OBJECTIVE To study the association of apoE genotypes with dementia and cerebrovascular disorders in a population based sample of 85 year old people. METHODS A representative sample of 85 year old people (303 non-demented, 109 demented) were given a neuropsychiatric and a medical examination and head CT. The apoE isoforms were determined. Dementia was diagnosed according to DSM-III-R. RESULTS At the age of 85, carriers of the apoE epsilon4 allele had an increased odds ratio (OR) for dementia (1.9; p<0.01) and its subtypes Alzheimer's disease (1.9; p<0.05) and vascular dementia (2.0; p<0.05). Among those categorised as having vascular dementia, the apoE epsilon4 allele was associated with mixed Alzheimer's disease-multi-infarct dementia (OR 6.5; p<0.05), but not with pure multi-infarct dementia (OR 1.5; NS). Only carriers of the apoE epsilon4 allele who also had ischaemic white matter lesions on CT of the head had an increased OR for dementia (OR 6.1; p=0.00003), and its main subtypes Alzheimer's disease (OR 6.8; p=0.002) and vascular dementia (OR 5.6; p=0.0007), whereas carriers of the apoE epsilon4 allele without white matter lesions had an OR for dementia of 1.0 (OR for Alzheimer's disease 1.8; NS and for vascular dementia 0.6; NS) and non-carriers of the apoE epsilon4 allele with white matter lesions had an OR for dementia of 2.2; NS (OR for Alzheimer's disease 2.7; NS and for vascular dementia 1.6; NS). The apoE allele variants were not related to mortality or incidence of dementia between the ages of 85 and 88. The epsilon2 allele was related to a higher prevalence of stroke or transient ischaemic attack at the age of 85 (OR 2.1; p<0.05) and a higher incidence of multi-infarct dementia during the follow up (OR 2.9; p<0.05). CONCLUSIONS Neither the apoE epsilon4 allele nor white matter lesions are sufficient risk factors by themselves for dementia at very old ages, whereas possession of both these entities increases the risk for Alzheimer's disease and vascular dementia substantially.