AbstractObjectives: To refine mild cognitive impairment (MCI) diagnostic criteria, we examined progression to dementia using two approaches to identifying MCI. Methods: A total of 1203 Framingham Heart Study participants were classified at baseline as cognitively normal or MCI (overall and four MCI subtypes) via conventional Petersen/Winblad criteria (single cognitive test impaired per domain, >1.5 SD below expectations) or Jak/Bondi criteria (two tests impaired per domain, >1 SD below norms). Cox proportional hazards models were constructed to examine the association between each MCI definition and incident dementia. Results: The Petersen/Winblad criteria classified 34% of participants as having MCI while the Jak/Bondi criteria classified 24% as MCI. Over a mean follow-up of 9.7 years, 58 participants (5%) developed incident dementia. Both MCI criteria were associated with incident dementia [Petersen/Winblad: hazards ratio (HR) = 2.64; p-value=.0002; Jak/Bondi: HR=3.30; p-value <.0001]. When both MCI definitions were included in the same model, only the Jak/Bondi definition remained statistically significantly associated with incident dementia (HR=2.47; p-value=.008). Multi-domain amnestic and single domain non-amnestic MCI subtypes were significantly associated with incident dementia for both diagnostic approaches (all p-values <.01). Conclusions: The Jak/Bondi MCI criteria had a similar association with dementia as the conventional Petersen/Winblad MCI criteria, despite classifying ~30% fewer participants as having MCI. Further exploration of alternative methods to conventional MCI diagnostic criteria is warranted. (JINS, 2016, 22, 937–943)