Data were collected from 445 children and adolescents (M = 13.1 years, SD = 2.6); 195 were participants from Bethesda, Maryland; 114 were from Ghent, Belgium; 81 were from Boston, Massachusetts; 47 were from Pittsburgh, Pennsylvania; and 8 were from St. Louis, Missouri. Seventy percent (n = 312) of the sample was overweight (BMI z score ≥ 1.64). Of the overweight children and adolescents, 189 (60.6%) were initiating weight loss treatment. Participant demographics by site are described in .
participant Demographics by Site
Twenty-four percent of participants reported OBEs, 15% endorsed SBEs, 24% described OOs, and 37% reported NEs. Groups did not differ with regard to sex, χ2(3, N = 445) = 2.0, p = .56; race, χ2(3, N = 445) = 3.6, p = .31; or socioeconomic status, F = 0.6, p = .65. Youths with OOs (M = 14.0 years, SD = 2.3) were significantly older than those reporting SBEs (M = 12.3 years, SD = 2.8) and NEs (M = 13.0 years, SD = 2.5), F(3, 444) = 6.3, p < .01. After sex and treatment status were controlled, OBE youths (BMI z score, M = 2.0, SD = 0.7) were heavier than participants with OOs (M = 1.6, SD = 1.0) or NEs (M = 1.6, SD = 1.1), and SBE youths (M = 2.0, SD = 0.7) were heavier than participants with NEs, F(3, 427) = 4.8, p < .01; η2 = .03. Those with OBEs were more likely to be involved in a weight loss treatment study than those in other groups, OBE = 58.5%, SBE = 34.3%, OO = 48.1%, NE = 37.3%; χ2(3, N = 445) = 15.2, p < .01.
On the basis of data from the EDE, the prevalence and frequency of compensatory behaviors were low. Twenty-four participants (5.4%) reported engaging in a compensatory behavior at least one time over the past month, 10 of whom also reported OBEs. The most commonly reported behavior was excessive exercise (n = 19), with an average, of 4.1 (SD = 3.4) episodes in the past month. Four participants reported self-induced vomiting in the past month, with a mean of 3.1 (SD = 2.4) episodes. Only 1 participant met criteria for bulimia nervosa (through use of appetite suppressants). All results remained the same when this participant was removed from the data set.
Of the 445 children and adolescents who completed the SPEEI, 95% of cases had complete data. Among the 28 incomplete cases, the most common missing data were where the episode took place, the amount of food the participant consumed compared to others, and with whom the participant was eating. Compared to cases with complete data, those with missing data had significantly lower scores on the EDE Eating Concern subscale (M
= 0.46, SD
= 0.68, vs. M
= 0.25, SD
= 0.34, respectively, p
= .01) and were more likely to be participants from NICHD (n = 15) than Ghent (no missing data; p
= .03). Only participants with complete data were included in the cluster analyses since all variables from the SPEEI are nominal, thus precluding the use of normal-theory-based likelihood methods for imputation of missing data. Furthermore, when cluster analyses are used, it is assumed that the population consists of heterogeneous subgroups that have not been identified, so the appropriate subgroup for calculating summary statistics is unknown. Although one method for using all available data has been described for cluster analyses (Everitt, Landau, & Leese, 2001
), this approach is not clearly superior to excluding data and was less robust when applied to the present data set.2
However, as a conservative measure, we included the participants with missing data in the nonclustered sample for the analyses of variance.
Convergent Validity of Binge and Loss of Control Eating
Significant differences among responses were revealed on the basis of episode type (see ). Compared to youths endorsing OO and NE, greater percentages of participants with OBEs and SBEs reported that the episode began after consumption of a “forbidden” food, χ2(3, N = 444) = 27.4, p < .01; that they experienced a negative emotion before eating, χ2 (6, N = 445) = 38.6, p < .01; that they ate despite a lack of hunger, χ2(3, N = 445) = 70.0, p < .01; and that the episode was a snack as opposed to a meal, χ2(12, N = 442) = 403, p < .01. These individuals also were more likely to report eating alone, χ2(6, N = 442) = 20.8, p < .01; and experiencing secrecy regarding eating, χ2 (3, N = 444) = 38.6, p < .01. Furthermore, participants with OBEs and SBEs were more likely to endorse experiencing a sense of “numbing out” while eating, χ2(3, N = 445) = 47.7, p < .01. After eating, greater percentages of youths with OBEs and SBEs reported experiencing a negative emotion, χ2(6, N = 445) = 53.5, p < .01; including feelings of shame and guilt, χ2(3, N = 445) = 46.2, p < .01. Compared to all other groups, youths describing OBEs were more likely to report hiding the food being eaten, χ2(3, N = 442) = 18.4,p < .01; eating quickly, χ2(3, N = 444) = 43.1, p < .01; eating more food than others, χ2(12, N = 441) = 57.8, p < .01; and feeling physically sick after the meal, χ2(3, N = 445) = 11.6, p < .01. SBE participants were more likely to eat while watching television compared to other groups, χ2(12, N = 439) = 38.2, p < .01.
Hierarchical cluster analysis revealed a small but cohesive subgroup of participants (3% of the sample; n
= 15), 87% of whom had reported engaging in an OBE or SBE (disordered eating cluster; see ). Five participants from the cluster were from NICHD, 5 were from Ghent, 4 were from Boston, and 1 was from Washington University. Most (87%) were adolescents (ages 12.7–17.7 years). Behaviors that were reported by this group included having an event (e.g., an argument) trigger the episode, eating alone, and eating while watching television. Additionally, the disordered eating cluster described the episode as a snack and reported eating despite a lack of hunger, experiencing secrecy and a sense of numbing out while eating, and experiencing negative emotions before and after eating, including feelings of shame and guilt. The disordered eating cluster also reported being unsure of how much they were eating. Seventy-five percent of these youths were participating in a weight loss treatment study, 80% were overweight, 80% were female, and 93% were Caucasian. According to their EDE interviews, 47% (95% confidence interval = 21%–73%) of these participants met full or subthreshold frequency criteria (on average, at least four OBEs or eight SBEs per month) for Diagnostic and Statistical Manual of Mental Disorders
(4th ed., text rev.; DSM-IV-TR
; APA, 2000
) BED, compared to 6% (95% confidence interval = 4%–9%) for the rest of the sample (p
< .001, Fisher’s exact test). After the contributions of sex, BMI z
score, and treatment status were accounted for, compared to the rest of the sample, the disordered eating cluster had greater EDE Eating Concern, F
(l, 444) = 8.2,p
< .01, r
= .34; Shape Concern, F
(2,444) = 3.9,p
= .05, r
= .31; and global, F
(l, 445) = 5.8, p
< .05, r
= .31, scores than the rest of the sample (see ). However, no differences were found with regard to EDE Dietary Restraint, F
(1, 427) = 1.1, p
= .3, r
= .20; or Weight Concern, F
(1, 426) = 1.8, p
= .2, r
= .23, subscales (see ). BMI z
scores did not differ between the disordered eating cluster group (BMI z
= 1.8, SD
= 0.60) and the rest of the participants (BMI z
= 1.8, SD
= 0.96), F
(1, 427) = 0.0, p
= 1.0, r
Common Factors Among Disordered Eating Cluster (n =15)
Eating Disorder Examination subscales and global score by disordered eating cluster (n = 15) and nondisordered eating (n = 430) cluster. Error bars represent standard errors of the mean, *p ≤ .05. **p < .01.
Post Hoc Analyses by Age, Sex, and Treatment Status
The data were analyzed separately for younger (12 years or younger; n = 156) and older (older than 12 years; n = 289) participants. On the basis of chi-square analyses, compared to younger children reporting OOs and NEs, younger children with OBEs and SBEs were more likely to report the following behaviors with regard to their episodes: eating despite a lack of hunger, χ2(3, N = 156) = 29.5, p < .01; eating quickly, χ2(3, N = 152) = 9.7, p < .05; and consuming more than others, χ2(12, N = 154) = 34.0, p < .01. Specific to the younger children was the likelihood of reporting feeling sick after eating, χ2(3, N = 156) = 11.6, p < .01. Similar to the older participants, younger children endorsed all of the emotion-related probes other than having an event trigger the episode. Hierarchical cluster analysis including only the younger children revealed a somewhat different set of characteristics from results generated by the entire sample. These 15 children, 12 of whom endorsed OBEs or SBEs, tended to report that the episode took place at a home other than their own and in the afternoon and that they were eating more than others. Similar to the cluster for the entire group, younger children also reported experiencing a negative emotion and a trigger occurring prior to the episode, eating in secret, and feeling numb.
An analysis of the data was also conducted separately for boys (n = 177) and girls (n = 268). Chi-square frequencies did not differ considerably by sex compared to findings of the entire sample, although boys with OBEs (78.4%) and SBEs (70.0%) were just as likely as those with OOs (68.1%) or NEs (60.3%) to be overweight, χ2(3, N = 177) = 2.2, p = .54. Among the boys, there were no reported differences by episode type with regard to experiencing a trigger prior to or feeling sick after the episode (ps > .10). Cluster analyses revealed a distinct subgroup of boys (n = 10), of whom 4 reported an OBE or SBE. These boys were likely to endorse being alone and watching television while eating. They were also likely to report that the episode was a snack and being unsure of how much they were eating. Other than experiencing a negative emotion prior to the episode, unlike the rest of the sample, this subgroup of boys did not endorse any other emotional experiences during their episode.
In a third set of analyses, we examined youths participating in weight loss studies (n = 198) and nonintervention participants (n = 247) separately. Chi-square analyses based on treatment status differed little from those for the entire sample. However, for participants in weight loss studies, there were no significant differences based on episode type regarding where the episode took place or whether they felt sick after the episode. Weight loss study participants with OBEs were more likely to feel full after eating, χ2(3, N = 198) = 10.6, p = .01. For nonintervention children and adolescents, there were no significant differences between episode type and feeling secretive about eating, hiding their food, or the amount they were eating compared to others. By contrast, non-treatment participants with OBEs were more likely to report feeling physically hungry before their episode, χ2(3, N = 247) = 7.8, p = .05. Cluster analyses for the weight loss treatment participants and nonintervention children and adolescents each identified a subgroup of 5 participants, all of whom were included in the cluster generated from the entire sample. As expected, these youths were very similar to the cluster from the entire sample. However, the weight loss treatment cluster was also likely to report hiding their food, eating more than others, and feeling sick afterward. The nontreatment cluster was additionally likely to report that their episode was in the afternoon and associated with feeling tired. Both groups differed from the entire sample in that they were likely to report eating quickly.