BackgroundThe Lancet Commission has highlighted 12 modifiable risk factors for dementia. However, the effects of addressing these risk factors among the heterogeneous older adult population is not fully understood. We compared the association between risk factors and dementia by frailty, conceptualized as an explanation for the underlying health heterogeneity in old age. MethodsData were from the UK Biobank, a cohort study with over 500,000 participants aged 37–73 between 2006 and 2010. Frailty was measured by five criteria: slowness, weakness, exhaustion, inactivity, and shrinking. Participants meeting 0, 1–2, and 3–5 criteria were considered non-frail, prefrail, and frail, respectively. We included 13 modifiable risk factors. We used logistic regression to determine the associations of risk factors with 10-year dementia among non-frail, pre-frail, and frail individuals, respectively. Additionally, we adopted a g-computation method to estimate the individual and combined population intervention effects (PIE) of the risk factors for dementia. ResultsOf 381,419 participants, 58.4 %, 38.2 %, and 3.4 % were classified as non-frail, pre-frail, and frail, respectively. Except for smoking, depression, and excessive alcohol use, the other 10 risk factors had a stronger association with dementia among frailer individuals. We observed the highest PIEs among frail individuals when considering hypothetical interventions targeting low education, physical inactivity, central obesity, social isolation, hearing impairment, hypertension, diabetes, high nitrogen dioxide (NO2) exposure, high exposure of particulate matter with a diameter less than or equal to 2.5 micrometers (PM2.5), and traumatic brain injury individually. For the hypothetical interventions targeting all 13 risk factors together, we found a graded increase in the PIE across frailty status. ConclusionThe associations between modifiable risk factors and dementia were stronger among the frail. We advocate for incorporating frailty assessments to pinpoint those most likely to benefit from targeted risk factor interventions.