Among 465,175 UK Biobank participants (mean age 56.5), latent class analysis identified five distinct chronic-disease clusters all significantly associated with elevated dementia risk. The cardiometabolic cluster carried the highest hazard (HR = 2.14), followed closely by the mental health cluster (HR = 1.99). Crucially, moderate physical activity reduced dementia risk only within the pain-dominated multimorbidity cluster (HR = 0.77), with no significant lifestyle moderation detected in the other four groups.
This large-scale analysis challenges the broadly promoted assumption that lifestyle interventions uniformly reduce dementia risk across disease profiles. The finding that physical activity's protective signal is cluster-specific — appearing only in the pain-dominated group — suggests that the biological pathways linking chronic disease combinations to neurodegeneration are heterogeneous and may not be equally amenable to behavioral modification. The cardiometabolic-dementia link aligns with decades of vascular dementia research, but the near-equivalent hazard in the mental health cluster is notable and underexplored in intervention design.
Key limitations include reliance on self-reported lifestyle data, the UK Biobank's well-documented healthy-volunteer bias, and an observational design that cannot establish causality. The analysis also cannot rule out reverse causation — early neurodegeneration may itself reshape disease clustering. As a preprint posted on medRxiv and not yet peer-reviewed, these findings require independent validation before informing clinical guidelines. Nonetheless, the cluster-specific moderation finding is a methodologically sound incremental advance that argues for precision-medicine approaches to dementia prevention rather than one-size-fits-all lifestyle prescriptions.