INTRODUCTION: Recent Food and Drug Administration guidance endorses cognitive assessment as a possible primary endpoint for early trials for Alzheimer's disease but emphasizes the need for certainty regarding the relationship with progression to dementia. METHODS: We compared the validity of the 2-year change (Y0-Y2) of 11 markers of neuropsychological and functional abilities for the prediction of incident dementia over the following 3 years (Y2-Y5), in 860 subjects aged 70 years or older, who consulted for memory loss and were included in the "GuidAge" prevention trial. RESULTS: The Free and Cued Selective Reminding Test-Free Recall (FCSRT-FR) score showed the most predictive 2-year change (area under the curve = 0.72 95% confidence interval = 0.64;0.81). Changes in other subscores of the FCSRT, verbal fluencies tasks, and composite cognitive score were also significantly predictive. Conversely, 2-year change of Mini-Mental State Examination, Trail Making test (TMT)-A, TMT-B, Clinical Dementia Rating Sum of Boxes, and Instrumental Activities of Daily Living scores did not significantly predict occurrence of dementia. CONCLUSION: The FCSRT, the Fluency Task, and the composite cognitive score appear to be good cognitive markers of progression toward dementia in early prevention trials.Read less <