Pooling CHARLS, ELSA and HRS (27,417 adults), CES-D items split into somatic symptoms, depressed affect and positive affect showed that each 1-point increase in somatic symptoms raised incident dementia odds by 43% (OR 1.43, 95% CI 1.32-1.55) and depressed affect by 31% (OR 1.31). With all three dimensions modeled together, only somatic symptoms stayed consistently significant in every cohort. Physical activity did not modify the associations and mediated just 5-11% of the effect in ELSA and HRS.
The dimensional split matters because total depression scores blur very different signals. Somatic items such as fatigue, poor sleep, appetite change and slowed movement overlap with early neurodegenerative changes, vascular disease and inflammation, so they may be a prodromal marker as much as a cause. Adjusting for baseline memory helps, but it cannot fully rule out reverse causation, and dementia was algorithmically defined from memory scores plus functional dependence rather than clinically diagnosed. The observational design also prevents causal claims, and the modest mediation figures are statistical, not proof that exercise interrupts the pathway. Practically, persistent fatigue, sleep disruption and physical slowing in midlife and later life deserve attention from clinicians, not just low mood. The finding fits prior work linking depression to dementia but sharpens it, and I rate it as incremental yet useful. It suggests risk-reduction efforts should target the bodily symptom cluster, though trials are needed to show that treating it lowers dementia incidence.