Introduction

The global public health impact of increased obesity prevalence and comorbid physical health complications is well known.1 The examination of the relationship of obesity comorbid with eating disorders is much less studied. However eating disorders similarly have high public and personal health impact.2 Understanding the comorbid occurrence of these two public health problems is important because large multi-site studies have found a higher prevalence of morbid obesity in individuals who experience binge eating episodes in comparison to individuals who do not experience any eating disorder behaviors.2 Also, the risk of eating disorders is reported to be influenced by a vulnerability to obesity as well as exposure to criticism about weight and shape.3 Individuals with overweight or obesity who are dissatisfied with their weight and have difficulties regulating their emotions4 may often engage in eating disorder behaviors in a dysfunctional attempt to manage their weight and reduce negative emotions. Conversely, it is also known that eating disorder behaviors contribute to obesity. For example, a longitudinal study with adolescents in the USA found that weight concerns, body dissatisfaction, dieting and unhealthy weight control practices were risk factors for the development of overweight in adolescents.5 Thus, it is likely that there is a reciprocal relationship between obesity and eating disorder behaviors, in which each contributes to the other.

Some studies have examined the occurrence of obesity and eating disorder behaviors over time. One study6 examined data from adolescents in the USA from 1995 to 2005 and found that during this period there was an increase in body mass index and dieting among female adolescents. These findings suggest that the dieting that these female adolescents were engaging in was not effective in controlling their weight. Further, this increase in dieting prevalence is potentially problematic, given that dieting is a risk factor for the occurrence of eating disorders.7 Increases in weight control behaviors, namely dieting, the use of diet products, purging and vigorous exercise, were also found among male adolescents in the USA from 1995 to 2005, however—unlike their female counterparts—no significant changes in body mass index were found in male adolescents during this period.6

Other studies have also examined the occurrence of obesity and eating disorder behaviors, albeit separately and not in the same cohorts. A systematic analysis of the prevalence of overweight and obesity in several countries found that the worldwide prevalence of overweight and obesity combined rose by 27.5% in adults from 1980 to 2013.8 According to that study,8 the increasing prevalence of overweight and obesity that was noted in developed countries around 1980 attenuated around 2006, but the rising prevalence of overweight and obesity is likely to continue in developing countries. That study8 did not examine the prevalence of eating disorder behaviors. A study of the Danish psychiatric health-care system found an increase in the incidence of eating disorder diagnoses from 1970 to 2008.9 The greatest increase in eating disorder diagnoses during this period was observed among females aged 15–19 years. However, the prevalence of obesity was not examined in that study.9

The occurrence of obesity and comorbid eating disorder behaviors may be a growing problem. A statewide epidemiological study from Australia found an increase in the prevalence of obesity and comorbid eating disorder behaviors from 1995 to 2005, and that this increase was higher than the increase in prevalence of either of these two conditions separately.10 This finding suggests that obesity and comorbid eating disorder behaviors may be a faster-growing public health problem than the occurrence of these conditions separately, and these data need to be updated to the present time. We hypothesize that the prevalence of obesity and comorbid eating disorder behaviors has continued to increase from 2005 to 2015.

Materials and methods

Design

Data from large, cross-sectional, representative community samples of adults from South Australia during the years of 1995, 2005 and 2015 were analyzed and compared. These years were specifically selected for analysis because they may be representative of specific decades and because of the regular intervals between them. Data collection for each of these time points was performed by independent researchers from Harrison Health Research, using the Health Omnibus Survey. The Health Omnibus Survey is a comprehensive survey that provides data that facilitate community research by health organizations and research institutes in South Australia and more broadly. The Health Omnibus Survey provides comprehensive data for research related to several aspects of health, but for the purposes of our study we examined only data regarding weight, height and eating disorder behaviors.

Sample selection and interview procedures

Sample selection and interview procedures were similar in 1995, 2005 and 2015. Metropolitan and rural 'collector districts' in South Australia were identified based on a probability proportional to size sampling procedure, according to the Australian Bureau of Statistics Census data. For the metropolitan sample, the number of collector districts in 1995, 2005 and 2015 was 320, 386 and 398, respectively. The metropolitan collector districts for these years were selected from those used by the Australian Bureau of Statistics for their Censuses in 1991, 2001 and 2011, respectively. For the rural sample, collector districts were all towns of 10 000 or more in population size, plus a sample of towns of at least 1000 people, which were selected with probability proportional to size.

For selection of which households to invite to participate in the survey within each collector district, the following procedure was used. A single household ('starting point') was randomly selected within each collector district. From that starting point, every fourth household was selected, until a total of 10 households had been identified from that collector district. Only one survey was to be conducted at each household. Where a household had more than one resident aged over 15 years, the person targeted for the survey was the person with the most recent birthday at the time of interview. Fifty interviews were conducted as pilot surveys in each of the years of 1995, 2005 and 2015, to ensure that the questionnaire was constructed appropriately, prior to conducting all interviews. As a quality check, the operations manager of Harrison Health Research contacted a sample of the households scheduled for survey administration by each interviewer, to ensure they had in fact been interviewed and were the person in the household over 15 years of age to have their birthday last at the time of the interview. The percent of households from each interviewer to be contacted in this way was 5% in 1995, and 10% in each of 2005 and 2015. The response rate (percent of interviews conducted relative to the number of households invited to participate) was 71.5% (n=3001) in 1995, 63% (n=3047) in 2005 and 53.7% (n=3005) in 2015.

The interview

The interview consisted of a structured, self-report interview that covered a range of health-related and demographic questions including weight, height and eating disorder behaviors. body mass index was calculated from height and weight, and obesity was defined as a body mass index⩾30 kg m2. Questions related to eating disorder behaviors were adapted from the 'gold standard' instrument in assessment of eating disorder features, namely the Eating Disorder Examination,11 and are listed below. In brief, three types of eating disorder behaviors were assessed: objective binge eating (eating an unusually large amount of food and experiencing a sense of loss of control overeating), very strict dieting/fasting in order to control weight or shape, and purging (use of laxatives, diuretics or vomiting in order to control weight or shape). As seen in the specific questions used in this study as listed below, the questions focused on eating disorder behaviors occurring at the frequency and period specified in the diagnostic criteria for eating disorders as defined by the Diagnostic and Statistical Manual of Mental Disorders, fifth edition (DSM-5)12 (that is, at least once weekly during the last 3 months).

The questions relating to eating disorder behaviors in the Health Omnibus Survey were:

  1. 1

    I would now like to ask you about episodes of overeating that you may have had recently. By overeating or binge eating, I mean eating an unusually large amount of food in one go and at the time feeling that your eating was out of control (that is, you could not prevent yourself from overeating, or that you could not stop eating once you had started). Over the past 3 months, how often have you overeaten in the way I have described? (Responses were on a set list of 'not at all', 'less often than once a week', 'once a week', or 'two or more times a week').

  2. 2

    This question is about various weight control methods some people use. Over the past 3 months, have you regularly used—that is, used at least once a week any of the following: laxatives, diuretics (water tablets), made yourself sick, gone on a very strict diet or eaten hardly anything at all for a time, in order to control your shape or weight?

If the response was yes, the interviewer then asked:

  • Have you used laxatives/diuretics/vomiting regularly over the past 3 months? (yes/no/refused)

  • Have you fasted and/or dieted very strictly on a regular basis over the past 3 months? (yes/no/refused)

Statistics

All data were weighted according to the Australian Census most proximal to the respective survey. The changes in population characteristics between 2 years were compared using independent sample z-tests when the statistic was a percentage or proportion, and using independent sample t-tests when the statistic was a mean (that is, age). All tests were adjusted for multiple comparisons using the Sidak’s method, which is slightly less conservative than the Bonferroni method. Since the outcomes of obesity and eating disorder behaviors, namely binge eating, very strict dieting/fasting and purging, or obesity with a specific comorbid eating disorder behavior, were all dichotomous, logistic regression was used to predict each outcome based on year, while also controlling for age, sex and educational qualifications. As education and income were correlated, income was not controlled for in the logistic regression analysis. Our exploratory models included income in addition to year, age, sex and education as predictors; but income was not a statistically significant predictor of outcome and hence it was excluded from the final models. However, a logistic regression to predict the outcome of obesity with a comorbid purging behavior is not presented because this model fitted poorly as the outcome was a very uncommon event (only 34 cases which is 0.4% of the sample). Following logistic regression analysis, post hoc tests for multiple comparisons between each pair of years were performed to investigate whether the change in the predicted prevalence between 2 years was statistically significant. All logistic regression analysis and multiple comparisons in this study were performed using a 5% level of significance. All analyses in this study were performed using Stata Statistical Software (Release 14, StataCorp LP, College Station, TX, USA). For performing logistic regression analysis the svy: logistic command was used.

Ethics

All participants provided verbal informed consent. Ethical approval for data collection in the years of 1995 and 2005 was obtained through the South Australian Department of Health Ethics Committee. Ethical approval for data collection in the year of 2015 was obtained through the University of Adelaide Human Research Ethics Committee.

Results

Population characteristics

The number of participants who responded to the survey questions relevant to the current study was 3001 in 1995, 3047 in 2005 and 3005 in 2015. The demographic characteristics of participants in these years are shown in Table 1. The percentage of participants for which the country of birth was Australia differed significantly between 1995 and 2015, and 2005 and 2015, but not between 1995 and 2005. Participant characteristics also differed significantly between sampling years with regard to mean age, percentage of participants reporting a household income of A$50 000 per year or more and the percentage with a graduate diploma or higher level of education between the years of 1995 and 2005, 1995 and 2015, as well as 2005 and 2015. The percentage of participants that were female or married or in a de facto relationship, were similar in the years of 1995, 2005 and 2015.

Table 1 Characteristics of participants in the years of 1995, 2005 and 2015.
Full size table

Changes in observed prevalence

These results are summarized in Table 2. The observed prevalence of obesity increased from 1995 to 2005 (from 19.3 to 25.7%) and continued to increase at a similar rate from 2005 to 2015 (from 25.7 to 33.2%), with a 1.7-fold increase from 1995 to 2015. There was a similar increasing trend from 1995 to 2015 for the observed prevalence of participants with recurrent binge eating episodes (at least one binge eating episode per week during the last 3 months) independent of weight, however, the increase in the observed prevalence from 1995 to 2015 was much higher than the increase in that for obesity, at a 3.5-fold. There was a large increase in the observed prevalence of strict dieting/fasting from 1995 to 2005 (from 1.6 to 4.2%), but from 2005 to 2015 the observed prevalence increased slowly, from 4.2 to 5.2%, producing a 3.3-fold increase from 1995 to 2015. The observed prevalence of participants who reported purging (use of laxatives, diuretics or self-induced vomiting at least once weekly, during the last 3 months in order to control their weight or shape) increased from 1995 to 2005 (from 0.8 to 1.8%), but decreased from 2005 to 2015 (from 1.8 to 0.8%). The observed prevalence of participants with obesity as well as comorbid recurrent binge eating episodes increased 3.1-fold from 1995 to 2005 (from 1.0 to 3.1%), and continued to increase but at a slightly slower rate (1.8-fold) from 2005 to 2015 (from 3.1 to 5.7%). There was thus a 5.7-fold increase in the observed prevalence of obesity with comorbid recurrent binge eating episodes from 1995 to 2015. The observed prevalence of participants with obesity as well as comorbid very strict dieting/fasting increased substantially from 1995 to 2005 (from 0.3 to 1.8%), but increased from 1.8 to only 2.4% from 2005 to 2015. Thus, overall there was an eight-fold increase in the observed prevalence of obesity with comorbid very strict dieting/fasting from 1995 to 2015. The observed prevalence of participants with obesity as well as comorbid purging was very low in all three time periods: it increased from 1995 to 2005 (from 0.2 to 0.6%), and declined slightly (from 0.6 to 0.4%) from 2005 to 2015. Overall, there was a two-fold increase in the observed prevalence of obesity and comorbid purging from 1995 to 2015.

Table 2 Observed prevalence of obesity, eating disorder behaviors (binge eating, very strict dieting/fasting and purging) or obesity with comorbid eating disorder behaviors in the years of 1995, 2005 and 2015
Full size table

Changes in prevalence predicted by logistic regression analysis

The observed prevalence rates of each of the seven outcomes listed above do not take into account the likelihood that the participants sampled in each of the 3 years (1995, 2005 and 2015) may be different with respect to confounders that influence obesity and eating disorder behaviors, namely age, sex and educational qualifications. As such, we controlled for possible differences in age, sex and educational qualifications between the samples from 1995, 2005 and 2015 by using logistic regression analysis to determine predicted prevalence rates of each of the seven outcomes for each of the 3 years. These predicted prevalence rates are discussed below and are shown in Table 3. In general, the magnitude of changes in predicted prevalence of an outcome between each pair of years was higher when the outcome was obesity and a particular comorbid eating disorder behavior, as compared to obesity alone or an individual eating disorder behavior alone (Table 3).

Table 3 Predicted prevalence of the seven outcomes (obesity, eating disorder behaviors (binge eating, very strict dieting/fasting, purging) or obesity with comorbid eating disorder behaviors) and comparisons by each pair of years
Full size table

Changes in the predicted prevalence of obesity

The predicted prevalence of participants with obesity increased significantly (P<0.001) from 1995 to 2005 (from 18.1 to 25.1%) and continued to increase significantly (P<0.001) from 2005 to 2015 (from 25.1 to 32.5%). Thus, there was close to a two-fold significant (P<0.001) increase in the predicted prevalence of obesity from 1995 to 2015.

Changes in the predicted prevalence of binge eating

The predicted prevalence of participants with recurrent binge eating episodes increased significantly (P<0.001) from 1995 to 2005 (from 2.4 to 6.3%), and continued to increase significantly (P<0.001) from 2005 to 2015 (from 6.3 to 12.7%). Thus, there was a 5.3-fold significant increase (P<0.001) in the predicted prevalence of binge eating from 1995 to 2015.

Changes in the predicted prevalence of very strict dieting/fasting

The predicted prevalence of participants who reported very strict dieting or eating hardly anything at all for a time (at least once weekly, or regularly during the last 3 months in order to control their weight or shape), increased significantly (P<0.001) from 1995 to 2005 (from 1.3 to 4%) but did not change significantly from 2005 to 2015 (when prevalence went from 4.0 to 5%). However, overall there was a 3.8-fold significant increase (P<0.001) in the predicted prevalence of very strict dieting/fasting from 1995 to 2015.

Changes in the predicted prevalence of purging

The predicted prevalence of participants who reported purging increased significantly (P<0.01) from 1995 to 2005 (from 0.5 to 1.3%), but decreased—albeit non-significantly—from 2005 to 2015 (from 1.3 to 0.7%). Overall, the increase in the predicted prevalence of purging between 1995 and 2015 was not significant.

Changes in the predicted prevalence of obesity with comorbid binge eating

The predicted prevalence of participants with obesity as well as comorbid recurrent binge eating episodes increased significantly (P<0.001) from 1995 to 2005 (from 0.8 to 2.9%), and continued to increase significantly (P<0.001) from 2005 to 2015 (from 2.9 to 5.8%). There was thus a 7.3-fold significant increase (P<0.001) in the predicted prevalence of participants with obesity as well as comorbid binge eating from 1995 to 2015.

Changes in the predicted prevalence of obesity with comorbid very strict dieting/fasting

The predicted prevalence of participants with obesity as well as comorbid very strict dieting/fasting increased significantly (P<0.001) from 1995 to 2005 (from 0.2 to 1.9%), but did not change significantly from 2005 to 2015 (when prevalence went from 1.9 to 2.3%). However, overall there was a 11.5-fold significant increase (P<0.001) in the predicted prevalence of obesity with comorbid very strict dieting/fasting from 1995 to 2015.

Discussion

This study aimed to update the last published study10 on the prevalence of obesity and eating disorder behaviors, as well as the prevalence of obesity with comorbid eating disorder behaviors, in statewide community representative samples from Australia in the years 1995, 2005 and 2015. Our findings show an increase during the 20 years from 1995 to 2015 in the independent prevalence of obesity, binge eating and very strict dieting/fasting. The comorbid occurrence of obesity with recurrent binge eating, or obesity with regular engagement in very strict dieting/fasting, also increased significantly from 1995 to 2015.

The prevalence of obesity, binge eating, as well as obesity with comorbid binge eating, all increased significantly during the decades from 1995 to 2005 and from 2005 to 2015. This is concerning because it shows a consistent increase in these conditions, indicating a potential upward trend for future years. The number of afflicted individuals is substantial; in 2015 there was a high (predicted) prevalence of obesity (32.5%), binge eating (12.5%) and obesity with comorbid binge eating (5.8%) in South Australia. However, of these three conditions that increased significantly from 1995 to 2005 and from 2005 to 2015, the prevalence of people with obesity as well as comorbid binge eating is of highest concern because it increased the most (7.3-fold after adjusting for age, sex and education in the analysis). Our study did not examine causal relationships, but it is possible that binge eating and obesity contribute to each other,2, 3, 4 and this may have contributed to the rapid increase in this comorbid condition during this period.

The prevalence of very strict dieting or eating hardly anything at all—both in the general population as well as in participants with obesity—increased from 1995 to 2005, and there was a substantial proportion of people engaging in this behavior (for example, 5% of the general population in 2015 after controlling for age, sex and educational qualifications). This is concerning because it indicates that a significant proportion of the population (with or without obesity) may be engaging in attempts to reduce their body weight using strategies that have been linked to the development of eating disorders, notably binge eating disorder (characterized by recurrent binge eating) and bulimia nervosa (characterized by periodic binge eating and compensatory weight loss behavior).11, 13, 14, 15 Notwithstanding the attenuation of the increase in the prevalence of obesity with comorbid very strict dieting/fasting from 2005 to 2015, of all the conditions examined in our study, the prevalence of obesity with comorbid very strict dieting/fasting is the condition that increased the most (11.5-fold) from 1995 to 2015. This high increase in the prevalence of obesity with comorbid very strict dieting/fasting may be related to social expectancy for people with obesity to lose weight and to be thin. Indeed the impact of anti-obesity campaigns on body image and eating disorder behaviors remains unknown. More support for healthy weight management strategies in the general population and for individuals with obesity engaging in extreme dietary restriction thus appear to be indicated.

Our study also examined purging (regular use of laxatives, diuretics or self-induced vomiting in an attempt to control weight or shape), as well as the occurrence of obesity comorbid with purging. These behaviors may be less of a public health concern than obesity, binge eating or very strict dieting/fasting, because of their low observed (crude) prevalence (0.2%, 0.6% and 0.4% for 1995, 2005 and 2015, respectively), and because of the lack of increase in their observed (crude) prevalence from 1995 to 2015.

This study has several strengths and limitations. Notable strengths include collection of data from population-representative samples over a long (20-year) period, regular (10-yearly) intervals between time points and the a priori selection of time points that may be representative of specific decades (for example, 1995 as a midpoint representative of the 1990’s). This study was limited, however, in that participants were not interviewed by clinicians or researchers trained in the diagnosis of eating disorders or eating disorder behaviors. Also, this study does not provide data on the prevalence of diagnosed eating disorders, nor does its cross-sectional design provide information regarding causal relationships.

Our findings enable us to make recommendations for future research. First, this field would benefit from longitudinal studies that further investigate the bi-directional relationship between obesity and binge eating (for example, if binge eating precedes obesity or vice versa) or other eating disorder features. This is important because both conditions can potentially contribute to the other,2, 3, 4 and temporal data would help to dissect any causality in these associations. Similarly, this field would also benefit from longitudinal studies examining any temporal relationships between binge eating and very strict dieting (for example, if binge eating precedes very strict dieting or vice versa). This is relevant due to continued debate about whether dieting is a causative factor in the development of binge eating.11, 13, 14, 15, 16, 17 Finally, future research investigating whether individuals with obesity and comorbid eating disorder behaviors have higher risk for other comorbidities, such as type 2 diabetes, than individuals with obesity but without eating disorder symptoms, could improve our understanding of the overall health of these people.

This study has public health and clinical implications. Our findings imply that improved preventative and treatment strategies are required for obesity, binge eating, very strict dieting/fasting and the comorbid occurrence of obesity with these eating disorder behaviors. Research in this domain is necessary, for example, because people with obesity and comorbid binge eating are at greater risk than the general population of multiple physical and mental complications related not only to obesity (for example, cardiovascular disease, type 2 diabetes, some musculoskeletal conditions, certain cancers18), but also to binge eating (for example, distress and depression).19, 20 Some studies have already been conducted in this field and have found that behavioral interventions for obesity can reduce binge eating,21 although with less effectiveness for weight reduction. Additionally, unsupervised very strict dieting/fasting can be a risk factor for the development of eating disorders.11, 13, 14, 15 As such, an alternative and safer procedure for weight loss in people with obesity may be the use of clinically supervised very-low-energy diets. A recent systematic review found that very-low-energy diets, achieved using meals replacement formulae under clinical supervision, can provide beneficial effects for both weight reduction and binge eating in individuals with obesity.17 Research in the sphere of treatments for people with obesity and comorbid eating disorder behaviors is continuing, with our team currently trialing an intervention developed for the treatment of individuals with obesity and comorbid eating disorders, namely binge eating disorder and bulimia nervosa (da Luz et al., submitted).22 On the other hand, less research has focused on the prevention of obesity with comorbid eating disorder behaviors, and this will be integral to reducing the burden of these conditions.