The Lancet Commission has recommended 14 modifiable risk factors for dementia, but their cumulative impacts and potential socioeconomic disparities remain underexplored. We aimed to investigate the association of individual risk factors and their weighted composite risk score with incident dementia in different socioeconomic status (SES).

This prospective cohort study included 344,793 participants without dementia at baseline (2006∼2010) from the UK Biobank. A weighted composite dementia risk score (0–14) was developed by assigning distinct weights to binary risk factors using log (relative risk) × (1 − communality) from the Lancet study. Cox models and restricted cubic splines were used to examine associations of interest across SES tertiles.

During a median 14.1-year follow-up, 5,515 individuals were diagnosed with dementia. Some risk factors showed SES-specific associations, e.g., hypertension was linked to dementia only in the low-SES group. The population attributable fraction (PAF) for all 14 risk factors was 40.70% (95% CI: 35.81%, 45.23%), with less education accounting for the largest preventable fraction in low- and medium-SES groups (24.6% and 18.3%) and smoking in the high-SES group (5.7%). Each 1-point increase in the composite risk score corresponded to a 19% (95% CI: 17%, 21%) higher risk of dementia, especially in early-onset cases, individuals younger at baseline, APOE ε 4 non-carriers, and low-SES population.

Our findings highlight that a substantial proportion of dementia cases may be preventable by addressing modifiable risk factors, with potential effects varying by SES. Targeted modifiable interventions, especially in disadvantaged populations, are substantial to reducing dementia burden and socioeconomic inequalities.