Computational phenotyping of effort-based decision-making in type-2 diabetes on and off semaglutide
Mehrhof SZ, Fleming H, Nord CL
Neuropsychopharmacology (2026) · n=228
In a pre-registered Cambridge MRC CBU study, adults with type-2 diabetes (N=112, on or off semaglutide) showed a blunted effort-acceptance bias on a computational decision-making task compared with matched controls (N=116) — a metabolic-cognitive shift toward energy conservation that was not restored by semaglutide treatment and was not driven by neuropsychiatric comorbidity or antidepressant use.
Mehrhof and colleagues at the MRC Cognition and Brain Sciences Unit (University of Cambridge) ran a pre-registered computational-psychiatry study examining whether motivation, indexed by effort-based decision-making parameters, differs in type-2 diabetes and whether semaglutide treatment alters those parameters. Effort-based decision-making is the cognitive process of trading off energetic cost for reward — long studied in animal motivation and linked in humans to dopaminergic signaling and to neuropsychiatric anhedonia.
The design compared four groups: type-2 diabetes participants on semaglutide (N=58), type-2 diabetes participants off semaglutide (N=54), and two matched control groups without diabetes (N=58 each). Participants completed a computational effort-reward task; researchers extracted the acceptance bias parameter — the bias to accept effort for reward — using a computational model.
Findings: participants with type-2 diabetes showed a blunted acceptance bias relative to matched controls. The effect was not driven by neuropsychiatric comorbidity or by antidepressant use. Across all participants, increasing diabetes risk linearly predicted reduced acceptance bias. Critically, participants with diabetes treated with semaglutide did not show restored motivation — the diabetes-associated motivational deficit was not normalised by semaglutide.
This is a cross-sectional comparison of pre-existing groups, not a randomised trial of semaglutide for motivation — participants were not randomised to on / off semaglutide, and any treatment-versus-no-treatment inference is confounded by the clinical reasons that selected one group into semaglutide therapy. The "semaglutide does not restore motivation" finding therefore cannot rule out reverse causation, selection effects, or insufficient duration of treatment.
The N per arm (~55–60) is modest for computational-psychiatry studies, which typically require larger samples to estimate small differences in latent decision-making parameters reliably. Effect sizes and confidence intervals are not in the abstract.
The result is in apparent tension with the Gill 2026 JAMA Psych RCT showing semaglutide improvements in effort-based decision-making in major depressive disorder. The discrepancy is likely explained by indication, baseline state, and design — MDD participants start with depression-driven effort deficits that may be more responsive to GLP-1 receptor agonism than the type-2-diabetes-only motivation deficits in this study, or by the difference between a randomised trial and a cross-sectional comparison.
Abstract-only extraction.
07·Member discussion
No member discussion yet.
Member-only conversation lives here — cycle notes, practitioner commentary, pattern-matching. Be the first paying member to start the thread.