Stimulant misuse and dependence contribute substantially to the global burden of disease. A major component of Stimulant Use Disorder (SUD) is maladaptive decision-making, whereby individuals persist in risky choices that are harmful to themselves and those around them. Decision-making is a complex process that requires an individual to determine which options are worth pursuing. To that aim, the values of possible rewards and costs must be calculated and compared. As most decisions encountered in everyday life present incomplete information about the outcomes of possible choices, individuals have to operate on this uncertainty, which can cause distortions in choice. Individuals with SUD may be especially susceptible to choice biases related to ambiguity, which arouses negative feelings that can motivate drug use. However, the neural basis of how uncertainty influences decision-making in SUD remains unknown. The goal of the studies presented in this dissertation was to address this question by combining neuroimaging with computational modeling of decision-making tasks. Studies were performed to compare the behavior and neural function of healthy control participants and those who chronically used stimulants on four decision-making tasks paired with brain imaging. The Balloon Analogue Risk Task (BART), a naturalistic decision-making task, was performed by participants with Methamphetamine Use Disorder (MUD) during functional magnetic resonance imaging (fMRI). A different subset of participants who performed the BART underwent positron emission tomography (PET) for estimation of dopamine D2-type (D2 and D3) receptor binding potential (BPND). Participants with SUD (cocaine and methamphetamine) performed the Loss Aversion Task (LAT) and also received PET scans. Some participants with SUD also performed the Choice under Risk and Ambiguity (CRA) task to compare aversion to risk (known outcome probabilities) and ambiguity (unknown outcome probabilities), and received fMRI scans for assessment of resting-state functional connectivity (RSFC). Lastly, a delay discounting task (DDT) was performed by participants who performed the CRA task to determine the contribution of risk and ambiguity to intertemporal choice.
The first two chapters present background and methods. Chapter 1 provides a general overview of the neurobiology of addiction and decision-making, with a focus on value computation and how it can be biased. Chapter 2 outlines the tasks, computational models, estimation procedures, and behavioral statistics used in the studies. Brain imaging methods are described in the chapter in which they were used.
Chapter 3 presents the results of using a cognitive model to decompose performance on the BART and associate the resulting parameters with neural function. We found a marked impairment in behavioral updating and adaptive risk-taking in participants with MUD. Risk-taking was negatively correlated with dopamine D2-type BPND in the striatum and midbrain only in healthy control participants, who also showed nonlinear associations between updating rate and dopamine D2-type BPND in the insula and medial OFC. No significant relationships between behavioral parameters and dopamine D2-type BPND were exhibited by MUD participants. However, behavioral updating was correlated with modulation of activation by risk in the dorsolateral prefrontal cortex in both groups, and in the anterior insula only in MUD participants. These findings linked cortical activity and D2-type binding potential to updating behavior during advantageous risk-taking in healthy control participants. In MUD, impairments in adaptive risk-taking and behavioral updating and their lack of association with striatal or cortical dopamine D2-type BPND suggest D2-type receptor-related deficits in accurately updating behavior under uncertain conditions.
Chapter 4 presents two studies related to loss aversion. The first tested the hypothesis that prefrontal cortical thickness would mediate age-related changes in loss aversion in healthy control participants. The relationship between age and loss aversion followed a quadratic function that was mediated by thickness of the posterior cingulate cortex. The U-shaped function reached a minimum around age 35 before increasing across middle-age, following the developmental trajectory of the cortex and suggesting that thinning of the posterior cingulate cortex may emerge as a contributing factor to loss aversion only once cortical thinning is underway. The second study tested whether loss aversion differed between healthy control and SUD participants and was related to striatal or amygdala D2-type binding potential. In SUD but not in healthy control participants, loss aversion and risk tolerance were positively related to striatal D2-type BPND, establishing a role for D2-type signaling in risky decisions involving loss in SUD, perhaps through neuroadaptations related to drug use.
Chapter 5 introduces ambiguity aversion as an important yet underexplored factor in SUD. Participants with SUD who were in inpatient treatment were more extreme in their aversion to ambiguity but not to risk, and ambiguity aversion was associated with stimulant (methamphetamine or cocaine) use in the 30 days prior to entering treatment. Ambiguity aversion in SUD participants was correlated positively with cortico-amygdalar RSFC and negatively with frontostriatal RSFC. To obtain an accurate assessment of differences in delay discounting between SUD and healthy control participants, a DDT was given in tandem with the CRA task. Group differences in delay discounting were eliminated when accounting for risk aversion. Further, ambiguity aversion and delay discounting showed a correlation that disappeared when risk tolerance was taken into account. These findings suggest that ambiguity aversion—and not the desire for immediate gratification—underlies delay discounting in SUD and that ambiguity aversion is related to frontostriatal function and cortico-amygdalar connectivity.
Taken together, these studies suggest that people with SUD have an impairment in advantageous risk-taking under uncertainty, perhaps due to a difficulty estimating ambiguous risk and an exaggerated response to ambiguity. Our methods demonstrate the advantages of pairing a visceral, naturalistic risk-taking task with computational modeling and economic choice tasks. In combination with brain imaging, these methods can clarify the neural substrates of complex behavior, including those that drive maladaptive choices in addiction. Characterizing the behavior of individuals who use drugs provides a stronger foundation for therapeutic strategies that address decision-making, for instance by introducing ambiguity aversion as a novel target. While clarifying suboptimal decision-making can refine policy toward addiction, understanding behavioral biases can be applied broadly to improve choices made in everyday life.
