BACKGROUND: The links between diet and the risk of dementia have never been studied considering the possibility of protopathic bias (i.e., reverse causation). OBJECTIVE: We aimed to examine the relationship between consumption frequency of meat, fish, fruits, and vegetables and long-term risk of dementia and Alzheimer's disease (AD), by taking into account this possibility. METHODS: We analyzed data of 5,934 volunteers aged 65 and over from the Three-city study who were followed every 2 to 4 years for 12 years. Dietary habits were assessed at inclusion using a brief food frequency questionnaire. The presence of symptoms of dementia was investigated at each follow-up visit. To limit the risk of protopathic bias, a 4-year lag window between exposure and disease assessment was implemented by excluding from the analyses all dementia cases that occurred during the first four years after inclusion. Analyses were performed using a Cox proportional hazard model and were adjusted for socio-demographic, lifestyle, and health factors. RESULTS: The average follow-up time was 9.8 years. During this period, 662 cases of dementia, including 466 of AD, were identified. After adjustment, only low meat consumption (</=1 time/week) was associated with an increased risk of dementia and AD compared with regular consumption (>/=4 times/week) (HR = 1.58 95% CI = [1.17-2.14], HR = 1.67 95% CI = [1.18-2.37], respectively). No association was found between the consumption of fish, raw fruits, or cooked fruits and vegetables and the risk of dementia or AD. CONCLUSION: These findings suggest very low meat consumption increases the long-term risk of dementia and AD, and that a protopathic bias could have impacted finding from previous studies.< Réduire