Table 3.

Examples of sub-group analyses assessing the potential modifiers for the intervention effect on cognition in multidomain prevention trials

Multidomain trials	Study	Potential intervention effect modifiers	Analyses	Results
FINGER	Rosenberg et al., 2018 [27]	Sex, age, education, socioeconomic status, cognition, cardiovascular factors, and cardiovascular comorbidity at baseline	Pre-specified	No significant differences in cognitive intervention benefits by sex, age, education, socioeconomic status, cognition, cardiovascular factors, and cardiovascular comorbidity.
	Solomon et al., 2018 [28]	APOE ε4 allele	Pre-specified	Intervention benefits were not significantly different between carriers and non-carriers. Clear benefit in APOE4 carriers in stratified analyses.
	Deckers et al., 2020 [29]	LIBRA index at baseline	Post hoc	Participants with a higher LIBRA index at baseline had overall less cognitive improvement over time, but this effect was not different between the intervention and control groups.
	Sindi et al., 2017 [30]	Leukocyte telomere length	Post hoc	More pronounced cognitive intervention benefits in individuals with shorter baseline leukocyte telomere length (higher-risk individuals).
	Stephen et al., 2019 [31]	Brain volumes and cortical thickness	Post hoc	More pronounced cognitive intervention effects in individuals with higher brain baseline cortical thickness and volumes.
MAPT	Andrieu et al., 2017 [19]	Cognition and functioning level at baseline	Pre-specified	No significant differences in intervention effects.
	Andrieu et al., 2017 [19]	APOE ε4 allele	Post.hoc	Intervention effects were not significantly different between carriers and non-carriers.
	Tabue-Teguo et al., 2018 [32]	Frailty status	Post hoc	Beneficial effects of multidomain intervention and n3 PUFA supplementation on cognition did not differ between frail and non-frail participants.
	Delrieu et al., 2019 [33]	Amyloid status	Post hoc	Multidomain intervention alone or in combination with omega-3 fatty acids was associated with improved primary cognitive outcomes in individuals with positive amyloid status.
	Chhetri et al., 2018 [34]	CAIDE score ≥ 6 points	Post hoc	High-risk subjects for dementia screened with CAIDE dementia score might benefit more from multidomain intervention.
preDIVA	Moll van Charante et al., 2016 [18]	Participants free from cardiovascular disease	Pre-specified	Participants with a history free from cardiovascular disease who were adherent to the intervention had a significantly lower risk of dementia compared to the control group.
	Moll van Charante et al., 2016 [18]	Untreated hypertension at baseline	Pre-specified	Participants with untreated hypertension who were adherent to the intervention had a significantly lower risk of dementia compared with the control group.
	van Middelaar et al., 2018 [35]	LIBRA index at baseline	Post hoc	LIBRA modifiable dementia risk score did not identify a (high-)risk group in whom the multidomain intervention was effective in preventing dementia or cognitive decline.

Subgroup analysis type (pre-specified and post hoc) was assessed from published trial protocols

APOE apolipoprotein E, CAIDE Cardiovascular Risk Factors, Aging and Dementia, FINGER Finnish Geriatric Intervention Study to Prevent Cognitive Impairment and Disability, LIBRA LIfestyle for BRAin health, MAPT Multidomain Alzheimer Preventive Trial, preDIVA Prevention of Dementia by Intensive Vascular Care, PUFA polyunsaturated fatty acids