This study sought to determine if the components of our cognitive model are related to the severity of GAD. The first hypothesis, that all model components would be related to GAD severity, was supported. Using zero-order correlations, all components of the model were related to diagnostic severity. It should be noted, however, that the importance of both positive beliefs about worry and cognitive avoidance received only modest support, as the zero-order correlations were the only analyses to show that these model components were related to the severity of GAD. It may be that positive beliefs about worry and cognitive avoidance showed modest relationships with the severity of GAD for different reasons. There is now evidence that positive beliefs about worry play a more important role in the early stages of the development of GAD than in the later stages (Holowka et al., 2000
). This is consistent with the metacognitive theory of GAD (Wells, 2004
), which posits that positive beliefs about worry contribute to the initial use of worry as a coping strategy but that negative beliefs about worry play a more important role in full-blown GAD. Anecdotally, our clinical experience suggests that patients with GAD present with varying degrees of positive beliefs about worry. For some, positive beliefs appear to play a key role in maintaining high levels of worry and anxiety, whereas others do not appear to hold these beliefs. Thus, it may be that positive beliefs about worry are only modestly associated with the severity of GAD because their role in full-blown GAD is quite variable (as opposed to their role in the early stages of the development of the disorder).
As for cognitive avoidance, its modest association with the severity of GAD may be the result of a measurement issue. As underscored by Borkovec’s avoidance theory of GAD (Borkovec et al., 2004
; Borkovec & Newman, 1999
), some forms of cognitive avoidance in GAD are not necessarily accessible to awareness. In fact, the avoidance of concrete mental images in patients with GAD is thought to be mainly an automatic or implicit cognitive process. Although the avoidance theory of GAD has received much empirical support (e.g., Borkovec & Hu, 1990
; Borkovec et al., 1993
), implicit cognitive avoidance remains an elusive process in terms of assessment. Thus, although the CAQ has shown strong psychometric properties in previous nonclinical studies (Gosselin et al., 2002
), like any self-report measure, it is not ideally suited for the assessment of processes that may not be entirely accessible to awareness. However, given that some cognitive avoidance strategies assessed by the CAQ are clearly under volitional control (e.g., thought suppression, distraction), it is our position that, at the very least, it is an appropriate measure for some forms of cognitive avoidance in GAD.
The second hypothesis, that the model components would be more closely related to worry severity than to somatic symptom severity, was not supported in the uncontrolled analyses, but received some support when potential confounds were statistically controlled. In the zero-order correlations, worry severity was only related to two model components (intolerance of uncertainty and negative problem orientation), whereas somatic symptom severity was related to three components (intolerance of uncertainty, negative problem orientation, and cognitive avoidance). This is surprising given that the components of the model are thought to contribute directly to worry, which is then theorized to contribute to GAD somatic symptoms (Dugas & Ladouceur, 1998
; Dugas & Robichaud, 2007
). In the present study, however, the results of the uncontrolled analyses suggest that the model is quite sensitive to GAD somatic symptoms, as the severity of these symptoms was related to all components but positive beliefs about worry.
When age, gender, and depressive symptoms were controlled, however, a different picture emerged. Specifically, the only significant relationships that remained were those between worry severity and both intolerance of uncertainty and negative problem orientation. Furthermore, the correlation between intolerance of uncertainty and worry severity was significantly higher than the correlation between intolerance of uncertainty and somatic symptom severity. Thus, the results of the partial correlations suggest that the model components are more specifically related to worry severity than to somatic symptom severity. This, however, may not actually be the case. Of the six GAD somatic symptoms, three are also found in the diagnostic criteria for major depressive disorder (MDD), namely, concentration problems, insomnia, and fatigue. Furthermore, a fourth GAD somatic symptom, restlessness, resembles the MDD symptom of psychomotor agitation. Because we controlled for scores on the BDI-II, which is specifically designed to assess the DSM-IV symptoms of MDD, the partial correlations may well have constituted an overly conservative test of the relationship between the model components and the severity of GAD somatic symptoms. In effect, the partial correlations controlling for depressive symptoms eliminated the shared variance of at least half of the GAD somatic symptoms with the model components. Thus, the lack of specificity observed in the relationship between the model components and the severity of GAD somatic symptoms may well have resulted from the fact that at least half of these symptoms are common to both GAD and MDD.
The final hypothesis, which stated that intolerance of uncertainty would be the best predictor of the severity of GAD, received moderate support. In the zero-order correlations, intolerance of uncertainty was related to all three indicators of GAD severity. When participants were divided into groups based on scores on the indicators of GAD severity, intolerance of uncertainty distinguished participants in the Moderate and Severe GAD groups from those in the Mild GAD group. And finally, when age, gender, and depressive symptoms were statistically controlled, intolerance of uncertainty continued to be related to the severity of worry and continued to distinguish the Moderate and Severe GAD groups from the Mild GAD group.
Overall, the results suggest that intolerance of uncertainty is a better predictor of the severity of GAD than either positive beliefs about worry or cognitive avoidance. However, given that intolerance of uncertainty and negative problem orientation produced a similar pattern of results, it could be argued that both of these model components are equally strong predictors of GAD severity. To further explore the relative contributions of intolerance of uncertainty and negative problem orientation to the severity of GAD, we conducted additional linear regressions in which the model components were entered simultaneously and the predicted variables were diagnostic severity, worry severity, and somatic symptom severity. In all three regressions, the model components significantly predicted the severity of GAD. However, in terms of individual contributions of the model components to the prediction of GAD severity, the only significant result was that intolerance of uncertainty made a unique contribution to worry scores, above and beyond the other model components.3
Thus, although negative problem orientation appears to be a strong predictor of the severity of GAD, when the contribution of intolerance of uncertainty is taken into account, negative problem orientation does not make a unique contribution to the prediction of GAD severity.
The present study had a number of limitations. First, because only participants receiving a diagnosis of GAD on the MINI were referred for the ADIS-IV, the assessors who administered the ADIS-IV were not completely blind to the results of the MINI (although they had no knowledge of the severity of GAD or the presence of comorbid conditions). Thus, the reported rates of agreement between the structured diagnostic assessments are undoubtedly overestimated. Second, the CAQ is a new measure that has yet to be validated in clinical samples of anxious patients. Although the CAQ has strong psychometric properties in nonclinical samples (Gosselin et al., 2002
), this is the first study to use the questionnaire in a clinical sample. As discussed previously, a related limitation is that the Transformation of Images into Thoughts subscale of the questionnaire taps a process that is not fully conscious (this may also be true for other subscales, but to a lesser extent). As a result, self-report only offers a partial assessment of cognitive avoidance as it relates to GAD. Finally, the study would have benefited from the inclusion of a measure of negative beliefs about worry such as the Consequences of Worrying Scale (Davey et al., 1996b
) or the Meta-Cognitions Questionnaire (Cartwright-Hatton & Wells, 1997
), both of which contain negative belief subscales. Although negative beliefs about worry are implicitly included in our model within the cognitive avoidance component (negative beliefs lead to various attempts to avoid worrisome thoughts), it could certainly be argued that a more direct measure of these beliefs would have been useful.
To our knowledge, this is the first study to test a model of GAD within a sample of patients with primary GAD. As mentioned earlier, by probing the relationship between the model components and various indicators of GAD severity, this study constitutes a relatively stringent test of the model’s validity. The findings show that intolerance of uncertainty and negative problem orientation were robustly related to the severity of GAD. For the most part, the relationship between the severity of GAD and these two model components was independent of sociodemographic and clinical factors, including the severity of depressive symptoms.