Public Health

Research into an effective dementia treatment is ongoing. Therefore, identifying individuals at risk of declining cognition and dementia is fundamental for initiating modifiable risk factor interventions that can delay dementia onset. Research into modifiable risk factors has almost exclusively been from high-income countries, despite 60% of individuals with dementia living in low- and middle-income countries (LMICs). Addressing this research inequality, the current study examines cross-sectional relationships between risk factors and cognitive performance in LMICs, with the aim of identifying modifiable risk factors particularly suitable for interventions in these regions.

Data were obtained from 15 members of the Cohort Studies of Memory in an International Consortium (COSMIC), representing 11 countries (Brazil, China, Colombia, Cuba, India, Indonesia, Malaysia, Nigeria, Republic of Congo, Tanzania, & Uganda) across 6 continents (participants: 53,136; M_age  = 70.75, SD_age  = 7.87; 57% female). We investigated (after harmonisation) the following risk factors: age, APOE ε4, anxiety, body mass index (BMI), blood pressure, cardiovascular disease, cholesterol, depression, diabetes, education, excessive alcohol, hypertension, hearing loss, physical activity, sex, smoking status, and stroke history. Harmonised global cognition was the outcome. On the AD Workbench, we performed linear regressions for each risk factor within each study at baseline. Cross-sectional results from all studies were pooled in a multivariate meta-analysis, with a random intercept for Country and study. We investigated age, sex and education as risk factors, and included them as covariates when analysing other factors.

Older age (β=-.231, p < .001), being male (β=.151, p < .001), less education (β=1.569, p < .001), history of angina (β=-.099, p < .001), anxiety (β=-.262, p < .001), depression (β=-.203, p < .001), stroke (β=-.375, p < .001), diabetes (β=-.060, p = .002), excessive alcohol consumption (β=-.190, p < .001), currently smoking (β=-.097, p < .001), and less physical activity (β=.160, p < .001) were significantly associated with worse global cognition. Heterogeneity in the relationship between risk factors and cognition exists across countries, warranting further analyses for each LMIC.

Findings suggest that risk factor interventions may help to reduce dementia in LMICs, though country level heterogeneity in the effects suggests that tailoring interventions to regions will be needed. Future research will explore how the factors we identified predict incidence of dementia in LMICs using Machine Learning models.