Effects of preventive interventions on neuroimaging biomarkers in subjects at-risk to develop Alzheimer's disease: A systematic review

1. Introduction

Alzheimer's disease (AD) is a devastating neurodegenerative disorder characterized by a complex and multifactorial physiopathology. Clinically, the typical form induces episodic memory deficit, progressively associated with language and behavioral troubles and leading to a loss of autonomy (DeTure and Dickson, 2019; Breijyeh and Karaman, 2020; Scheltens et al., 2021). Brain amyloidosis and neurodegenerative processes remain the main therapeutic targets as they occur many years prior to cognitive and clinical symptoms appearance (Sperling et al., 2013; Makin, 2018).

Two interventional strategies have been developed, one focusing on drugs targeting specific molecules such as the amyloid-beta (Aβ) peptide or Tau protein, and the other focusing on holistic non-specific targets such as epidemiological and/or and exposome risk factors. For the first strategy, Aduhelm has recently been FDA-approved for US AD patients as a specific anti-amyloid drug (Cummings et al., 2021). The readout for other anti-amyloid phase 3 drugs are coming and other Tau, neuroinflammation and APOE ε4 targets are studied. For the second category, multimodal preventive interventions are promoted by governmental health organizations (WHO) based on epidemiological and interventional clinical trials data (World Health Organization, 2019). Modifiable risk factors such as low education, midlife hypertension, midlife obesity, diabetes, physical activity (PA), smoking or depression have been linked to AD (Norton et al., 2014; Serrano-Pozo and Growdon, 2019). By acting on those risk factors before the apparition of clinical symptoms, one third of AD cases could be potentially reduced with reasonable costs (Livingston et al., 2017).

Healthy lifestyles (PA, nutrition, cognitive stimulation…) are associated with lower incidence of AD (Dhana et al., 2020). The amount of PA has clearly been inversely associated with the risk of cognitive decline and AD (Paillard, 2015). Aerobic physical exercise (PE) induces the release of neurotrophic factors and reduces the production of free radicals, both phenomena participating in improving memory and cognitive function while limiting the alteration of specific neuronal populations (Paillard et al., 2015). Nutrition and diet might be significant modifiable risk factors of AD and multiple antioxydants, vitamins, polyphenols, fish, or dietary patterns (Japanese, Mediterranean) have been reported to decrease the risk of AD (Hu et al., 2013). Cognitive training (CT) seems also very promising (Sitzer et al., 2006) by targeting several domains of cognition such as memory, executive or visuospatial functions (Nguyen et al., 2019). While preventive multimodal interventions for AD including PE (Erickson et al., 2011; Zhu et al., 2020; López-Ortiz et al., 2021), nutrition/diet (Cremonini et al., 2019) or CT (Buschert et al., 2010) have shown promising results, the majority of these trials have small sample sizes and evidences from large single-domain lifestyle interventions (PE, LIFE study [Longitudinal Impact of Fitness and Exercise]; dietary, OPAL study [Older People And n-3 Long-chain polyunsaturated fatty acid]; CT, ACTIVE [Advanced Cognitive Training in Vital Elderly], IHAMS [Iowa Healthy and Active Minds Study]) are limited (Kivipelto et al., 2018).

As AD is multifactorial, multi-domains (MD) interventions would be more relevant than individual factors or even have a superadditive effect on clinically meaningful outcomes (Coley et al., 2008; Scarmeas, 2009; Kivipelto et al., 2018). The combination of interventions has been addressed by clinical trials (FINGER, MAPT, PreDIVA; Richard et al., 2009; Vellas et al., 2014; Ngandu et al., 2015), and different hypotheses have been made about the potential synergistic effects between interventions. For instance, PE could increase the potential for neuro-, synapto-, and angiogenesis while CT would guide it to the stimulated brain regions (Bamidis et al., 2014). Nutrition, including omega-3 (ω3) intake, could fuel structural changes associated with these interventions (Köbe et al., 2016). Interestingly, multimodal interventions may be more effective before clinical symptoms, especially for at-risk of AD populations such as the carriers of the ϵ4 allele of the apolipoprotein E (APOE ϵ4; Berkowitz et al., 2018). The primary outcomes to define the efficacy of these interventions are on cognitive performances. The effect on surrogate biomarkers is less described or as exploratory analyses (Rolandi et al., 2016).

Thus, we aimed to evaluate the effect of multimodal interventions alone or combined such as PE, CT, and nutrition/diet on a large variety of brain imaging outcomes analyzed globally and regionally in participants that may develop AD. We also assessed whether this effect may be dependent on the cognitive status of the population included in the studies.

2. Materials and methods 2.1. Studies selection

A search of the PUBMED database was performed on May the 17th of 2021. The query included the following terms: (“elderly” OR “frail elderly” OR “risk factors” OR “MCI” OR “alzheimer”) AND (“PET” OR “brain imaging” OR “MRI” OR “structural MRI” OR “functional MRI”) AND (“training” OR “nutrition” OR “diet” OR “physical activity” OR “cognitive training” OR “cognitive stimulation” OR “exercise”). The full search strategy is provided in the Supplementary material. We selected preventive studies involving older adults susceptible to convert to AD and evaluated the effect of CT, nutrition/diet, PE or MD interventions on brain imaging outcomes. The description of the aim of this review, using the PICO framework (Huang et al., 2006), is available in Supplementary Table 1. As our goal was to examine the effect of interventions on brain regions, including areas located in subcortical structure, we did not select studies using electrophysiological techniques such as electroencephalography (EEG), which have a relatively low spatial resolution (Krishnaswamy et al., 2017). Older adults were considered susceptible to convert to AD if the studies in which they were included stated that they exhibited risk factors for AD (e.g., APOE ϵ4) and/or cognitive impairments (subjective or objective) and/or biological biomarkers of AD (e.g., elevated amyloid load). The participants at risk for AD were defined into two groups with either no objective cognitive impairment (nCI) or objective cognitive impairment (CI). The nCI category groups together different type of participants: participants with risk factors for AD such as hypertension are included in this category, as well as participants with subjective cognitive decline. For the participants with CI, they could also present positive neuroimaging or cerebrospinal fluid (CSF) biomarkers.

We excluded (1) observational studies, (2) studies including exclusively healthy older adults (HOA) not predisposed to conversion to AD, (3) studies including exclusively patients already diagnosed with AD, (4) studies including populations with additional neurodegenerative diseases or vascular cognitive impairment, (5) articles not written in English, (6) studies for which only the abstract was available. We did not include yoga and dance interventions as we consider them special types of exercises. Yoga is a spiritual activity that encompasses physical exercise, controlled breathing, and meditation training. Dance is an artistic expression requiring memorization and execution of a series of movements according to the rhythm of a type of music and to the movements of a partner. The brain processes involved in these activities may be different from those involved in a simpler form of physical exercise. It has indeed been suggested that yoga and dance could have different effects on the brain from traditional forms of physical exercise (Rehfeld et al., 2018; van Aalst et al., 2021; Kaur et al., 2022). Analysis was not restricted to studies including a control for the interventions (i.e., a “placebo” or “sham” intervention), and no additional restriction was applied to the control condition when included (i.e., for two studies evaluating the effect of PE, control condition could either be resistance or balance training). When multiple brain imaging outcomes were evaluated in a study, they were all reported and equally considered. The distinct types of brain imaging data are referred to as imaging “modalities.” Article's abstract screening was performed by LP and full-text review was performed by GB and LP independently and validated by AG and ELB. Any discrepancy was resolved through discussion until a consensus was obtained. The inclusion of the studies was described by a PRISMA flowchart generated by the PRISMA 2020 Shiny application (https://estech.shinyapps.io/prisma_flowdiagram/; Haddaway et al., 2022).

2.2. Criteria used to assess the quality of selected studies

Multiple criteria were used to evaluate and compare the relative quality of the studies included. Part of these criteria are listed as the first thirteen items of Table 1 and were previously defined by Pitkälä et al. (2013), and used in a review specifically focussed on brain imaging outcomes (Haeger et al., 2019). To fit with the specificity of this review, we added/adapted some criteria. First, studies without a sham group were included and, in those cases, criteria #6 and #9 were applied to the comparisons between types of populations [e.g., HOA vs. mild cognitive impairment (MCI); Table 1]. Second, for studies evaluating only imaging outcomes, blinding criterion (#10) was not taken into consideration. Third, criterion concerning studies' power (#4) was reconsidered acknowledging a recent review on neuroimaging studies: in 2017–2018, only 3–4% of them did an a priori power calculation (Szucs and Ioannidis, 2020). As most intervention trials are expected to have at least two groups (intervention and sham), we estimated that studies were sufficiently powered if a proper power calculation was made or if there were at least 20 participants per group—considering that 11–56% of clinical studies with a single group and published between 1990–2012 had at least 40 participants (Szucs and Ioannidis, 2020). Study power was assessed as part of the overall quality assessment. However, power itself was not a criterion for exclusion. All studies were examined and included in this review, regardless of whether they were sufficiently powered or not. Additional quality criteria specific to neuroimaging studies were assessed: (1) imaging protocol and analyses had to be adequately described and (2) appropriate correction for multiple testing had to be implemented (e.g., voxel-wise analysis). An additional point was attributed to studies performing extra quality controls for raw imaging data or analyses' outputs. All criteria were assessed independently by GB and LP, differences in notation were discussed until a consensus was met.

www.frontiersin.org

Table 1. Notation criteria used to rank studies' methodology.

Studies were attributed a percentage of validated criteria and classified into limited, good and high quality if they respectively had < 50%, between 50 and 80% and more than 80% of validated items.

2.3. Criteria to assess the different types of results according to cognitive profile

The effects of interventions on neuroimaging modalities was assessed for nCI and CI participants (Supplementary Figure 1). For cases with multiple modalities tested, the effect was considered for each type. We define “k” as the number of “results” for all modalities and for N studies (a result relates to either the effect (positive result) or lack of effect (null result) of an intervention for one modality). We did not take into account results on pooled participants (e.g., mixed population of MCI and HOA), unless a distinction was made between the populations. Similarly, we discussed separately results associated with sub-analyses for specific subgroups (e.g., participants with APOE ϵ4 status), or that were reflecting correlations. Measures obtained at intermediate time-points during the intervention or long-term follow-up measures were discussed separately. Only measures obtained directly after the end of the intervention were examined.

A score (called “s”) quantifying the effect of interventions for each neuroimaging modality on participants was computed as:

s=∑x∈E(control(x)+quality(x))-∑x∈A(control(x)+quality(x))    (1)

where E were the results reporting an effect of intervention on neuroimaging biomarkers and A was the set of results associated with an absence of effect of the interventions. Control and quality functions were defined as:

control(x)=,,,]},,,]},,,]},,,,,]}],"socialLinks":[,"type":"Link","color":"Grey","icon":"Facebook","size":"Medium","hiddenText":true},,"type":"Link","color":"Grey","icon":"Twitter","size":"Medium","hiddenText":true},,"type":"Link","color":"Grey","icon":"LinkedIn","size":"Medium","hiddenText":true},,"type":"Link","color":"Grey","icon":"Instagram","size":"Medium","hiddenText":true}],"copyright":"Frontiers Media S.A. All rights reserved","termsAndConditionsUrl":"https://www.frontiersin.org/legal/terms-and-conditions","privacyPolicyUrl":"https://www.frontiersin.org/legal/privacy-policy"}'>

留言 (0)

沒有登入
gif