Anxiety and depression subtypes and their psychotherapy response: A network analysis of 33,675 patients
Zoë Mermin, Donald J Robinaugh, Thomas D Hull, Kristin L Szuhany, Naomi Simon, Matteo MalgaroliAbstract
Background
Major depressive disorder and generalized anxiety disorder frequently co-occur, leading to heterogeneous presentations that complicate diagnosis and treatment. While diagnostic categories often obscure symptom variability, network approaches offer a powerful way to examine how symptoms relate to one another within and across conditions. By identifying symptom clusters, these methods can provide insights into potential mechanisms of comorbidity and symptom maintenance and may ultimately help guide more personalized treatment.
Methods
We examined patients ( N = 33,675) seeking treatment for anxiety or depression through a message-based psychotherapy platform. Participants completed standardized measures of anxiety (GAD-7) and depression (PHQ-9) at baseline. We estimated a Gaussian Graphical Model of anxiety and depressive symptoms and identified overlapping symptom clusters using a modified Walktrap algorithm. We assigned patients to clusters based on presenting symptoms. In a subsample ( n = 10,718), we examined the relationship between baseline cluster probabilities and three outcome trajectories over 12 weeks of treatment.
Results
We found four baseline symptom clusters relating to Affective Dysregulation, Worries, Neurovegetative symptoms, and Hyperarousal. Affective Dysregulation and Neurovegetative clusters were associated with poorer outcomes, whereas Worries and Hyperarousal were associated with recovery over partial improvement; Hyperarousal also differentiated recovery from non-response.
Conclusions
Symptom clusters derived from standard screening measures were associated with differential treatment trajectories. Targeting these symptom configurations may be particularly effective at disrupting multiple symptom pathways. Future research should examine whether personalized treatment strategies that consider symptom clusters based on simple screening in practice settings yield greater clinical improvement than traditional diagnostic categories.