IMPORTANCECognitive behavioral therapy (CBT) for obsessive-compulsive disorder (OCD) is a highly specialized treatment that is in short supply worldwide. OBJECTIVES To investigate whether both therapist-guided and unguided internet-based CBT (ICBT) are noninferior to face-to-face CBT for adults with OCD, to conduct a health economic evaluation, and to determine whether treatment effects were moderated by source of participant referral. DESIGN, SETTING, AND PARTICIPANTS This study is a single-blinded, noninferiority, randomized clinical trial, with a full health economic evaluation, conducted between September 2015 and January 2020, comparing therapist-guided ICBT, unguided ICBT, and individual face-to-face CBT for adults with OCD. Follow-up data were collected up to 12 months after treatment. The study was conducted at 2 specialist outpatient OCD clinics in Stockholm, Sweden. Participants included a consecutive sample of adults with a primary diagnosis of OCD, either self-referred or referred by a clinician. Data analysis was performed from June 2019 to January 2022. INTERVENTIONS Guided ICBT, unguided ICBT, and face-to-face CBT delivered over 14 weeks. MAIN OUTCOMES AND MEASURESThe primary end point was the change in OCD symptom severity from baseline to 3-month follow-up. The noninferiority margin was 3 points on the masked assessor-rated Yale-Brown Obsessive Compulsive Scale.RESULTS A total of 120 participants were enrolled (80 women [67%]; mean [SD] age, 32.24 [9.64] years); 38 were randomized to the face-to-face CBT group, 42 were randomized to the guided ICBT group, and 40 were randomized to the unguided ICBT group. The mean difference between therapist-guided ICBT and face-to-face CBT at the primary end point was 2.10 points on the Yale-Brown Obsessive Compulsive Scale (90% CI, −0.41 to 4.61 points; P = .17), favoring face-to-face CBT, meaning that the primary noninferiority results were inconclusive. The difference between unguided ICBT and face-to-face CBT was 5.35 points (90% CI, 2.76 to 7.94 points; P < .001), favoring face-toface CBT. The health economic analysis showed that both guided and unguided ICBT were costeffective compared with face-to-face CBT. Source of referral did not moderate treatment outcome.The most common adverse events were anxiety (30 participants [25%]), depressive symptoms (20 participants [17%]), and stress (11 participants [9%]). CONCLUSIONS AND RELEVANCEThe findings of this randomized clinical trial of ICBT vs face-toface CBT for adults with OCD do not conclusively demonstrate noninferiority. Therapist-guided ICBT could be a cost-effective alternative to in-clinic CBT for adults with OCD in scenarios where traditional (continued)

Major depressive disorder and the anxiety disorders are highly prevalent, disabling and moderately heritable. Depression and anxiety are also highly comorbid and have a strong genetic correlation ( r g ≈ 1). Cognitive behavioural therapy is a leading evidence-based treatment but has variable outcomes. Currently, there are no strong predictors of outcome. Therapygenetics research aims to identify genetic predictors of prognosis following therapy. We performed genome-wide association meta-analyses of symptoms following cognitive behavioural therapy in adults with anxiety disorders ( n = 972), adults with major depressive disorder ( n = 832) and children with anxiety disorders ( n = 920; meta-analysis n = 2724). We estimated the variance in therapy outcomes that could be explained by common genetic variants ( h 2 SNP ) and polygenic scoring was used to examine genetic associations between therapy outcomes and psychopathology, personality and learning. No single nucleotide polymorphisms were strongly associated with treatment outcomes. No significant estimate of h 2 SNP could be obtained, suggesting the heritability of therapy outcome is smaller than our analysis was powered to detect. Polygenic scoring failed to detect genetic overlap between therapy outcome and psychopathology, personality or learning. This study is the largest therapygenetics study to date. Results are consistent with previous, similarly powered genome-wide association studies of complex traits.