, in particular according to the type of measure used (measures of intensity, or DSM-IV-based) as well as according to whether or not gender differences are sought. In France, using criteria for depression according to a standardized assessment (CIDI short form), the prevalence of depression for the age of 15 to 24 years old as been reported to range from 6.3% for boys, to 11.2% for girls . A review of recent epidemiological studies showed prevalence rates ranging from 1.8% to 5.9% in the United States [33–35]. For the years 2005-2006, the Center for Disease Control found a prevalence of depression between 4% and 6.4% for ages 12 to 17, using the PHQ-9, but without testing for gender differences . Many epidemiological and clinical studies have shown that girls have typically been found to display higher levels of depressive symptoms than boys [37, 38]. This has been attributed to genetics, increased prevalence of anxiety disorders in females, biological changes associated with puberty, cognitive predisposition and socio-cultural factors .
The most important result of our study concerns the non-linear and significant association between depression and BMI, with no effect of parental working and marital status. In addition, the shape of the association is different for boys and girls. We have a U- shaped association for boys, which shows that boys that are too thin or too fat have higher levels of depression. For girls, we have a beginning of a U shape, and then an inversely convex curve, which means, first, that depression levels in underweight girls are higher than in normal weight girls, and second, that depression levels are higher in overweight adolescent girls than in obese girls. This can be related to the "fat and jolly" hypothesis, which has been discussed for older women [40–42]. Our results recall the study by Wardle et al , who found, in a small sample of adolescents, that regardless of moderators reports of depressive symptoms were more frequent in overweight than in obese adolescent girls of 14, and who found no difference for depressive symptoms between normal weight and obese girls. Using splines, Cortese  in a small sample found curves of different shape for boys and a U-shaped curve for girls aged 11 to 14, showing that underweight girls had an equivalent level of depression to that of obese girls, this not being true for the boys. While some previous studies have explored the association between obesity and depression without questioning the gender difference [23, 43, 44], according to the recent meta-analysis of community-based studies by de Wit et al  there is some support for the hypothesis that there is a difference between males and females in the association.
The "fat and jolly" hypothesis, which we suggest for the adolescent girls of 17 years old, is generally thought to be related to hormone and neuro-endocrine issues, but it could be envisaged as a possible dynamic in adolescent girls. It can indeed be thought that the difficult struggle against excess weight could let up when the subject reaches the obesity stage, as if, once they give up the struggle, they may feel less depressed. In all events, obesity, despite creating negative social experiences in adolescence, does not in itself cause depression. The effect of moderators needs to be explored in the association between weight and depression, and in particular the complex association between weight and the perception of weight, as a risk factor of depression [45, 46].
Several limitations of the study must be considered. BMI was calculated from self-report, with the attendant risk of under-estimation of weight. The degree of divergence between subjective and objective BMI has been described for different ages, different countries and according to gender . However, most studies use self-report data. The underestimation tends to be greater among obese adolescents, so that the prevalence in our study may well be underestimated. Secondly, as our study is cross-sectional, we cannot demonstrate the mechanisms of the association found. Only longitudinal studies are able to explore the different effects of the potential moderators of the association, and investigate the possible predictive relationship between overweight and depression. Indeed, longitudinal studies can provide evidence for depression predicting future obesity or weight gain or the reverse. Published research to date has shown a mixed picture [11, 48, 49]. Thirdly, we have no data available about race or ethnicity because no studies in France are allowed to ask questions about theses major factors in public health. Moreover, the present data only concerns French nationals from metropolitan France (i.e. excluding overseas territories). However the sample nevertheless represents 5.2% of the French metropolitan population aged 17, and is representative of it.
Overall, since the results are derived in part from the visual examination of regression splines, they should be considered basically as exploratory, and therefore as requiring replication. Despite these limitations, this study provides valuable information on account of the exceptionally large size of the sample. To our knowledge, this is the very first survey worldwide on such a large representative sample of a total population of adolescents. And another important point is that while most studies use depression measures comprising items concerning the body, food, or weight, which are confounders in the exploration of the links between depression and weight, in the present study we used the ADRS measure of depression, which is the only measure that does not contain these confounders.