Multidomain lifestyle interventions for the risk reduction of cognitive impairment and dementia: a systematic review and meta-analysis

Background: Multidomain lifestyle interventions targeting multiple risk factors have been proposed to reduce cognitive impairment and dementia risk. However, mixed findings hamper their application. More evidence is needed to optimise approaches in different settings. The increasing number of clinical trials being conducted warrants an up-to-date synthesis. Methods: We conducted a systematic review and meta-analysis of randomised controlled trials (RCTs) testing multidomain interventions (three or more components) on cognition or dementia incidence. Maximum-likelihood random-effect models were applied. Sensitivity analyses were conducted to explore source of heterogeneity Meta-regression analyses were conducted, including by intervention duration and intensity. Risk of bias on cognition was assessed using the revised Cochrane risk-of-bias tool for RCTs (RoB-2) Heterogeneity was estimated using Chi2 test, I2 statistics, and 95% prediction intervals GRADE was used for evidence certainty assessment. Results: After screening 4759 and full-text reading 128 publications, 43 RCTs were eligible and 41 included in the meta-analysis (N=23209). Risk of bias was generally low, with most concerns in older studies Small but statistically significant intervention benefits were found for global cognition (composite score of validated neuropsychological tests; SMD=0,28; 95% CI: 0,10 to 0,45), and most of the other cognitive measures. High heterogeneity was observed for global cognition composite scores and could be only partially explained by three smaller RCTs. Intervention effect-size was significantly associated with shorter duration (P-value=0,009) and higher observed intensity (P-value=0,008). Interpretation: Multidomain interventions have small but consistent beneficial effects on cognitive measures, suggesting the potential to reduce cognitive impairment and dementia risk. High heterogeneity across RCTs can hinder data pooling. More evidence on longer-term effect is needed. Future research should prioritise harmonisation of methodologies and reporting, long-term extended follow-up data, clinically relevant dementia-risk surrogate outcomes, and evidence from more diverse cultural, geographical, and socio-economic contexts.