Objectives: Several epidemiological studies suggest declining trends in dementia over the last three decades with both decreasing age-specific prevalence and incidence. There is limited data on whether this delayed clinical onset is accompanied by a shorter postdiagnosis survival. Methods: A total of 5,205 participants from the Framingham Original and Offspring cohorts were studied. Four epochs were considered from 1977-1984 to 2004-2008. Gender and education adjusted 5-year mortality risks were estimated using delayed entry Cox models with the earliest epoch as reference category. Stratified analyses by sex, education, and age were undertaken. A nested case control study of 317 dementia cases and 317 controls matched on age, gender and epoch was initiated. Results: In the whole sample, 5-year mortality risk has decreased with time, it was 33% lower in the last epoch compared to the earliest. In the 317 persons who developed dementia, age at onset increased (1.5 years/epoch), and years alive with dementia decreased (1 year/epoch) over time. We observed however, a decreased adjusted relative mortality risk (by 18%) in persons with dementia in 1986-1991 compared to 1977-1983 and no significant change from then to the latest epoch. The nested case control study suggested in matched controls that 5-year mortality relative risk had increased by 60% in the last epoch compared to Epoch 1. Discussion: In the FHS, in the last 30 years, disease duration in persons with dementia has decreased. However, age-adjusted mortality risk has slightly decreased after 1977-1983. Consequences of such trends on dementia prevalence should be investigated.< Réduire