METABOLIC SIGNALS SHAPE LEARNING AND DECISION-MAKING

To support long-term homeostasis, reward-related behavior is tuned according to the metabolic state of the body. Thus, altered metabolic signaling may contribute to mental and metabolic disorders. Despite evidence in animals, the influence of metabolic state over reward-related learning in humans is unproven. Moreover, the effect of differences in metabolic traits (e.g., glycemic stability) that may alter this relationship is not yet clear, although they could offer mechanistic insight into disorders.
To close this gap, we collected observational and interventional data concerning metabolic scaling of reinforcement learning in a sample (N=84, M_age=24±3.5) with a wide range of BMI (18–36 kg/m²). First, participants completed up to 60 runs (4 weeks) of our App, Influenca, assessing reinforcement learning and ecological momentary assessment of metabolic state. At the same time, we continuously monitored glucose levels (every 5 min) to differentiate ‘objective’ and self-reported metabolic state. Using Bayesian mixed-models, we found that wins and reward sensitivity for money were higher when hungry (wins: 95%CI=[0.001,1.28]) and with low glucose levels (reward sensitivity: 95%CI=[-0.47,-0.07]). Second, in an experimental session, we manipulated metabolic state via a caloric load (milkshake) vs. water, finding alterations in reward sensitivity and neural responses depending on trait glycemic stability (p=0.019).
We show that caloric intake can affect how we learn and decide. Crucially, metabolism plays a role in scaling reward-related signals according to energy intake. Ultimately, inter-individual differences in the integration of metabolic state related to differences in metabolic traits will provide insights into adaptive and maladaptive behavioral changes.