Of the 324 participants with BED, 60% (n = 195) had clinical overvaluation and 40% (n = 129) with subclinical overvaluation. As expected, per the DSM-IV criteria determined using the SCID-I/P, the BED, BN, and sub-threshold BN groups differed significantly in the proportion of participants meeting overvaluation thresholds on the EDE (Dχ2 (2, N=436)) = 37.56, p < .001).Clinical overvaluation was nearly universal in the BN patients: 95%(58 of 61) participants with BN and 86% (44 of 51) participants with sub-threshold BN had this feature.
Chi-square tests of independence performed on the four study groups on ethnicity (White versus non-White) reveal Eed significant differences (χ2 (df = 3, N = 324) = 16.79, p < .001). While the majority of participants within all groups were White (84.9%), lower percentages of BED participants were White (76.7% of the clinical and 84.1% of the subclinical overvaluation groups) in comparison to the BN (96.7%) and sub-threshold BN (94.1%) groups. A general linear model (GLM) analysis of variance (ANOVA) revealed a significant difference among the four groups on age (F(3, 432) = 74.21, p < .001, η2 = 0.34); participants within the BED clinical overvaluation (M = 44.0, SD = 9.2) and subclinical overvaluation (M = 44.8, SD = 9.7) groups did not differ from each other but were significantly older than BN (M = 28.1, SD = 10.3) and EDNOS (M = 29.1, SD = 10.6) groups, which did not differ from each other.
summarizes descriptive statistics and findings from ANOVAs comparing the four groups on BMI and the study measures. also shows partial η2
, an effect size (ES) measure representing the proportion of variation in the criterion measure accounted for by group membership. Cutoff conventions for this ES measure are as follows: small (.01 - .09), medium (.10 - .24), and large (> .25). When ANOVAs revealed significant overall group differences, Scheffe post-hoc tests were performed to determine which specific groups differed2
Comparison of BED Clinical Overvaluation (n=195), BED Subclinical Overvaluation (n=129), BN (n=61), and BN Subthreshold (n=51) Groups on Study Measures
The groups differed significantly, as expected, in BMI: the two BED groups, which did not differ significantly in BMI, had significantly higher BMI than the BN and sub-threshold BN groups, which did not differ significantly from each other. There were no significant within group correlations between BMI and overvaluation: 0.05 (for BED clinical overvaluation group), -0.06 (for BED subclinical overvaluation group), 0.02 (for BN group), and 0.11 (for EDNOS group). Restricting these correlational analyses to just white participants within each of the four study groups resulted in similar pattern of near-zero correlations between BMI and overvaluation.
The groups differed significantly in the frequency of OBEs. Scheffe post-hoc tests revealed that the sub-threshold BN group (per definition) had significantly fewer OBEs than the three other groups; in addition, the two BED groups did not differ significantly from one another but both had significantly fewer OBEs than the BN group. The groups differed significantly in levels of overvaluation and well as on the four EDE subscales; Scheffe post-hoc tests revealed a number of significant specific differences. In overvaluation levels, the BED subclinical overvaluation group differed significantly (i.e., had lower levels) from the three other groups (BED clinical overvaluation group, the BN group, and the subthreshold BN group) which did not differ significantly from each other. On the re straint scale, the two BED groups did not differ from each other but had significant lower scores than the BN and sub-threshold BN groups which did not differ from each other. For three EDE scales (eating concern, shape concern, and weight concern scales), the BED clinical overvaluation group had significantly higher scores than the three other groups which did not differ significantly from each other.
Since the BED groups had a significantly greater proportion of minority participants than the other groups, we repeated the ANOVAs using data from White participants only. shows the partial η2 values for these ANOVAs; these effect sizes are similar to those for the overall study group. Since the groups differed significantly in age and in BMI, we performed two sets of ANCOVAs (one set controlling for age and one controlling for BMI). As summarized in , controlling for age and BMI, respectively, altered the overall findings little.