Effect of Internet vs Face-to-Face Cognitive Behavior Therapy for Health Anxiety: A Randomized Noninferiority Clinical Trial

Key Points

Question  In the treatment of health anxiety, can internet-delivered cognitive behavior therapy lead to comparable effects as conventional face-to-face cognitive behavior therapy while using less resources?

Findings  This randomized noninferiority clinical trial in a primary care setting included 204 adults with a principal diagnosis of health anxiety. Internet-delivered cognitive behavior therapy was found to be noninferior to face-to-face cognitive behavior therapy while generating lower net societal costs.

Meaning  As internet-delivered cognitive behavior therapy is not inferior to face-to-face cognitive behavior therapy and requires little resources, the online treatment format should be considered a first-line intervention for health anxiety.

Importance  Health anxiety is a common and often chronic mental health problem associated with distress, substantial costs, and frequent attendance throughout the health care system. Face-to-face cognitive behavior therapy (CBT) is the criterion standard treatment, but access is limited.

Objective  To test the hypothesis that internet-delivered CBT, which requires relatively little resources, is noninferior to face-to-face CBT in the treatment of health anxiety.

Design, Setting, and Participants  This randomized noninferiority clinical trial with health economic analysis was based at a primary care clinic and included patients with a principal diagnosis of health anxiety who were self-referred or referred from routine care. Recruitment began in December 10, 2014, and the last treatment ended on July 23, 2017. Follow-up data were collected up to 12 months after treatment. Analysis began October 2017 and ended March 2020.

Interventions  Patients were randomized (1:1) to 12 weeks of internet-delivered CBT or to individual face-to-face CBT.

Main Outcomes and Measures  Change in health anxiety symptoms from baseline to week 12. Analyses were conducted from intention-to-treat and per-protocol (completers only) perspectives, using the noninferiority margin of 2.25 points on the Health Anxiety Inventory, which has a theoretical range of 0 to 54.

Results  Overall, 204 patients (mean [SD] age, 39 [12] years; 143 women [70%]) contributed with 2386 data points on the Health Anxiety Inventory over the treatment period. Of 204 patients, 102 (50%) were randomized to internet-delivered CBT, and 102 (50%) were randomized to face-to-face CBT. The 1-sided 95% CI upper limits for the internet-delivered CBT vs face-to-face CBT difference in change were within the noninferiority margin in the intention-to-treat analysis (B = 0.00; upper limit: 1.98; Cohen d = 0.00; upper limit: 0.23) and per-protocol analysis (B = 0.01; upper limit: 2.17; Cohen d = 0.00; upper limit: 0.25). The between-group effect was not moderated by initial symptom level, recruitment path, or patient treatment preference. Therapists spent 10.0 minutes per patient per week in the online treatment vs 45.6 minutes for face-to-face CBT. The net societal cost was lower in the online treatment (treatment period point difference: $3854). There was no significant group difference in the number of adverse events, and no serious adverse event was reported.

Conclusions and Relevance  In this trial, internet-delivered CBT appeared to be noninferior to face-to-face CBT for health anxiety, while incurring lower net societal costs. The online treatment format has potential to increase access to evidence-based treatment for health anxiety.

Trial Registration  ClinicalTrials.gov Identifier: NCT02314065

Health anxiety, also referred to as hypochondriasis, is a common, debilitating, and often chronic psychiatric condition^1-3 characterized by an excessive and persistent fear or worry about serious illness. Throughout the health care system,⁴ health anxiety leads to excessive medical investigations and substantial societal costs.² As medical reassurance does not lead to enduring improvement,^5-7 health anxiety needs to be targeted by specific interventions.

Although there is some evidence in support of antidepressants in the treatment of health anxiety,⁸ the criterion standard treatment is cognitive behavior therapy (CBT).⁹ In CBT, the patient meets with a therapist, typically once a week over approximately 3 months. About 66% of patients respond to CBT,¹⁰ but given the prevalence of health anxiety and scarcity of mental health professionals, the need for treatment far exceeds the availability of evidence-based face-to-face therapy.

Internet-delivered CBT (ICBT) is a text-based online treatment in which the patient works with conventional CBT strategies and communicates regularly with a therapist through an email-like system.¹¹ The patient can access the treatment at any time of the day, the treatment can be delivered regardless of geographical distances, and little time is required from the therapist. Thus, ICBT holds promise for increasing the outreach of psychological treatment.¹²

In the treatment of health anxiety, ICBT has been found to be more effective than a rudimentary attention control (d = 1.62; 95% CI, 1.10-2.10)¹³ and active comparator (pooled d = 0.81; 95% CI, −0.30 to 1.92)^14,15 but has not yet been compared head to head with face-to-face CBT. Therefore, we conducted a randomized noninferiority clinical trial of the 2 treatment formats, within a primary care setting. Our primary hypothesis was that ICBT would be noninferior to individual face-to-face CBT, while requiring less therapist time and generating lower net societal costs.

This was a randomized noninferiority clinical trial comparing ICBT to face-to-face CBT for health anxiety at a primary care clinic in Stockholm, Sweden. In Sweden, primary care clinics provide access to care for most common disease states, including anxiety disorders, and are intended to serve as first-line gatekeepers in authorizing access to specialty care. Adults with principal health anxiety were recruited through routine care referral and online self-referral. The trial protocol is available in Supplement 1.

We informed local clinics about the study and advertised the study under the heading “Do you worry a lot about your health?” Recruitment began in December 10, 2014, and the last treatment ended on July 23, 2017. Applicants provided informed consent via the secure online platform and completed a comprehensive online screening battery. The study was approved by the Stockholm Regional Ethics Review Board (2014/1530-31/2). A face-to-face psychiatric interview was held with a clinical psychologist, as based on the Mini-International Neuropsychiatric Interview,¹⁶ the hypochondriasis module of the Anxiety Disorders Interview Schedule for DSM-IV,¹⁷ and the Health Preoccupation Diagnostic Interview.¹⁸ Based on the Health Preoccupation Diagnostic Interview, we included adults (≥18 years) with a principal diagnosis of DSM-5 somatic symptom disorder or illness anxiety disorder (these have replaced DSM-IV hypochondriasis and the diagnostic procedure has been found to be reliable¹⁸). Applicants who met criteria for a bipolar disorder or psychosis, severe depression (DSM-5 criteria), or recurrent clinically significant suicidal ideation (based on clinical judgement aided by a structured clinical interview and item 9 of the self-rated Montgomery-Åsberg Depression Rating Scale, rated from 0: “I have a normal appetite for life” to 6: “I am convinced that my only way out is to die, and I think a lot about how to best go about taking my own life”), a personality disorder likely to severely interfere with treatment, an alcohol or substance use disorder in the past 6 months, or a serious somatic condition that required immediate or extensive care, were excluded and referred to routine care. We required patients taking antidepressant medication to have been taking a stable dose for at least 2 months. Patients were also required not to have any other ongoing psychological treatments for health anxiety and not to have undergone CBT for health anxiety during the past year (eTable 1 in Supplement 2).

Patients who met all eligibility criteria and completed the baseline assessment were randomly allocated (1:1) to ICBT or face-to-face CBT in consecutive even-numbered blocks. On each starting week, therapists were assigned an equal number of patients per treatment (ICBT vs face-to-face CBT) to balance therapist characteristics over the treatment arms. Randomization was based on a true random number service (http://www.random.org) and conducted by a member of the research group who was blinded to patient characteristics and otherwise not involved in the main phase of the trial. Because inclusion was a requirement for randomization, clinicians who conducted the eligibility interview could not anticipate treatment allocation.

The ICBT and face-to-face CBT protocols were equivalent except for the delivery format (eTable 2 in Supplement 2). Both treatments lasted 12 weeks and included standard cognitive behavioral techniques for health anxiety. Typically, the first 2 weeks focused on education about the CBT model of health anxiety, and subsequent weeks focused on exposure to health anxiety–provoking stimuli. Patients were encouraged to confront situations that trigger intrusions about illness and to refrain from behaviors believed to maintain health anxiety over time. For example, a patient with an excessive fear of having a weak heart, despite frequent medical reassurance, could be encouraged to exercise while refraining from pulse checking. The treatment also included mindfulness training as a means of increasing the patient’s willingness and ability to experience unwanted thoughts and emotions. To reduce dropout rates, we instructed therapists to telephone patients who canceled appointments or did not contact the therapist in the online platform. The therapists did not use these telephone calls to convey treatment content.

In ICBT, patients gained access to their treatment through a secure website. Here, patients could communicate freely with their therapist by way of an email-like system and expect a reply within 48 hours on weekdays. The treatment was communicated via a self-help text divided into 12 chapters (ie, modules). We encouraged patients in ICBT to complete about 1 module per week, and patients received feedback from the therapist after each module (eAppendix 1 and eFigures 1-3 in Supplement 2).

Patients randomized to face-to-face CBT were expected to attend 12 weekly in-person sessions with their therapist and to work with daily homework assignments. The treatment followed a written manual, and exposure and response prevention was tailored to suit the needs of each patient. The first session lasted 80 minutes and the remaining were approximately 50 minutes (eAppendix 2 in Supplement 2).

Therapists in ICBT and face-to-face CBT were the same 5 psychologists (3 of which were authors: E. Axelsson, D.B., and M. H.-L.), employed at the primary care clinic. All therapists were psychologists with training in CBT (eTable 3 in Supplement 2) who were introduced to the study by one of the authors (E. Axelsson) and given scheduled reading time, detailed manuals with case studies, session-by-session adherence checklists, examples of email correspondence in ICBT, and an audio recording of a competently delivered CBT session. Every other week, therapists attended 40 to 60 minutes of supervision with 1 of 2 authors (E. Axelsson or E. H.-L.). At the beginning of the trial, 3 of 5 therapists preferred noninferiority as the outcome of the trial, and 3 of 5 expected this outcome (eTable 3 in Supplement 2).

In accordance with research recommendations,¹⁹ audio from all face-to-face CBT sessions were recorded and a random 10% (109 of 1059) of these files, stratified by period and therapist, were anonymized and assessed for fidelity by trained clinicians (eAppendix 5 in Supplement 2). Ratings were primarily based on a structured 27-item instrument developed to capture key aspects of therapist adherence from 0% (very poor) to 100% (excellent) and competence from 0 (very poor) to 5 (excellent). An initial concordance check based on 4 recordings indicated excellent interrater reliability (adherence: perfect agreement; competence: intraclass correlation coefficient = 0.77).

Primary outcome was change (slope) in health anxiety over the treatment period, as measured using the clinical reference standard 18-item Health Anxiety Inventory (HAI), which has a theoretical range of 0 to 54.²⁰ For the primary analysis, patients completed the HAI at baseline, each week during treatment, and at posttreatment. Follow-up assessments were also conducted 6 and 12 months after treatment completion. The HAI has excellent psychometric properties²¹ and is widely used. In this study, response was operationalized as an HAI score reduction more than 7.74 points, remission as an end point score below 23.55 points, and clinically significant improvement was the combination of response and remission (eAppendix 7 in Supplement 2).²²

Secondary self-rated outcome questionnaires were completed online at baseline, immediately after treatment, and 6 and 12 months after treatment (eTable 4 in Supplement 2). We measured general anxiety using the Beck Anxiety Inventory,²³ depression using the self-rated Montgomery-Åsberg Depression Rating Scale²⁴ and disability using the Sheehan Disability Scale.²⁵ We also measured health anxiety using the Illness Attitude Scales²⁶ and Whiteley Index,²⁷ anxiety sensitivity using the Anxiety Sensitivity Index,²⁸ sleep disturbance using the Insomnia Severity Index,²⁹ alcohol use using the Alcohol Use Disorders Identification Test,³⁰ and disability using the World Health Organization Disability Assessment Schedule 2.³¹ At week 2, treatment credibility was assessed using the C scale³² and therapeutic alliance using the Working Alliance Inventory.³³ Adverse events and treatment satisfaction³⁴ were assessed at posttreatment. Data on resource utilization in the preceding month³⁵ (used to estimate costs) and health-related quality of life³⁶ were collected at baseline, posttreatment, and at follow-up (eAppendix 8 in Supplement 2). At posttreatment and follow-up, we also asked about parallel treatments.

The primary noninferiority and power analyses were done by an independent statistician, and the analysis plan for the primary outcome was finalized in August 2016. Analysis began October 2017. Change over time was modeled using mixed-effects linear regression (patients at level 2) with a random intercept and slope (time). That is, the HAI was regressed on time (baseline to end point), group (ICBT vs face-to-face treatment), and the time × group interaction. We considered ICBT to be noninferior to face-to-face CBT if the upper bound of the 1-sided confidence interval for the time × group interaction (α = .05) was within the a priori noninferiority margin of 2.25 points on the HAI (Cohen d of approximately 0.3). We decided on this noninferiority margin based on expert consensus, the confidence interval for the effect of CBT vs nonactive controls,³⁷ and what most clinicians and patients consider a minimally important difference on patient-reported outcomes.³⁸ Monte Carlo simulations indicated that a sample size of 200 would be required for 80% power to confirm noninferiority in the primary analysis (Supplement 1). As the intention-to-treat approach may indicate noninferiority simply due to poor adherence to the study protocol, in a sensitivity analysis, we also investigated noninferiority within a per-protocol framework with data from patients who had initiated at least 6 modules in ICBT, or attended at least 6 sessions in face-to-face CBT. We also estimated Cohen d by dividing the time × group coefficient by the pooled observed end point standard deviation. The main outcome of this trial was reported in a doctoral dissertation in 2018.³⁹

We based all secondary analyses on 2-sided tests and the α level of 5%. In moderator analyses, we tested if the primary outcome was moderated by treatment preference (0 indicates strong preference for face-to-face CBT and 5, strong preference for ICBT), path of recruitment (routine care vs not routine care), and baseline health anxiety. As this trial lacked a nonactive control group, we also tested for dose-response relationships, ie, the effect of the number of sessions or modules on the slope of the HAI. For all secondary analyses of change, data were imputed using multiple imputation by chained equations (20 samples). In the analysis of long-term outcomes, piecewise linear mixed models were fitted with a spline at the posttreatment assessment. Furthermore, we explored dichotomous outcomes in terms of response, remission, and clinically significant improvement. In health economic analyses, we compared ICBT to face-to-face CBT with regard to societal costs in relation to clinical efficacy (cost effectiveness) and quality-adjusted life-years (cost utility). Costs included costs of therapies and medications (including the cost of ICBT or face-to-face CBT), costs of nonmedical help and services, and indirect costs such as those of unemployment (eAppendix 3 and 4 in Supplement 2).

Between December 2014 and March 2017, 503 applicants completed the online screening battery, and 159 (32%) of these were referred from routine care (Figure 1). In total, 204 patients were included and randomized to either ICBT (102 [50%]) or face-to-face CBT (102 [50%]). We collected 2386 data points on the HAI during the treatment phase (eTable 5 in Supplement 2). Sociodemographic and clinical characteristics were evenly distributed between the treatment groups (Table 1). The mean (SD) age was 39 (12) years. Most patients were women (143 [70%]), married (168 [82%]), had a postsecondary or university education (154 [75%]), were currently employed (137 [67%]), and had experienced health anxiety for 9 years but never received psychological treatment (171 [84%]). Forty-six patients reported having a somatic condition out of the ordinary (eTable 6 in Supplement 2). Most commonly, this was hypothyroidism (n = 12), followed by hypertension (n = 8), irritable bowel syndrome (n = 5), and osteoarthritis (n = 4).

In the intention-to-treat analysis, the estimated difference in effect over the treatment period was 0.00 points on the primary outcome (HAI), and the upper limit of the 1-sided 95% CI was 1.98. The per-protocol analysis was indicative of a difference of 0.01 points in favor of individual face-to-face CBT, with the upper limit of the 95% CI being 2.17. Thus, both models converged with confidence interval upper bounds within the noninferiority margin of 2.25, indicating that ICBT was noninferior to individual face-to-face CBT (Figure 2).

The main outcome was not significantly moderated by the patient’s baseline treatment preference (B = −1.28; 95% CI, −2.90 to 0.34), health anxiety (HAI; B = 0.23; 95% CI, −0.15 to 0.62), or whether the patient was referred from routine care (B = 2.33; 95% CI, −2.63 to 7.30). Recruitment path was also not a significant predictor of preference. Significantly larger reductions in health anxiety were seen for patients who completed more modules in ICBT (B = −0.44; 95% CI, −0.88 to −0.01) or attended more sessions in face-to-face CBT (B = −0.92; 95% CI, −1.57 to −0.26).

At the primary end point, responder rates did not differ significantly between the treatments: 78 of 102 (76%) in ICBT vs 75 of 102 (74%) in face-to-face CBT (eTables 9-14 in Supplement 2). No treatment had significantly higher effects on secondary symptom outcomes at the primary end point (Table 2; eTables 7 and 8 and eAppendix 6 in Supplement 2). Within-group effects on health anxiety were large for both ICBT (d = 1.76) and face-to-face CBT (d = 1.76) with sustained effects at follow-up.

The main outcomes of the health economic evaluation are presented in Figure 3. Evidently, it is highly probable that ICBT incurs lower net societal costs than face-to-face CBT, mainly owing to lower therapist costs (mean [SD]: $454 [257] vs $2059 [595]) (eTables 15, 16, and 17 in Supplement 2). As can be seen in Figure 3A, from baseline to the primary end point, ICBT was associated with a similar rate of clinically significant improvement as face-to-face CBT but at a lower net societal cost. eTable 18, eAppendices 8-11, and eFigure 4 in Supplement 2 include a comprehensive presentation of the health economic evaluation.

Face-to-face therapist adherence (mean [SD] percentage score, 98 [0.1]) and competence (mean [SD], 4.5 [0.6]) were good to excellent, indicating high treatment fidelity. Therapists spent a mean (SD) of 10.0 (5.7) minutes per patient per week in ICBT and 45.6 (13.1) minutes in face-to-face CBT. There was a smaller proportion of treatment completers in ICBT than in face-to-face CBT (81 of 102 [79.4%] vs 92 of 102 [90.2%]). The median (interquartile range) number of initiated ICBT modules was 11 (6-12), and the median (interquartile range) number of attended face-to-face CBT sessions was 12 (11-12).

There was no significant difference between patients in ICBT and face-to-face CBT with regard to changes in psychotropic medication over the treatment period (6 of 97 [6.2%] vs 5 of 97 [5.1%]) or the follow-up period (7 of 92 [7.6%] vs 10 of 89 [11.2%]), or other psychological treatments during the treatment period (3 of 97 (3.1%) vs 0 of 97 [0%]) or follow-up period (10 of 92 [10.9%] vs 8 of 89 [9.0%]). Therapeutic alliance was rated significantly lower in ICBT than in face-to-face CBT (mean [SD], 32.3 [7.9] vs 36.3 [5.9]), but there was no significant difference in treatment credibility (mean [SD], 35.3 [8.4] vs 37.5 [6.6]; 95% CI 0.0-4.2; P = .05).

At least 1 adverse event was reported by 19 of 97 patients in ICBT (20%) and 17 of 97 patients in face-to-face CBT (18%). This difference was not significant. The most commonly reported type of adverse event in both treatments was increased anxiety or stress (16 [16%] vs 14 [14%]). No serious adverse events were reported (eAppendix 12 in Supplement 2).

Results from this large-scale noninferiority trial conducted in a primary care setting showed that ICBT, while incurring lower net societal costs, was not inferior to individual face-to-face CBT for patients with health anxiety. The robustness of this conclusion was indicated by intention-to-treat and per-protocol analyses converging on the same outcome. Compared with face-to-face CBT, ICBT therapists spent 78% less time per patient. Yet, both treatments achieved large effects on health anxiety and moderate to large effects on general anxiety and depression.

To our knowledge, this is the first noninferiority trial to compare ICBT with criterion standard face-to-face CBT for health anxiety. It is also, to our knowledge, the largest randomized controlled comparison between ICBT and face-to-face therapy for any mental health condition.⁴⁰ Four previous randomized clinical trials also corroborate that ICBT has beneficial effects on health anxiety.^13-15,41 We have found ICBT to have specific effects beyond those of an active, structured, and credible treatment in the form of internet-delivered behavioral stress management.¹⁴ Another research group has also found ICBT to be more efficacious than the fortnightly provision of psychoeducational material.¹⁵

Health anxiety is prevalent, not least in medical clinics,⁴ tends to persist over time without treatment,³ and is associated with increased health care consumption.² ICBT reduces the time needed per treatment, does not require patients to work with their treatment during any particular time of the day, and allows clinics to have a nationwide outreach. Tentatively, there is typically no need for clinicians to select specific patients for ICBT. That being said, just like certain patients are not able to participate in face-to-face CBT for example owing to geographical barriers, others are not able to participate fully in text-based ICBT for example owing to difficulties reading and writing. It should also be mentioned that exploratory analyses pointed to face-to-face CBT being associated with slightly higher adherence and moderately larger effects on general anxiety and depression in the long term. At 12 months, the between-group effect on health anxiety (2.4 points on the HAI) was clinically but not statistically significant. In summary, although differences appear to be rather small also in the long term, they could be clinically meaningful, and more data are needed to draw firm conclusions in this regard.

Quality markers of this trial were a well-powered sample, minimal data loss, the predetermined primary statistical model, the high degree of treatment fidelity and therapist competence, and also that treatment content and therapist assignment were balanced over the treatment groups. The within-group effects of face-to-face CBT were large and similar to those seen in previous studies, indicating that the comparison group in this trial was valid.

This trial also had limitations. First, the sample had a relatively high mean level of education, which may affect the generalizability of our findings. However, baseline health anxiety was high and most patients had been ill for years. Second, there were no posttreatment psychiatric interviews. Still, we believe that the choice of the HAI as primary outcome was ideal because health anxiety is a dimensional phenomenon and there appears to exist no clear boundary between moderate and severe worry about health.⁴² The HAI has also been used as primary outcome in previous randomized clinical trials^15,43 and is a valid instrument²¹ that enabled a repeated measurements design with high statistical power. Third, we did not ask our patients about all specific putative parallel treatments, eg, interval treatment, throughout the study. However, we did establish that the treatment groups did not differ with regard to psychotropic medication or parallel psychological interventions. Also, based on free text questions that prompted patients to report all health care visits and medical expenses, medical costs were comparable in the 2 groups (eTable 15 in Supplement 2). Fourth, although we followed up patients up to 12 months after treatment, this trial was not powered to assess noninferiority over the follow-up period. However, at least in the short term, it seems that when face-to-face CBT works, ICBT is a noninferior intervention with potential to markedly increase the availability of efficacious treatment for individuals with health anxiety.

Based on the findings of this trial and previous evidence, we conclude that therapist-guided ICBT is noninferior to individual face-to-face CBT for patients with health anxiety. Given the low societal costs of ICBT for health anxiety, our findings highlight the potential benefits of implementing this treatment on a wider scale.

Corresponding Author: Erik Hedman-Lagerlöf, PhD, Division of Psychology, Department of Clinical Neuroscience, Karolinska Institutet, Nobels väg 9, SE-171 65 Stockholm, Sweden (kire.hedman@ki.se).

Accepted for Publication: March 3, 2020.

Published Online: May 13, 2020. doi:10.1001/jamapsychiatry.2020.0940

Author Contributions: Drs Axelsson and E. Hedman-Lagerlöf had full access to all of the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.

Concept and design: Axelsson, Andersson, Ljótsson, E. Hedman-Lagerlöf.

Acquisition, analysis, or interpretation of data: All authors.

Drafting of the manuscript: Axelsson, Andersson, Ljótsson, E. Hedman-Lagerlöf.

Critical revision of the manuscript for important intellectual content: All authors.

Statistical analysis: Axelsson, Ljótsson.

Obtained funding: Axelsson, E. Hedman-Lagerlöf.

Administrative, technical, or material support: All authors.

Supervision: Axelsson, Andersson, Ljótsson, E. Hedman-Lagerlöf.

Conflict of Interest Disclosures: Drs Axelsson and E. Hedman-Lagerlöf report grants from Karolinska Institutet and Stockholm County Council (ALF Medicine) during the conduct of the study. Drs Axelsson, Andersson, Ljótsson, and E. Hedman-Lagerlöf have coauthored a self-help book for severe health anxiety available in the public marketplace, published between the start of the trial and the writing of the present manuscript. Drs Ljótsson and E. Hedman-Lagerlöf are shareholders of DahliaQomit Inc, a company specializing in online psychiatric symptom assessment, and Hedman-Lagerlöf och Ljótsson psykologi AB, a company that creates commercialized internet-based cognitive behavior therapy manuals. No other disclosures were reported.

Funding/Support: This research was funded by Karolinska Institutet, Psykiatrifonden, and Region Stockholm.

Role of the Funder/Sponsor: The funding sources took no part in the design or conduct of the study. This includes the collection, management, analysis, and interpretation of data, the preparation, review and approval of the manuscript, and the decision to submit the manuscript for publication.

Data Sharing Statement: See Supplement 3.

Additional Contributions: We thank the staff of Gustavsberg Academic Primary Care Clinic and all who took part in the development of the internet cognitive behavior therapy protocol in previous trials. We especially thank Louise Andersson, MSc (assessed treatment fidelity; employed specifically for this purpose); Anna Granlund, MSc (assessed treatment fidelity; employed specifically for this purpose); Anna-Clara Hellstadius, MSc (therapist; standard salary); Sofia Hultkrantz, MSc (therapist; standard salary); Anna Lövgren, MSc (assessed treatment fidelity; employed specifically for this purpose); Greta Wester, MSc (reminded patients about assessments; employed specifically for this purpose); and Julia Widoff, MSc (reminded patients about assessments; employed specifically for this purpose), at Gustabsverg Academic Primary Care Clinic.