Seven randomised controlled trials (RCTs) met the inclusion criteria.
Within-group comparisons: 4 of the studies reported parameters for calculating the effect sizes of CBT for binging, while 6 studies reported data on purging frequency.
Between-group comparisons: the parameters necessary for calculating the effect sizes for binge eating and purging were reported in 4 and 5 studies, respectively.
Within-group comparisons.
The weighted effect size (Zr) for binge eating (4 studies, n=219) was 0.61; this corresponded to an effect size (r) of 0.55 or alternatively, a Cohen's d of 1.32. The Zr for purging (6 studies, n=317) was 0.70; this corresponded to a r-value 0.61 or alternatively, a Cohen's d of 1.5. The combined probability for the effect of CBT on purging frequency (3 studies) was calculated (p=0.00000001).
Between-group comparisons.
The Zr of CBT for binge eating (3 studies, n=197) was 0.23; this corresponded to a r-value of 0.23 or alternatively, a Cohen's d of 0.47. The Zr of CBT for purging (5 studies, n=334) was Zr 0.29; this corresponded to a r-value of 0.28, or alternatively, a Cohen's d of 0.58. The combined probability for the effect of CBT on purging frequency (3 studies) was calculated (p=0.00001).
The analysis of homogeneity showed that in the within-group comparisons, there was significant heterogeneity in the effect sizes concerning binge eating frequency (chi-squared 11.68, d.f.=3, p<0.01), but not in those concerning purging frequency. The test of homogeneity for the between-group comparisons showed no significant results in the effect sizes for either binging or purging frequency.
The fail-safe N values for the combined likelihood of the effects of CBT on purging in within- and between-group comparisons were 24.5 and 19.4, respectively.