Prediction of response to cognitive-behavioral therapy in obsessive-compulsive disorder: a multivariate analysis of resting state functional connectivity

Jamie D Feusner, MD1; Nicco Reggente, MA2; Teena D Moody, PhD1; Francesca Morfini, MA1; Jesse Rissman, PhD1,2; Joseph O’Neill, PhD1

Affiliation:

1Department of Psychiatry and Biobehavioral Sciences, David Geffen School of Medicine at UCLA, Los Angeles, California 2Department of Psychology, UCLA, Los Angeles, California

Background: Cognitive-behavioral therapy (CBT) is an effective treatment for reducing symptoms of obsessive-compulsive disorder (OCD). Although many with OCD will benefit from CBT, the response still varies significantly between individuals. In addition, specialized CBT for OCD has limited availability, can be an expensive treatment, and by its nature is stressful and often time-consuming. This underscores the importance of developing reliable predictors of response to treatment to help with clinical decision-making. Although several studies have examined clinical and neurobiological features pre-treatment that are correlated with response to treatment, only one has examined functional connectivity as a predictor, and none have applied multivariate approaches. We used a multivariate pattern recognition approach applied to resting state functional connectivity pre-CBT in order to make predictive inferences on the individual patient level, as to their degree of response to treatment. In addition, we applied the same approaches to pre-treatment symptomatology in order to further elucidate mechanisms of functional connectivity associated with obsessions and compulsions, in a data-driven manner.

Methods: We acquired resting state functional magnetic resonance image BOLD data in 25 medicated and unmedicated adults with OCD before 4 weeks of intensive daily exposure and response prevention, a form of CBT. Core OCD symptomatology was measured using the Yale-Brown Obsessive Compulsive Scale (YBOCS). Image preprocessing included parcellation of the brain into 264 regions of interest, each belonging to one of 14 functional networks previously derived from meta-analyses of functional studies. We computed a pairwise Pearson-correlation matrix for each mean time course resulting in a 264 x 264 matrix containing the pairwise functional connectivity values (r-values) across all ROIs. Matrix cells corresponding to each functional network were identified to create feature sets. We implemented a leave-one-patient-out cross-validation to assess the predictive power of our feature sets in regards to our behavioral measures of interest: change in YBOCS scores from pre- to post-CBT. Specifically, we built a least absolute shrinkage and selection operator (LASSO) regression model on n-1 patients using their feature sets. We correlated the predicted values with the actual values in order to yield a multiple R2 as a measure of our model’s feature-dependent predictivity. Additionally, we applied the same analysis to the pre-CBT (baseline) YBOCS scores.

Results: OCD participants showed significant clinical symptom improvements pre- to post-CBT (YBOCS scores X±Y pre-CBT; Z±Q post-CBT; t26=P, p<.R). Connectivity strength in the ventral attention network predicted greater/lesser reduction of YBOCS scores pre- to post-CBT ( =.185, P=.01. Connectivity strength in the cingulo-opercular network at baseline was predictive of baseline severity of YBOCS scores ( =.35, P=.0009).

Conclusions: This represents the first study in OCD to use multivariate pattern recognition approaches to determine neurobiological markers predictive of response to treatment. Strength of resting state functional connectivity in the ventral attention network was associated with a better response to treatment. This may signify that those with better inherent ability to attend to perceptually-driven stimuli in their environment (perhaps also reflecting that they are less internally distracted by obsessive thoughts) may respond better to treatment. In addition, the phenomenology of obsessions and compulsions, specifically before treatment, is associated with connectivity in the cingulo-opercular network. Given the function of this network, those with weaker connectivity may be less able to maintain control over behaviors and thought patterns in the face of emotional arousal, and hence have higher degree of obsessions and compulsions. Results have clinical implications for identifying individual OCD patients who will maximally benefit from treatment with intensive CBT, and have implications for further understanding the pathophysiology of OCD.

View the Poster, Presented at ACNP (2016)

The Method of Loci revisited: Memory enhancement by way of virtually augmented memory palaces

Reggente, N., Essoe, J., Mehta, P.*, Ohno, A.*, Rissman, J.

Humans have long appreciated that visuospatial cues can serve as a scaffolding for the encoding of non-spatial content. The Method of Loci (MoL), which binds objects to a spatial context in one’s mental imagery, has been the favored mnemonic strategy of memory champions since Ancient Greece. In this work, we created a virtual reality implementation of the MoL to tease apart the factors that contribute to the MoL’s undeniable efficacy as a memory enhancement technique. We crafted three distinct virtual environments where subjects could view objects. Subjects that were told to place items at locations of their choosing recalled significantly more objects than subjects who only viewed the objects. We also addressed the contributions of volition and contextual richness to recall strength.

View the poster, presented at ICOM in Budapest (2016)

Neural correlates of fluid intelligence via functional and structural network connectivity measures

Connectivity across regions in the brain can be characterized as either functional (correlated fluctuations in activity as measured by resting-state fMRI data) or structural (white matter pathways as measured by diffusion MRI data). Emerging studies suggest that the connections across brain regions that make up distinct cognitive networks can partially explain individual differences in behavioral traits. Some theorize that a reliable benchmark of intelligence is the ability to identify subtle patterns across distantly related ideas. The Raven’s Progressive Matrices (RPM), a pattern completion task, is one widely used measure of general fluid intelligence. Here, we use a combination of functional and structural connectivity metrics derived from a large MRI dataset [n=127] to examine the relationship between neural connectivity and RPM scores. We used a Support Vector Regression cross-validation procedure to assess the degree to which we could predict a subject’s intelligence based on these connectivity values. We were able to account for 14% of the variance in individuals’ intelligence scores when using specific combinations of functional and structural connectivity values.

You can view our poster here:

Vuong, Reggente, Rissman Poster Presented at UCLA PURC 2016