Atrial fibrillation (AF) independently increases dementia risk, but whether accelerometer-measured physical activity (PA) modifies this association remains unquantified, particularly against self-reported PA limitations.

Prospective analysis of 91,795 UK Biobank participants with valid accelerometer data (median age 57, 42.9 % male; 2800 with baseline AF). We categorized whether measured activity met the standard recommendation [moderate-to-vigorous physical activity (MVPA) >_150 min/week]. Questionnaire-derived MVPA data from 353,643 UK Biobank participants (median age 57, male: 46.7 %) between 2006 and 2010 were used for validation. The primary outcome was the diagnosis of incident all-cause dementia. We also assessed correlation between accelerometer-derived and self-reported activity.

Over 7.6-year median follow-up, AF was significantly associated with a higher dementia risk [Adjusted hazard ratio (aHR): 1.76, 95 % confidential interval (CI): 1.51–2.05]. Guideline-adherent PA was associated with a lower AF-related dementia risk to non-significance (aHR: 1.36, 95 %CI: 0.96–1.91). Moreover, PA may be associated with higher protection effect on dementia risk in AF patients (aHR: 0.55, 95 % CI: 0.33–0.92) than in non-AF (aHR: 0.81, 95 % CI: 0.69–0.96), although without statistical difference (Pinteraction = 0.213). Correlation between accelerometer-derived and selfreported MVPA was weak (Spearman r = 0.155, 95 % CI: 0.148–0.162). Self-reported activity was not associated with a decreased risk of dementia in both AF and non-AF participants.

Higher accelerometer-measured PA is associated with lower AF-associated dementia risk. Future prospective studies with extended follow-up and serial activity monitoring are needed to confirm these findings.