tVNS and Approach-Avoidance Behavior in Anhedonia and Anxiety

This study investigates if anhedonia and anxiety symptoms are associated with alterations in reinforcement learning, effort trade-offs for wins vs. punishments, and foraging behavior under threat. Moreover, it will investigate whether these processes can be influenced by a metabolic load and/or transcutaneous vagus nerve stimulation (tVNS). The project consists of (a) an online reinforcement learning study, used to characterize learning, reward sensitivity, and meta-cognition, and (b) a laboratory study in which participants first undergo fMRI while completing an effort-based decision-making task. Second, participants will complete two sessions in VR with randomized active or sham tVNS during a foraging task before and after a caloric load with concurrent physiological recordings.

The overarching goal of this project is to investigate the potential of modulating internal signals in patients with mood and anxiety disorders to improve the balance between approach and avoidance behavior. To this end, tVNS (vs. sham) will be used to alter foraging behavior under threat. To address the inherent heterogeneity of symptoms, it is planned to recruit participants with anxiety symptoms (ANX), anhedonia symptoms (ANH), both anxiety and mood symptoms (ANX+ANH), and healthy control participants (HCP). The groups will be matched for the group-average and distribution of age, sex, and BMI (age: 18 to 40 years, BMI 18.5 - 30 kg/m²). After an online assessment, diagnostic visit, and MRl-based phenotyping, internal signals are targeted with tVNS and a caloric load to shift the approach-avoidance behavior. In a randomized crossover study, participants will receive stimulation (tVNS or sham) in a hungry state (\>4 h after the last meal at a time when they would typically have their next meal) and complete a VR-based foraging task under threat. They will then receive a standardized caloric load (milkshake containing \~400 kcal) and repeat the task with the same stimulation in a different metabolic state (postprandial). Crucially, participants can move freely during the task so that behavioral adaptations in response to threats (e.g., escaping a panther) as well as physiological adaptations (e.g., heart rate) and their recovery can be measured. The study is split into three parts: 1. Online screening and behavioral characterization At least 250 participants (≥150 with clinically relevant symptoms; STICSA \> 43 and/or SHAPS \> 29.5) will complete baseline questionnaires (anhedonia (SHAPS, TEPS, and DARS), anxiety (STICSA and STAI), depression (BDI), substance abuse) and an approx. two weeks online reinforcement learning task (10 runs). Behavioral indices (choices, response times) and computational modeling parameters (e.g., learning rates, reward sensitivity, decision noise) will be derived. At the end of each run, participants provide metacognitive performance ratings, enabling assessment of deviations between subjective and objective performance. 2. Neural correlates of approach-avoidance (phenotyping) A subset of participants (high anxiety, high anhedonia, combined, and healthy controls (STICSA \< 40 and/or SHAPS \< 23.5); \~26 per group) will undergo laboratory-based testing. Assessments include diagnostic interviews, fasting blood samples (glucose, insulin, triglycerides, cortisol), and fMRI during an effort trade-off task comparing the effort to gain rewards and to avoid punishments. 3. tVNS intervention The same participants will then complete a randomized, sham-controlled crossover trial. Participants will perform a VR foraging task under threat in two metabolic states: hungry (\>4 h fasted) and postprandial (following a standardized caloric load \~400kcal). During both states, participants will receive either active or sham tVNS. Behavioral outcomes will be combined with physiological measures (e.g., heart rate) to assess the effects of vagal stimulation on approach-avoidance trade-offs. Hypotheses: 1. Patients with more severe anxiety symptoms are more sensitive to punishments. Thus, they spend less time foraging under threat and gain fewer rewards. In contrast, patients with more severe anhedonia symptoms are less sensitive to rewards. Thus, their overall foraging rate will be lower, leading to fewer rewards independent of threat. 2. Behavioral and self-rated differences in learning from rewards and punishments are reflected in altered brain responses when making effort-based choices to either approach rewards or avoid punishments. 3. A hungry state will reduce anxiety-like behavior and increase the approach of rewards. 4. tVNS will enhance the anxiolytic effects of hunger by increasing the weight on internal signals of metabolic demand, facilitating the approach of rewards. In contrast, it will facilitate avoidance in a postprandial state. 5. Inter-individual differences in the balance between motivational and threat-related circuits during the effort trade off task will predict tVNS effects when foraging. Enrollment: In the online assessment, at least 250 participants will be included to investigate differences in learning from wins and losses and how well behavioral shifts align with self-evaluation of performance. The sample size allows detecting even small effects that are likely in psychiatric research (r = .20) with a high power (1-β = 0.89). If necessary, further participants will be recruited for the online sample until the planned sample of N=104 participants has completed the tVNS intervention. For the phenotyping and the subsequent intervention study, 104 participants from 4 groups combining low/high anxiety and low/high anhedonia will be reinvited. With this transdiagnostic approach, the effects of anhedonia and anxiety on approach-avoid behavior can be disentangled. Crucially, oversampling participants with either high or low symptoms will maximize the expected effect size. The sample allows for evaluation of medium correlations (r = .30) between symptoms and approach-avoid trade-offs with a power of 1-β = 0.89. For the tVNS intervention, a sample of 104 participants allows for the conclusive study of medium-sized effects (dz = .40) observed in previous work (Neuser et al., 2020) with very high power (1-β = 0.98) across the sample. Differences in the tVNS response between participants (medium effect size, r = .30) can be evaluated with a power of 1-β = 0.89.