The impact of APOE ε4 on mild cognitive impairment (MCI) and its progression to dementia remain controversial. We aimed to examine the association of APOE ε4 with MCI, and to verify the hypothesis that ε4 accelerates progression from MCI to dementia. In the Kungsholmen project, 756 cognitively healthy participants and 212 people with MCI aged ≥75 years were identified at baseline. Amnestic MCI (aMCI) and other cognitive impairment no dementia (oCIND) as two subtypes of MCI were assessed based on standard definitions. The two cohorts were followed for 9 years to detect incident cases of MCI and dementia following international criteria. APOE genotypes were assessed at baseline. Data were analyzed using Cox models. During the follow-up, in the cognitively healthy cohort, 165 people developed MCI (40 aMCI and 125 oCIND) and 176 developed dementia; in the MCI cohort, 118 persons progressed to dementia. Compared with APOE ε3ε3, the hazard ratios (HRs) (95% CIs) of ε2ε4/ε3ε4 were 2.24 (1.10-4.57) for aMCI and 1.78 (1.15-2.75) for oCIND, while the ε4ε4 was related to dementia with a HR of 4.35 (1.97-9.63) in the cognitively healthy cohort. In the MCI cohort, the ε4ε4 genotype led to a multi-adjusted HR of 2.89 (1.12-7.48) for dementia and accelerated the progression to dementia by 3.36 years. The APOE ε4 heterozygotes are associated with an increased risk of aMCI and oCIND. The ε4 homozygote substantially accelerates progression from MCI to dementia, and anticipate dementia occurrence by more than 3 years in people with MCI.