First, a power analysis was conducted to determine whether the sample size of 114 was sufficient to detect differences on the TSI.44
If one hypothesizes that rape victims with a child sexual abuse history will not respond to treatment as well as rape victims without a child sexual abuse history then setting the effect size at .50 is a reasonable estimate. With the alpha set at .01 and power set at .95, only 76 participants would be needed to find differences that exist. Given the overall findings of the parent study, one would also predict large effect sizes from pre- to posttreatment; therefore, the samples are also sufficient to detect pre- to posttreatment differences.
The child sexual abuse and non-child sexual abuse samples were compared with regard to their demographic and other trauma history variables aside from child sexual abuse (). The child sexual abuse sample reported significantly more robberies, kidnappings, rapes, attempted rapes, other sexual assaults, physical assaults with minor injuries, attempted murder, child physical abuse, and number of crime victimizations. There were no significant differences in the number of assaults with serious injuries or the criminal/vehicular death of a loved one between the two samples. There were no differences between the child sexual abuse and non-child sexual abuse samples regarding which treatment they received or whether they completed treatment.
Comparison of non-child sexual abuse versus child sexual abuse survivors on demographic and crime history variables
Because, for these analyses, the delayed-treatment samples were folded into the samples that were treated immediately, preliminary analyses were conducted comparing these collapsed samples first on the CAPS and BDI, then the TSI subscales. The three TSI factors were subjected to two 2 × 3 multivariate analysis of variances (MANOVAs), representing treatment type (CPT or prolonged exposure), and assessments (pretreatment, posttreatment, and 9-month posttreatment). In the first analysis, CAPS total score and BDI scores were the dependent variables. The analysis resulted in a significant assessment effect, F(4,472)=61.1, Pillai’s trace=.68, P=.000, but no treatment-type effect or interaction. Both CAPS and BDI changed significantly from pre to posttreatment and there were no changes from the posttreatment to the 9-month assessment on either measure. In the second analysis, the three TSI factors were the dependent variables. This analysis also resulted in an assessment effect, F(6,446)=35.0, Pillai’s trace = .64, P=.000, but no treatment type effect or interaction. Significant assessment effects were found for all three TSI factors with significant decreases in scores occurring from pre to posttreatment, and no change from the posttreatment to follow-up assessments. Because there were no differences between the two types of therapy on any measure at any time point, the data were collapsed for the comparison of rape victims with or without a child sexual abuse history.
The means and standard deviations of the CAPS, BDI, and TSI subscale raw scores for the child sexual abuse and non-child sexual abuse samples are located in . The two groups were then compared by means of repeated measures MANOVA on the CAPS total score and BDI. This analysis produced a significant assessment period effect, F(4,472)=60.3, Pillai’s trace=.68, P=.000, but no group main effect or interaction. The participants improved on both the CAPS and BDI from pre- to posttreatment and then maintained their improvement through the 9-month follow-up.
Means, standard deviations, and pre-post hedges’ g effect sizes of non child sexual abuse versus child sexual abuse survivors on clinical scales.
An initial MANOVA comparing the two groups at pretreatment on the three TSI factors was significant overall, F(3,111)=3.7, Pillai’s trace=.09, P=.02, indicated that the child sexual abuse participants reported significantly higher scores at pretreatment on two of the three factors, the Self factor, F(1,113)=7.3, P=.008 and the Trauma factor, F(1,113)=8.8, P=.004. There was only a trend for the two groups to be different on the Dysphoria factor. A repeated measures MANOVA of the three TSI factors resulted in significant child sexual abuse status main effect, F(3,110)=2.9, Pillai’s trace=.07, P=.039, an assessment main effect, F(6,446)=34.8, Pillais trace=.64, P=.000, but no assessment by child sexual abuse status interaction. Post hoc univariate analyses indicated that, while the two groups improved significantly on all three factors from pre- to posttreatment and maintained their improvement, there were also significant differences between the two groups on 2 of the 3 factors. The two groups differed overall on the Trauma, F(1,112)=5.8, P=.02, and Self factors, F(1,112)=6.9, P=.01, but not on the Dysphoria factor. The findings are depicted in .
Pretreatment, posttreatment, and 9-month follow-up comparison of rape victims with and without a history of CSA on three factors of the Trauma Symptom Inventory: Trauma, Self, and Dysphoria.
In order to explore which of the subscales contributed to the group differences, the MANOVAs were repeated for the constituent subscales of the Trauma and Self factors. The Trauma factor includes intrusive experiences, defensive avoidance, dissociation, and impaired self-reference. These subscales were subjected to a repeated measures MANOVA, which resulted in a significant assessment period effect, F(8,444)=24.9, Pillai’s trace=.62, P=.000, but no interaction and only a trend for a group main effect (P=.07). Post hoc pairwise comparisons indicated that the sample improved from pre- to posttreatment on all four subscales and these improvements were maintained through the 9-month follow-up.
On the Self scale, consisting of sexual concerns, dysfunctional sexual behavior, and tension reduction behaviors, the MANOVA resulted in a significant group main effect, F(3,111)=2.7, Pillai’s trace=.07, P=.05, a significant assessment period effect, F(6,450)=24.8, Pillai’s trace=.50, P=.000, but no group by assessment interaction. Post hoc analyses on the assessment periods indicated that all three subscale scores improved significantly from pre- to posttreatment and were then maintained over the 9-month follow-up period. Post hoc analyses of the group effect indicated significant overall group differences on all three subscales. Pre- to post-effect sizes calculated with Hedge’s G on all TSI sub-scales as well as the CAPS and BDI are located in . There were large effect sizes for all subscales with one exception. In the child sexual abuse group, there was a medium effect size on the impaired self-reference scale. That subscale had a very large standard deviation compared with the other subscales, indicating greater variability of responses to treatment on that scale.
In order to determine how much of the overall group effect was determined by pretreatment differences, a MANOVA was conducted on the three Self subscales at posttreatment using the pretreatment scores as covariates. The MANOVA was nonsignificant, indicating that there were no differences between the child sexual abuse and non-child sexual abuse groups at posttreatment, once pretreatment scores had been accounted for. The same thing occurred with a MANOVA on the Trauma factor subscales. Once the pretreatment differences were accounted for there were no differences between the child sexual abuse and non-child sexual abuse groups at posttreatment. depicts the subscale scores plotted on a TSI profile. Both groups are within normal limits at the posttreatment and follow-up periods.
Pretreatment, posttreatment, and 9-month follow-up comparison of rape victims with and without a history of CSA on the 10 subscales of the Trauma Symptom Inventory.