Functional connectivity of dorsolateral prefrontal cortex predicts cocaine relapse: implications for neuromodulation treatment

Abstract

Relapse is one of the most perplexing problems of addiction. The dorsolateral prefrontal cortex is crucially involved in numerous cognitive and affective processes that are implicated in the phenotypes of both substance use disorders and other neuropsychiatric diseases, and has become the principal site to deliver transcranial magnetic stimulation for their treatment. However, the dorsolateral prefrontal cortex is an anatomically large and functionally heterogeneous region, and the specific dorsolateral prefrontal cortex locus and dorsolateral prefrontal cortex-based functional circuits that contribute to drug relapse and/or treatment outcome remain unknown. We systematically investigated the relationship of cocaine relapse with functional circuits from 98 dorsolateral prefrontal cortex regions-of-interest defined by evenly sampling the entire surface of bilateral dorsolateral prefrontal cortex in a cohort of cocaine dependent patients (n = 43, 5 F) following a psychosocial treatment intervention. Cox regression models were utilized to predict relapse likelihood based on dorsolateral prefrontal cortex functional connectivity strength. Functional connectivity from only 3 of the 98 dorsolateral prefrontal cortex loci, one in the left and two in the right hemisphere, significantly predicted cocaine relapse with an accuracy of 83.9%, 84.6% and 85.4%, respectively. Combining all three loci significantly improved prediction validity to 87.5%. Protective and risk circuits related to these dorsolateral prefrontal cortex loci were identified that have previously been implicated to support “bottom up” drive to use drug and “top down” control over behavior together with social emotional, learning and memory processing. Three dorsolateral prefrontal cortex-centric circuits were identified that predict relapse to cocaine use with high accuracy. These functionally distinct dorsolateral prefrontal cortex-based circuits provide insights into the multiple roles played by the dorsolateral prefrontal cortex in cognitive and affective functioning that affects treatment outcome. The identified dorsolateral prefrontal cortex loci may serve as potential neuromodulation targets to be tested in subsequent clinical studies for addiction treatment and as clinically relevant biomarkers of its efficacy.

Cox Regression Based Modeling of Functional Connectivity and Treatment Outcome for Relapse Prediction and Disease Subtyping in Substance Use Disorder

“…The prediction modeling we proposed here will grant the final results of (1) a specific ROI (e.g., dlPFC) that is identified predictive of certain behavior (e.g., cocaine relapse), with prediction accuracy evaluated with the AUC value of the ROC curve; and (2) a set of protective circuits and risk circuits that are underlying the prediction. In our previous investigation on the dlPFC ROIs across the entire surface area of bilateral dlPFC, three dlPFC loci were identified significantly predictive of cocaine relapse with their corresponding protective and risk functional circuits ( Zhai et al, 2021 ). Here we choose the predictive ROI on the left dlPFC to demonstrate the anticipated results of our prediction modeling pipeline ( Figure 3 ).…”

“…Summary of the demographic information of the cohort is presented in Table 1 ( n = 43 after excluding two participants for excessive head motion during fMRI scanning, see section “Computational Pipeline of Relapse Prediction” in the section “Methods”). More detailed information on participants’ inclusion/exclusion, treatment/assessment procedures have been described previously ( Zhai et al, 2021 ).…”

“… Schematic diagram of analytical pipeline. Illustration of our analytical procedures using one exemplar seed ROI, adapted from Zhai et al (2021) . First, an ROI is selected as a seed (A) ; the whole-brain rsFC of this seed is calculated for each participant (B) ; a voxel-wise Cox regression is conducted using rsFC and days until relapse to generate beta maps (C) ; beta maps of Cox regression is thresholded (D) ; generation of indexP and indexR by linearly summation of the rsFC values within the thresholded beta maps for the negative and positive beta voxels, respectively (E) ; construction of the final prediction model by fitting indexP and indexR into the Cox model (F) ; procedures B to F were organized in a nested cross-validation loop, and after each participant is left out once, an ROC analysis evaluates the final prediction model (G) .…”

“…The dlPFC has been utilized as stimulation target for high frequency rTMS treatment that reduced craving for nicotine and cocaine ( Politi et al, 2008 ; Li et al, 2013 ; Pripfl et al, 2014 ). Here, we chose an ROI on the left dlPFC for demonstration due to its promising role in the treatment of cocaine dependence as stimulation target of high frequency rTMS ( Terraneo et al, 2016 ), as well as the high validity in predicting cocaine relapse with its downstream functional circuits ( Zhai et al, 2021 ). Whole-brain functional connectivity maps were obtained by calculating the cross-correlation (CC) between the time series of the seed and that of each voxel in the whole-brain ( 3dDeconvolve, AFNI ).…”

“…Since the initial demonstration of blood-oxygen-level-dependent (BOLD) signal in vivo in early 1990s ( Ogawa et al, 1990 ), functional magnetic resonance imaging (fMRI) has become a noninvasive neuroimaging technique widely used in research of cognitive neurosciences, as well as in understanding of neural mechanisms of neuropsychiatric/neurological diseases. For example, in exploring neurobiological mechanisms of substance use disorder, a highly relapsing chronic brain disease ( Dutra et al, 2008 ; Koob and Volkow, 2016 ; Volkow et al, 2016 ) currently without effective treatments, imaging biomarkers based on both resting-state fMRI ( Zhai et al, 2021 ) and task-evoked fMRI ( Luo et al, 2013 ) have shown prediction validity of relapse to cocaine use following treatment. Such neuroimaging-based studies provide mechanistic understanding of relapse to drug use and suggest neural targets for the development of neuromodulation (e.g., transcranial magnetic stimulation/TMS) treatment protocols.…”