Polygenic risk scores for predicting symptom change after internet-delivered cognitive behavioral therapy for depression and anxiety disorders

Internet-delivered cognitive behavioral therapy (ICBT) is an effective and accessible alternative to face-to-face delivered CBT for depression and anxiety disorders. However, not all patients achieve clinically significant reduction in symptom severity following treatment. Individual genetic variation may partially explain differential symptom change. Yet, evidence in favor of using polygenic risk scores (PRS) for prediction of treatment outcome remains scarce. We previously reported poorer response to ICBT for depression in patients with higher genetic load for autism spectrum disorder (PRS-ASD) in a sample of 894 patients with depression. Here we aimed to identify whether higher PRS for psychiatric and personality traits were associated with less improvement and whether utilizing PRS adds predictive power above and beyond clinical and socioeconomic predictors. Data on 2668 patients who received ICBT at the Internet psychiatry clinic in Stockholm, Sweden, for major depressive disorder (MDD) (n=1330), panic disorder (PD) (n=727), and social anxiety disorder (SAD) (n=641) were used. The patients were genotyped pre-treatment, detailed clinical data were collected, and register linkage was performed. Eight PRS were constructed for MDD, ASD, ADHD, Bipolar Disorder, Schizophrenia (SZ), Educational Attainment, IQ, and Cross Disorder. Symptom change was operationalized as percentage change between pre- and post-treatment score on the scale of each respective disorder (Montgomery–Åsberg Depression Rating Scale for MDD, Panic Disorder Severity Scale for PD, and Liebowitz Social Anxiety Scale for SAD). Measurements were harmonized and combined into a single score. Associations of each of the eight PRS with symptom change was assessed through (i) bivariate linear regression, (ii) multiple linear regression with all PRS, and (iii) a full model including all PRS and 38 clinical and register-based socioeconomic predictors (e.g., self-reported pre-treatment symptom severity, comorbidities, symptom duration, prior psychiatric diagnoses, medication use, education, employment, and income). Mean relative pre- to post-treatment symptom reduction in the sample was 41.55% (SD 39.51). In bivariate analyses, higher PRS-ASD was significantly associated with less reduction in symptom severity: β = -1.88 (p =.045, 95% CI [-3.73, -0.04]). In multiple regression with PRS for eight traits, three were positively associated with less symptom reduction (PRS-ASD: β = -2.27 (p =.036, 95% CI [-4.40, -0.15]); PRS-MDD: β = -2.71 (p =.025, 95% CI [-5.07, -0.35]); PRS-SZ: β = -3.90 (p =.019, 95% CI [-7.15, -0.65]), while higher Cross Disorder PRS was associated with more symptom reduction: β = 4.59 (p =.027, 95% CI [0.52, 8.66]). The association remained statistically significant for PRS-ASD even in the full model with inclusion of multiple clinical and socioeconomic data covariates: β = -2.29 (p =.042, 95% CI [-4.49, -0.08]). Our results suggest that psychiatric PRS may play a role in differential symptom reduction during ICBT for depression and anxiety disorders. This study supports previous findings suggesting that higher genetic liability to autistic traits may be associated with lower symptom reduction following ICBT for depression, while expanding to a three times larger sample and across three psychiatric disorders.