Among 151,832 UK Biobank participants aged 60–69, six distinct clusters of co-occurring modifiable risk factors were identified using latent class analysis, all associated with elevated dementia incidence over a median 13-year follow-up during which 5,304 cases developed. The highest-risk cluster — characterised by multiple simultaneous risk factors — carried a hazard ratio of 2.38 (95% CI 2.11–2.68) versus those with zero or one risk factor. Even milder clusters conferred meaningful risk: hearing loss combined with unhealthy lifestyle (HR 1.37), sleep and psychosocial factors (HR 1.34), poor cardiometabolic health (HR 1.32), heavy drinking with unhealthy lifestyle (HR 1.20), and vitamin D deficiency with sedentary and psychosocial factors (HR 1.18).
What distinguishes this work is its cluster-level framing. Prior dementia-prevention literature — including the Lancet Commission's 14-factor model — treats risk factors additively and independently. This analysis shows they naturally co-occur in patterned ways that may warrant targeted, multi-domain interventions rather than single-factor approaches. The practical implication is that adults harbouring combinations like untreated hearing loss, poor sleep, social isolation, or cardiometabolic dysfunction face compounding risk that dwarfs any single factor alone. Limitations include the observational design (causality cannot be confirmed), UK Biobank's healthier-than-average sampling bias, and reliance on hospital records for dementia ascertainment. Being a preprint not yet peer-reviewed, the cluster assignments and effect sizes should be treated as preliminary. Still, the multi-domain prevention message is broadly consistent with existing evidence and clinically actionable.