Background: Dementia is a worldwide concern. Its global prevalence is increasing. Currently, no effective medical treatment exists to cure or to delay the onset of cognitive decline or dementia. Up to 40% of dementia is attributable to potentially modifiable risk factors, which has led to the notion that targeting these risk factors might reduce the incidence of cognitive decline and dementia. Since sporadic dementia is a multifactorial condition, thought to derive from multiple causes and risk factors, multi-domain interventions may be more effective for the prevention of dementia than those targeting single risk factors.

Objectives: To assess the effects of multi-domain interventions for the prevention of cognitive decline and dementia in older adults, including both unselected populations and populations at increased risk of cognitive decline and dementia.

Search methods: We searched ALOIS, the Cochrane Dementia and Cognitive Improvement Group's register, MEDLINE (Ovid SP), Embase (Ovid SP), PsycINFO (Ovid SP), CINAHL (EBSCOhost), Web of Science Core Collection (ISI Web of Science), LILACS (BIREME), and ClinicalTrials.gov on 28 April 2021. We also reviewed citations of reference lists of included studies, landmark papers, and review papers to identify additional studies and assessed their suitability for inclusion in the review.

Selection criteria: We defined a multi-domain intervention as an intervention with more than one component, pharmacological or non-pharmacological, but not consisting only of two or more drugs with the same therapeutic target. We included randomised controlled trials (RCTs) evaluating the effect of such an intervention on cognitive functioning and/or incident dementia. We accepted as control conditions any sham intervention or usual care, but not single-domain interventions intended to reduce dementia risk. We required studies to have a minimum of 400 participants and an intervention and follow-up duration of at least 12 months.

Data collection and analysis: We initially screened search results using a 'crowdsourcing' method in which members of Cochrane's citizen science platform identify RCTs. We screened the identified citations against inclusion criteria by two review authors working independently. At least two review authors also independently extracted data, assessed the risk of bias and applied the GRADE approach to assess the certainty of evidence. We defined high-certainty reviews as trials with a low risk of bias across all domains other than blinding of participants and personnel involved in administering the intervention (because lifestyle interventions are difficult to blind). Critical outcomes were incident dementia, incident mild cognitive impairment (MCI), cognitive decline measured with any validated measure, and mortality. Important outcomes included adverse events (e.g. cardiovascular events), quality of life, and activities of daily living (ADL).  Where appropriate, we synthesised data in random-effects meta-analyses. We expressed treatment effects as risk ratios (RRs) and mean differences (MDs) with 95% confidence intervals (CIs).

Main results: We included nine RCTs (18.452 participants) in this review. Two studies reported incident dementia as an outcome; all nine studies reported a measure for cognitive functioning. Assessment of cognitive functioning was very heterogeneous across studies, ranging from complete neuropsychological assessments to short screening tests such as the mini-mental state examination (MMSE). The duration of the interventions varied from 12 months to 10 years. We compared multi-domain interventions against usual care or a sham intervention. Positive MDs and RRs <1 favour multi-domain interventions over control interventions. For incident dementia, there was no evidence of a difference between the multi-domain intervention group and the control group (RR 0.94, 95% CI 0.76 to 1.18; 2 studies; 7256 participants; high-certainty evidence). There was a small difference in composite Z-score for cognitive function measured with a neuropsychological test battery (NTB) (MD 0.03, 95% CI 0.01 to 0.06; 3 studies; 4617 participants; high-certainty evidence) and with the Montreal Cognitive Assessment (MoCA) scale (MD 0.76 point, 95% CI 0.05 to 1.46; 2 studies; 1554 participants), but the certainty of evidence for the MoCA was very low (due to serious risk of bias, inconsistency and indirectness) and there was no evidence of an effect on the MMSE (MD 0.02 point, 95% CI -0.06 to 0.09; 6 studies; 8697participants; moderate-certainty evidence). There was no evidence of an effect on mortality (RR 0.93, 95% CI 0.84 to 1.04; 4 studies; 11,487 participants; high-certainty evidence). There was high-certainty evidence for an interaction of the multi-domain intervention with ApoE4 status on the outcome of cognitive function measured with an NTB (carriers MD 0.14, 95% CI 0.04 to 0.25, noncarriers MD 0.04, 95% CI -0.02 to 0.10, P for interaction 0.09). There was no clear evidence for an interaction with baseline cognitive status (defined by MMSE-score) on cognitive function measured with an NTB (low baseline MMSE group MD 0.06, 95% CI 0.01 to 0.11, high baseline MMSE group MD 0.01, 95% CI -0.01 to 0.04, P for interaction 0.12), nor was there clear evidence for an effect in participants with a Cardiovascular Risk Factors, Aging, and Incidence of Dementia (CAIDE) score > 6 points (MD 0.07, 95%CI -0.00 to 0.15).

Authors' conclusions: We found no evidence that multi-domain interventions can prevent incident dementia based on two trials. There was a small improvement in cognitive function assessed by a NTB in the group of participants receiving a multi-domain intervention, although this effect was strongest in trials offering cognitive training within the multi-domain intervention, making it difficult to rule out a potential learning effect. Interventions were diverse in terms of their components and intensity.