Among 779 Japanese adults aged 65 and older with diabetes, those scoring 8 or higher on a cognitive activity scale (reading, games, hobbies) had a 38% lower hazard of incident dementia over five years than those scoring 0-7 (HR 0.615, 95% CI 0.414-0.913). Of the 112 participants who developed dementia, the association held after adjusting for moderate-to-vigorous physical activity (HR 0.624) and when death was treated as a competing event (subdistribution HR 0.667).
Diabetes roughly doubles dementia risk, so identifying low-cost behaviours that offset it matters. The result fits the cognitive reserve hypothesis and echoes broader cohorts linking mentally stimulating activity to slower decline, but this paper narrows the question to a high-risk group where evidence has been thin. The robustness to physical activity adjustment is a real strength, since exercise is the usual confounder in lifestyle-dementia work. Still, caution is warranted. The design is observational, and reverse causation is plausible: people in the earliest, undiagnosed stages of dementia often drop reading and puzzles before any diagnosis, even with baseline cognition adjusted for. Dementia was ascertained from long-term care insurance records, which can miss milder cases. A binary split of a self-reported scale, a modest event count, and a single Japanese cohort limit generalisability. Randomised trials of cognitive training have shown mixed, domain-specific benefits. Editorially, this is confirmatory rather than paradigm-shifting: it supports keeping cognitively engaging hobbies alongside glucose control and exercise, without proving that taking them up reduces dementia.