Introduction. binge episodes could contribute to weight gain, SA French 1 *, RW Jeffery 1, NE Sherwood 1 and D Neumark-Sztainer 1

Transcription

1 International Journal of Obesity (1999) 23, 576±585 ß 1999 Stockton Press All rights reserved 0307±0565/99 $ Prevalence and correlates of binge eating in a nonclinical sample of women enrolled in a weight gain prevention program SA French 1 *, RW Jeffery 1, NE Sherwood 1 and D Neumark-Sztainer 1 1 Division of Epidemiology, University of Minnesota, Minneapolis, USA OBJECTIVES: The aims of the present study were to examine the prevalence and correlates of binge eating in a nonclinical sample of women and to examine whether associations differed by overweight status. DESIGN: Cross-sectional comparison of women based on self-reported binge eating status (large amount of food eaten and feelings of lack of control during these eating episodes) and overweight status (measured body weight: overweight de ned as body mass index (BMI) 27.3 kg=m 2 ). PARTICIPANTS: Subjects were 817 women aged 20 ± 45 y from the community who enrolled in a three year prospective intervention study to examine methods for preventing age-related weight gain. MEASURES: Body weight was measured at baseline and three-year follow-up. Self-report measures of binge eating, dieting practices, eating and exercise behaviours, depression, self-esteem and stressful life events were collected at the three-year follow-up. RESULTS: The prevalence of binge eating in the past six months was 9% among normal weight women and 21% among overweight women. The frequency of binge eating was low ( > 50% of binge eaters binged less than once per week) and did not signi cantly differ by body weight status. Compared to non-binge eaters, binge eaters reported more dieting practices, more extreme attitudes about weight and shape, and higher levels of depression and stressful life events. was not related to habitual eating and exercise behaviours. In multivariate models, weight=shape importance (odds ratio (OR) ˆ 3.33; 95% con dence intervals (95% CI) ˆ 2.10, 5.29), depression (OR ˆ 1.73; 95% CI ˆ 1.07, 2.79) and history of intentional weight loss episodes (OR ˆ 1.68; 95% CI ˆ 1.03, 1.13) were independently associated with increased odds of binge eating. CONCLUSIONS: is about twice as prevalent among overweight women, compared to normal weight women, in a nonclinical sample, but has similar correlates (that is, dieting, depression, weight=shape preoccupation). Prospective research is needed to determine whether there are causal associations between binge eating, depression, dieting and weight gain. Keywords: binge eating; binge eating disorder; dieting; obesity; weight loss Introduction behaviour is important to study because of its potential role in the development of obesity and or eating disorders. 1±4 Yet surprisingly little is known about the psychosocial and behavioural correlates of binge eating in nonclinical samples, despite its purported high prevalence in the community. 5 Prevalence estimates in the community vary according to the de nition of binge eating used and whether the assessment is based on self-report or interview. 6±8 Prevalence estimates from college populations average 35.8% (range 7 ± 79% based on 16 studies). 6 Community-based prevalence estimates are 6.3% for binge eating in the past six months and 4.6% for binge eating disorder. 9 *Correspondence: Dr Simone A. French, Division of Epidemiology, School of Public Health, University of Minnesota, Suite 300, 1300 South Second St, Minneapolis, MN, USA. french@epivax.epi.umn.edu Received 5 October 1998; revised 15 December 1998; accepted 7 January 1999 Examination of the correlates of binge eating may provide clues about its etiology. 2,9 Overweight is perhaps the most consistent correlate of binge eating, with past month binge eating reported by 6.3% of the non-overweight and 15.9% of the overweight women in one community-based sample. 5,10 Furthermore, binge eating is more prevalent among obese individuals in weight loss treatment programs. A recent review estimated that between 25 ± 50% of obese people seeking treatment for weight loss binge eat. 11 is associated with a history of severe obesity, dieting and frequent weight losses, 9,12,13 and with negative psychological variables such as depression, weight and shape preoccupation. 9,13 ± 19 Little is known about the eating and exercise behaviours of binge eaters and whether they differ from non-binge eaters. 20 A better understanding of the eating and exercise behaviour patterns of binge eaters is needed to evaluate the potential impact of binge eating on body weight and the development of obesity. 1±4 The contribution of excess energy intake from binge episodes could contribute to weight gain,

2 especially among binge eaters who do not purge by vomiting. In addition, dietary intake during binge eating episodes may be more likely to be comprised of high fat foods. 20 ± 26 Physical activity patterns, both usual, and speci cally in response to, binge eating, have not been systematically examined among binge eaters. To summarize, binge eating is an important behaviour in the general population. However, potentially important correlates of binge eating, such as physical activity and eating patterns, and psychological factors, have not been extensively examined in nonclinical samples. Descriptive studies have largely focused on treatment-seeking populations of women with eating disorders or obesity. Additional data are needed to further describe the distribution and correlates of binge eating in the broader community. This will help evaluate the potential role that binge eating may play in the development of obesity and its effects on psychological well-being. The purpose of the present study is to provide additional descriptive information on the prevalence and correlates of binge eating in a nonclinical sample of healthy adult women who enrolled in a weight gain prevention trial. Although not representative, the sample was heterogeneous and similar in demographic characteristics to a general community sample of healthy middle-aged women. It was hypothesized that independent of body weight, binge eaters would have more negative body image attitudes, more extreme dieting and weight control practices and greater depression. Interactions between binge eating and obesity were also examined to determine whether potentially negative behavioural or psychosocial correlates of binge eating were more pronounced in either obese or normal weight binge eaters. Methods Study overview and sample description Data for this report were collected as part of a three year longitudinal study on methods to prevent weight gain with age among young adults in the community. 27 The Pound of Prevention (POP) study participants were recruited through newspaper and radio advertisements, and direct mail to employees of a large university. Advertisements and iers requested participants between the ages of 20 ± 45 y, who were interested in enrolling in a study to prevent weight gain. Interested women and men telephoned the study or returned a ier and were contacted by study staff for further screening. Using these methods 594 women and 228 men were recruited. A subsample of 404 low-income women were recruited in-person at local community clinics administering the Supplemental Nutrition Program for Women, Infants and Children (WIC). Eligibility criteria included 1) age 20 ± 45 y; 2) not currently pregnant or pregnant in the previous 12 months; 3) absence of current serious medical (for example, diabetes, heart disease, cancer) or psychological conditions (clinical depression, eating disorders) requiring treatment; 4) not currently following a physician-prescribed special diet; 5) willingness to participate in the study for three years. People with a history of depression, eating disorders, or other illness were not excluded. Few people seeking enrolment reported either current or a history of serious medical or psychological illnesses and no formal exclusion rates were calculated. There were no body weight criteria for study enrolment because prevention of weight gain was believed to be important for people of any body weight. Participants completed questionnaires and were measured for body weight annually (at baseline, and annual follow ups at year 1, year 2, and year 3). Half the participants were randomized to a mail-based intervention program and half to a no-contact control group. Intervention consisted of mailed monthly educational=motivational newsletters. Newsletters emphasized ve major themes: 1) regular self-weighing (at least once per week); 2) eat more fruits (at least two servings per day); 3) eat more vegetables (at least three servings per day); 4) reduce intake of high fat foods; 5) increase exercise, with an emphasis on walking (at least 20 min per day). Reduction in energy intake was not speci cally recommended. Additional details of the study are reported elsewhere. 27 Women who did not have weight or binge eating data at follow up (n ˆ 74) and women who became pregnant during the study (n ˆ 106) were excluded from analyses, leaving 817 women for the analysis sample. Men were not examined, because too few men reported binge eating (n ˆ 13). Women in the study sample at the end of year 3, were on average aged 37.3 y (s.d. ˆ 6.6), weighed lb (s.d. ˆ 38.4 lb.), had an average body mass index (BMI) of 27.4 kg=m 2 (s.d. ˆ 6.4 kg=m 2 ), and about 42% were overweight ( 27.3 kg=m 2 ). 28 ± 30 Of the female subjects 86% were white race=ethnicity, 85% were employed for pay, and 28% were low-income ( $ annual household income). About half of the women were currently married, about half had college degrees or more education, 18% were current smokers, and 36% had previously participated in a weight loss program. Measures The following measures were collected at the third follow-up clinic visit.. This was measured using the Questionnaire on Eating and Weight Revised. 2 Questions assessed the prevalence and frequency of binge eating during the past six months. Subjects were classi ed as binge eaters if they answered yes to two questions: 577

3 578 1) during the past six months, did you ever eat within any 2 h period what most people would regard as an unusually large amount of food; and 2) during the times when you ate this way, did you feel you couldn't stop eating or control what or how much you were eating. Frequency of binge eating was assessed for those who answered yes to these two questions. disorder (BED). DSM IV criteria 31 for BED were also self-reported (that is, symptoms experienced during a binge eating episode: eating rapidly, eating until uncomfortably full, eating large amounts of food when not hungry, embarassment or self-disgust after overeating), as well as degree of distress over eating, importance of weight and shape in self-evaluation, and compensatory behaviours during the past three months, speci cally in response to binge eating (for example, vomiting, laxative or diuretic use, diet pill use, fasting, exercise speci cally to compensate for binge episodes). Subjects were classi ed as having BED if they reported binge eating at a frequency of two or more days per week and 1) three or more of the ve items listed above for symptoms experienced during binge eating episodes; 2) reported being greatly=extremely upset by their overeating; or 3) reported being greatly=extremely upset by their lack of control over their eating. Body weight and weight loss history Body weight. Weight in street clothing without shoes was measured at each of the annual clinic visits, using a calibrated balance beam scale. Height was measured at the rst clinic visit with a wallmounted ruler. BMI was calculated as follow-up three measured body weight (kg)=height (m) 2. Overweight was de ned as follow-up three BMI 27.3 kg=m ± 30 Weight change was calculated as weight at follow-up three minus baseline weight. Desired weight. This was measured by asking subjects to report how much they would like to weigh. Desired weight loss was calculated as the difference between current and desired weight. History of weight loss and weight gain episodes. These were each assessed at the baseline clinic visit, by asking subjects to report the number of times (since the age of 15 y) they had lost or gained each of the following amounts: 5 ± 9 lb, 10 ± 19 lb, 20 ± 49 lb, 50 ± 79 lb, 80 ± 99 lb and > 100 lbs. Response options were 0, 1 ± 2 times, 3 ± 4 times, 5 ± 6 times, > 7 times. A continuous measure of intentional weight loss episodes was created by assigning the lower bound value of the range for the frequency category and summing across amounts. A continuous measure of pounds intentionally lost was created by multiplying the lower value of the frequency range by the midpoint of the pound range and summing across amounts (100 was used as the pound amount for the > 100 category). Intentional weight loss episodes and pounds, unintentional weight loss episodes and pounds, and weight gain episodes and pounds were calculated in this manner. Intentional weight loss was speci cally de ned as weight loss on purpose (not including pregnancy or illness). Unintentional weight loss was de ned speci cally as weight loss when not trying to lose weight (for example, because of illness). Weight gain was de ned as weight gain (not including pregnancy). In a separate study, nine month test-retest reliability for the composite measure of lifetime intentional weight loss was r ˆ History of participation in organized weight loss programs. This was also assessed (ever vs never participated). Dieting practices Speci c weight control behaviours. These were assessed using a 23-item checklist developed by the researchers. Participants were asked to report whether they had engaged in each of 23 speci c weight control behaviours during the past year. Two scales were derived from this list, based on factor analysis of the items. 33 A healthy weight control practices score was comprised of the sum of nine items (reducing calories, increasing exercise, increasing fruit and vegetable intake, decreasing fat intake, eliminating sweets and junk food, reducing food amounts, changing type of food eaten, eating less meat, and eating low-calorie diet foods; Cronbach alpha ˆ 0.86). An unhealthy weight control practices score consisted of the sum of ve items (laxative use, diuretic use, appetite suppressants, diet pills and liquid diet supplements; Cronbach alpha ˆ 0.60). The remaining items did not load on any factor and were not included in the two summary scales. Use of organized weight loss programs in the past year is reported for descriptive purposes. Weight control practices that were speci cally in response to binge eating episodes, were assessed separately, as described above, under the binge eating measure. In addition to the 23 behaviours, subjects were asked to report the frequency of self-weighing, the number of pounds they gained before noticing, and the number of pounds they gained before taking action to reverse the weight gain. A difference score was calculated for the latter two variables as a threshold measure for action against weight gain. Restrained eating. This was measured using the Cognitive Restraint subscale of the Three Factor Eating Questionnaire (TFEQ-R) of Stunkard and

4 Messick. 34 The TFEQ-R measures dietary restraint, or the extent to which one engages in cognitive and behavioural efforts to limit food intake. Items focus on speci c eating behaviours, such as eating smaller portions, keeping track of caloric intake and avoiding speci c foods. The TFEQ-R is valid and reliable, and identi es those who consume fewer calories in naturalistic 34 ± 37 settings. during the previous seven days and is sensitive to subclinical changes in mood or affect. The BDI is reliable and valid, and has been used previously with diverse populations, including people with lower incomes. Cutpoints for the study were based on the literature: 44 0 ± 9, not depressed; 10 ± 17, mild depression; 18 ± 29, moderate depression; 30 and above, severe depression. 579 Eating and exercise behaviours Dietary intake. This was measured using the 60- item Block et al 38 Food Frequency Questionnaire. Dietary variables included estimates for total daily energy, percent energy from fat, sweets, alcoholic beverages, and daily servings of fruits and vegetables. An additional question was included that asked subjects to report the number of meals per week eaten at fast food restaurants. Physical activity. Measurement was undertaken using an instrument adapted from Jacobs et al, 39 that has been used in several large epidemiological studies, including those involving low-income populations. The questionnaire describes 13 physical activities, with a range of intensity levels. For the present analyses, a physical activity score was calculated as the reported frequency per week of each activity multiplied by its estimated intensity level in metabolic equivalents (METS). These products were summed across 13 activity items. As an indirect measure of sedentary behaviour, subjects were asked to report the number of hours per day spent watching television. Kristal low fat eating behaviours. This measured the extent to which individuals adopted food preparation methods related to a low-fat diet. 40 This 18-item scale assesses ve theoretically based dimensions of eating behaviour and includes behaviours such as eating bread without butter or margarine, choosing low fat dairy products, removing skin from chicken, and using low-calorie salad dressing. The measure is valid and reliable, and sensitive to interventions targeting low fat eating behaviours. 40 Psychological variables Self-esteem. Measurement was made using the Rosenberg Self Esteem scale. 41 The scale consists of ten items assessing feelings of self-worth and has been used in a wide variety of populations, including those of low income. 42 Depression. This was measured using the Beck Depression Inventory (BDI). 43 This 20-item selfreport inventory, assesses depressive symptoms Life events. These were measured using the Holmes and Rahe Life Change scale. 45 Changes in each of the following ve domains, during the past six months, were reported as having occurred or not (yes or no): personal health, family events, personal events, work, and nancial. There are a total of 67 possible life changes. Each life event has an associated stress score. Stress scores are summed across reported life events during the past six months. Demographic variables. These were self-reported and included age in years; current marital status (currently married vs other); education ( high school degree, some college, college degree); ethnicity (white vs other); annual household income ( $25 000=y vs > $25 000=y); employment status (currently employed vs not currently employed for pay) and smoking status (current smoker vs not current smoker). Statistical analysis All analyses were conducted using SAS 6.12 statistical software. 46 Women from treatment and control groups were combined for the purpose of analysis, because the intervention did not have a signi cant impact on body weight, eating or exercise behaviours. 27 Analyses including treatment group as a covariate yielded very similar results and are thus not shown. To examine differences in weight history, dieting practices, eating and exercise behaviours, and psychological variables by binge eating and overweight status, general linear modeling was used for continuous dependent variables, and logistic regression for dichotomous dependent variables. Weight status, binge eating status and their interaction were entered as independent variables. Income and ethnicity were included as covariates, because they were signi cantly associated with overweight and binge eating in univariate analyses (described below). For retrospective analyses of weight change (year 3 minus baseline) over three years, weight at baseline was included as a covariate. Effects were interpreted as statistically signi cant at P < Four groups of interest were de ned for illustrative purposes in the tables below: overweight binge eaters (n ˆ 71), normal weight binge eaters (n ˆ 43), overweight non-binge eaters (n ˆ 271) and normal weight non-binge eaters (n ˆ 432).

5 580 Within binge eaters, differences among normal weight and overweight women on binge eating variables were examined using Chi square analyses. To examine the relative importance of each of the independent variables on binge eating, while simultaneously controlling for the other independent variables, a multivariate logistic regression was run. Independent variables were selected based on the literature and included depression (none vs mild or more severe; scores of 10 on BDI), weight at baseline, weight change (follow-up three - baseline), restrained eating, history of intentional weight loss episodes, and importance of weight=shape to selfevaluation in the past six months (main or most important). Income and ethnicity were included as covariates. Feelings of upset due to lack of control when overeating in the past six months and feelings of upset by overeating in the past six months were not included in this model, because they were believed to be outcomes of the binge eating itself, rather than potential causes. Results Demographic characteristics by binge eating and weight status Table 1 shows demographic characteristics of participants by binge eating and weight status. Overall, the prevalence of binge eating was 14.0% (9.1% among normal weight women and 20.8% among overweight women (P < 0.001). Thirteen individuals (1.6%) met the criteria for BED, based on self-reports to the Binge Eating questionnaire. Of these 13, 11 were overweight and two were normal weight. Thus, the prevalence of BED was 0.4% among normal weight women and 3.2% among overweight women in this sample. There were signi cant main effects for overweight status and binge eating status on several demographic variables. Compared to normal weight women, overweight women were more likely to report annual household income < $ and more likely to be of nonwhite ethnicity. Binge eaters were more likely to be white, weighed more and had higher BMIs than those who did not binge eat. No other demographic variables differed signi cantly by weight or binge eating status. Weight history and dieting behaviours by binge eating and weight status Table 2 shows weight history and dieting practices by binge eating and weight status. Overall, there were signi cant main effects for both weight status and binge eating for most weight history and dieting practice variables. Binge eaters gained more weight over three years compared to non-binge eaters. Overweight women gained signi cantly more weight over the three years study period than their normal-weight counterparts. The interaction between binge eating and overweight status for weight change was not statistically signi cant. Overweight women and binge eaters were also further above their desired weight, compared to normal weight women and non-binge eaters, respectively. Overweight women and binge eaters reported a greater number of weight gain and intentional weight loss episodes and pounds. For example, overweight women reported 2.78 more lifetime intentional weight loss episodes than normal weight women (P < ). Binge eaters reported 2.17 more lifetime intentional weight loss episodes compared to non-binge eaters (P < ). Unintentional weight loss did not differ by weight status or binge eating. Dieting practices differed signi cantly by binge eating and weight status, and there were no signi cant interactions for any of the dieting variables. Compared to non-binge eaters and normal weight women, binge eaters and overweight women reported a greater number of both healthy and unhealthy weight loss practices. In addition, binge eaters were more likely Table 1 Demographic variables by binge eating and overweight status in a community sample of 819 women enrolled in a weight gain prevention program Normal weight Overweight Non-bingers Bingers Non-bingers Bingers n (%) 432 (52.9) 43 (5.3) 271 (33.2) 71 (8.7) BMI (kg/m 2 ; m; s.e.m.) 23.1 (.2) 23.8 (.6) 33.2 (.2) 34.6 (.5) a *** b ** Weight (lbs; m; s.e.m.) (1.2) (3.9) (1.5) (3.0) a *** b ** Age (y; m; s.e.m.) 34.2 (.3) 34.7 (1.0) 35.0 (.4) 34.9 (.8) Education (% college) Income (% $25000) a ** White ethnicity (%) a *** b * Currently married (%) Not employed (%) Current smoker % a ˆ main effect for overweight status. b ˆ main effect for binge eating. **P < 0.01; ***P < BMI ˆ body mass index; m; s.e.m. ˆ mean; standard error.

6 Table 2 Adjusted means (s.e.m.) for weight history, dieting practices, and eating and exercise behaviours, by binge eating and overweight status in a community-based sample of 819 women 581 Normal weight Overweight Non-bingers Bingers Non-bingers Bingers n Weight history Weight change (lbs) a (0.8) (2.1) 13.4 (1.0) 17.9 (1.8) b ** c *** Current-desired weight (lbs) 12.1 (1.0) 15.0 (3.1) 51.6 (1.2) 60.4 (2.4) b ** c *** Intentional weight loss (lifetime): Episodes 3.8 (0.2) 6.4 (0.7) 6.6 (0.3) 8.7 (0.6) b *** c *** lb 27.9 (3.0) 58.6 (9.5) 65.6 (3.8) 92.0 (7.4) b *** c *** Unintentional weight loss (lifetime): Episodes 1.9 (0.1) 2.5 (0.4) 2.2 (0.2) 2.4 (0.3) lb 13.2 (1.8) 18.2 (5.9) 17.3 (2.3) 24.6 (4.6) Weight gain (lifetime): Episodes 4.3 (0.2) 6.0 (0.7) 7.2 (0.3) 9.7 (0.6) b *** c *** lb 33.1 (4.0) 46.9 (12.8) 77.7 (5.1) (10.0) b ** c *** Dieting practices Weighing frequency (weekly) 4.8 (0.4) 5.1 (1.3) 3.8 (0.5) 6.2 (1.0) Weight loss practices (past year) (n) 4.0 (0.2) 6.2 (0.6) 4.6 (0.2) 6.1 (0.4) b *** c * Healthy practices (past year) (n) 2.1 (0.1) 3.5 (0.4) 2.6 (0.2) 4.0 (0.3) b *** c * Unhealthy practices (past year) (n) 0.07 (0.02) 0.19 (0.1) 0.14 (0.03) 0.21 (0.1) b * c * Pounds gained to notice (n) 6.0 (0.5) 5.3 (1.6) 8.8 (0.7) 7.3 (1.3) c ** Pounds gained to act (n) 9.5 (0.8) 8.1 (2.5) 16.4 (1.0) 14.3 (2.0) c *** Pounds gained to act-notice (n) 3.4 (0.5) 2.7 (1.7) 7.5 (0.7) 7.0 (1.3) c *** Restraint score 8.5 (0.2) 9.4 (0.7) 8.2 (0.3) 8.4 (0.5) Joined weight loss program (past year) (%) b ** Ever joined weight loss program (%) b ** c *** Eating and exercise behaviours Physical activity score 46.8 (1.5) 41.1 (4.6) 39.7 (1.9) 37.0 (3.6) c ** Television viewing (h/d) 1.8 (0.1) 1.9 (0.2) 2.2 (0.1) 2.0 (0.2) c ** Fast food meals (weekly, n) 1.1 (0.1) 1.6 (0.2) 1.8 (0.1) 1.8 (0.2) c *** Total energy (kcals/d) (34.5) (109.7) (43.7) (85.6) c *** Fat energy (%) 31.9 (0.4) A 33.6 (1.2) AB 33.9 (0.5) B 32.5 (0.9) AB bc * Sweets energy (%) 13.3 (0.5) A 11.2 (1.6) A 13.8 (0.6) AB 16.0 (1.3) Bbc * Alcohol energy (%) 3.0 (0.2) 2.1 (0.6) 2.1 (0.3) 1.4 (0.5) c ** Vegetables (servings/d) 1.7 (0.1) 1.8 (0.2) 1.8 (0.1) 1.7 (0.1) Fruit (servings/d) 1.0 (0.04) 0.94 (0.1) 0.94 (0.1) 1.0 (0.1) Kristal Lowfat Eating Behaviours 51.4 (0.4) 50.7 (1.4) 49.8 (0.6) 49.7 (1.0) c * Psychosocial variables Past 6 Months: Upset about overeating (% greatly) b *** Upset about lack of control (% greatly) b *** Importance of weight or shape (% main or most important) b *** c ** Beck Depression Score 4.7 (0.3) 8.6 (0.9) 6.6 (0.4) 10.6 (0.7) b *** c *** Rosenberg Self Esteem Score 34.0 (0.2) 31.7 (0.8) 33.5 (0.3) 29.9 (0.6) b *** Life Events Score (5.2) (16.6) (6.6) (13.0) b ** Means adjusted for income and ethnicity. a Adjusted for weight at baseline, income and ethnicity. b main effect for binge eating. c ˆ main effect for overweight status. b c ˆ signi cant interaction. Different superscripts ( A,B,C,D ) indicate signi cantly different means by post hoc comparison (P < 0.05). *P < 0.05; **P < 0.01; ***P < than non-binge eaters to have taken part in an organized weight loss program in the past year (odds ratio (OR) 95% con dence interval (95% CI): 2.76 (1.48, 5.12) or in their lifetime (OR (95% CI) ˆ 1.78 (1.16, 2.74). Binge status was not signi cantly associated with number of pounds needed to gain before noticing or taking action. Compared to normal weight women, overweight women were more likely to have participated in an organized weight loss program (OR (95% CI) ˆ 3.44 (2.51, 4.73) and to have larger weight gain thresholds to notice and to take action. There were no signi cant main effects or interactions for frequency of self-weighing or restrained eating score. Eating and exercise behaviours Table 2 shows eating and exercise behaviours by binge eating and weight status. Overall, there were several main effects for weight status, but few signi cant effects for binge eating status. Compared to normal weight women, overweight women reported less physical activity, a greater number of hours of weekly television viewing, a greater frequency of fast food meals, higher total energy intake, and fewer low fat eating behaviours (lower Kristal scores). There was a signi cant interaction between weight and binge eating status for percent energy from fat and percent energy from sweets. Normal weight non-binge eaters consumed the lowest percent energy from fats and overweight non-binge eaters consumed the highest. Normal-weight and overweight binge eaters did not differ from either of the other two groups. Percent energy from sweets consumption was highest among overweight binge eaters, while the other three groups did not differ from each other. Normal weight women reported higher intake of alcohol and a greater number

7 582 of low fat eating behaviours than overweight women. Fruit and vegetable intake did not signi cantly differ by binge eating or weight status. Psychological characteristics Table 2 shows psychological variables by binge eating and weight status. In general, binge eaters scored higher than non-binge eaters on general measures of psychological distress, as well as on measures speci cally focusing on body image. There were no signi cant interactions between binge eating and overweight status on any of the psychological measures. Results of logistic regression analyses showed that binge eaters were about eight times more likely than non-binge eaters to report in the past six months being greatly upset about overeating (OR (95% CI) ˆ 8.06 (4.74, 13.71)) and were thirteen times more likely to report being upset by their lack of control over eating (OR (95% CI) ˆ (7.13, 24.94)). Weight and shape was over three times more likely to be the main or most important aspect of self-evaluation among binge eaters compared to non-binge eaters (OR (95% CI) ˆ 3.82 (2.50, 5,83). Binge eaters scored higher on the BDI, reported more stressful life events and lower self-esteem compared to non-binge eaters. Binge eaters were over twice as likely to meet clinical cutpoints for mild or more severe depression compared to non-binge eaters (OR (95% CI) ˆ 2.47 (1.58, 3.86). Compared to normal weight women, overweight women were more likely to report that weight and shape were central to selfevaluation (OR (95% CI) ˆ 1.66 (1.20, 2.30) and were more likely to meet clinical cutpoints for mild or more severe depression (OR (95% CI) ˆ 2.23 (1.56, 3.20). Binge characteristics by weight status Table 3 shows binge frequency and binge characteristics among binge eaters by weight status. In general, normal weight and overweight women who binged resembled each other on binge frequency and binge characteristics. Over half reported binge eating less than one day per week. A higher proportion of overweight binge eaters compared to normal weight binge eaters reported binging more than four days per week, but the difference was not statistically signi cant. Normal weight binge eaters were signi cantly more likely than overweight binge eaters to report exercising for an hour or more to avoid weight gain after binge eating. No other signi cant differences in compensatory behaviours were observed between normal weight and overweight binge eaters. Cross-sectional multivariate models examining correlates of binge eating Table 4 shows ORs and 95% CIs for the multivariate model predicting binge eating status. Variables with OR > 1 indicate a positive correlation with binge Table 3 Frequency and characteristics of binge eating and purging behaviors among binge eaters (normal weight and overweight) eating status or depression. Statistically signi cant associations are those in which the CI does not include 1.0. Compared to women who did not binge eat, binge eaters were more than three times as likely to report that weight and shape were the main or most important aspect of self-evaluation. Depression and history of intentional weight loss episodes were also independently positively associated with binge eating, although associations were weak. Discussion Normal Weight Overweight (n ˆ 43) (n ˆ 71) Binge frequency (past six months): < 1 d/week d/week ± 3 d/week d/week Usually experience during binge episode: Rapid eating Eat until uncomfortable Eat large amount when not physically hungry Eat alone due to embarrassment Feel self-disgust after overeating Upset about overeating (% greatly) Upset about lack of control (% greatly) Compensatory behaviours (past three months ± to avoid weight gain after binge eating): Vomiting Laxatives (twice recommended dose) 0 0 Diuretics (twice recommended dose) Fast (24 h) Exercise ( 1 h) * Diet pills (twice recommended dose) *P < Table 4 Multivariate model predicting binge eating among women enrolled in a weight gain prevention program 95% CI Variable OR Lower Upper P Weight (baseline) Weight change (follow up baseline) Low-income group White ethnicity Depressed Restraint score Intentional weight loss episodes Importance of weight/shape (main or most important) Overweight (follow up) % CI ˆ 95% con dence interval; OR ˆ odds ratio. The present study is unique in its inclusion of both overweight and normal weight binge eating women recruited from the community, and its inclusion of standardized measures of dietary intake, physical

8 activity, eating behaviours and psychological factors. Overall, overweight and normal weight binge eaters resembled each other in level of weight and shape concerns, dieting practices, depression, self-esteem and life events. Binge eaters were about three times as likely as non-binge eaters to believe that weight and shape are of central importance to their self-evaluation. Multivariate models showed this to be the strongest correlate of binge eating, although depression and history of weight loss episodes were weakly associated with binge eating, independent of weight= shape importance. These ndings are consistent with previous studies showing greater levels of maladaptive attitudes about weight and shape, and higher levels of depression and psychological distress among binge eaters compared to non-binge 1,2,9,12 ± 19,47 eaters. Interestingly, normal weight binge eaters resembled overweight non-binge eaters in weight loss history, suggesting that normal weight binge eaters are as involved in weight control efforts as overweight women. In addition, normal weight binge eaters were over twice as likely as overweight non-binge eaters to have joined an organized weight loss program in the past year. Normal weight binge eaters were signi cantly more likely than overweight binge eaters to use exercise to compensate for a binge; within binge eaters, this was the only compensatory behaviour that signi cantly differed by weight status. Whether their `excessive' preoccupation with body weight and shape, and their active dieting practices, prevent normal weight binge eaters from gaining weight over time is an interesting question that deserves further study. If these women have a history of weight gain or are otherwise predisposed to weight gain, perhaps their weight concerns and weight control efforts act to prevent weight gain that would otherwise occur. On the other hand, the nding that excessive importance placed on weight and shape is strongly associated with binge eating implies that this attitude may be a harmful one to hold, possibly leading to binge eating. This is an area deserving additional research attention to further de ne and evaluate preventive weight control strategies practiced by normal-weight women and their effects on psychological well-being. Longitudinal studies are needed to further understand the causal association, if any, between depression, binge eating and maladaptive attitudes about weight and shape, and whether these associations differ among normal weight and overweight women. Depression and binge eating could result from failure to meet unrealistic standards for body weight or shape, or binge eating could contribute to higher depression and unrealistic attitudes about weight and shape. 4 In the present study, few differences were found among binge eaters and non-binge eaters in energy intake and physical activity levels. Overall, little is known about the eating and exercise habits of women who binge eat. Previous studies have found few macronutrient differences between binge and nonbinge eaters on usual intake at meals. 20 However, there is some evidence of higher fat intake during binge episodes. 21 ± 26 Eating and exercise data are dif cult to interpret among women who binge eat, because habitual patterns of eating are typically assessed. Episodes of binge eating may not be re ected on measures of usual dietary intake, especially if binge frequency is low (for example, less than once a week in the present study). In addition, it is unknown how subjects are reporting eating and exercise behaviours that surround their binge eating. For example, do they incorporate these episodes (of eating, exercise) into their average behaviour reports, or are these episodes exempt from the `average' behaviour, because they are viewed as unusual and not representative of their typical patterns? In addition, shame or embarrassment could contribute to the omission of these episodes from self-reports of eating and exercise behaviour. Few data are available that address these methodological issues. Data are needed on this issue, because one binge episode per week might result in 52 d of unusually high energy intake, which could substantially affect average estimates of energy intake or expenditure. Even less is known about the role of physical activity, and whether binge eaters use physical activity to regulate weight to a greater extent than non-binge eaters. Data from the present study suggest that normal weight binge eaters are more likely than overweight binge eaters to use physical activity speci cally to prevent weight gain following a binge episode. This is an intriguing area in need of further research. Independent of initial weight, women who were binge eaters at follow-up had gained more weight over the three year study period than women who were not binge eaters at follow-up. The ndings from the present study suggest that the effects of overweight and binge eating may be independent sources of in uence on weight change over time. However, since binge eating was assessed only at the three year follow-up and not at the baseline, the causal association between binge eating and weight gain is not clear in the present study. Future studies should include both normal weight and overweight binge eaters and assess binge eating prospectively to determine the unique contributions of binge eating and overweight status to weight gain over time. Strengths of the present study include its large and heterogeneous nonclinical sample of women, standardized measures of binge eating and dieting practices, and comprehensive assessment of other important behavioural and psychological variables. While previous studies have examined the prevalence of binge eating in the community, the present study included standardized measures of other important behavioural and psychological correlates of binge eating, including dietary intake, eating behaviours, physical activity, weight and shape concerns, and depression. These measures were collected using a 583

9 584 standardized protocol on a large and demographically heterogenous nonclinical sample, that included overweight and normal weight women. Limitations are those shared by most cross-sectional epidemiological survey studies, including lack of information about temporal associations between binge eating, dieting and psychological factors, and reliance on self-reports for assessment of binge eating, purging behaviours, and eating and exercise behaviours. In addition, although community-based, the sample was not a random or representative one, thus results on the prevalence and correlates of binge eating may not be completely generalizable. For example, the exclusion of women currently in treatment for depression, eating disorders, or other serious medical or psychological problems, may have resulted in an underestimate of the prevalence of binge eating. Very few interested people were excluded from study participation, so it is believed that this particular source of potential selection bias was not strong. However, the sample may have been overrepresented by individuals who were generally healthy and who were interested in taking an active role to maintain their health. Conclusion To summarize, about 9% of normal weight women and 21% of overweight women reported binge eating in the past six months in this nonclinical sample of adult women. These ndings show that binge eating may not be as prevalent in the community, as suggested by earlier studies which used less well-de ned questions to assess binge eating. 5±7 Recent community studies, that used more precise measures of binge eating, yielded similar prevalence estimates to the present study. 9 In the present nonclinical sample of healthy women, binge eating was about nine times more prevalent than eating disorders such as bulimia nervosa 5 and it was associated with signi cant psychological distress. It is unclear whether psychological distress leads to binge eating, whether binge eating leads to distress, or whether both are caused by other factors such as extreme attitudes toward weight and shape. is not associated with unhealthy eating and exercise habits; although measures of habitual patterns may not capture practices associated with binge episodes. Further prospective studies are needed to understand the potential causal associations between binge eating, weight gain, dieting practices and psychological distress, among women in the community. Acknowledgements This research was supported by grant DK45361 from the National Institute of Diabetes and Digestive and Kidney Diseases with additional funding from the Centers for Disease Control and Prevention. References 1 Marcus, MD. in obesity. In: Fairburn CG, Wilson GT (eds). : Nature, assessment and treatment. Guilford Press: NY, 1993; pp 77 ± Yanovski SZ. Binge Eating Disorder: Current knowledge and future directions. Obes Res 1993; 1: 306 ± Polivy J, Herman CP. Dieting and binging: A causal analysis. Am Psychol 1985; 40: 193 ± Polivy J, Herman CP. Etiology of binge eating: Psychological mechanisms. In: Fairburn CG, Wilson GT (eds). : Nature, assessment and treatment. Guilford Press: NY, 1993; pp 173 ± Fairburn CG, Hay PJ, Welch SL. and bulimia nervosa: Distribution and determinants. In: Fairburn CG, Wilson GT (eds). : Nature, assessment and treatment. Guilford Press: NY. 1993; pp 123 ± Fairburn CG, Beglin SJ. Studies of the epidemiology of bulimia nervosa. Am J Psychiatry, 1990; 147: pp 401 ± Fairburn CG, Beglin SJ. Assessment of eating disorders: Interview or self-report questionnaire? Int J Eat Dis 1994; 16, pp 363 ± French SA, Peterson CB, Story M, Anderson N, Mussell MP, Mitchell JE. Agreement between survey and interview measures of weight control practices in adolescents. Int J Eating Dis 1998; 23: pp 45 ± Spitzer RL, Yanovski S, Wadden T, Wing R, Marcus MD, Stunkard A, Devlin M, Mitchell J, Hasin D, Horne RL. Binge eating disorder: Its further validation in a multisite study. Int J Eating Dis 1993; 13: pp 137 ± Fairburn CG, Beglin SJ, Davies B. Eating habits and disorders amongst young adult women: An interview-based study (Unpublished manuscript). 11 Devlin MJ, Walsh BT, Spitzer RL, Hasin D. Is there another binge eating disorder? A review of the literature on overeating in the absence of bulimia nervosa. Int J Eat Dis 1992; 11: pp 333 ± Spitzer RL, Devlin M, Walsh BT, Hasin D, Wing R, Marcus M, Stunkard A, Wadden T, Yanovski S, Agras S, Mitchell J, Nonas C. Binge Eating Disorder: A multisite eld trial of the diagnostic criteria. Int J Eat Dis 1992; 11: pp 191 ± Marcus MD, Smith D, Santelli R, Kaye W. Characterization of eating disordered behavior in obese binge eaters. Int J Eating Dis 1992: 12: pp 249 ± Mussell MP, Peterson CG, Weller CL, Crosby RD, dezwaan M, Mitchell JE. Differences in body image and depression among obese women with and without binge eating disorder. Obes Res 1996: 4: pp 431 ± dezwaan M, Nutzinger DO, Schoenbeck G. in overweight women. Comp Psychiatry 1992; 33: pp 256 ± Fichter MM, Quad ieg N, Brandl B. Recurrent overeating: An empirical comparison of binge eating disorder, bulimia nervosa and obesity. Int J Eat Dis 1993; 14: pp 1 ± Marcus MD, Wing RR, Ewing L, Kern E, Gooding W, McDermott M. Psychiatric disorders among obese binge eaters. Int J Eat Dis 1990; 9: pp 69 ± Marcus MD, Wing RR, Hopkins J. Obese binge eaters: Affect, cognitions and response to behavioral weight control. J Consult Clin Psychol 1988; 3: pp 433 ± Wilson GT, Nonas CA, Rosenblum GD. Assessment of binge eating in obese patients. Int J Eat Dis 1993; 13: pp 25 ± French SA, Jeffery RW. Consequences of dieting to lose weight: Effects on physical and mental health. Health Psychol 1994; 13: pp 195 ± Mitchell JE, Crow S, Peterson CB, Wonderlich S, Crosby RD. Feeding laboratory studies in patients with eating disorders: A review. Int J Eat Dis 1998; 24: pp 115 ± Yanovski SZ, Leet M, Yanovski JA, Flood M, Gold PW, Kissileff HR, Walsh BT. Food selection and intake of obese women with Binge Eating Disorder. Am J Clin Nutr 1992; 56: pp 975 ± 980.

10 23 Hetherington MM, Altemus M, Nelson ML, Bernat AS, Gold PW. Eating behavior in bulimia nervosa: Multiple meal analysis. Am J Clin Nutr 1994; 60: pp 864 ± Davis R, Freeman RJ, Garner DM. A naturalistic investigation of eating behavior in bulimia nervosa. J Consult Clin Psychol 1988; 56: pp 273 ± Rosen JC, Leitenberg H, Fisher C, Khazam C. episodes in bulimia nervosa: The amount and type of food consumed. Int J Eat Dis 1986; 5: pp 255 ± Kaye WH, Weltzin TE, McKee M, McConaha C, Hansen D, Hsu LKG. Laboratory assessment of feeding behavior in bulimia nervosa and healthy women: Methods for developing a human feeding laboratory. Am J Clin Nutr 1992; 55: 372 ± Jeffery RW, French SA. Prevention of weight gain in adults: the Pound of Prevention Study. Am J Public Health 1999; (in press). 28 Najjar MF, Rowland M. Anthropometric reference data and prevalence of overweight, United States, 1976 ± 80. Vital Health Stat 1987; Metropolitan Life Insurance Company. Metropolitan height and weight tables. Stat Bull Metrop Insur Co. 1983; 64: pp 2 ± National Institutes of Health. Health implications of obesity: National Institutes of Health Consensus Development Conference statement. Ann Intern Med 1985; 103: pp 1073 ± American Psychiatric Association. Diagnostic and statistical manual of mental disorders (4th ed). American Psychiatric Association: Washington, DC, French SA, Jeffery RW, Folsom AR, Willamson DF, Byers T. Weight variability in a population-based sample of older women: Intercorrelation and reliability of measures. Int J Obes 1995; 19: pp 22 ± French SA, Jeffery RW. Current dieting, weight loss history, and weight suppression: Behavioral correlates of three dimensions of dieting. Addictive Behav 1997; 22: pp 31 ± Stunkard AJ, Messick S. The Three Factor Eating Questionnaire to measure dietary restraint, disinhibition and hunger. J Psychosom Res 1985; 29: pp 71 ± Laessle RG, Tuschl RJ, Kotthaus BC, Pirke KM. A comparison of the validity of three scales for the assessment of dietary restraint. J Abnorm Psychol 1989; 98: pp 504 ± Tuschl RJ, Platte P, Laessle RG, Stichler W, Pirke KM. Energy expenditure and everyday eating behavior in healthy young women. Am J Clin Nutr 1990; 52: pp 81 ± Heatherton TF, Herman CP, Polivy J, King GA, McGree ST. The (mis)measurement of restraint: An analysis of conceptual and psychometric issues. JAbnormPsychol1988; 97: pp 19 ± Block G, Hartman AM, Dresser CM, Carroll MD, Gannon J, Gardner L. A data-based approach to diet questionnaire design and testing. Am J Epidemiol 1986; 124: pp 453 ± Jacobs DR, Hahn LP, Haskell WL, Pirie P, Sidney S. Validity and reliability of short physical activity history: CARDIA and the Minnesota Heart Health Program. J Cardiopulmon Rehabil 1989; 9: pp 448 ± Kristal AR, White E, Shattuck AL, Curry S, Anderson GL, Fowler A, Urban N. Long-term maintenance of a low-fat diet: Durability of fat-related dietary habits in the Women's Health Trial. J Am Dietetic Assoc 1992; 92: 553 ± Rosenberg M. Society and the adolescent self-image. Princeton University Press: Princeton, Gortmaker SL, Must A, Perrin JM, Sobol AM, Dietz WH. Social and economic consequences of overweight in adolescence and young adulthood. New Engl J Med 1993; 329: pp 1008 ± Beck AT, Ward CH, Mendelson M, Mock J, Erbaugh J. An inventory for measuring depression. Arch Gen Psychiat 1961; 4: pp 53 ± Beck AT, Steer RA, Garbin MG. Psychometric properties of the Beck Depression Inventory: Twenty- ve years of evaluation. Clin Psychol Rev 1988; 8: pp 77 ± Holmes T, Rahe R. The Social Readjustment Rating Scale. Psychomet Res 1967; 11: pp 213 ± SAS Institute, Inc. SAS=STAT software: Changes and enhancements through Release SAS Institute: Cary, Loro AD, Orleans CS. in obesity: Preliminary ndings and guidelines for behavioral analysis and treatment. Addict Behav 1981; 6: pp 155 ±