Objective: To explore the extent to which binge eating in the absence of compensatory behaviors (BE) is associated with psychiatric and medical symptoms in men and women and to control for the independent effects of BMI.
Research Methods and Procedures: A series of regression models was applied to questionnaire data on 8045 twins, 18 to 31 years old, from a population-based Norwegian registry.
Results: BE was significantly associated with elevated obesity, overweight, symptoms of eating disorders, symptoms of anxiety and depression, panic attacks, depressive episodes, and reduced life satisfaction in both men and women. In women, BE was independently associated with insomnia and early menarche. In men, BE was independently associated with specific phobia, daily smoking, alcohol use, use of pain medication, impairment due to mental health, neck-shoulder, lower back, and chronic muscular pain, and impairment due to physical health. Both men and women with BE reported higher rates of psychiatric treatment.
Discussion: Our results indicate that there is substantial comorbidity between BE and psychiatric symptoms independently of BMI for both men and women. Medical symptoms co-occur less frequently than previously reported from treatment-seeking populations in women. Across all domains, the array of symptoms exhibited by men with BE was broader than that observed in women with BE. This observation suggests the importance of considering gender differences in future studies of psychiatric and medical morbidity, binge eating, and obesity.
Binge eating disorder (BED)1 is currently listed in the appendix of the DSM-IV as a disorder requiring further study. The process of providing scientific validation for a new diagnostic category is complex and includes the evaluation of concurrent symptoms (1), an approach that has been undertaken with regard to BED.
Obese individuals with binge eating have a higher lifetime prevalence of mood, anxiety, alcohol use, and personality disorders than obese individuals without binge eating (2, 3, 4, 5). Individuals with BED report elevated prevalences of lifetime mood disorder, anxiety disorders, substance use disorders, insomnia, and psychosocial distress (6, 7).
BED has been shown to be common in patients with type 2 diabetes (8). BED in women seeking treatment for medical problems has been found to be associated with higher rates of diabetes, somatic symptoms such as limb or joint pain, headache, gastrointestinal problems, menstrual problems, and disability, even when controlling for the presence of comorbid psychiatric disorders (7). Obese women with binge eating have reported higher rates of major medical disorders than obese women without binge eating, but the differences were not significant, although the obese women with binge eating reported significantly greater health dissatisfaction (5).
These studies lend support to the notion that BED carries with it substantial risk for psychiatric and medical symptoms and disorders. However, the literature has three important limitations. First, the associations have been identified primarily in clinical samples (i.e., samples recruited for treatment trials). Therefore, the results may be influenced by sampling bias and, thus, not representative of persons with the disorder in the general population. It has been shown that people with more than one problem are more likely to seek treatment, resulting in an overestimation of comorbidity in clinical samples (9, 10). Second, given that BED is commonly associated with obesity, it is critical to estimate the impact of BED independently of BMI. Third, the literature has focused almost entirely on women. Although there are many similarities in the clinical features, etiology, and outcome for men and women with eating disorders, differences have been identified with regard to both eating disorder attitudes and comorbidity (11). Gender differences in biology and sociocultural environment and differing patterns of symptom expression also suggest that results from studies of women samples cannot automatically be generalized to male populations.
In the present investigation, we sought to provide more representative estimates of psychiatric and medical symptoms associated with the core features of BED by studying individuals with binge eating in the absence of compensatory behaviors from a population-based sample of young-adult Norwegian twins of both genders. By controlling for BMI, we explored the extent to which the observed associations were accounted for by body mass.
Research Methods and Procedures
The data used in this report derive from an ongoing longitudinal study of mental and physical health in twins from the Norwegian Institute of Public Health Twin Panel, a population-based twin registry (12). In 1998, 8045 subjects (3443 men and 4602 women) responded to a questionnaire sent to all twins born in Norway between 1967 and 1979, where both members of the pairs were still alive and for whom an address could be obtained (response rate 63%). The mean age of the responders was 25.52 ± 3.70, with a range of 18 to 31 years. The sample consisted of 3334 complete pairs and 1377 single responders. The numbers of twins in each zygosity group in this sample were: 1240 monozygotic men, 1068 dizygotic men, 1713 monozygotic women, 1517 dizygotic women, and 2507 dizygotic opposite sex twins.
The questionnaire included nine questions assessing symptoms during the past 6 months directly related to the DSM-IV eating disorders criteria, as well as self-rated height and weight. The binge eating item “Have you lost control while eating and were unable to stop before you had eaten too much?” was designed to capture the two key features of binge eating as outlined in DSM-IV: eating an unusually large amount of food and experiencing a sense of loss of control (possible responses: at least twice a week, 1 to 4 times a month, and seldom or never). The frequency indicated by the first option reflects DSM-IV criterion D for BED. The same response categories were used for each of the items regarding inappropriate compensatory behaviors: “Have you used: 1) vomiting or 2) laxatives or 3) fasting or 4) excessive physical exercise to control your weight?”. The DSM-IV criteria B and C, which deal with behavioral aspects of the binge eating episodes and attitudes toward binge eating, were not assessed.
Our criteria for binge eating in the absence of compensatory behaviors (BE) required binge eating with a feeling of loss of control at least 2 times a week and the absence of regular use of any of the above-mentioned inappropriate compensatory behaviors. Individuals who fulfilled criteria for anorexia nervosa or bulimia nervosa were excluded. The overall number of individuals with missing data for BE was 91 (2.6%) for men and 123 (2.7%) for women. Missing scores were evenly distributed among zygosity groups. BMI was calculated based on self-rated height and weight. Overweight was defined as BMI > 25 and obesity as BMI > 30 (World Health Organization; http:www.who.intnutobs.htm).
A five-item version (SCL-5) of the Hopkins Symptom Check List (SCL)-25 was used to assess symptoms of anxiety and depression (13). Self-report symptom scales like the SCL have considerable temporal stability and seem to approximate the “trait”- like aspects of psychological functioning (14). The questionnaire also included screening questions for lifetime depressive episodes (“I have had periods when I have been depressed and at the same time had sleeping problems and diminished ability to concentrate”); panic attacks (“I have suddenly had intense fear or panic without a reason”); specific phobias (“I am very scared of specific things, e.g., animals, heights, deep water, blood, flying”); daily smoking, alcohol problems, use of psychiatric treatment, insomnia, impairment due to mental health problems (“How much have mental health problems impaired you in work or in your daily social activities with family or friends during the last 4 weeks?”); use of tranquilizers, sleep medication, or pain medication during the last month; and current life satisfaction (“When you think of how you are doing currently, how pleased are you in general with your life?”).
The twins were asked about the lifetime occurrence of physical symptoms and disorders that they found to be sufficiently severe to be classified as an “illness or a health problem,” including neck-shoulder pain, low back pain, chronic muscular pain, migraine, other types of frequent headaches, irritable bowel syndrome, diabetes, and asthma. Questions on exercise habits [“How often do you do exercise (e.g., hiking, skiing, or training/sports)?”] and impairment due to physical health problems in the last month were also included, as well as three questions related to menstruation.
Logistic regressions were used to examine the association between BE and eating disorder symptoms, psychiatric and behavioral symptoms, and medical symptoms. BE was entered as an independent variable in a number of analyses with each of the symptom variables as dependents. The analyses were then repeated while controlling for BMI. One difficulty with the analysis of twin data is that the members of a twin pair are not statistically independent. In our sample, the correlations between twins ranged from 0.45 for monozygotic women to 0.12 for dizygotic opposite sex twins. Although “clustering” of twins does not alter regression coefficient estimates, it can yield inappropriately small SEs. We used Generalized Estimating Equations (15) to account for clustering. The Generalized Estimating Equation estimates the within-pair correlation as a nuisance parameter and yields asymptotically correct SEs.
The SCL-5 variable was highly skewed and was dichotomized for the regression analyses. We used a cut-off point that selected individuals with the highest 26% of the scores (30.3% of the women and 20.3% of the men). Other variables with several response categories were dichotomized. Linear regression was used with continuous variables. All significance tests were two-tailed.
The intent of these analyses was to gain insight into comorbidity patterns associated with BE independently of BMI. Given that this was the first study of its kind to use a large population-based register of both men and women, these analyses were explicitly exploratory in nature and ultimately hypothesis generating. The different associations in this study are of interest on a purely one-at-a-time basis (i.e., no joint hypothesis was tested). Therefore, no stringent corrections for multiple tests were applied (16). However, in recognizing that we performed a large number of statistical tests, which increases the risk of false positive findings, we discuss only those results that are significant at a p ≤ 0.01 level.
The 6-month prevalence of BE was 5.2% for women and 3.8% for men (χ21 = 9.65; p = 0.002). There were no significant differences between individuals with and without BE with regard to age, total years of education, and, for men, number of children. Women with BE had significantly more children than women without BE, but the difference was not significant after controlling for BMI (t = 1.53; df = 4244; p = 0.13) (Table 1). Men with BE were significantly more often married or cohabiting than men without BE, but this difference was not significant after controlling for BMI (χ21 = 3.17; p = 0.075). There was no significant difference in marital status between the two groups in women. Life satisfaction was significantly lower in both men and women with BE compared with individuals without BE. This remained significant after controlling for BMI (men, χ21 = 7.50, p = 0.003; women, χ21 = 17.00, p < 0.0001). Men with BE exercised significantly less than men without BE (χ21 = 14.82, p < 0.0001 after controlling for BMI). In women, exercise frequency was not significantly associated with BE. Age at menarche was significantly lower in women with BE even after controlling for BMI (t = 4.17; df = 4206; p < 0.0001).
Table 1. Demographic and health-related characteristics of men and women with and without BE
|Males||(N = 3226)||(N = 126)|
| Age [mean (SD)]||25.60 (3.67)||26.16 (3.71)|
| BMI [mean (SD)]*||23.81 (2.81)||26.44 (4.22)|
| Marital status [% (N)]|| || |
| Not married or cohabiting||59.7 (1918)||44.0 (55)|
| Married or cohabiting||39.7 (1274)||54.4 (68)|
| Widower||0.2 (6)||0 (0)|
| Separated/divorced||0.5 (15)||1.6 (2)|
| Total years of education [mean (SD)]||10.64 (5.02)||10.26 (4.90)|
| Number of children [mean (SD)]||0.31 (0.67)||0.36 (0.66)|
| Life satisfaction (satisfied) [% (N)]†||85.6 (2754)||72.0 (90)|
| Exercise frequency (once a week or more) [% (N)]‡||66.8 (2150)||45.2 (57)|
|Females||N = 4244||N = 235|
| Age [mean (SD)]||25.47 (3.72)||25.21 (3.71)|
| BMI [mean (SD)]*||22.30 (3.32)||24.76 (4.09)|
| Age at menarche (SD)‡||13.08 (1.38)||12.66 (1.42)|
| Marital status [% (N)]|| || |
| Not married/cohabiting||44.9 (1894)||37.3 (87)|
| Married/cohabiting||53.5 (2256)||60.9 (142)|
| Widow||0 (2)||0 (0)|
| Separated/divorced||1.6 (66)||1.7 (4)|
| Total years of education [mean (SD)]||9.90 (5.08)||9.27 (5.03)|
| Number of children (mean)||0.55 (0.88)||0.73 (0.94)|
| Life satisfaction (satisfied) [% (N)]‡||83.8 (3547)||71.9 (169)|
| Exercise frequency (once a week or more) [% (N)]||70.8 (2996)||69.7 (163)|
Eating- and Weight-Related Variables
The presence of BE in both men and women was associated with significantly elevated odds ratios (ORs) for both overweight and obesity (Table 2). Attitudes commonly associated with eating disorders, such as feeling too fat, intense fear of gaining weight, and undue influence of weight on self-evaluation, were significantly associated with BE in both men and women and remained significant after controlling for BMI (Table 2).
Table 2. Association between BE and obesity, overweight, and eating disorder symptoms before and after adjusting for BMI
|Males||N = 3226||N = 126|| || || || || || |
| Obesity||2.8||16.7||7.06||4.22 to 11.82||<0.0001|| || || |
| Overweight||28.9||55.6||3.07||2.14 to 4.40||<0.0001|| || || |
| Feel too fat||4.6||29.4||8.66||5.72 to 13.12||<0.0001||4.68||2.80 to 7.82||<0.0001|
| Intense fear of weight gain||3.3||18.3||6.38||3.94 to 10.34||<0.0001||3.97||2.29 to 6.88||<0.0001|
| Undue influence of weight on self-evaluation||8.8||20.0||2.76||1.76 to 4.32||<0.0001||2.59||1.62 to 4.14||<0.0001|
|Females||N = 4244||N = 235|| || || || || || |
| Obesity||2.9||12.3||4.64||3.02 to 7.13||<0.0001|| || || |
| Overweight||16.0||37.9||3.20||2.42 to 4.21||<0.0001|| || || |
| Feel too fat||23.4||63.0||5.00||3.78 to 6.60||<0.0001||3.93||2.91 to 5.30||<0.0001|
| Intense fear of weight gain||20.6||54.5||4.33||3.30 to 5.68||<0.0001||3.67||2.78 to 4.86||<0.0001|
| Undue influence of weight on self-evaluation||23.6||41.4||2.19||1.67 to 2.87||<0.0001||2.36||1.79 to 3.10||<0.0001|
Psychiatric and Behavioral Symptoms
Symptoms of anxiety and depression (SCL-5) and lifetime occurrence of panic attacks and depressive episodes were significantly more frequent in both men and women with BE compared with individuals without BE after controlling for BMI (Table 3). For some of the associations (e.g., depression in men and women), the relationship was slightly stronger after controlling for BMI. This is explained by the fact that when BMI has a negative effect on the dependent variable, the association with BE becomes more positive because BE and BMI are highly correlated.
Table 3. Association between BE and psychiatric and behavioral symptoms before and after adjusting for BMI
|Males||N = 3226||N = 126|| || || || || || |
|Symptoms of anxiety and depression (SCL-5>8)||19.5||36.4||2.34||1.58 to 3.45||<0.0001||2.76||1.85 to 4.13||<0.0001|
| Panic attacks*||3.7||11.5||3.38||1.89 to 6.06||<0.0001||3.68||1.95 to 6.93||<0.0001|
| Specific phobia*||15.3||25.2||1.77||1.17 to 2.67||0.006||1.71||1.12 to 2.59||0.01|
| Depressive episodes*||26.1||36.3||1.55||1.07 to 2.25||0.02||1.69||1.15 to 2.47||0.007|
| Smoking (daily)*||40.0||60.3||2.17||1.51 to 3.13||<0.0001||2.20||1.51 to 3.19||<0.0001|
| Alcohol problems*||22.0||38.8||2.20||1.51 to 3.20||<0.0001||2.07||1.40 to 3.05||0.0003|
| Insomnia†||5.5||8.7||1.55||0.77 to 3.24||0.21||1.58||0.77 to 3.24||0.21|
| Tranquilizers‡||0.9||2.1||2.31||0.54 to 9.89||0.26||1.95||0.44 to 8.58||0.38|
| Sleep medication‡||0.3||2.1||7.43||1.56 to 35.48||0.01||5.32||0.94 to 30.10||0.06|
| Pain medication‡||3.1||10.3||3.57||1.91 to 6.68||0.0001||3.56||1.82 to 6.98||0.0002|
| Impairment due to mental health‡||4.8||15.0||3.47||2.07 to 5.83||<0.0001||3.89||2.23 to 6.79||<0.0001|
| Psychiatric treatment*||6.0||11.9||2.09||1.20 to 3.64||0.009||2.04||1.13 to 3.68||0.02|
|Females||N = 4244||N = 235|| || || || || || |
| Symptoms of anxiety and depression (SCL-5>8)||28.5||47.1||2.09||1.59 to 2.74||<0.0001||2.16||1.63 to 2.86||<0.0001|
| Panic attacks*||9.8||16.2||1.64||1.12 to 2.42||0.01||1.69||1.14 to 2.52||0.0095|
| Specific phobia*||32.2||40.4||1.35||1.02 to 1.78||0.03||1.28||0.97 to 1.69||0.08|
| Depressive episodes*||34.2||46.1||1.59||1.21 to 2.09||0.0009||1.65||1.25 to 2.18||0.0004|
| Smoking (daily)*||44.0||53.2||1.33||1.03 to 1.71||0.03||1.28||0.99 to 1.65||0.06|
| Alcohol problems*||20.1||27.7||1.42||1.05 to 1.93||0.02||1.37||1.00 to 1.87||0.05|
| Insomnia†||5.7||10.6||1.90||1.23 to 2.94||0.004||1.97||1.25 to 3.13||0.004|
| Tranquilizers‡||1.6||2.7||1.70||0.66 to 4.36||0.27||1.28||0.48 to 3.42||0.62|
| Sleep medication‡||0.9||2.8||0.65||0.09 to 4.71||0.67||0.44||0.06 to 3.27||0.42|
| Pain medication‡||7.4||9.4||1.19||0.72 to 1.98||0.49||1.15||0.70 to 1.89||0.58|
| Impairment due to mental health†||6.6||10.7||1.64||1.03 to 2.62||0.04||1.61||1.00 to 2.59||0.05|
| Psychiatric treatment*||12.5||19.2||1.57||1.12 to 2.21||0.009||1.56||1.11 to 2.21||0.01|
In men, the presence of BE was also significantly associated with elevated ORs for specific phobia, daily smoking, alcohol problems, use of pain medication, and impairment due to mental health after controlling for BMI. Use of sleep medication was only significantly associated with BE before adjusting for BMI. In women, higher rates of insomnia were associated with BE even after controlling for BMI. Both men and women with BE reported higher rates of psychiatric treatment. This association remained significant in women, but not men, when adjusting for BMI.
In men, significant associations were observed between BE and neck-shoulder pain, low back pain, and chronic muscular pain, independently of BMI (Table 4). Men with BE also reported significantly more impairment due to physical health problems independently of BMI. In women, there were no significant associations between BE and any medical symptoms. The prevalence of type 2 diabetes was below 1% and was not included in the analyses.
Table 4. Association between BE and self-reported lifetime medical symptoms before and after adjusting for BMI
|Males||N = 3226||N = 126|| || || || || || |
| Neck-shoulder pain||6.9||18.3||2.88||1.78 to 4.66||<0.0001||2.72||1.63 to 4.55||0.0001|
| Low back pain||16.9||29.4||2.04||1.38 to 3.02||0.0004||1.93||1.29 to 2.89||0.001|
| Migraine||5.1||7.1||1.52||0.78 to 2.96||0.21||1.50||0.77 to 2.95||0.24|
| Other frequent headaches||5.5||10.3||1.99||1.11 to 3.58||0.02||1.76||0.95 to 3.26||0.07|
| Chronic muscular pain||2.5||9.5||4.15||2.19 to 7.86||<0.0001||3.50||1.76 to 6.98||0.0004|
| Irritable bowel syndrome||2.8||5.6||1.79||0.76 to 4.17||0.18||1.66||0.69 to 3.98||0.26|
| Asthma||7.3||9.5||1.31||0.71 to 2.41||0.38||1.17||0.63 to 2.19||0.62|
| Impairment due to physical health*||5.5||16.0||2.98||1.81 to 4.93||<0.0001||3.57||2.08 to 6.12||<0.0001|
|Females||N = 4244||N = 235|| || || || || || |
| Neck-shoulder pain||18.1||19.1||1.08||0.77 to 1.52||0.66||1.03||0.73 to 1.46||0.85|
| Low back pain||20.3||25.1||1.31||0.96 to 1.78||0.09||1.20||0.87 to 1.64||0.26|
| Migraine||9.8||11.5||1.12||0.73 to 1.72||0.60||1.21||0.78 to 1.87||0.40|
| Other frequent headaches||15.2||14.5||0.95||0.65 to 1.38||0.78||0.89||0.61 to 1.31||0.55|
| Chronic muscular pain||5.2||7.7||1.51||0.92 to 2.48||0.10||1.35||0.82 to 2.23||0.24|
| Irritable bowel syndrome||6.7||10.6||1.59||1.04 to 2.43||0.03||1.51||0.98 to 2.33||0.06|
| Asthma||8.7||10.6||1.21||0.78 to 1.88||0.40||1.10||0.70 to 1.71||0.68|
| Impairment due to physical health*||7.5||9.6||1.31||0.83 to 2.06||0.24||1.10||0.70 to 1.74||0.69|
Most previous studies of comorbidity in BED and related phenotypes have used samples recruited from clinical settings, although most individuals with eating disorders do not enter into treatment (17). In our sample, ∼20% of the women and 12% of men with BE had received psychiatric treatment.
A previous investigation in which only 15% of the participants had received clinical service indicated that psychiatric comorbidity in women with BED was lower than in clinical samples (only major depression and any lifetime axis I disorder were significantly associated with BED) (18). A more recent study comparing a clinical sample and a sample recruited from the community (<10% of whom had received mental health treatment), however, concluded that the use of treatment-seeking samples does not lead to an overestimation of psychiatric comorbidity (19). The participants from the community in both studies were recruited mainly by advertisement. Our results in women support thelatter study, indicating that symptoms of both anxiety disorders and depression are elevated in individuals with BE regardless of treatment seeking. Our finding of higher frequency of eating disorder symptoms in individuals with BE is in accordance with the results from previous studies in both clinical and community samples (18).
The comorbidity with medical symptoms in women with BE was lower than previously reported, for example by Johnson et al. (7), who reported on a population of women seeking treatment because of medical problems (primary care, gynecology). For reasons mentioned above (9), one can assume that the results from that study are biased because of the way the BED patients were selected. Furthermore, the participants in that study were older than in our study. Our results are, however, in accordance with those reported from a population-based study of obese women with binge eating, in which neither the individual medical conditions examined, nor impairment due to physical health, were significantly elevated (5). This suggests that binge eating in the absence of compensatory behaviors may be less associated with medical problems in population-based studies than in treatment-seeking populations, at least for young-adult women.
The rates of psychiatric and behavioral symptoms were increased in both men and women with BE, although the associations were somewhat stronger in men. The results are in accordance with a previous study that found similar patterns of psychiatric comorbidity in men and women with eating disorders (20). Men showed broader elevations in areas beyond depression and anxiety including smoking, alcohol use, use of sleep and pain medications, and impairment. Both men and women with BE reported higher psychiatric treatment use than those without BE.
The relative frequencies of medical symptoms were substantially higher in men with BE than in women with BE. It has previously been reported that recurrent binge eating during adolescence is associated with chronic or frequent pain during early adulthood in a representative community sample (51% women) (21). The literature suggests that women display greater pain sensitivity than men (22). The elevated rates of comorbidity between BE and musculoskeletal pain in men are, therefore, probably not explained by sex differences in pain sensitivity.
Exercise frequency was significantly lower in men with BE compared with men without BE, whereas no such difference was found for women. The reason for this remains unclear, although it may be related to higher levels of pain and physical impairment. In addition, the decreased exercise might explain why the proportion of men with BE who were overweight or obese (55.6%) was substantially higher than that of women with BE (37.9%).
With the exception of use of sleep medication in men, the associations with psychiatric symptoms in BE were independent of BMI. This indicates that the observed comorbidity was not caused by the higher rates of obesity and overweight in the BE group and is in accordance with findings from a previous study of women with BED where psychopathology was significantly related to binge eating but not to obesity (23). The increased risk for medical symptoms associated with BE was also independent of BMI.
Early age at menarche was associated with BE independently of BMI in women. A previous study with a small sample found no association between age at menarche and BED (24), but early age at menarche has recently been found to be a risk factor for development of body image and dieting concerns (25). Further studies are needed to clarify whether early age at menarche is a risk factor for binge eating independently of BMI.
Concurrent symptoms have been suggested to be an important external validator for psychiatric disorders together with factors such as family history, biological and psychological tests, demographic correlates, and course of illness (1). From this perspective, our finding that the core symptoms of BED are associated with increased symptoms of anxiety, depression, and disordered eating behavior independently of BMI supports the validity of BED as a diagnostic category. However, it is important to underscore that the validation process is complex and that the different validators must be considered together and be related to the defining features of the disorder.
To what degree our sample is representative of the general Norwegian population depends on whether attrition was associated with any of the variables studied and whether twins differ from singletons in their risk for psychiatric disorders. The second question has been addressed in several studies, and the available evidence suggests that rates of psychopathology do not differ substantially between twins and singletons (26). With a response rate of 63%, a selection bias may be present. In samples of twins, this can be tested, given that familial/genetic factors are of significant etiologic importance to the symptoms or disorders of interest (27). If levels of symptoms predict cooperation, then differences should be observed in individuals whose co-twin did and did not return the questionnaire. We found no significant differences with regard to SCL-5, binge eating, BE, or BMI. Therefore, it is not likely that response bias has substantially influenced our results.
It is theoretically possible that the wording of our item for binge eating could have resulted in the inclusion of individuals with both “objective” binge episodes and “subjective” episodes (i.e., eating an amount of food that made the person feel uncomfortable but was not “an unusually large amount”). The validity of our binge eating item is, however, strongly supported by the finding of significant associations with the other DSM criteria for bulimia nervosa and a substantial heritability (51%) (28).
The use of diet pills and diuretics was not assessed. In 1998, when our survey was conducted, no diet pills were legally available in Norway. The use of diuretics requires a prescription from a medical doctor and, therefore, is not likely to be utilized for compensatory purposes.
Our criteria for binge eating without compensatory behaviors based on questionnaire data may have led to broader inclusion than if all diagnostic criteria for BED as outlined in the Appendix B of the DSM-IV had been assessed. The specific items, thresholds, and duration of the proposed new categories listed in the appendix are, however, tentative, and researchers are encouraged to experiment with alternative phenotypic definitions. Our definition covers the core features of BED, but the absence of assessment of criteria B and C may have influenced the severity of our sample. In the DSM-IV field trials, only a minority of individuals with binge eating (overeating with loss of control) also fulfilled criteria B and C (overeating with loss of control plus behavioral indicators of loss of control and subjective distress) (29). The effect of lowering the inclusion threshold on comorbidity is uncertain. It has not been possible to demonstrate clear differences between full and subsyndromal BED based on frequency of binge eating (30), and individuals with full and subthreshold BED do not differ significantly with respect to psychiatric distress, weight and shape concern, and history of treatment seeking (31). The term “subthreshold” in these studies, however, refers to binge eating fewer than two times a week, and no comparisons are made between individuals with binge eating who fulfill criteria B and C and those who do not. We are not aware of any study in which this has been done; therefore, it is possible that binge eating with and without behavioral indicators of loss of control and subjective distress could be different with regard to comorbidity. However, our results with regard to psychiatric comorbidity in women are in accordance with findings from most of the previous studies of women with BED diagnosed according to the DSM criteria (19). This suggests that the inclusion of criteria B and C in the diagnostic definition does not seem to affect the co-occurrence with psychiatric problems.
Supporting the generalization of our results, the gender ratio in our sample (∼1.5 to 1) is similar to that previously reported in the literature for BED (32), indicating that gender ratio was not influenced by altering the threshold. The proportion of women seeking psychiatric treatment was higher in our sample than in two other community samples using full DSM diagnoses (18, 19). However, because these studies were done in the U.S., where the health care system differs in many respects from that in Norway, it is difficult to draw conclusions regarding the similarities of the samples by comparing treatment rates. If our results are biased, one would expect the estimates of comorbidity to be lower in a broadly diagnosed sample than for more severely ill individuals.
Collecting information on a broad range of variables in samples as large as ours, it is necessary to rely on self-report questionnaire data with unknown validity. Preliminary results from a subsample of the twins in this study currently undergoing structured psychiatric interviews (n = 2490) support the validity of the questionnaire items for major depressive episode and panic disorder (tetrachoric correlations 0.46 and 0.66, respectively). Our data suggest that individuals who exhibited the low back pain and neck-shoulder pain items suffer from relatively severe, clinically significant pain (33). It is, however, possible that the use of self-report data may lead to an overestimation of some symptoms and, therefore, increased comorbidity. Some physical health problems (e.g., type 2 diabetes) usually emerge later in life. The relatively low age range of our sample (18 to 31 years) may have lead to underestimation of the co-occurrence between BE and medical symptoms. It may also have contributed to the relatively low weight distribution in our sample compared with most clinical samples because binge eaters tend to gain weight as they age. The main goal of this study was to examine psychiatric and medical comorbidity independently of BMI. As noted above, population-based samples have substantial advantages for that purpose. However, the unadjusted comorbidity rates, especially with regard to medical problems, are probably lower than what would be encountered in clinical settings because BMI is usually higher in such samples (29), and several of the medical problems examined would be expected to emerge or worsen with increasing weight. Therefore, clinical samples would have advantages if specific questions regarding the interaction between binge eating and obesity, e.g., whether binge eating amplifies obesity-related medical problems, were to be addressed.
When a large number of statistical comparisons are conducted, the probability of false positive findings (Type I errors) increases. Given the exploratory nature of this study, we found it inappropriate to use stringent corrections for multiple tests (e.g., Bonferroni); hence, our decision to limit our discussion to results that exhibited p values ≤ 0.01. If such corrections had been applied, the significant relationships with the highest p values would have been rejected. Other studies with which we compare our results (5, 7, 19, 20) have not applied corrections and have reported results at the p = 0.05 level. The sample size in this study gives us adequate statistical power to minimize false negative results (Type II errors).
The results from this population-based sample of young adults indicate that there is substantial comorbidity between BE and symptoms of psychiatric disorders independently of BMI. Medical symptoms co-occur less frequently with BE than previously reported from treatment-seeking populations in women. This can, to some extent, be due to the relatively low age in our sample. Across all domains, the array of symptoms exhibited by men with BE was broader than that observed in women with BE. This observation suggests the importance of considering gender differences in future studies of psychiatric and medical morbidity, binge eating, and obesity.
The Norwegian Institute of Public Health Program of Twin Research is supported by grants from The Norwegian Research Council, The Norwegian Foundation for Health and Rehabilitation, The Norwegian Council for Mental Health, The European Commission under the program “Quality of Life and Management of the Living Resources” of Fifth Framework Programme (QLG2-CT-2002-01254), and Sommer's Foundation.