Obesity rates among youth are high and factors implicated in their aetiology need to be identified. Eating behaviour traits have been related to body weight(Reference Provencher, Drapeau and Tremblay1, Reference Gallant, Tremblay and Pérusse2), weight gain over time(Reference Drapeau, Provencher and Lemieux3), reduced weight loss in a weight loss intervention(Reference Dalle Grave, Calugi and Corica4) and reduced weight loss maintenance(Reference Wing and Phelan5). Recently, we reported that the eating behaviour traits of rigid control and disinhibition, as measured by the Three-Factor Eating Questionnaire (TFEQ)(Reference Stunkard and Messick6), were positively associated with body weight in adolescents from the Québec Family Study (QFS)(Reference Gallant, Tremblay and Pérusse2). Dieting to lose weight is also prevalent among youth, with proportions reaching up to 57 % and 25·3 % for girls and boys, respectively (mean age 12·8 (sem 0·8) years)(Reference Neumark-Sztainer, Wall and Guo7). Additionally, dieting in this age group has been shown to be related to weight gain(Reference Neumark-Sztainer, Wall and Guo7, Reference Field, Austin and Taylor8) and binge eating(Reference Field, Austin and Taylor8, Reference Neumark-Sztainer, Wall and Haines9) and is important in the prevention of obesity, particularly since dieting and disordered eating are known to persist from adolescence into adulthood(Reference Neumark-Sztainer, Wall and Larson10). In adults of the QFS, past dieters had greater cognitive restraint, rigid control and disinhibition compared to non-dieters(Reference Provencher, Drapeau and Tremblay11). On the contrary, Turkish adolescents who were chronic dieters (dieting ≥ 5 times/year) had lower cognitive restraint, disinhibition and hunger scores compared to adolescents with no past dieting(Reference Bas, Bozan and Cigerim12). Thus, the relationship between dieting or past dieting and eating behaviour traits is not clear. Moreover, to our knowledge, no study has looked at dieting history and the specific TFEQ subscales among adolescents. The aim of the present study is to examine the relationship between past dieting and eating behaviour traits among adolescents from the QFS.
Methods
Participants for this study belonged to Phase 2 of the QFS. Details of the QFS are described elsewhere(Reference Bouchard13). At the end of Phase 2 (1989–95), sixty adolescents completed the TFEQ(Reference Stunkard and Messick6) which measures cognitive restraint, disinhibition and hunger as well as their respective subscales(Reference Westenhoefer, Stunkard and Pudel14, Reference Bond, McDowell and Wilkinson15). They also completed a questionnaire pertaining to their eating habits. Participants who responded positively to the question ‘do you remember following a diet in the past 5 years?’ were categorised as past dieters. Participants who responded positively to the question ‘are you currently following a diet?’ were categorised as current dieters. Body weight and height were measured using standardised laboratory methods. BMI z-scores were calculated from the Centers for Disease Control and Prevention (CDC) growth charts to create an age-normalised body weight variable(Reference Kuczmarski, Ogden and Grummer-Strawn16). In addition, body weight categories, e.g. normal weight, overweight and obese, were determined by cut-offs suggested by Cole et al. (Reference Cole, Bellizzi and Flegal17). Median split was used to categorise high and low eating behaviour traits, creating groups of equal sizes. Statistical analyses were performed using JMP 7.0 statistical software (SAS Institute). Descriptive statistics consist of means with their standard errors and medians with inter-quartile ranges (TFEQ scores). ANOVA and Fisher's exact test were used to test for significant differences between groups, i.e. past dieters and non-dieters. The χ2 test was used to test for independence between high and low eating behaviour trait categories and past dieting status. Bivariate analyses were adjusted for age, sex and BMI. Relative risk was used to quantify the risk of dieting between eating behaviour trait categories. This study was conducted according to the guidelines laid down in the Declaration of Helsinki and all procedures involving human subjects/patients were approved by the ethics committee of Laval University. Written informed consent was obtained from all subjects/patients.
Results
Mean age and BMI of the participants were 15·0 (sem 2·4) years and 25·9 (sem 8·0) kg/m2, respectively. Overall, 55 % of the sample was categorised as either overweight or obese. Few adolescents were currently following a diet (n 3 or 5·0 %). More adolescents reported a past dieting experience (n 14 or 23·3 %). Analyses were not performed on current dieters due to the low number of participants expressing this behaviour. Past dieters were older (16·9 v. 14·4 years, P < 0·001) and constituted a greater proportion of females (78·6 % v. 41·3 %, P < 0·05), compared to those who had no past dieting experience. There was a trend for past dieters to have a higher BMI z-score (P = 0·10), higher waist circumference (P = 0·06) and greater percentage body fat (P = 0·09). Significantly more overweight/obese adolescents reported past dieting than normal-weight adolescents (36·4 v. 7·4 %, P = 0·008). Most TFEQ scores were higher among past dieters than non-past dieters (Table 1). After correcting for sex, age and BMI, a past dieting experience predicted rigid control (P = 0·05), attitude to self-regulation (P = 0·03) and disinhibition and its subscales (P < 0·05–0·001). A greater proportion of past dieters was found among those with high rigid control and high disinhibition compared to those with low rigid control and low disinhibition (P < 0·01; Fig. 1). In addition, relative to adolescents reporting both low rigid control and low disinhibition, adolescents reporting both high rigid control and high disinhibition had an increased relative risk of being a past dieter of 15·6 (95 % CI 2·13, 113·8). Overall, 10 % of the sample (n 6) was considered past dieters with high rigid control and high disinhibition and all six were either overweight (n 3) or obese (n 3).
Table 1 Eating behaviour traits separated by dieting status among adolescents (Median values and interquartile ranges (IQR); mean values with their standard errors, n 60)
* P value for Three-Factor Eating Questionnaire scores corrected for possible confounding factors (sex, age, BMI).
Fig. 1 Proportion of adolescents with a past dieting history in each eating behaviour trait category (high–low categories determined by median split); χ2, P < 0·002.
Discussion
The main finding of the present study was that past dieting is related to rigid control and disinhibition in adolescents. More specifically, the highest prevalence of past dieters was found among those characterised as having both high rigid control and high disinhibition. Similar findings have been recorded among adults of the QFS(Reference Provencher, Drapeau and Tremblay11). The eating behaviour profile of the dieter is informative as it may suggest the types of dieting behaviour adopted by the individual.
Past adolescent dieters of the QFS reported high levels of rigid control and attitude to self-regulation. Rigid control is an extreme form of dietary restraint(Reference Westenhoefer, Stunkard and Pudel14), which has been related to dieting(Reference Provencher, Drapeau and Tremblay11, Reference Timko and Perone18), disinhibition, binge eating and increased body weight in adults(Reference Westenhoefer, Stunkard and Pudel14). For these reasons, rigid control is considered an unhealthy form of dietary restraint. Moreover, attitude to self-regulation, which is associated with an all-encompassing perspective on eating and weight control(Reference Bond, McDowell and Wilkinson15), is another form of strict control. It is not surprising that these behaviour traits were prevalent among adolescent dieters, as unhealthy dieting methods, e.g. laxative use, fasting, diet pills, eating little food, are prevalent among this age group(Reference Neumark-Sztainer, Wall and Story19). In addition, flexible control, which is considered a more healthful approach to dietary restriction and inversely related to BMI in adults(Reference Westenhoefer, Stunkard and Pudel14), was not different between past dieters and those without a dieting history, after correcting for sex, age and BMI. Therefore, past dieters in this sample are not characterised by healthy solutions to restrict their dietary intake. Overall, this finding indicates that the dieting approach adopted by these adolescents is unlikely to lead to successful long-term body weight control.
Not only did past dieters have higher scores of rigid control, but they also had higher scores of disinhibition. Due to the cross-sectional nature of the data, the direction of the past dieting–disinhibition relationship, i.e. whether disinhibited eating was a result of previous dieting, cannot be determined. In other studies, dieting has previously been shown to lead to binge eating among adolescents(Reference Field, Austin and Taylor8, Reference Neumark-Sztainer, Paxton and Hannan20). Therefore, it is possible that dieting could lead to disinhibited eating through mechanisms described by the restraint theory(Reference Polivy and Herman21), i.e. the chronic restriction of food alters the ability to self-monitor satiety signals and may lead to overeating once the restriction is removed. It would be interesting to know if increased disinhibition scores were also found among current dieters or if this is a behaviour found among restrained eaters not currently on a diet. Unfortunately, this could not be examined because of the few adolescents reporting currently dieting.
The lack of a significant difference in body weight measures between past dieters and non-dieters is perplexing because past dieting has been shown to lead to weight gain in adolescents(Reference Field, Austin and Taylor8, Reference Neumark-Sztainer, Paxton and Hannan20) and high rigid control and high disinhibition have previously characterised the heaviest adolescents in this sample(Reference Gallant, Tremblay and Pérusse2). However, the trends for higher measures in the present study suggest a possible sample size issue, a clear limitation of our study. Furthermore, the types of dieting strategies were not documented in our study. Past dieters could have adopted healthy means of weight control. Indeed, one participant had a history of dieting, and yet presented low levels of both disinhibition and rigid control. Other factors could also play a part in weight gain among dieters, such as reduced physical activity and reduced breakfast and fruit/vegetable consumption(Reference Neumark-Sztainer, Wall and Haines9).
Besides sample size, one main limitation of this study is the cross-sectional nature of the data. Thus, the direction of the eating behaviour–dieting relationship cannot be addressed by our analyses. As such, we can only speculate on the nature of this relationship. With so few adolescents reporting currently dieting, we cannot determine which eating behaviours characterise the former as opposed to past dieters. In addition, the present results may not be generalisable to all adolescent populations due to the high percentage of overweight/obese participants. For example, a recent Canadian Survey reported adolescent overweight and obesity rates to be 27 %(22), which is less than that reported in this study. Nevertheless, this observational study suggests that dieting and unhealthy eating behaviour co-exist among adolescents.
In conclusion, past dieting adolescents from the QFS report high scores of disinhibition and rigid control. This profile suggests that these adolescents may not be adopting a healthy weight control behaviour which may create a challenge for successful weight control in the future. In the context of the obesity epidemic, acknowledging these behaviours, specifically among dieters, should be incorporated into obesity interventions.
Acknowledgements
A. R. G. is funded by the Cardiology and Pneumology University Research Institute of Quebec. The authors specially thank G. Fournier, Dr G. Thériault, L. Allard, M. Chagnon and C. Leblanc for their contributions to the recruitment and data collection of the QFS. The present work was funded by the Canadian Institutes of Health Research (MCG-15187). The QFS was supported over the years by multiple grants from the Medical Research Council of Canada and the Canadian Institutes for Health Research (PG-11811, MT-13960 and GR-15187) as well as other agencies. C. B. is partially funded by the John W. Barton Sr. Chair in Genetics and Nutrition. There are no conflicts of interest to report in relation to the present study. A. R. G. completed the data analyses and results interpretation and wrote the manuscript. V. D. helped with data interpretation and reviewed the manuscript. A. T., J.-P. D., L. P. and C. B. were all involved with the QFS and reviewed the manuscript.
