Neurocomputational Training to Improve Reward Responsiveness in Veterans With Anhedonia

Anhedonia, a reduction in motivation and in the ability to feel positive emotions, is widespread among Veterans across a range of psychiatric disorders. This impairment is associated with poorer psychosocial functioning and is resistant to first-line medication and psychotherapy treatments. However, interventions to improve anhedonia are sorely lacking. Reward learning supports motivation to seek rewards, and this mechanism can be enhanced by engaging in reward exploration. Our preliminary findings suggest that this approach can help improve reward responsiveness and positive mood among depressed and anxious individuals. Thus, the proposed research will test the preliminary efficacy of a computerized reward learning training in improving reward sensitivity and reducing anhedonia in Veterans. This new intervention could offer a novel remotely accessible treatment and help enhance the efficacy of existing evidence-based treatments for depression and other psychiatric conditions in Veterans.

Anhedonia, which is characterized by reduced reward responsiveness and blunted positive affect, is widespread among Veterans across a range of psychiatric conditions, including depression and PTSD. Anhedonia manifests as reduced exploration of rewards in a range of domains, including social connectedness and workforce engagement, and is associated with poorer psychosocial functioning and reduced response to psychiatric treatment. Despite the prominence of anhedonia and these associated deficits, the treatment of impaired reward responsiveness and underlying mechanisms has received relatively little attention compared to punishment sensitivity and negative affect, particularly in Veteran populations. Moreover, pharmacological and behavioral treatments for depression have shown mixed results for anhedonia. Thus, there is a significant need for mechanistically informed interventions targeting anhedonia in Veterans.

Recent advances in computational psychiatry have shown that reward learning is a key top-down mechanism of reward exploration and foraging, which could be leveraged to develop cognitive training interventions for anhedonia. For instance, Bayesian reinforcement learning models have been particularly useful in disentangling various aspects of reward responsiveness, including reward prediction updating (learning) and reward maximization strategy. Within this framework, subjective anhedonia, has been robustly linked to lower learning-based reward maximization in the multi-arm bandit (MAB), a reward exploration task in which one must sample from various unknown options with the goal of maximizing rewards. Critically, a recent randomized clinical trial (RCT) from our group suggests that a MAB experimentally tuned to provide high variance of reward delivery can increase reward learning and associated reward maximization among clinically depressed and anxious individuals with anhedonia.

This proposal will capitalize on this approach to test the effectiveness of a novel neurocomputational cognitive training aimed at boosting reward sensitivity in Veterans with anhedonia. The specific aims are to : 1) examine the degree to which a reward learning training protocol (reward learning intensive training/ReLIT) changes the neurocomputations supporting reward sensitivity in Veterans with anhedonia (Aim 1); 2) test the effectiveness and optimal dosage of ReLIT in improving subjective markers of anhedonia, depression, and psychosocial functioning (Aim2), and 3) assess the degree to which the effects of ReLIT generalize to parallel measures of automatic affective processing (Aim 3).

In a 3-arm double-blind RCT, Veterans (N=120) with moderate to high anhedonia will be randomized to an 8-week protocol of: 1) sham training (non-active), 2) active ReLIT of moderate intensity, or 3) active ReLIT of high intensity. At baseline (pre-treatment), 8-weeks (post-treatment), and 16-weeks follow-up (8 weeks post-treatment), participants will complete an assessment MAB and 2 automatic affective processing tasks (while undergoing functional MRI for a subset of 75 participants). Bayesian computational modeling will be used to derive individual parameters of reward learning and reward maximization and to extract neural activity associated with dynamic updating of reward predictions. Clinical and affective measures will be collected at all 3 assessments and each treatment session.

This proposal aims to assess the preliminary efficacy of a novel intervention for improving healthy reward responsiveness and anhedonic pathology in Veterans, which is currently an underrepresented area of mental health treatment for Veterans. This mechanistically informed intervention could a) provide a scalable and cost-effective treatment for Veterans with anhedonia, relying on technology interface rather than clinician-guided administration, b) be used as an adjunctive training to boost the effectiveness of existing evidence-based treatments for Veterans, such as cognitive behavioral therapy.