Cognitive Endpoint

A cognitive endpoint is a measurement instrument with a defined administration procedure, a bounded score range and a stated direction of benefit, nominated in the protocol before the data exist. It may be a single-domain task such as delayed recall, or a composite summing subtests across several domains. Its usefulness is psychometric: test-retest reliability, sensitivity to change in the population studied, and whether that population sits near a ceiling or floor where the scale cannot move.

The instruments used in dementia trials are well characterised. The cognitive subscale of the Alzheimer's Disease Assessment Scale runs to seventy points with higher scores worse, the Mini-Mental State Examination has thirty, and the Clinical Dementia Rating sum of boxes has eighteen and mixes cognition with function. Regulators have generally expected cognition and function together rather than a cognitive score alone. Recent anti-amyloid antibody trials reported group differences on these scales of well under one point, statistically robust in large samples and small in absolute terms.

An endpoint is therefore uninterpretable without two further numbers: the placebo group's drift over the same period, and a minimal clinically important difference established independently. A significant difference on a multi-point scale can be a fraction of what a clinician or family could detect.

Where this goes wrong is small studies with large batteries. A dozen subtests give a dozen chances at significance, and reporting the one that moved without multiplicity correction is the standard failure in nootropic research. Healthy-volunteer studies compound it, because intact adults sit near the ceiling of most clinical instruments and improve on retesting through practice alone, which an uncontrolled design records as benefit.