In 2020, the Lancet Commission identified 12 modifiable factors that increase population-level dementia risk. It is unclear if these risk factors co-occur among individuals in a clinically meaningful way. Using latent class analysis, we identified profiles of modifiable dementia risk factors in dementia-free adults aged 60-64 years from the UK Biobank. We then estimated associations between these profiles with incident dementia, cognition, and neuroimaging outcomes, and explored the differences across profiles in the effects of a polygenic risk score for Alzheimer's disease on outcomes. In 55,333 males and 63,063 females, three sex-specific latent profiles were identified: cardiometabolic risk, substance use-related risk, and low risk. The cardiometabolic risk profile in both males and females was associated with greater incidence of all-cause dementia (male: OR [95% CI] = 2.33 [2.03, 2.66]; female: OR [95% CI] = 1.44 [1.24, 1.68]), poorer cognitive performance, greater brain atrophy, and greater white matter hyperintensity volume compared to the low risk profile. The substance use-related risk profile in males was associated with poorer cognitive performance and greater white matter hyperintensities compared to the low risk profile, but no difference in all-cause dementia incidence was observed (OR [95% CI] = 1.00 [0.95, 1.06]). In females, the substance use-related risk profile demonstrated increased dementia incidence (OR [95% CI] = 1.58 [1.57, 1.58]) and greater brain atrophy but smaller white matter hyperintensity volume compared to the low risk profile. The polygenic risk score had larger effects among females, and differentially influenced outcomes across profiles; for instance, there were larger effects of the polygenic risk score on atrophy in the cardiometabolic profile vs. the low risk profile among males, and larger effects of the polygenic risk score on loss of white matter integrity in the cardiometabolic profile vs. the low risk profile among females. These results reveal three modifiable dementia risk profiles, their unique cognitive/neuroimaging outcomes, and their interactions with genetic risk for Alzheimer's disease. These differences highlight the need to consider population heterogeneity in risk prediction tools and in planning personalized prevention strategies.