Abstract
The efficacy of cognitive-behavioral therapy (CBT) for anxiety is well established. Investigations into the mechanisms of change in CBT report changes in cognitive vulnerabilities mediating improvements over the course of treatment. As anxiety disorders share certain risk factors, there is a trend toward CBT emphasizing these vulnerabilities, including negative affectivity (NA) and also more specific constructs such as anxiety sensitivity (AS) and intolerance of uncertainty (IU). The purpose of this investigation was to analyze potential mediators of anxiety reduction over the course of transdiagnostic group CBT. NA, AS, and IU all decreased over the course of treatment. Among the potential mediators, change in NA had a significant relationship with change in anxiety but change in AS and change in IU did not. Neither the main effect of primary diagnosis nor the interactions between potential mediators and primary diagnoses were significant, indicating that there were no differential changes in anxiety or the potential mediators across primary diagnoses. Results strongly point toward NA as an overarching mediator of anxiety reduction during transdiagnostic group CBT.
The criterion variables, session-by-session STAI scores, were modeled over time using a mixed-effect linear regression using a Maximum Likelihood estimator. The data for the STAI, as well as the potential mediator variables, are shown in Fig. 1. Results suggested good fit to the data, RMSEA = 0.05, SRMR = 0.08, CFI = 0.95, with an average session 1 intercept of 47.76 (se = 0.59) and a significantly decreasing slope parameter (−1.01, se = 0.06, p < .001). Similarly, the potential mediator variables were modeled at pre-treatment, mid-treatment, and post-treatment and showed adequate fit to the data, RMSEA = 0.09, SRMR = 0.08, CFI = 0.95. Each of the potential mediator variables showed expected estimated intercepts (PANAS-NA: 29.29, se = 0.49; ASI: 31.68, se = 0.76; IUS: 79.73, se = 1.56) and significantly decreasing slope parameters (PANAS-NA: −4.47, se = 0.32, p < .001; ASI: −6.56, se = 0.53, p < .001; IUS: −8.60, se = 1.07, p < .001). Each of the slope parameters of the potential mediator variables showed positive simple correlations with the STAI slope parameter, rs = 0.50–0.87, ps < .001, indicating that decreases in each of the potential mediator variables were independently associated with decreases in anxiety severity.
Full-size image (45 K)
Fig. 1.
Changes in criterion and proposed mediators scores across treatment. Note: Lines represent mixed-effect linear regression modeled trends, while markers (■○▴♦) represent actual observed values. Specific values for estimated and raw scores on each variable are presented in the table.
Figure options
Each of the slope parameters for the potential mediator variables were then regressed together onto the slope parameter of the STAI to identify the extent to which change on the each of the potential mediator variables uniquely predicts change in anxiety severity. Additionally, primary diagnosis was included in the model to identify possible differential change in anxiety severity by diagnosis, as were interaction terms of each potential mediator by diagnosis (i.e., PANAS-NA × Dx; ASI × Dx; IUS × Dx) to examine for differential relationships between change in potential mediators and change in anxiety severity by diagnosis.
Results indicated a significant overall model, R2 = 0.84, F(11,256) = 123.67, p < .001. Inspection of the main effects of the potential mediator variables, revealed that change on the PANAS-NA variable was positively, uniquely, and strongly related to change in STAI scores, r2 = 0. 74, B = 0.27 (Bootstrapped 95% CI: 0.22–0.35), p = .001, after accounting for variability associated with the other variables in the model. Neither change on ASI nor IUS showed significant unique relationships with change in STAI, ASI: r2 = 0.05, B = −0.02 (Bootstrapped 95% CI: −0.05 to 0.03), p = .27; IUS: r2 = 0.05, B = −0.02 (Bootstrapped 95% CI: −0.05 to 0.003), p = .20, after controlling for the other variables in the model. Similarly, the main effect of primary diagnosis, r2 < 0.001, p = .98, was not significant, indicating no differential change in anxiety severity among participants with primary diagnoses of panic disorder, social phobia, or GAD. Finally, none of the potential mediator by diagnosis interaction terms was significant, r2s < 0.01, p > .36, suggesting no differential association between potential mediators and change in anxiety severity among participants with primary diagnoses of panic disorder, social phobia, or GAD.
