Background: Despite an urgent need for multi-domain lifestyle interventions to reduce dementia risk there is a lack of interventions which are informed by theory and evidence based behaviour change strategies and no interventions in this domain have investigated the feasibility or effectiveness of behaviour change maintenance. Objectives: We tested the feasibility, acceptability, and cognitive effects of MedEx-UK, a personalised theory-based 24-week intervention to improve Mediterranean-diet (MD) adherence alone, or in combination with physical activity (PA), in older-adults at risk of dementia, defined using a cardiovascular risk-score of >10%. Design: 104 participants (74% female, 57 to 76 years) were randomised to three parallel intervention arms: 1) control, 2) MD, or 3) MD+PA for 24-weeks and invited to an additional 24-week follow-up period with no active intervention. Behaviour change was supported using personalised targets, a web-based intervention, group sessions and food provision. Results: The intervention was feasible and acceptable with the intended number of at risk participants completing the study. Participant engagement with group sessions and food provision components was high. There was improved MD adherence in the two MD groups compared with control at 24-weeks (3.7 points on a 14-point scale (95% CI 2.9, 4.5, p <0.01) and 48-weeks (2.7 points (95% CI 1.6, 3.7) p<0.01). Intervention did not change objectively measured PA. Improvements in general cognition (0.22 (95% CI 0.05, 0.35, p =0.01), memory (0.31 (95% CI 0.10, 0.51, p <0.01), and select cardiovascular outcomes captured as underpinning physiological mechanisms were observed in the MD groups at 24-weeks. Conclusion: The intervention was successful in initiating and maintaining dietary behaviour change for up to 12-months, which was likely due to the intense early support provided which resulted in cognitive benefits. These results will inform larger-scale complex behaviour change interventions with health and well-being endpoints.