Why cognition is hard to study, and easy to sell
Cognitive outcomes need validated instruments, long follow-up and large samples because effects are small and noisy. That makes real trials expensive and rare, and it makes rodent maze studies — cheap, fast and superficially impressive — the basis for most marketing.
The serious trial that came back negative
Observational data and animal work had suggested that GLP-1 receptor agonists might protect against neurodegeneration, and Novo Nordisk ran two large phase 3 trials (evoke and evoke+) of oral semaglutide in early Alzheimer's disease. In November 2025 the company reported that the trials did not meet their primary endpoint of slowing progression on the CDR-SB scale. This is a useful lesson: a plausible mechanism plus encouraging observational signals still failed the definitive test. We grade semaglutide unsupported for cognition on that basis, while noting that its metabolic and cardiovascular indications are unchanged.
Everything else
BPC-157, growth-hormone secretagogues and various "nootropic peptides" have no adequately powered human cognitive trials that we can find. Rodent behavioural results — forced swim, elevated plus maze, novel object recognition — do not translate reliably to human cognition, and we treat them as hypothesis-generating only.
What would change our grades
Registered, placebo-controlled trials with validated cognitive endpoints. We monitor ClinicalTrials.gov for peptide interventions with cognitive primary outcomes and will update profiles when results post.
