Del Pozo Cruz · JAMA neurology 2022 · prospective cohort study · n=78430

Association of Daily Step Count and Intensity With Incident Dementia in 78 430 Adults Living in the UK.

Cited 184 times in the scientific literature.

Level 3 - non-randomized controlled study

Prospective cohort study with registry-based follow-up

PubMed 36066874 · doi:10.1001/jamaneurol.2022.2672 · record verified 2026-08-28

What was done

Data from 78,430 UK Biobank participants aged 40 to 79 years who wore a wrist accelerometer between 2013 and 2015 were analyzed. Daily step count, incidental steps (<40 steps/min), purposeful steps (≥40 steps/min), and peak 30-minute cadence were derived from accelerometer data. Incident all-cause dementia was tracked via hospital records, primary care data, and death registers over a median follow-up of 6.9 years. Spline Cox proportional hazards models were used to evaluate dose-response relationships.

What was found

Over a median of 6.9 years, 866 participants developed dementia. A non-linear dose-response was observed: the optimal dose for total steps was 9,826 steps/day (hazard ratio [HR], 0.49; 95% CI, 0.39-0.62), while a minimal dose of 3,826 steps/day yielded half the maximum benefit (HR, 0.75; 95% CI, 0.67-0.83). For walking intensity, the optimal dose was 3,677 steps/day for incidental walking (HR, 0.58; 95% CI, 0.44-0.72), 6,315 steps/day for purposeful walking (HR, 0.43; 95% CI, 0.32-0.58), and a peak 30-minute cadence of 112 steps/minute (HR, 0.38; 95% CI, 0.24-0.60).

Why it matters

This study provides specific, objective daily step targets (~9,800 total steps and brisk cadence) for dementia risk reduction, demonstrating that even lower volumes (~3,800 steps) offer measurable protective associations.

Limits

The study is observational, leaving open the possibility of reverse causation (early undiagnosed cognitive decline reducing physical activity) and unmeasured confounding. The cohort is predominantly White (96.7%), limiting generalizability to other populations. Accelerometer wear was captured at a single baseline window rather than tracked longitudinally.

Cited by