Predicting Response for Pharmacotherapy Using Neuroanatomical Single-Subject Pattern Recognition in Obsessive-Compulsive Disorder

Abstract

Background: Primary pharmacotherapy regimen for obsessive-compulsive disorder (OCD) named Serotonin reuptake inhibitors (SRIs) does not attain sufficient symptom improvement in 40-60% of OCD. We aimed to decode the differential profile of OCD-related brain pathology per subject in the context of cortical surface area (CSA) or thickness (CT)-based individualized structural covariance (ISC) and to demonstrate the potential of which as a biomarker of treatment response to SRI-based pharmacotherapy in OCD using the support vector machine (SVM).

Methods: T1-weighted magnetic resonance imaging was obtained at 3T from 56 unmedicated OCD subjects and 75 healthy controls (HCs) at baseline. After 4 months of SRI-based pharmacotherapy, the OCD subjects were classified as responders (OCD-R, N=25

≥35% improvement) or nonresponders (OCD-NR, N=31

<35% improvement) according to the percentage change in the Yale-Brown Obsessive Compulsive Scale total score. Cortical ISCs sustaining between-group difference (p < .001) for every run of leave-one-out group-comparison were packaged as feature set for group classification using the SVM.

Results: An optimal feature set of the top 12 ISCs including a CT-ISC between the dorsolateral prefrontal cortex versus precuneus, a CSA-ISC between the anterior insula versus intraparietal sulcus, as well as perisylvian area-related ISCs predicted the initial prognosis of OCD as OCD-R or OCD-NR with an accuracy of 89.0% (sensitivity 88.4%, specificity 90.1%). Extended sets of ISCs distinguished the OCD subjects from the HCs with 90.7-95.6% accuracy (sensitivity 90.8-96.2%, specificity 91.1-95.0%).

Conclusion: We showed the potential of cortical morphology-based ISCs, which reflect dysfunctional cortical maturation process, as a possible biomarker that predicts the clinical treatment response to SRI-based pharmacotherapy in OCD.