Multidomain lifestyle interventions have great potential to delay cognitive aging, but the effects of previous trials are small and mixed. Furthermore, previous studies mainly focused on cognitive functioning as the primary outcome. Consequently, underlying neurocognitive, central mechanisms, and involved peripheral pathways remain poorly understood. The HELI (Hersenfuncties na LeefstijlInterventie) study is a 6-month multidomain lifestyle intervention in older adults at risk of cognitive decline based on modifiable lifestyle factors, focusing on involved central and peripheral mechanisms by using a broad range of neuroimaging, blood- and gut-related outcome measures. Below, we explain the rationale of our objectives and hypotheses.

In our aging population, the incidence of cognitive decline and incurable neurodegenerative diseases like Alzheimer dementia (AD) and other types of dementia is increasing drastically [1]. A substantial part of the risk factors for dementia, such as obesity, hypertension, type II diabetes, and physical inactivity [2-6], is modifiable by changes in lifestyle. Indeed, observational research shows a relationship between a healthy lifestyle, such as a healthy diet, regular physical and cognitive exercise, not smoking, and maintaining social contacts, with a lower dementia risk [7,8]. As the signs of neurodegenerative diseases, such as amyloid beta pathophysiology and tauopathy in AD, begin decades before the first symptoms become apparent [9,10], paired with the fact that long-term healthy lifestyles show inverse relations with dementia risk, interventions addressing cognitive decline need to start at an early stage [11]. The World Health Organization (WHO) has therefore designated preventive multidomain lifestyle interventions as having the greatest potential to reduce the risk of cognitive decline and dementia within its 2019 guidelines [12].

Previous multidomain lifestyle intervention trials in aging individuals show small but significant effects on both cognitive functioning and dementia risk, as concluded by a meta-analysis from 2022 [13]. Especially when compared to single-domain interventions, multidomain lifestyle programs showed stronger effects [14,15]. However, previous studies are heterogeneous in population characteristics, intervention design, components, and duration and outcome measures. Moreover, when zooming in on individual studies and considering the latest published trials [16,17], the reported effects can be considered mixed. The 24-month Finnish Geriatric Intervention Study to Prevent Cognitive Impairment and Disability (FINGER) study [18] was the first large, longitudinal randomized controlled trial (RCT) in older adults at risk of cognitive decline (n=1260) and found subtle but robust positive effects of a multidomain lifestyle intervention on cognitive functioning. Other lifestyle interventions such as the Prevention of Dementia by Intensive Vascular Care (preDIVA) study (n=2994) [3], Multidomain Alzheimer Prevention Trial (MAPT; n=1680) [19], Japan-Multimodal Intervention Trial (J-MINT; n=531) [16], and AgeWell.de (n=1030) [17] did not find differences in cognition or dementia risk. However, specific subgroups within these trials with higher dementia risk based on modifiable cardiovascular lifestyle factors [3,19] did show improvements in cognitive functioning. Altogether, these findings point to specific at-risk older adults who are most likely to benefit from multidomain lifestyle interventions.

Crucially, these previous lifestyle intervention studies all had behavioral cognitive tests or dementia incidence as primary outcomes. We still have a limited understanding via which brain mechanisms and potential underlying peripheral pathways, the multidomain lifestyle interventions might exert their effects on cognition. To elucidate underlying brain mechanisms, different neuroimaging techniques are essential to analyze brain health and function in vivo in a longitudinal, intervention setting. Moreover, to understand involved peripheral pathways, we need extensive measures of inflammatory, metabolomic, and microbiome markers in blood and feces.

Among the previous lifestyle intervention studies, only FINGER (n=132, 2 years) and MAPT (n=68, 6 months) used (structural) neuroimaging in a subsample of participants. Within FINGER, no structural (ie, anatomical) differences were found between intervention groups [20], whereas within MAPT, the lifestyle intervention group showed decreased deformation of periventricular areas and temporal lobes, and decreased whole-brain atrophy compared to placebo [19]. In addition, positron emission tomography (PET) scans within MAPT also demonstrated elevated levels of cerebral glucose metabolism within, for example, the hippocampus and the adjacent parahippocampal gyrus [21]. These limited and heterogeneous neuroimaging findings highlight the poor understanding we currently have of underlying neurocognitive mechanisms of lifestyle changes. In addition, some of the previous studies did assess blood, cardiovascular [3], inflammatory [19], or metabolomic markers [18,19], which confirmed that these pathways are influenced by lifestyle changes. However, none of the preceding lifestyle interventions have combined extensive blood and fecal measures to fully understand underlying peripheral(-central) mechanisms.

Therefore, within the HELI study, we will combine a broad neuroimaging protocol, including task-related functional magnetic resonance imaging (fMRI) and quantitative imaging of cerebral blood flow (CBF) as primary central outcomes, with peripheral measures in blood and feces, including blood inflammatory profile and fecal microbiota composition as primary peripheral outcomes. Below, we will explain the rationale of our chosen primary and secondary outcome measures.

The longitudinal process of aging impacts multiple facets of the brain, causing changes in both brain structure and function and, subsequently, in cognition. Age-related structural brain changes are characterized by changes in gray and white matter volume, leading to loss of overall brain volume [22]. Importantly, certain brain structures such as the hippocampus and dorsolateral prefrontal cortex (dlPFC) appear to shrink at a faster rate compared to other brain regions, possibly explained by their increased vulnerability to vascular risk factors such as hypertension and hypercholesterolemia [23-25]. These findings indicate a possible link between lifestyle-modifiable cardiovascular risk factors and aging-related cognitive decline. In addition, the large interindividual differences reported in age-related structural and functional brain deterioration have been proposed to be partially explained by differences in adherence to healthy lifestyle patterns [26].

Earlier studies in pathological aging, specifically in mild cognitive impairment (MCI) and AD, have broadly established key patterns of functional brain deterioration, even in prodromal stages of disease before structural deterioration becomes apparent [27]. These findings indicate that measures of brain function may be more sensitive to interventions targeting age-related cognitive decline and are more strongly modifiable by nature as opposed to brain structure. The most notable cognitive changes associated with normal and pathological aging can be found in the more complex cognitive domains, such as executive functioning and working memory [28]. Working memory is a fundamental cognitive process, supporting higher-order cognitive functions such as planning, problem-solving, and decision-making. Although the overall effects of age on working memory, which shows gradual decreases with advancing age, have been well-established, multiple studies have also identified increased activity in regions of the prefrontal cortex (PFC) in older adults who show comparable working memory performance to young adults [29,30]. These studies indicate that in aging, differences in functional activation patterns, measured with fMRI, may already show a—perhaps compensatory—reorganization before detriments in cognitive performance are observed [28,31]. However, as current literature remains scarce, it remains unclear how this potential functional reorganization is affected by a multidomain lifestyle intervention, especially in at-risk older adults. Another brain function measure, which is a potential cause underlying age-related decline in cognitive functioning, is decreased (local) CBF [32-35]. Multiple studies have reported the predictive value of quantified CBF on cognitive performance in both normal and pathological cognitive aging [32-35]. Moreover, increased local CBF in specific brain regions, such as the hippocampus and PFC, has been directly linked to increased cognitive performance [32]. Although previous studies have found short-term positive effects on CBF after lifestyle improvements such as improved diet or increased physical exercise, it remains unclear whether these effects retain after a multidomain lifestyle intervention and how the increase in regional CBF translates to cognitive functioning in aging [36].

Other promising brain measures in cognitive aging research are the quantification of local iron deposition and neuroinflammation. The accumulation of iron across the lifespan (especially in the deep gray matter), although conventionally found in healthy aging adults [37-39], has been directly linked with increased local oxidative stress leading to neurodegeneration [40-43] and decreased cognitive functioning in older adults with amyloid-β pathology [44]. Similarly, and in addition to intracranial iron deposition, neuroinflammation contributes to neurodegenerative processes in the normally aging brain [45], for example, by affecting the cerebrovascular endothelium leading to neurovascular dysfunction [46]. These impairments could eventually contribute to decreased brain functioning and cognitive decline [47]. We now know that increased regional neuroinflammation—for example, measured by quantifying intracranial myo-inositol concentrations with magnetic resonance spectroscopy (MRS)—is strongly associated with memory dysfunction in normal aging [48,49], in MCI and in AD [49,50]. These findings emphasize the effect of local neurodegenerative effects on specific cognitive domains associated with cognitive aging and dementia.

We currently know that age-related structural and functional brain changes consequently cause cognitive decline, and that brain perfusion, neuroinflammation, and iron accumulation are possible underlying mechanisms. A small number of lifestyle intervention studies, most of which target one specific domain, have reported positive associations between improved diet, exercise, and cognitive training, and neuroimaging measures of cerebral perfusion [51-53] and functional connectivity [54]. However, it is currently unknown to what extent a multidomain lifestyle intervention influences the involved mechanisms of age-related changes in brain activation. Applying a broad neuroimaging protocol could provide important insights into the effects of multimodal lifestyle improvements on the underlying brain mechanisms of age-associated cognitive decline and provide direct mechanistic targets for future interventions.

Growing evidence indicates an important role for intestinal health and other peripheral factors, like inflammation, metabolic, and vascular health, in the development of cognitive decline [55,56]. Furthermore, a substantial part of lifestyle effects on brain functioning is expected to arise via mechanisms in the periphery. An overview of these hypothesized pathways and the link with brain health can be found in Figure 1. These pathways are expected to be influenced by multidomain lifestyle interventions at different levels. Below, we will describe the rationale behind these pathways.

The gut microbiome, in particular, has an increasingly recognized impact on brain functioning and behavior via different routes captured in the gut-brain axis. These include the modulation of immune signaling to the brain, affecting the hypothalamic—pituitary—adrenal axis, and transmission via the vagus nerve [57]. Microbiota-derived bioactive metabolites, such as short-chain fatty acids (SCFAs), tryptophan metabolites, (endo)toxins, and secondary bile acids, could mediate the effects of the gut microbiome on brain health [58]. They reflect microbiome function [59] and have demonstrated health effects on the host (eg, [anti]inflammatory and metabolic processes) [60,61], also in aging [62].

We know that aging is associated with changes in microbial composition and diversity and gut health [63]. Older adults more frequently show gut microbiome dysbiosis (imbalance) compared to younger individuals, potentially due to prolonged exposure to an unhealthy lifestyle [64,65]. Particularly, microbial diversity and uniqueness seem to correlate with aging in Western societies [66]. A lower fecal microbial diversity predicted poorer cognitive function in healthy older adults [67]. Microbial metabolites are also affected by aging, as older adults show reduced fecal total SCFA levels [68]. SCFAs, especially butyrate, are known for their beneficial anti-inflammatory effects [69,70] and have been linked to improved cognition in mice [71,72]. Importantly, these microbial changes become more pronounced in the case of pathological cognitive aging like AD. Compared to healthy age-matched individuals, patients with AD show a decreased microbial diversity and stability over time [73], differences in abundance of individual taxa [56,74], and more often have small-intestinal bacterial overgrowth (SIBO) [75]. In addition, a further decrease in levels of fecal SCFAs and tryptophan metabolites was found, which correlated with cognitive decline [76]. However, as these findings mainly come from cross-sectional studies, causal evidence from intervention studies to provide a stronger link between these gut-brain mechanisms is still limited.

Among the proposed gut-brain pathways, inflammation is assumed to play a pivotal role in brain health. Dysbiosis of the gut microbiome can potentially impair the integrity of the intestinal barrier [77], thereby provoking a local inflammatory reaction that could eventually lead to systemic inflammation [78]. In addition, microbiota dysbiosis can also lead to the dysregulation of abdominal adipose tissue, affecting immune and insulin pathways, promoting systemic inflammation even more [79,80]. Low-grade, systemic inflammation is commonly seen in aging [81] and is characterized by a mild and sustained increase in immune mediators in the systemic circulation. For example, the acute-phase response marker C-reactive protein (CRP) and its major regulators interleukin (IL)-6 and tumor necrosis factor (TNF)-α are increased [81,82]. In turn, low-grade, systemic inflammation due to impaired intestinal health can lead to elevated levels of neuroinflammation [83] and is linked with neurodegenerative processes in the normally aging brain.

Importantly, a direct causal role for the gut-immune-brain axis in aging has been demonstrated in mice using fecal microbial transplantation (FMT) [84-87]. These findings emphasize that microbial modulation in humans might be of significant therapeutic benefit in preventing age-related inflammation and cognitive decline [88]. Altogether, the gut microbiome and its interaction with the host can be considered a crucial modulator of (un)healthy aging, determining the rate of physical and cognitive decline [89].

There are numerous studies showing that gut microbiome composition and immune-metabolic health effects on the host are modified by diet and nutritional components [90-92], including the Mediterranean-type diets [93,94]. In addition, other lifestyle components such as physical activity [95,96], stress [97], and sleep [98,99] are known to affect the microbiota composition and intestinal health, as well as immune, vascular, and metabolic health [100-107]. Importantly, we hypothesize that multiple lifestyle domains can synergistically affect peripheral markers and should therefore be targeted simultaneously. However, it is still largely unclear how gut microbiome, immune, and metabolic factors link to brain health in aging, and which mechanisms are involved. By investigating how lifestyle-induced changes in these peripheral factors relate to changes in brain health, we can make stronger conclusions about the role of these peripheral-brain links in cognitive aging. Moreover, elucidating peripheral-brain mechanisms might also explain individual differences seen in lifestyle interventions for healthy cognitive aging.

We designed the HELI study (derived from the Dutch “HErsenfuncties na LeefstijlInterventie,” meaning “Brain function after lifestyle intervention”) to better understand the neurocognitive effects of a multidomain lifestyle intervention and underlying peripheral (central) mechanisms. The HELI study is a 6-month multidomain lifestyle intervention in older adults at risk of cognitive decline (based on lifestyle-modifiable cardiovascular risk factors) with a high and low-intensity intervention arm. We focus on 5 different lifestyle domains within the intervention period: diet, physical activity, stress management and mindfulness, cognitive training, and sleep. We will assess effects on multiple neuroimaging brain measures (primary: working memory task-related fMRI responses and performance, and quantitative imaging of CBF), peripheral measures related to the gut-immune-brain axis (primary: blood inflammatory profile and fecal microbiota composition), as well as their connections. We propose that, compared to the low-intensity group, the high-intensity group will show greater improvement in central neuroimaging outcomes (eg, fMRI and CBF) as well as peripheral outcomes (eg, blood inflammatory profile and fecal microbiota composition). Moreover, we hypothesize that individual changes in central outcomes correlate with changes in peripheral outcomes, as summarized in Figure 1.

The HELI study is a multicenter randomized controlled trial, aiming to investigate the effects of a 6-month multidomain lifestyle intervention on brain functioning and its relation with immunometabolic markers in aging. Powered to include 104 total participants, HELI has two parallel-arm treatment groups: (1) a supervised “high-intensity” coaching intervention group with weekly group meetings, lifestyle-related information and exercises, and (2) a nonsupervised “low-intensity” coaching intervention group, receiving information leaflets every 2 weeks. Because the participants of the high-intensity intervention arm were supervised in a group context, we recruited and subsequently randomized participants in 4 different inclusion waves (between May 2022 and October 2023), consisting of 19‐32 participants. Participants partook in outcome measure visits at baseline (T0) and at 6 months follow-up (T1) (see Figure 2).

The HELI study has been reviewed and accepted according to the Medical Research Involving Human Subjects Act (“Wet Medisch-wetenschappelijk Onderzoek met mensen” in Dutch) by an accredited Medical Research Ethics Committee (MREC), the MREC Oost-Nederland on December 2, 2021 (ToetsingOnline filenumber NL78263.091.21, ClinicalTrials.gov ID NCT05777863).

In the HELI multidomain lifestyle intervention, five distinct lifestyle domains were combined, namely (1) diet, (2) physical activity, (3) stress management and mindfulness, (4) cognitive training, and (5) sleep. The intervention period was 6 months (26 weeks) in total, which was subsequently followed by the follow-up outcome measure visits (see Figure 3 for an overview).

The high-intensity intervention group received a structured and personalized intervention program consisting of weekly group sessions, active guidance from an intervention supervisor, and lifestyle domain–related individual and group exercises. The low-intensity intervention group received access to general lifestyle-related health information in an individual context, provided through information leaflets sent by email once every 2 weeks. To promote intervention adherence and blinding, we adopted the FINGER-NL [108] and US POINTER approach [114] of framing both groups as lifestyle intervention recipients, instead of using the label “control group.” The groups differ in intervention structure, engagement, and intensity rather than treatment versus control.

During communication with participants, we refrained from using the terms “high-intensity” and “low-intensity” to promote group blinding as much as possible. Instead, the intervention arms were called “weekly group sessions” for the high-intensity intervention group, and “independently working on your lifestyle” for the low-intensity intervention group. An extensive description and overview of the multidomain lifestyle intervention groups is provided below and depicted in Figure 3.

Participants randomized to the low-intensity intervention group received general lifestyle-related health information, addressing the aforementioned 5 lifestyle domains. Participants received this general information through information leaflets (1‐2A4), through email, once every 2 weeks. The information leaflets discussed the lifestyle domains but exclusively offered general health information without providing participants any individual or personal guidance based on their personal circumstances. For example, participants were generally told that a healthy diet promotes a healthy brain, and which kind of food components are promoted by the guidelines for a healthy diet, but participants were not guided or advised how their own personal diet should therefore be adjusted to fit these diet recommendations and guidelines.

The information of these leaflets was also handed to the high-intensity intervention group to ensure both groups were provided with the same general health advice.

All outcome measures were measured at baseline (T0) and after participation in the 6-month multidomain lifestyle intervention (T1), during visits at both research centers. Blood drawing and breath tests were performed in a fasted state. Some outcome measures, such as lifestyle domain or adherence-related questionnaires, were also collected throughout the intervention period. For a detailed overview of all outcomes and the collection schedule, see S4 in Multimedia Appendix 1.

The primary outcomes are the changes between baseline (T0) and follow-up after 6 months (T1) in:

In addition, we will investigate intervention-induced gut-immune-brain links by assessing the relation between the effects found in the abovementioned primary peripheral markers and brain outcome measures.

The secondary outcomes are the changes between baseline (T0) and follow-up after 6 months (T1) in:

Other study outcomes include the following:

The expected effect size for this study was determined based on other, mostly single-domain, lifestyle intervention studies with similar outcomes as our primary outcomes in the brain (eg, cerebral perfusion, working memory assessments) and in the periphery (eg, immunometabolic biomarkers, gut microbiome diversity). These studies generally found medium-to-large effect sizes (see Table S4 in Multimedia Appendix 1 [52,53,95,106,107,153-161]). Therefore, we deemed a slightly above-medium effect size of 0.30 (Cohen f[V]) suitable for our primary outcomes from baseline (T0=week 0, before start intervention) to follow-up (T1=week 26, after end intervention). The sample size calculation was performed using G*Power (version 3.1.9.6; Universität Kiel [162]). Using a multivariate ANOVA (with repeated measures and within-between interaction) F test with a power of 80% and a 2-tailed α-error probability of .05, we calculated that a total of 90 study participants would be required. However, based on our experience with previous fMRI and intervention studies, we expect a potential loss of data of 15% due to poor fMRI data quality (eg, motion artifacts) and potential intervention drop-outs. Therefore, to reach sufficient power, we would require a total sample size of 104 (90 participants×15% potential data loss=103.5=104) participants in this study.

Participants were randomized per inclusion wave, using stratified (2,4) block-randomization to provide a 1:1 allocation between our two intervention arms. Randomization was stratified by the factors sex (male vs female), risk factor profile (medium risk 2‐3 points vs high risk ≥4 points), and a combined factor of education level +age (70+ yearsor low education vs 60‐69 years and high education, specified as university of applied science associate degree (Dutch: HBO associate degree or higher). As with our sample size, (2,4) block-randomization was only possible with up to 3 strata, age, and education were combined in one stratification factor as they are both important covariates of cognitive functioning and aging-related cognitive decline [163].

An independent researcher from the DCCN was authorized to manually perform the randomization. This authorized researcher had no affiliation with the project and therefore had no bias in the allocation procedure.

Given the nature of the intervention arms, the participants and intervention supervisor were unable to remain blinded to the assignment to the intervention arms. Researchers conducting the follow-up outcome measure visits remained blinded and unaware of the intervention arm allocation. Only in case of unforeseen circumstances would deviations be made (eg, an unblinded researcher conducting the MRI protocol). Neuropsychological tests and task trainings (ie, n-back task) were conducted by a blinded researcher in all cases. Participants were asked to refrain from speaking about the intervention contents and experience during the follow-up outcome measure visits to prevent accidental unblinding.

The requirement for emergency unblinding was not applicable to this study, as participants were unblinded to the group allocation and the nature of the study setting.

We will test for change between T0 and T1 in the primary and secondary outcome variables, both within-group and between-group, using linear mixed-effect models with random effects for intercept (individuals). The intervention group and time will be included as fixed effects, as well as the interaction term between intervention group and time, to model the difference in change between T0 and T1 as a function of group allocation.

In addition, we will test for correlations in change (T0 minus T1) between (primary) brain outcomes and (primary) peripheral outcomes. We will select an appropriate correlation test for our data, based on whether assumptions are met. We will include both the high- and low-intensity groups if there is enough variance in change across both groups; else, we will only include the high-intensity group.

The level of significance will be set at 0.05 (2-sided). FDR correction will be used to account for multiple comparisons of secondary and exploratory outcomes. FWE correction will be used for secondary whole-brain analyses.

As part of exploratory analyses to assess the domain-specific effects on our primary outcome measures, we will take the pre- and posteffects of the primary outcome measures and use domain-related questionnaire scores and smartwatch-derived physiological data as a measure of improvement for each respective domain as predictors in linear mixed-effect models.

We will perform an intention-to-treat analysis. Participants who withdrew from the study during the 6-month intervention period, but had not withdrawn consent, were invited to the T1 (after week 26) follow-up outcome assessments. This means that we will use all available collected data, including collected (follow-up) data of participants who dropped out of the intervention. Missing data or data with insufficient quality, however, will not be used in the analysis.

The HELI study has been reviewed and approved by the MREC Oost-Nederland on December 2nd 2021 (ToetsingOnline filenumber NL78263.091.21, ClinicalTrials.gov ID NCT05777863). All participants provided written informed consent before study inclusion, and participants were free to opt out of participation at any stage of the study. Obtained study data were only accessible to the research team directly involved in the study, and personal privacy was protected by removing personal identifying information from the dataset. Participants received compensation for travel expenses and trial participation, which were evaluated and approved by the MREC to avoid financial motivation in study participation.

We described the design of the HELI study, an RCT focusing on the mechanisms of a 6-month multidomain lifestyle intervention in Dutch older adults at risk of cognitive decline, by assessing functional and structural MRI-related brain measures as well as peripheral measures related to the gut-immune-brain axis. The study design is based on the design of the FINGER-NL trial [108] and optimized for assessment of mechanisms instead of effectiveness.

The HELI study is characterized by its unique multimodal approach, which enables us to obtain mechanistic insights of a multidomain lifestyle intervention. We are using a broad range of neuroimaging techniques (see Table S3 in Multimedia Appendix 1 for the MRI sequence protocol) and neurocognitive tests to acquire a comprehensive profile of brain health. In addition, we also measure an extensive set of peripheral intestinal and immunometabolic markers, allowing us to investigate gut-immune-brain interactions and discover mediators of lifestyle effects on brain health. Combined with questionnaire data on lifestyle compliance, these brain, blood, and fecal markers have the potential to provide an elaborate overview of the pathways involved in lifestyle effects on cognitive aging and to explain individual differences in effectivity.

Compared to previous 24-month multidomain lifestyle interventions in cognitive aging [3,18,19,164] and the FINGER-NL trial [108], the HELI study intervention of 6 months is significantly shorter. However, our multidomain lifestyle intervention was carefully modified from the 2-year FINGER-NL intervention into a more condensed and intensified 6-month program. Especially, the high-intensity intervention group was characterized by more frequent meetings and group contact, requiring higher weekly effort. Within the field of functional neuroimaging, the duration of HELI (26 weeks) is relatively long compared to other (single-domain) lifestyle intervention studies (6‐9 weeks), which reported significant central mechanistic intervention effects [52,54]. The main focus of our study is on the underlying mechanisms of lifestyle adaptations on cognitive aging, rather than on trial effectiveness. Therefore, we consciously chose to use multiple specific brain and peripheral parameters as primary outcomes, instead of cognitive functioning. We expect these parameters to be more sensitive to lifestyle changes, and therefore to be affected in an earlier stage than cognitive functioning.

We ultimately managed to include a total of 102 participants within our study. Our study population shares a similar baseline lifestyle modifiable risk factor profile (CAIDE) compared with the FINGER study, based on the total number of participants with hypertension, hypercholesterolemia, and diabetes (see Table S5 in Multimedia Appendix 1 [165-167]). Other previous multidomain lifestyle intervention studies in older adults, such as the pre-DIVA, MAPT, and J-MINT studies, used inclusion criteria different from the CAIDE-related lifestyle modifiable risk factors, or included participants solely based on age. Compared with previous studies, our study population is relatively highly educated, and we were unable to include a comparable number of participants with a lower education level. However, many of these earlier studies completed participant inclusion over a decade ago, and education levels have generally increased over time, which may reflect cohort effects. Because education level is often used as a covariate in task-related neuroimaging outcomes and neuropsychological test data, this is an important limitation to consider during data interpretation and comparison of results with other studies. For a more detailed description of the baseline characteristic comparison between the HELI study and previous multidomain lifestyle intervention studies, see Table S5 in Multimedia Appendix 1.

To summarize, the most important strength of the HELI study is its unique combination of multimodal neuroimaging and peripheral phenotyping with a multidomain lifestyle intervention. Moreover, the study has several additional strengths. First, the design has two active arms, which promote blinding and intervention adherence, while preserving between-group comparisons of lifestyle interventions. Second, the group approach within the high-intensity group is a strength, as it encourages peer support and intervention adherence. Finally, the online-onsite combination establishes a future-oriented intervention framework that ensures both quality and broad applicability. Limitations of the study comprise the limited number of objective domain-specific adherence measures and the selection bias toward more highly educated individuals, potentially due to the digital intervention format.

The HELI study offers a novel mechanistic approach by integrating multimodal neuroimaging and peripheral phenotyping with a multidomain lifestyle intervention in older adults at risk of cognitive decline. Results of the HELI study can help explain the previously observed small effect sizes of multidomain lifestyle interventions to prevent cognitive decline in aging, paving the way for personalized prevention strategies in the future.