Effects and Mechanisms of Change of Internet Delivered Cognitive Behavior Therapy for Generalized Anxiety Disorder

A randomized trial will be conducted within the Internet Psychiatry Clinic in Stockholm. An option will be added to the online sign-up form giving potential patients the option to participate in a treatment for GAD as part of this study. Based on information provided in the sign-up-form and an assessment via video-call with a psychologist, patients will be included/excluded.

Included patients will be randomized (1:1) to two different ICBT treatment programs (IU-ICBT and Meta-ICBT). These programs will be implemented into regular care at the Internet Psychiatry Clinic in Stockholm, meaning that each patient will undergo the regular assessment process at the clinic. Both programs are ICBT programs of similar scope, consisting of 8 modules to be completed during 10 weeks. During this time patients read written material which instructs them to do specific exercises, answer specific questions and can message a psychologist at the clinic via the web-platform. Based on similar treatments on the Internet Psychiatry Clinic we expect around 10% drop-out.

The effectiveness of both forms of ICBT (meaning within-group change in symptoms during the ten-week treatment) will be compared to other ICBT studies on GAD. Our criteria for studies to compare are:

• ICBT as guided self-help, i.e. treatment where the patient is meant to read or otherwise take in pre-written material and is communicating with a psychologist. Not included here are studies on video-call-therapy or chat-bot based treatments.
• Patients are diagnosed with GAD by a professional, either MD or psychologist.
• Either GAD-7 or PSWQ exist as an outcome measure.

When comparing the treatment programs to each other the threshold for a clinically relevant difference will be 3 points on the GAD-7 scale. A simulation based power analysis based on estimates from a pilot study of the IU-ICBT protocol has shown that we have well above 95% power when using mixed models, for that equivalence bound and n = 200 per group. Note however that the parameter estimates that goes into that power-analysis is based on only n = 22 patients. For comparison a more simple power-analysis for a post-treatment t-test show n = 200 gives 95% power (rounded up) for an equivalence bound of d = 0.36. Note that equivalence bounds should normally be justified based on clinical grounds rather than statistical; this latter power analysis is thus for illustrative purposes.

Hypothesis 1: IU-ICBT and Meta-ICBT will be effective and equivalent in a benchmarking analysis. (i.e. the estimated outcome at end of treatment will be equivalent, as defined by our equivalence bounds, and the pre-post change will be comparable to other ICBT studies on GAD).

Data-analysis: We plan to use multi-level modeling to utilize the increased power/precision we get from within-participant repeated measures.

Hypothesis 2: Changes in intolerance of uncertainty will statistically mediate the treatment effect in the IU-ICBT group. Changes in negative metacognitive beliefs about worry will statistically mediate the treatment effect in the Meta-ICBT group. The effect of both mediators will be moderated by level of intolerance uncertainty and negative metacognitive beliefs at baseline. We will also compare the putative mediators to each other with the hypothesis that the mediation-effect will be stronger for the specific mediator that a treatment protocol targets. This last analysis is to be regarded as the most severe test of our mediation analysis.

Data-analysis: Participants will fill out weekly questionnaires which will include our putative mediators (MCQ-30 and IU-12 questionnaires) and their current symptom level (GAD-7 questionnaire). The mediation-effect of both treatment protocols will be analyzed using latent growth curve modeling which is considered suitable for the study of moderation and mediation in clinical trials.

Additional hypotheses: In addition to the primary hypotheses, we may also investigate alternative moderators proposed by the literature. Worry behaviors is hypothesised to moderate treatment effect in the IU-ICBT group since that treatment partially has a behavioral treatment focus. Emotional contrast avoidance will also be investigated as a moderator. We will also explore demographic moderators such as gender and education level.

We will also investigate treatment satisfaction using the client satisfaction questionnaire, and treatment credibility using the treatment credibility scale, as well as patient global impressions of improvement.