Largely Intact But Less Reliable and Distributed Neural Representations of Subjective Value in Human Opioid Addiction

Abstract

Addiction, particularly opioid use disorder (OUD), is often characterized by heightened propensity for risk-taking. While tolerance for risk and uncertainty varies across individuals, the elevated risk-taking in people with OUD is assumed to stem from altered cognitive decision-making processes beyond differences due to idiosyncratic yet lawful tolerances. Specifically, the prevailing assumption is that people with addiction exhibit impairments in the internal representation and integration of information that should guide decisions and judgments about what is valuable. Using model-based functional magnetic resonance imaging, we examined how the choice behavior of treatment-engaged male and female participants with chronic OUD aligns with the neural encoding of their inferred subjective value (SV) of uncertain (risky and ambiguous) rewards and the evidence for impairment in this neural process. Using both univariate and multivariate analyses, we found that canonical value regions [ventromedial prefrontal cortex (vmPFC), striatum, and posterior cingulate cortex] track the SV of uncertain choice options in both participants with OUD and comparison controls, irrespective of their tolerance for uncertainty. This speaks against a fundamentally impaired subjective valuation process in OUD. However, value representations were less reliably decodable in people with OUD in some value regions (vmPFC) and throughout the brain, especially within the limbic and salience/ventral-attention networks. Thus, while people with OUD engage a neurocomputationally similar process during risky decision-making to controls, they may differ in the fidelity and distribution of SV signals across brain networks.