Cognitive Training Modulates Brain Hypersynchrony in a Population at Risk for Alzheimer’s Disease

Abstract

Background:

Recent studies demonstrated that brain hypersynchrony is an early sign of dysfunction in Alzheimer’s disease (AD) that can represent a proxy for clinical progression. Conversely, non-pharmacological interventions, such as cognitive training (COGTR), are associated with cognitive gains that may be underpinned by a neuroprotective effect on brain synchrony.

Objective:

To study the potential of COGTR to modulate brain synchrony and to eventually revert the hypersynchrony phenomenon that characterizes preclinical AD.

Methods:

The effect of COGTR was examined in a sample of healthy controls (HC, n = 41, 22 trained) and individuals with subjective cognitive decline (SCD, n = 49, 24 trained). Magnetoencephalographic activity and neuropsychological scores were acquired before and after a ten-week COGTR intervention aimed at improving cognitive function and daily living performance. Functional connectivity (FC) was analyzed using the phase-locking value. A mixed-effects ANOVA model with factors time (pre-intervention/post-intervention), training (trained/non-trained), and diagnosis (HC/SCD) was used to investigate significant changes in FC.

Results:

We found an average increase in alpha-band FC over time, but the effect was different in each group (trained and non-trained). In the trained group (HC and SCD), we report a reduction in the increase in FC within temporo-parietal and temporo-occipital connections. In the trained SCD group, this reduction was stronger and showed a tentative correlation with improved performance in different cognitive tests.

Conclusion:

COGTR interventions could mitigate aberrant increases in FC in preclinical AD, promoting brain synchrony normalization in groups at a higher risk of developing dementia.

Keywords

Cognitive decline functional neuroimaging intervention study longitudinal studies magnetoencephalography

INTRODUCTION

Alzheimer’s disease (AD) is a growing healthcare concern with a current prevalence estimated at 50 million worldwide [1]. The increase in life expectancy, advancements in diagnosis, and the absence of effective pharmacological treatments could raise the rates to 152 million cases by 2050 [2]. These figures could drop in almost 9 million cases if early treatment were to succeed in delaying the onset of the disease or in slowing down its progression by 1 year [3]. At present, pharmacological research is focused on disease-modifying therapies with several agents undergoing clinical trials [4]. The approval of aducanumab by the US Food and Drug Administration (FDA) in June 2021 closed a period of 18 years without new drug approvals for the treatment of AD [5], albeit its efficacy is still unresolved [6]. In this context, the limited impact of pharmacological therapies has brought attention to non-pharmacological interventions seen as a promising therapeutic strategy for the prevention of cognitive decline.

On the other hand, research in the last decade has shown that the toxicity of soluble amyloid-β (Aβ) oligomers and plaques in the proximity of GABAergic synapses [7 –9] drives the appearance of neuronal hyperexcitability [10]. At the circuit level, neuronal hyperexcitability triggers the disruption of the excitatory/inhibitory (E/I) synaptic balance, likely leading to large-scale network dysfunction and cognitive impairment in AD [11]. In this respect, AD patients often exhibit profiles of aberrant hypersynchrony [12, 13] and epileptiform activity [14] that could constitute a macroscopic sign of underlying overexcitation [11]. Importantly, brain hypersynchrony is already detected in vulnerable populations, such as individuals with subjective cognitive decline (SCD) [15,16, 15,16] and first-degree relatives of AD patients [17], which has motivated its study as an early biomarker of synaptic dysfunction.

Non-pharmacological interventions, and specifically cognitive training (COGTR), are cost-effective strategies that aim to preserve the mental abilities, daily living performance, and cognitive well-being of individuals in the early stages of the disease [18]. Although the efficacy of COGTR has been somewhat controversial [18, 19], multiple studies have described an association between COGTR and cognitive improvement in individuals with SCD (reviewed in [20]), mild cognitive impairment (MCI) (reviewed in [21]), and mild-to-moderate AD (reviewed in [22]). At the neuronal level, it is conceivable that the cognitive improvement promoted by COGTR could be mediated by a neuroprotective effect associated with the restoration of basal excitability. This view is consistent with previous findings of COGTR-related functional changes in normal aging (e.g., [23, 24]) and MCI [25 –30], which are often accompanied by gains in cognitive performance. In general, these changes reflected increases in functional connectivity (FC), which were also topologically specific to the degree of disease severity [25]. In our study, we focused on preclinical patients presenting with SCD, which is an intermediate state between normal cognition and MCI. This condition frequently predicts clinical progression to objective impairment and dementia [31], and therefore, represents a fitting working point in the AD continuum to attempt preventive care. As far as we are aware, our study is the first using electrophysiology to explore the neural correlates of COGTR in this population. Furthermore, no previous study has examined the potential of COGTR to modulate brain synchrony and to eventually revert the hypersynchrony phenomenon that characterizes the early stages of the disease.

Here, magnetoencephalographic (MEG) activity and neuropsychological scores were acquired twice for a sample of healthy controls (HC) and SCD participants. This sample was blindly randomized into a non-experimental group and an experimental group that completed a ten-week COGTR intervention. Brain synchrony was analyzed with a measure of FC based on the phase coupling between neuronal oscillations. Changes in FC and cognitive performance over time were compared between groups. We tested: 1) whether the COGTR intervention could modulate FC and 2) whether its outcome would be influenced by being in a group at a higher risk of developing dementia (i.e., SCD). We hypothesized that the cognitive improvement promoted by COGTR should be underpinned by a neutralizing effect over the electrophysiological signature of the disease, particularly concerning aberrant hypersynchronous connections.

MATERIALS AND METHODS

Participants

The sample of this study was recruited from the Center for Prevention of Cognitive Impairment (Madrid Salud), the Faculty of Psychology of the Complutense University of Madrid (UCM), and the Hospital Clínico San Carlos (HCSC) (Madrid, Spain) between January 2014 and December 2015. Two hundred thirty seniors were sequentially enrolled as part of a bigger study. All participants were aged 60–80 years, right-handed [32], and native Spanish. Research was performed following current guidelines and regulations. The study was approved by the Ethics Committee of the HCSC, and every participant signed informed consent prior to enrolling. A detailed list of the general inclusion and exclusion criteria can be found in [15].

During the screening session, a neuropsychological evaluation was conducted for every candidate including Rey-Osterrieth Complex Figure B (ROCF-B) [33], Digit Span Test (forward and backward), Texts of Verbal Memory of the Wechsler Memory Scale-III-R (WMS-III, Spanish version) [34], Boston Naming Test (BNT) [35], Phonemic and Semantic Fluency Tests [36], Trail Making Test parts A and B (TMT-A and TMT-B) [37], Mini-Mental State Examination (MMSE) [38], 7 Minutes Test [39], Rivermead Behavioral Memory Test (RBMT) [40], Memory Failures in Everyday Questionnaire (MFE, Spanish version) [41], Functional Activities Questionnaire (FAQ) [42], and the Geriatric Depression Scale-Short Form (GDS-SF) [43]. The research criteria proposed by the Subjective Cognitive Decline Initiative (SCD-I) [31] were applied to assign participants to the SCD group. Candidates 1) reported self-experienced persistent cognitive concerns (mainly associated with memory) in an interview with an expert clinician, 2) performed within the normal range on standardized cognitive tests that discriminate MCI and prodromal AD, 3) felt that their cognitive decline affected their daily activities, 4) had requested medical consultation, and 5) were≥60 years at the onset of SCD, having it occurred within the last 5 years. The cognitive decline was also confirmed by a reliable informant. Definite inclusion in the SCD group was agreed on by multidisciplinary consensus after discarding possible confounders of SCD in preclinical AD (e.g., psychiatric or neurological disease, medical disorders, medication, or substance use).

Out of the initial 230 candidates, 154 were apt for inclusion after discarding candidates with MCI (n = 43), candidates that did not attend the first MEG session (n = 18), and candidates without a valid first MEG recording (n = 15). 34 participants did not attend the second MEG session, and 22 participants presented deficiencies in the recorded or source-reconstructed signals. Eight participants were excluded because a T1-weighted MRI scan was not available.

The resulting sample of this study (n = 90) included 41 HC and 49 SCD participants, who had been randomly split into an experimental group (n = 46, 24 of which SCD) and a non-experimental group (n = 44, 25 of which SCD). The experimental group (hereafter, trained) participated in the COGTR intervention, while no active control paradigm or placebo activity were appointed for the non-experimental group (hereafter, non-trained). To keep the non-trained group engaged, participants were offered the same COGTR intervention once the study was completed (wait-list). All groups were adjusted for age (mean (standard deviation): 70.81 (3.59) (trained HC), 71.58 (5.02) (trained SCD), 69.42 (4.42) (non-trained HC), 73.24 (5.84) (non-trained SCD); p = 0.0762) and sex (male/female: 9/13 (trained HC), 10/14 (trained SCD), 6/13 (non-trained HC), 6/19 (non-trained SCD); p = 0.5244). The neuropsychological, MEG, and MRI sessions were repeated for all participants after around 6 months. A schematic representation of the study timeline can be found in Fig. 1.

Fig. 1

Schematic representation of the study design. The participants assigned to the experimental group engaged in 30 COGTR sessions (28 regular sessions and two maintaining-booster sessions) of 90 minutes scheduled three times per week, while no active control paradigm or placebo activity were appointed for the non-experimental group. Before and after the COGTR interval, all participants underwent a MEG recording, a T1-weighted MRI scan, and a complete neuropsychological evaluation.

Cognitive training

The COGTR intervention applied to the trained group was designed in 1994 at the Memory Training Unit of the City Council of Madrid (UMAM) (Madrid, Spain). The COGTR consisted of 30 sessions (28 regular sessions and two maintaining-booster sessions) of 90 minutes, organized in groups of 12 to 18 people. Assistance was controlled for all participants. All sessions took place in the morning, three times per week, and under professional instruction. The ‘UMAM’ method is thoroughly detailed in [44]. Altogether, it follows a multifactorial organization, targeting cognitive stimulation (language, perception, attention, etc.), learning of cognitive strategies (association, categorization, etc.), intervention in daily-living performance (role-playing, prospective-retrospective memory training, automatic actions rehearsal, etc.), and analysis of the meta-memory to inspect the causes and variables of one’s own cognitive failures. The intervention also includes several activities related to the pursuit of a healthy life (eating properly, engaging in physical exercise, participating in social and leisure activities, etc.), and encourages homework hobbies such as crossword puzzles and the use of new technologies. A schematic representation of the COGTR timeline can be found in Fig. 1.

Magnetic resonance imaging

Three-dimensional T1-weighted anatomical MRI scans were acquired for each participant with a 1.5 T MRI scanner (GE Healthcare, Chicago, IL) using a high-resolution antenna and a homogenization PURE filter (fast spoiled gradient echo sequence, with parameters: repetition time/echo time/inversion time = 11.2/4.2/450 ms, flip angle 12°, slice thickness = 1 mm, 256×256 matrix, and field of view = 256 mm).

Magnetoencephalography

Data acquisition and pre-processing

MEG data was acquired at the Laboratory of Cognitive and Computational Neuroscience (UCM-UPM), in the Center for Biomedical Technology (CTB) (Madrid, Spain) using a 306 channel (102 magnetometers, 204 planar gradiometers) Vectorview MEG system (Elekta AB, Stockholm, Sweden) placed within a magnetically shielded room (VacuumSchmelze GmbH, Hanau, Germany). Preceding the MEG recording, every participant was informed of the MEG routine and voluntarily agreed to participate. We digitized the head shape of the participants using a Fastrak 3D digitizer (Polhemus, Colchester, Vermont). More specifically, we registered the coordinates of three fiducial points (nasion and bilateral preauricular points) and of approximately 400 scalp points. In addition, we placed four head position indicator (HPI) coils in the participant’s scalp (two in the mastoids and two in the forehead). The position of these coils was also digitized in order to monitor head motion during the recording. To capture 1) eye blinks and ocular movements, we used two electrooculogram (EOG) electrodes (set above and below the left eye), and 2) cardiac activity, two electrocardiogram (EKG) electrodes (set in a diagonal across the chest) in bipolar montages.

After the initial setup, the participants were asked to sit quietly in the MEG system. We recorded four minutes of resting-state electrophysiological activity while the participants rested awake with their eyes closed. MEG data was acquired using a sampling rate of 1,000 Hz and an online anti-alias band-pass filter between 0.1 and 330 Hz. Recordings were processed offline using the temporally extended signal space separation method [45] (Maxfilter software v2.2, correlation limit = 0.9, buffer length = 10 s) to eliminate external magnetic noises. The same algorithm was used to compensate for head motion during the recording.

Ocular, muscular, and flux jump artifacts were identified using 1) the automatic artifact-detection algorithm of the Fieldtrip toolbox [46] for Matlab (R2017b, Mathworks, Inc.) and 2) the visual confirmation of a MEG expert. An independent component-based algorithm was used to remove the effects of EOG and EKG signals from the data, together with external noises. The remaining data was split into 4 s epochs of non-overlapping artifact-free activity. The number of clean epochs did not differ across groups or conditions. Due to the high redundancy found in MEG data after spatial filtering, only magnetometers’ signals were used in the subsequent analyses [47].

Source reconstruction

The epoched data was band-pass filtered in the classical alpha-band [8 –12] Hz using a 1,800^th order FIR filter designed with a Hamming window. We constrained our analyses to the alpha-band since numerous M/EEG studies on SCD agree in the observation of anomalies within this range [15 , 49]. To avoid edge effects, the epochs were padded with 2 s (2,000 samples) of real data on each end, which was removed upon filtering. The source model consisted of 2,459 sources placed in a homogeneous grid of 1 cm in a Montreal Neurological Institute (MNI) template that was converted to subject-space by an affine transformation. Then, each source was anatomically mapped to one of 74 regions of interest defined in the Automated Anatomical Labeling (AAL) atlas [50] (only sources falling into cortical gray matter were considered, i.e., 1,160 of 2,459 sources). The lead field was calculated with a single-shell head model [51] with a unique boundary defined by the inner skull (the combination of white matter, gray matter, and cerebrospinal fluid) generated from the individual T1-weighted MRI using Fieldtrip. Source reconstruction was performed for each participant using a Linearly Constrained Minimum Variance (LCMV) beamformer [52]. Beamformer filters were obtained using the lead field, the epoch-averaged covariance matrix, and a 1% regularization factor.

Connectivity analysis

FC was estimated under the hypothesis of phase synchronization using the phase-locking value (PLV) [53], which demonstrates high reliability across MEG recordings [54]. This measure is based on the assumption that the degree of non-uniformity of phase differences between two time series should be a good estimator of their coupling. To reduce the dimensionality of the FC matrices (1,160 by 1,160, sources by sources), PLVs were averaged following the AAL parcellation to obtain a single PLV between each pair of regions A and B (74 by 74, regions by regions):

$PL V_{A, B} = \frac{1}{N_{A} N_{B}} \sum_{N_{A}} \sum_{N_{B}} | \frac{1}{T} \sum_{t} e^{- j (φ_{A_{k}} (t) - φ_{B_{l}} (t))} |$ (1)

where φ_{A
_k} (t) and φ_{B
_l} (t) are the instantaneous phases of signal A_k and signal B_l at instant t, T is the number of temporal points per epoch, j is the imaginary unit, N_A is the number of sources in region A, and A_k is the kth source inside this region.

Statistical analysis

We performed a mixed-effects ANOVA model with two between-subjects factors, training (trained/non-trained) and diagnosis (HC/SCD), and one within-subject factor time (pre-intervention/post-intervention) indicating the time-point of the MEG recording. We tested the main effects of all factors and every second- and third-order interaction (time^*training, time^*diagnosis, training^*diagnosis, time^*training^*diagnosis). In total, the model was applied over 2,701 FC values (i.e., the symmetrical PLV between the 74 AAL regions). To attain a manageable amount of information, we performed a first analysis focused only on main effects. Its outcome was corrected using a false discovery rate (FDR) [55] (q = 0.05) to address the multiple comparisons problem. Then, the interactions were inspected only for those FC links surviving after correction for multiple comparisons in any of the main effects. This analysis was repeated adjusting for GDS-SF score, educational level, and sex (see Supplementary Material).

The same model was applied to eleven neuropsychological variables with a complete pre- and post-intervention assessment: ROCF-B (memory), Digit Span Test (forward and backward), Texts of Verbal Memory (immediate and delayed), BNT, Phonemic and Semantic Fluency Tests, TMT-A and TMT-B (completion times), and the 7 Minutes Test. Also, we included the MFE to investigate the effect of COGTR on the subjective perception of memory. The ROCF-B (copy), TMT-A (hits), TMT-B (hits), and MMSE were not included as 50% or more of the participants had achieved the best or second-best score at baseline, precluding post-intervention improvement. In this case, we limited the analysis to the interaction between time and training (time^*training) to test whether the COGTR influenced the cognitive performance of the trained group.

Finally, to explore whether the changes in FC observed in the trained groups (HC and SCD) were associated with an improvement in cognitive performance, we searched for a relationship between the relative change in FC and neuropsychological scores using Pearson’s linear correlation coefficients. The same neuropsychological variables outlined above were included in the analysis. The significance level was established at 0.05. All the statistical analyses were performed using Matlab (R2017b, Mathworks, Inc.).

RESULTS

Connectivity analysis

The mixed-effects ANOVA model applied to the FC values showed a significant main effect of time (FDR corrected). Alpha-band FC was increased in the post-intervention stage in 18 links mainly involving temporo-parietal (6), fronto-parietal (4), and temporo-occipital (4) connections (lIFGor – rMOccL: p < 0.0001; lSTG – rCu: p = 0.0001; lSTG – rSOccL: p = 0.0001; lHeschl – rCu: p = 0.0001; lHeschl – rSOccL: p < 0.0001; rCu – rIFGor: p < 0.0001; lCu – lIFGor: p = 0.0002; rPrecu – rIFGt: p = 0.0002; rPrecu – rIFGor: p = 0.0001; rPrecu – rMTG: p = 0.0001; rPrecu – rITG: p < 0.0001; rPrecu – rFusiG: p < 0.0001; rPrecu – rHip: p = 0.0002; rSPG – rMFG: p = 0.0001; rSPG – rIFGor: p = 0.0003; rSPG – rITG: p < 0.0001; rSPG – rFusiG: p = 0.0001; rITG – rMTG: p < 0.0001) (see Fig. 2). The main effects of training and diagnosis were non-significant or did not survive after correction for multiple comparisons. The interaction between time and training (time^*training) was significant (p < 0.05) in nine of the aforementioned 18 links, mainly within temporo-parietal and temporo-occipital connections (lSTG – rCu: p = 0.0218; lSTG – rSOccL: p = 0.0408; lHeschl – rCu: p = 0.0427; lHeschl – rSOccL: p = 0.0461; rPrecu – rMTG: p = 0.0053; rPrecu – rITG: p = 0.0028; rPrecu – rFusiG: p = 0.0021; rPrecu – rHip: p = 0.0001; rITG – rMTG: p = 0.0404) (see Fig. 3). This interaction showed that the main effect of time was more attenuated in the trained group (see Fig. 5 and Supplementary Figure 1).

Fig. 2

FC changes over time. Lines represent significant post-intervention increases in FC (FDR corrected). Left panel: coronal (A), axial (B), and sagittal views (C and D). Right panel: Circular FC diagram. CV, Coronal view; AV, Axial view; SV, Sagittal view; RH, Right hemisphere; LH, Left hemisphere; R, Right; L, Left. For ROI abbreviations, see Table 1.

Fig. 3

Interaction between time and training. Lines represent the links where the interaction between time (pre-intervention or post-intervention) and training (trained or non-trained) is significant (p < 0.05). Left panel: coronal (A), axial (B), and sagittal views (C and D). Right panel: Circular FC diagram. CV, Coronal view; AV, Axial view; SV, Sagittal view; RH, Right hemisphere; LH, Left hemisphere; R, Right; L, Left. For ROI abbreviations, see Table 1.

Table 1

Summary of the FC results

Link	Abbreviation	Main effect	Double interaction	Triple interaction
L Inferior Frontal Gyrus pars opercularis – R Middle Occipital Lobe	lIFGor – rMOccL	^*
L Superior Temporal Gyrus – R Cuneus	lSTG – rCu	^*	^*
L Superior Temporal Gyrus – R Superior Occipital Lobe	lSTG – rSOccL	^*	^*
L Heschl Area – R Cuneus	lHeschl – rCu	^*	^*
L Heschl Area – R Superior Occipital Lobe	lHeschl – rSOccL	^*	^*
R Cuneus – R Inferior Frontal Gyrus pars opercularis	rCu – rIFGor	^*
L Cuneus – L Inferior Frontal Gyrus pars opercularis	lCu – lIFGor	^*		^*
R Precuneus – R Inferior Frontal Gyrus pars triangularis	rPrecu – rIFGt	^*
R Precuneus – R Inferior Frontal Gyrus pars opercularis	rPrecu – rIFGor	^*
R Precuneus – R Middle Temporal Gyrus	rPrecu – rMTG	^*	^*	^*
R Precuneus – R Inferior Temporal Gyrus	rPrecu – rITG	^*	^*	^*
R Precuneus – R Fusiform Gyrus	rPrecu – rFusiG	^*	^*	^*
R Precuneus – R Hippocampus	rPrecu – rHip	^*	^*	^*
R Superior Parietal Gyrus – R Medial Frontal Gyrus	rSPG – rMFG	^*
R Superior Parietal Gyrus – R Inferior Frontal Gyrus pars opercularis	rSPG – rIFGor	^*
R Superior Parietal Gyrus – R Inferior Temporal Gyrus	rSPG – rITG	^*
R Superior Parietal Gyrus – R Fusiform Gyrus	rSPG – rFusiG	^*
R Inferior Temporal Gyrus – R Middle Temporal Gyrus	rITG – rMTG	^*	^*	^*

Asterisks indicate the significant links for the main effect of time, the double interaction between time and training, and the triple interaction between time, training, and diagnosis.

Additionally, the interaction between time, training, and diagnosis (time^*training^*diagnosis) was significant (p < 0.05) in six of the 18 links. These links mainly involved associations between the right precuneus and nodes of the right temporal lobe (lCu – lIFGor: p = 0.0383; rPrecu – rMTG: p = 0.0101; rPrecu – rITG: p = 0.0110; rPrecu – rFusiG: p = 0.0236; rPrecu – rHip: p = 0.0083; rITG – rMTG: p = 0.0328) (see Fig. 4). This interaction showed that the main effect of time followed a separate trend in the trained SCD group. Trained SCD participants showed lower FC in the post- rather than in the pre-intervention stage, which was not observed in the trained HC group (see Fig. 5 and Supplementary Figure 1).

Fig. 4

Interaction between time, training, and diagnosis. Lines represent the links where the interaction between time (pre-intervention or post-intervention), training (trained or non-trained), and diagnosis (HC or SCD) is significant (p < 0.05). Left panel: coronal (A), axial (B), and sagittal views (C and D). Right panel: Circular FC diagram. CV, Coronal view; AV, Axial view; SV, Sagittal view; RH, Right hemisphere; LH, Left hemisphere; R, Right; L, Left. For ROI abbreviations, see Table 1.

Fig. 5

Selected boxplots of FC values in the two MEG recordings. Boxplots of the FC values in the two MEG recordings for the links with the right precuneus where the interaction between time (pre-intervention or post-intervention), training (trained or non-trained), and diagnosis (HC or SCD) was significant (p < 0.05). For each significant link, the upper panel represents pre- and post-intervention FC values for the whole sample. The middle panel represents pre- and post-intervention FC values separately for non-trained participants (left) and trained participants (right). The lower panel represents pre- and post-intervention FC values separately for the non-trained HC group (upper row, left), non-trained SCD group (lower row, left), trained HC group (upper row, right), and trained SCD group (lower row, right). The boxplots inform of each participant’s FC values, the group median, and the boundaries of the quartiles. The colored squares next to the abbreviated names of the links emphasize the significant results for that connection: red (main effect of time), green (interaction between time and training), purple (interaction between time, training, and diagnosis).

The results described here are summarized in Table 1.

Cognitive outcomes

The mixed-effects ANOVA model applied to the neuropsychological scores showed that the interaction between time and training (time^*training) was non-significant for all the neuropsychological variables included in the analysis (ROCF-B (memory score): F = 0.510, p = 0.477; ROCF-B (memory time): F = 0.459, p = 0.500; Digit Span Test (forward): F = 1.282, p = 0.261; Digit Span Test (backward): F = 2.382, p = 0.127; Texts of Verbal Memory (immediate): F = 3.917, p = 0.051; Texts of Verbal Memory (delayed): F = 1.565, p = 0.215; BNT: F = 0.294, p = 0.589; Phonemic Fluency Test: F = 1.070, p = 0.304; Semantic Fluency Test: F = 0.808, p = 0.371; TMT-A (completion time): F = 0.010, p = 0.919; TMT-B (completion time): F = 2.940, p = 0.090; 7 Minutes Test: F = 0.111, p = 0.740; MFE: F = 1.361, p = 0.247). As such, we did not find any significant changes across sessions based on the neuropsychological scores of our sample.

Correlation analysis

The correlation analysis between the relative change in FC and neuropsychological scores was carried out for each trained group (HC and SCD) independently. The aim of this analysis was to explore whether the change in FC found in the trained groups (i.e., maintained/lower post-intervention FC) was significantly associated with an improvement in cognitive performance, particularly in the case of SCD participants for whom the effect of COGTR was stronger. The relative change in FC was calculated for the links in which the interaction between time, training, and diagnosis was significant (see Table 1). We found significant (p < 0.05) correlations in the trained SCD group in five of the neuropsychological variables included in the analysis. It is important to mention that none of these correlations survived after correction for multiple comparisons (q = 0.05), however, we consider that they convey relevant information for the cautious understanding of the FC results, given the consistency of the signs and the quality of the scatter plots (see Supplementary Figure 2). Changes in FC in the trained SCD group were significantly associated with improved scores in the following cognitive tests: 1) ROCF-B (memory score) (lCu – lIFGor (ρ= –0.4901; p = 0.0151)); 2) Digit Span Test (forward) (rPrecu – rITG (ρ= –0.4472; p = 0.0324)); 3) BNT (rPrecu – rHip (ρ= –0.4914; p = 0.0237)); rPrecu – rITG (ρ= –0.4487; p = 0.0413)); 4) TMT-A (completion time) (rPrecu – rITG (ρ= 0.4785; p = 0.0180)); 5) TMT-B (completion time) (rPrecu – rFusiG (ρ= 0.5024; p = 0.0124)). No significant associations were found for the trained HC group.

DISCUSSION

In this study, we evaluated the potential of COGTR to modulate FC in a sample of HC and SCD participants. We found an average increase in alpha-band FC over time, but the effect was different in each group (trained and non-trained). In the trained group (HC and SCD), we report a reduction in the increase in FC within temporo-parietal and temporo-occipital connections. In the trained SCD group, this reduction was stronger and showed a tentative correlation with improved performance in different cognitive tests. Our results suggest that COGTR interventions could mitigate aberrant increases in FC in preclinical AD, promoting brain synchrony normalization in groups at a higher risk of developing dementia.

Our statistical analysis of FC showed a main effect of time so that FC was higher after an experimental interval in which half of the participants completed a COGTR intervention. This result implied that there was a direct relationship between the increase in FC and the time interval between MEG recordings, which was close to half a year. Other factors, such as the more relaxed experience of a second MEG recording, could also contribute to the change in FC. Nevertheless, if this prevailed, we would not expect any interactions, and therefore, we assume that this factor is not driving our results. On the other hand, the observation of increasing FC as a function of time was fairly consistent with recent findings of FC trajectories in normal aging [56] and early-stage AD [57,58, 57,58]. For instance, Staffaroni showed that healthy seniors demonstrate an early hyperconnectivity phase followed by a hypoconnectivity phase from age 74 onward [56], while others found that normal aging (up to age 75) was characterized by ubiquitous age-related increases in FC [59]. This agreement between our results and FC trends in early aging could indicate that the increase in post-intervention FC was partially influenced by different age-related changes at the neural, vascular, and metabolic levels. On the other hand, several authors reported mixed and reduced patterns of FC in aging [60 –62], although these discrepancies could be partly explained by the more limited strength of cross-sectional designs to infer all the complexity of FC trajectories [56].

At the same time, previous studies showed that hyperconnectivity was associated with several risk factors for AD in cognitively normal individuals [16,17,63–65 , 16,17,63–65]. During the very early silent stages of AD, there is a progressive accumulation of various neurotoxic forms of Aβ (i.e., oligomeric assemblies and fibril), which are strongly associated with the loss of GABAergic synapses and the subsequent phenomena of neuronal hyperexcitability [7 –10]. It is assumed that such context of increased excitability could exacerbate the likelihood of large-scale synchrony between brain regions, hence producing observations of abnormally increased FC [57]. Brain hypersynchrony is thought to represent an early event in the disease that is already detected, e.g., in intact adults with a first-degree family history of AD [17]. Likewise, in individuals with SCD, several studies have reported hyperconnections within regions of the parietal and occipital lobe [66], the default mode network (DMN) [16], the medial visual network [16], and between the posterior DMN and the medial temporal lobe [65]. In this group, increasing FC might also be evaluated under the framework of the ‘X’ model [57], which proposes that prodromal patients manifest patterns of hyperconnectivity until their clinical progression to dementia.

Based on the above, it is conceivable that hyperconnectivity in SCD participants could be mediated by early disease mechanisms of network disruption, such as Aβ pathology. This view is consistent with the specific topology of our results, which included some of the earliest regions to demonstrate signs of AD neuropathology. In particular, the right precuneus showed increased FC in one-third of the significant links and was the most hyperconnected node in our analyses. This node is a key region in AD due to its early vulnerability to Aβ deposition [67], which was found to induce local FC increases in individuals at risk [68,69, 68,69], as well as reduced self-confidence in memory abilities [70]. The increased FC between the right precuneus and the right medial temporal lobe (i.e., hippocampus) was also in line with the archetypical course of AD progression [67] and reproduced previous findings in SCD [16, 65]. In contrast, the increased FC of the inferior frontal gyrus was more intriguing but could relate to some neuroprotective role of frontal regions in the context of incipient cognitive decline [71]. Other results affected core structures of the DMN (i.e., pre/cuneus, superior parietal gyrus, and superior temporal gyrus), a critical brain network in aging due to its association with age- and disease-related changes and its role in memory processing [67]. These results comprised increases in DMN FC that matched previous observations in normal [56] and SCD individuals [16]. Finally, we found hyperconnectivity in some temporal loci (i.e., inferior temporal gyrus, middle temporal gyrus, and fusiform gyrus) of early tauopathy [72], which has been strongly associated with impaired alpha-band synchrony in AD cases [73], and therefore, could reflect a different pathway of incipient neurodegeneration.

The main finding in our study revealed that the increase in FC between MEG recordings followed separate trends in two cases. First, we found a reduction in the increase in FC in the case of the trained group. In this group, the increase in FC was attenuated in one-half of the original links, specifically between temporal, parietal, and occipital regions. Second, we found that this reduction was stronger in the case of SCD participants. In the trained SCD group, FC was lower in the post- rather than in the pre-intervention stage, which was not observed in HC counterparts. Region-wise, this result revealed lower FC between the right precuneus and nodes of the right medial (hippocampus) and lateral temporal lobe (inferior temporal gyrus, middle temporal gyrus, and fusiform gyrus). Taking into account the distinct vulnerability of these regions, we speculate that the COGTR may have helped to mitigate some abnormal hyperconnections that could ensue from the precuneus as a hub of incipient neuropathology. Furthermore, since the effect of COGTR was stronger in the SCD group, we might assume that SCD participants provided ‘more room’ for COGTR to neutralize early signs of the disease, which was substantiated by a greater reduction in FC. Apart from that, we found that the effect of COGTR was predominantly lateralized to the right hemisphere, which was intriguing as there is some evidence of early laterality in the development of the disease within some of our regions of interest, namely the right precuneus, medial temporal lobe, and lateral temporal lobe [74, 75]. Accordingly, it is possible that a greater neuropathological burden in the right hemisphere could also result in a greater effect of COGTR on this side in order to counteract impaired neural functioning and cognitive decline. Although some authors have described COGTR-related decreases as mostly right-lateralized [76], the specific reason for this is still unclear, and extra research, especially including AD biomarkers, is required to clarify the exact mechanisms underlying the COGTR modulation of FC and its relationship with potential asymmetries.

The aforementioned FC results also remained fairly unchanged after adjustment for additional covariates (i.e., GDS-SF, educational level, and sex). After adjusting for these variables, some effects lost significance but retained borderline significance in most cases (see Supplementary Material for details). Although the inclusion of covariates in the model provides better coverage of potential confounders, it might also compromise the statistical power of the adjusted analyses, which could explain the slight divergences found in our results.

It must be pointed out that there were some differences in baseline FC between the SCD groups (i.e., higher baseline FC in the trained SCD group) in three links in which the triple interaction was significant (rPrecu – rMTG, rPrecu – rITG, rITG – rMTG). One could argue that those differences could contribute to the results of the trained SCD group (i.e., lower post-intervention FC); however, we would dissent. As it has been discussed, brain synchrony (and, in turn, FC) is envisaged to increase throughout the early stages of the disease [11]. In this framework, it seems unlikely that FC in trained SCD participants could have reached some upper limit that would revert its hyperconnectivity trend without external intervention, even if its value at baseline was higher than in non-trained counterparts. This consideration is also supported by the fact that trained SCD and non-trained HC participants showed comparable values of FC at baseline, and yet non-trained HC participants show increased post-intervention FC. Based on this, it is reasonable to assume that the reduction in post-intervention FC found in the trained SCD group was driven by the effect of COGTR. Furthermore, since for half of the links there were no differences in baseline FC between the SCD groups, this argument seems to provide a more comprehensive explanation of our results.

Finally, we carried out a correlation analysis to clarify whether the COGTR-related changes in FC were indicative of cognitive improvement. In the trained SCD group, we found some associations between the reduction in FC and improved execution in different cognitive tests. Nevertheless, as none of them survived after correction for multiple comparisons, their interpretation should be treated with some caution. Specifically, better BNT performance correlated with reduced FC between the right precuneus and the right hippocampus/inferior temporal gyrus. These regions are known to play a role in some of the abilities tested with the BNT, i.e., semantic processing and retrieval [77, 78] and confrontational naming [79], hence this result aligns with our consideration that COGTR exerted its positive effect through a reduction in FC. Likewise, it is conceivable that the gains in visual attention (measured with the TMT) could have been mediated by the normalization of FC between the right precuneus and the inferior temporal visual association cortex, since the visual scanning component of the TMT may rest upon the functional integrity of this pathway [80]. Moreover, the change in ROCF-B scores showed a positive correlation with reduced hyperconnectivity in some brain regions involved in important abilities for accurate visual recall, such as attention [81] and visuospatial memory [82]. The aforementioned domains had been addressed during the COGTR intervention, although we could not find evidence of cognitive improvement from the neuropsychological data. This result was not fully unexpected, because although the ‘UMAM’ method had been validated in a large cohort of individuals with mixed degrees of disease severity [83], there was less evidence of its efficacy on SCD individuals [84]. In this respect, it is important to remember that SCD individuals exhibit a self-perceived cognitive decline that virtually goes unnoticed in standardized cognitive testing [31]. This singularity could challenge the detection of objective improvement following COGTR, as evidenced by the generalized small effect sizes found after non-pharmacological interventions in SCD samples [20]. At the same time, we could not find a COGTR-related change in the subjective perception of memory (measured with the MFE), which could imply that short-term effects might be too subtle to overcome long-lasting feelings of one’s own memory performance. Ultimately, electrophysiological evidence seems to precede detectability by cognitive testing, a finding that highlights the interest in pursuing neuroimaging studies to assess the efficacy of cognitive interventions.

Overall, our study offered novel evidence suggesting that COGTR interventions could contribute to normalizing brain synchrony in individuals at a higher risk of developing dementia. As with the majority of research, our design was also subjected to some limitations. First, it could have been interesting to include another assessment immediately after the COGTR intervention. This time-point could give insight into the development of short-term outcomes that might evolve differently or be no longer detected after the 4-month interval that separates the conclusion of the COGTR and the post-intervention appointment. At the same time, some of the effects of COGTR could occur on even longer time scales (> 1 year), and therefore, it is also plausible that we missed out on certain electrophysiological and cognitive changes that might only become discernible after a longer period of time [20]. Moreover, future studies should ideally include an active control group alongside the passive control (wait-list) group. Passive control accounts for practice effects in the sessions (e.g., data acquisition), but misses out on non-specific confounds of the COGTR intervention that could play a role in the study outcomes (experimenter contact, increased socialization, motivation, etc.). In this respect, well-designed active control paradigms could target such confounding effects without interfering with relevant study variables (here, the use of non-cognitive training designs would be proper). Although the debate on effective control is ongoing, there is some support for the combination of active-passive control groups to better measure extraneous effects [85]. Finally, the inclusion of cerebrospinal fluid biomarkers and amyloid PET information could greatly improve our understanding of the precise effects of COGTR on functional electrophysiology, but unfortunately, such information was not available for this cohort.

Footnotes

ACKNOWLEDGMENTS

This work was supported by the Spanish Ministry of Economy and Competitiveness under the grant PSI2015-68793-C3-1-R. The authors would also like to acknowledge the European Union’s Horizon 2020 Research and Innovation Action project Virtual Brain Cloud (826421), in which FM and IS-M participate, and which IS-M acknowledges as financial support; and the Community of Madrid project Neurocentro (B2017/BMD-3760). DL-S acknowledges financial support from the program Juan de la Cierva-Formación of the Spanish Ministry of Science, Innovation, and Universities under the grant FJC2018-037401-I.

Authors’ disclosures available online ().

The supplementary material is available in the electronic version of this article: .

References

Prince

, Wimo

, Guerchet

, Ali

G-C

, Wu

Y-T

, Prina

(2015) World Alzheimer Report 2015. The global impact of dementia: An analysis of prevalence, incidence, cost and trends. Alzheimer’s Disease International, London.

Alzheimer’s Disease International (2019) World Alzheimer Report 2019. Attitudes to dementia. Alzheimer’s Disease International, London.

Brookmeyer

, Johnson

, Ziegler-Graham

, Arrighi

(2007) Forecasting the global burden of Alzheimer’s disease. Alzheimers Dement 3, 186–191.

Cummings

, Lee

, Ritter

, Sabbagh

, Zhong

(2020) Alzheimer’s disease drug development pipeline: 2020. Alzheimers Dement 6, e12050.

Cummings

, Morstorf

, Zhong

(2014) Alzheimer’s disease drug-development pipeline: Few candidates, frequent failures. Alzheimers Res Ther 6, 37.

Mahase

(2021) FDA approves controversial Alzheimer’s drug despite uncertainty over effectiveness. BMJ Open 373, n1462.

García-Marín

, Blázquez-Llorca

, Rodríguez

, Boluda

, Muntane

, Ferrer

, DeFelipe

(2009) Diminished perisomatic GABAergic terminals on cortical neurons adjacent to amyloid plaques. Front Neuroanat 3, 28.

Ambrad Giovannetti

, Fuhrmann

(2019) Unsupervised excitation: GABAergic dysfunctions in Alzheimer’s disease. Brain Res 1707, 216–226.

Ulrich

(2015) Amyloid-β impairs synaptic inhibition via GABAA receptor endocytosis. J Neurosci 35, 9205–9210.

10.

Busche

, Konnerth

(2016) Impairments of neural circuit function in Alzheimer’s disease. Philos Trans R Soc B Biol Sci 371, 20150429.

11.

Maestú

, de Haan

, Busche

, DeFelipe

(2021) Neuronal excitation/inhibition imbalance: Core element of a translational perspective on Alzheimer pathophysiology. Ageing Res Rev 69, 101372.

12.

Maestú

, Peña

, Garcés

, González

, Bajo

, Bagic

, Cuesta

, Funke

, Mäkelä

, Menasalvas

, Nakamura

, Parkkonen

, López

, Del Pozo

, Sudre

, Zamrini

, Pekkonen

, Henson

, Becker

(2015) A multicenter study of the early detection of synaptic dysfunction in mild cognitive impairment using magnetoencephalography-derived functional connectivity. Neuroimage Clin 9, 103–109.

13.

Bajo

, Maestú

, Nevado

, Sancho

, Gutiérrez

, Campo

, Castellanos

, Gil

, Moratti

, Pereda

, Del-Pozo

(2010) Functional connectivity in mild cognitive impairment during a memory task: Implications for the disconnection hypothesis. J Alzheimers Dis 22, 183–193.

14.

Larner

(2010) Epileptic seizures in AD patients. Neuromolecular Med 12, 71–77.

15.

López-Sanz

, Bruña

, Garcés

, Martín-Buro

, Walter

, Delgado

, Montenegro

, López-Higes

, Marcos

, Maestú

(2017) Functional connectivity disruption in subjective cognitive decline and mild cognitive impairment: A common pattern of alterations. Front Aging Neurosci 9, 109.

16.

Hafkemeijer

, Altmann-Schneider

, Oleksik

, van de Wiel

, Middelkoop

HAM

, van Buchem

, van der Grond

, Rombouts

SARB

(2013) Increased functional connectivity and brain atrophy in elderly with subjective memory complaints. Brain Connect 3, 353–62.

17.

Ramírez-Toraño

, Bruña

, de Frutos-Lucas

, Rodríguez-Rojo

, Marcos de Pedro

, Delgado-Losada

, Gómez-Ruiz

, Barabash

, Marcos

, López-Higes

, Maestú

(2021) Functional connectivity hypersynchronization in relatives of Alzheimer’s disease patients: An early E/I balance dysfunction? Cereb Cortex 31, 1201–1210.

18.

Clare

, Woods

, Moniz Cook

, Orell

, Spector

(2003) Cognitive rehabilitation and cognitive training for early-stage Alzheimer’s disease and vascular dementia. Cochrane Database Syst Rev, CD003260.

19.

Butler

, McCreedy

, Nelson

, Desai

, Ratner

, Fink

, Hemmy

, McCarten

, Barclay

, Brasure

, Davila

, Kane

(2018) Does cognitive training prevent cognitive decline? A systematic review. Ann Intern Med 168, 63–68.

20.

Smart

, Karr

, Areshenkoff

, Rabin

, Hudon

, Gates

, Ali

, Arenaza-Urquijo

, Buckley

, Chételat

, Hampel

, Jessen

, Marchant

, Sikkes

SAM

, Tales

, van der Flier

, Wesselman

, and the Subjective Cognitive Decline Initiative (SCD-I) Working Group (2017) Non-pharmacologic interventions for older adults with subjective cognitive decline: Systematic review, meta-analysis, and preliminary recommendations. Neuropsychol Rev 27, 245–257.

21.

Belleville

(2008) Cognitive training for persons with mild cognitive impairment. Int Psychogeriatrics 20, 57–66.

22.

Kallio

, Öhman

, Kautiainen

, Hietanen

, Pitkälä

(2017) Cognitive training interventions for patients with Alzheimer’s disease: A systematic review. J Alzheimers Dis 56, 1349–1372.

23.

Chapman

, Aslan

, Spence

, Hart

, Bartz

, Didehbani

, Keebler

, Gardner

, Strain

, Defina

, Lu

(2015) Neural mechanisms of brain plasticity with complex cognitive training in healthy seniors. Cereb Cortex 25, 396–405.

24.

Mozolic

, Hayasaka

, Laurienti

(2010) A cognitive training intervention increases resting cerebral blood flow in healthy older adults. Front Hum Neurosci 4, 16.

25.

Barban

, Mancini

, Cercignani

, Adriano

, Perri

, Annicchiarico

, Carlesimo

, Ricci

, Lombardi

, Teodonno

, Serra

, Giulietti

, Fadda

, Federici

, Caltagirone

, Bozzali

(2017) A pilot study on brain plasticity of functional connectivity modulated by cognitive training in mild Alzheimer’s disease and mild cognitive impairment. Brain Sci 7, 50.

26.

Belleville

, Clément

, Mellah

, Gilbert

, Fontaine

, Gauthier

(2011) Training-related brain plasticity in subjects at risk of developing Alzheimer’s disease. Brain 134(Pt 6), 1623–1634.

27.

Hampstead

, Stringer

, Stilla

, Deshpande

, Hu

, Moore

, Sathian

(2011) Activation and effective connectivity changes following explicit-memory training for face-name pairs in patients with mild cognitive impairment: A pilot study. Neurorehabil Neural Repair 25, 210–222.

28.

Hampstead

, Stringer

, Stilla

, Giddens

, Sathian

(2012) Mnemonic strategy training partially restores hippocampal activity in patients with mild cognitive impairment. Hippocampus 22, 1652–1658.

29.

Rosen

, Sugiura

, Kramer

, Whitfield-Gabrieli

, Gabrieli

(2011) Cognitive training changes hippocampal function in mild cognitive impairment: A pilot study. J Alzheimers Dis 26 Suppl 3, 349–357.

30.

, He

, Qiao

, Xu

, Lu

, Cui

, Ling

, Yan

, Tang

, Chen S

(2019) Computerized cognitive training for Chinese mild cognitive impairment patients: A neuropsychological and fMRI study. Neuroimage Clin 22, 101691.

31.

Jessen

, Amariglio

, van Boxtel

, Breteler

, Ceccaldi

, Chételat

, Dubois

, Dufouil

, Ellis

, van der Flier

, Glodzik

, van Harten

, de Leon

, McHugh

, Mielke

, Molinuevo

, Mosconi

, Osorio

, Perrotin

, Petersen

, Rabin

, Rami

, Reisberg

, Rentz

, Sachdev

, de la Sayette

, Saykin

, Scheltens

, Shulman

, Slavin

, Sperling

, Stewart

, Uspenskaya

, Vellas

, Visser

, Wagner

, Subjective Cognitive Decline Initiative (SCD-I) Working Group (2014) A conceptual framework for research on subjective cognitive decline in preclinical Alzheimer’s disease. Alzheimers Dement 10, 844–852.

32.

Oldfield

(1971) The assessment and analysis of handedness: The Edinburgh inventory. Neuropsychologia 9, 97–113.

33.

Luzzi

, Pesallaccia

, Fabi

, Muti

, Viticchi

, Provinciali

, Piccirilli

(2011) Non-verbal memory measured by Rey-Osterrieth Complex Figure B: Normative data. Neurol Sci 32, 1081–1089.

34.

Wechsler

(2004) Escala de memoria de Wechsler-III: manual técnico, TEA Ediciones.

35.

Kaplan

, Goodglas

, Weintraub

(2001) The Boston Naming Test, Lippincott Williams & Wilkins.

36.

Benton

, Hamsher

, Sivan

(1983) Multilingual aphasia examination: manual of instructions, Department of Neurology and Psychology, University of Iowa.

37.

Reitan

(1958) Validity of the Trail Making Test as an indicator of organic brain damage. Percept Mot Skills 8, 271–276.

38.

Folstein

, Folstein

, McHugh

(1975) “Mini-mental state”. A practical method for grading the cognitive state of patients for the clinician. J Psychiatr Res 12, 189–198.

39.

Solomon

, Hirschoff

, Kelly

, Relin

, Brush

, DeVeaux

, Pendlebury

(1998) A 7 minute neurocognitive screening battery highly sensitive to Alzheimer’s disease. Arch Neurol 55, 349–355.

40.

Wilson

, Cockburn

, Baddeley

(1985) The Rivermead Behavioural Memory Test, Thames Valley Test Co., Reading, UK.

41.

Montejo Carrasco

, Montenegro Peña

, Sueiro

(2012) The memory failures of everyday questionnaire (MFE): Internal consistency and reliability. Span J Psychol 15, 768–776.

42.

Pfeffer

, Kurosaki

, Harrah

, Chance

, Filos

(1982) Measurement of functional activities in older adults in the community. J Gerontol 37, 323–329.

43.

Sheikh

, Yesavage

(1986) Geriatric Depression Scale (GDS): Recent evidence and development of a shorter version. Clin Gerontol J Aging Ment Health 5, 165–173.

44.

Montejo Carrasco

, Montenegro Peña

, Reinoso García

, de Andrés Montes

, Claver Martín

(2001) Programa de Memoria. Método UMAM, Ediciones Díaz de Santos, S.A.

45.

Taulu

, Hari

(2009) Removal of magnetoencephalographic artifacts with temporal signal-space separation: Demonstration with single-trial auditory-evoked responses. Hum Brain Mapp 30, 1524–1534.

46.

Oostenveld

, Fries

, Maris

, Schoffelen

J-M

(2011) FieldTrip: Open source software for advanced analysis of MEG, EEG, and invasive electrophysiological data. Comput Intell Neurosci 2011, 156869.

47.

Garcés

, López-Sanz

, Maestú

, Pereda

(2017) Choice of magnetometers and gradiometers after signal space separation. Sensors (Basel) 17, 2926.

48.

López-Sanz

, Bruña

, Garcés

, Cámara

, Serrano

, Rodríguez-Rojo

, Delgado

, Montenegro

, López-Higes

, Yus

, Maestú

(2016) Alpha band disruption in the AD-continuum starts in the subjective cognitive decline stage: A MEG study. Sci Rep 6, 37685.

49.

Babiloni

, López

, Del Percio

, Noce

, Pascarelli

, Lizio

, Teipel

, González-Escamilla

, Bakardjian

, George

, Cavedo

, Lista

, Chiesa

, Vergallo

, Lemercier

, Spinelli

, Grothe

, Potier

, Stocchi

, Ferri

, Habert

, Fraga

, Dubois

, Hampel

(2020) Resting-state posterior alpha rhythms are abnormal in subjective memory complaint seniors with preclinical Alzheimer’s neuropathology and high education level: The INSIGHT-preAD study. Neurobiol Aging 90, 43–59.

50.

Tzourio-Mazoyer

, Landeau

, Papathanassiou

, Crivello

, Etard

, Delcroix

, Mazoyer

, Joliot

(2002) Automated anatomical labeling of activations in SPM using a macroscopic anatomical parcellation of the MNI MRI single-subject brain. Neuroimage 15, 273–289.

51.

Nolte

(2003) The magnetic lead field theorem in the quasi-static approximation and its use for magnetoencephalography forward calculation in realistic volume conductors. Phys Med Biol 48, 3637–3652.

52.

Van Veen

, van Drongelen

, Yuchtman

, Suzuki

(1997) Localization of brain electrical activity via linearly constrained minimum variance spatial filtering. IEEE Trans Biomed Eng 44, 867–880.

53.

Lachaux

, Rodríguez

, Martinerie

, Varela

(1999) Measuring phase synchrony in brain signals. Hum Brain Mapp 8, 194–208.

54.

Garcés

, Martín-Buro

, Maestú

(2016) Quantifying the test-retest reliability of magnetoencephalography resting-state functional connectivity. Brain Connect 6, 448–460.

55.

Benjamini

, Hochberg

(1995) Controlling the false discovery rate: A practical and powerful approach to multiple testing. J R Stat Soc Series B Methodol 57, 289–300.

56.

Staffaroni

, Brown

, Casaletto

, Elahi

, Deng

, Neuhaus

, Cobigo

, Mumford

, Walters

, Saloner

, Karydas

, Coppola

, Rosen

, Miller

, Seeley

, Kramer

(2018) The longitudinal trajectory of default mode network connectivity in healthy older adults varies as a function of age and is associated with changes in episodic memory and processing speed. J Neurosci 38, 2809–2817.

57.

Pusil

, López

, Cuesta

, Bruña

, Pereda

, Maestú

(2019) Hypersynchronization in mild cognitive impairment: The “X” model. Brain 142, 3936–3950.

58.

Chiesa

, Cavedo

, Vergallo

, Lista

, Potier

M-C

, Habert

M-O

, Dubois

, Thiebaut de Schotten

, Hampel

, INSIGHT-preAD study group, Alzheimer Precision Medicine Initiative (APMI) (2019) Differential default mode network trajectories in asymptomatic individuals at risk for Alzheimer’s disease. Alzheimers Dement 15, 940–950.

59.

Ferreira

, Regina

ACB

, Kovacevic

, Martin

MDGM

, Santos

, Carneiro

CDG

, Kerr

, Amaro

, Mcintosh

, Busatto

(2016) Aging effects on whole-brain functional connectivity in adults free of cognitive and psychiatric disorders. Cereb Cortex 26, 3851–3865.

60.

Tomasi

, Volkow

(2012) Aging and functional brain networks. Mol Psychiatry 17, 549–558.

61.

Geerligs

, Renken

, Saliasi

, Maurits

, Lorist

(2015) A brain-wide study of age-related changes in functional connectivity. Cereb Cortex 25, 1987–1999.

62.

Sala-Llonch

, Bartrés-Faz

, Junqué

(2015) Reorganization of brain networks in aging: A review of functional connectivity studies. Front Psychol 6, 663.

63.

Sheline

, Morris

, Snyder

, Price

, Yan

, D’Angelo

, Liu

, Dixit

, Benzinger

, Fagan

, Goate

, Mintun

(2010) APOE4 allele disrupts resting state fMRI connectivity in the absence of amyloid plaques or decreased CSF Aβ42. J Neurosci 30, 17035–17040.

64.

Lim

, Nebes

, Snitz

, Cohen

, Mathis

, Price

, Weissfeld

, Klunk

, Aizenstein

(2014) Regional amyloid burden and intrinsic connectivity networks in cognitively normal elderly subjects. Brain 137, 3327–3338.

65.

Verfaillie

SCJ

, Pichet Binette

, Vachon-Presseau

, Tabrizi

, Savard

, Bellec

, Ossenkoppele

, Scheltens

, van der Flier

, Breitner

JCS

, Villeneuve

, PREVENT-AD Research Group (2018) Subjective cognitive decline is associated with altered default mode network connectivity in individuals with a family history of Alzheimer’s disease. Biol Psychiatry 3, 463–472.

66.

Kawagoe

, Onoda

, Yamaguchi

(2019) Subjective memory complaints are associated with altered resting-state functional connectivity but not structural atrophy. Neuroimage Clin 21, 101675.

67.

Buckner

, Snyder

, Shannon

, LaRossa

, Sachs

, Fotenos

, Sheline

, Klunk

, Mathis

, Morris

, Mintun

(2005) Molecular, structural, and functional characterization of Alzheimer’s disease: Evidence for a relationship between default activity, amyloid, and memory. J Neurosci 25, 7709–7717.

68.

Jack

, Wiste

, Weigand

, Knopman

, Lowe

, Vemuri

, Mielke

, Jones

, Senjem

, Gunter

, Gregg

, Pankratz

, Petersen

(2013) Amyloid-first and neurodegeneration-first profiles characterize incident amyloid PET positivity. Neurology 81, 1732–1740.

69.

Nakamura

, Cuesta

, Kato

, Arahata

, Iwata

, Yamagishi

, Kuratsubo

, Kato

, Bundo

, Diers

, Fernández

, Maestú

, Ito

(2017) Early functional network alterations in asymptomatic elders at risk for Alzheimer’s disease. Sci Rep 7, 6517.

70.

Perrotin

, Mormino

, Madison

, Hayenga

, Jagust

(2012) Subjective cognition and amyloid deposition imaging: A Pittsburgh compound B positron emission tomography study in normal elderly individuals. Arch Neurol 69, 223–229.

71.

Lin

, Ren

, Lo

, Chapman

, Jacobs

, Baran

, Porsteinsson

, Foxe

, Alzheimer’s Disease Neuroimaging Initiative (2017) Insula and inferior frontal gyrus’ activities protect memory performance against Alzheimer’s disease pathology in old age. J Alzheimers Dis 55, 669–678.

72.

Johnson

, Schultz

, Betensky

, Becker

, Sepulcre

, Rentz

, Mormino

, Chhatwal

, Amariglio

, Papp

, Marshall

, Albers

, Mauro

, Pepin

, Alverio

, Judge

, Philiossaint

, Shoup

, Yokell

, Dickerson

, Gomez-Isla

, Hyman

, Vasdev

, Sperling

(2016) Tau positron emission tomographic imaging in aging and early Alzheimer disease. Ann Neurol 79, 110–119.

73.

Ranasinghe

, Cha

, Iaccarino

, Hinkley

, Beagle

, Pham

, Jagust

, Miller

, Rankin

, Rabinovici

, Vossel

, Nagarajan

(2020) Neurophysiological signatures in Alzheimer’s disease are distinctly associated with TAU, amyloid-β accumulation, and cognitive decline. Sci Transl Med 12, eaaz4069.

74.

Scheef

, Spottke

, Daerr

, Joe

, Striepens

, Kölsch

, Popp

, Daamen

, Gorris

, Heneka

, Boecker

, Biersack

, Maier

, Schild

, Wagner

, Jessen

(2012) Glucose metabolism, gray matter structure, and memory decline in subjective memory impairment. Neurology 79, 1332–1339.

75.

Dong

Q-Y

, Li

T-R

, Jiang

X-Y

, Wang

X-N

, Han

, Jiang

J-H

(2021) Glucose metabolism in the right middle temporal gyrus could be a potential biomarker for subjective cognitive decline: A study of a Han population. Alzheimers Res Ther 13, 74.

76.

Duda

, Sweet

(2020) Functional brain changes associated with cognitive training in healthy older adults: A preliminary ALE meta-analysis. Brain Imaging Behav 14, 1247–1262.

77.

Burianova

, Grady

(2007) Common and unique neural activations in autobiographical, episodic, and semantic retrieval. J Cogn Neurosci 19, 1520–1534.

78.

Glikmann-Johnston

, Oren

, Hendler

, Shapira-Lichter

(2015) Distinct functional connectivity of the hippocampus during semantic and phonemic fluency. Neuropsychologia 69, 39–49.

79.

Obler

, Rykhlevskaia

, Schnyer

, Clark-Cotton

, Spiro

, Hyun

, Kim

D-S

, Goral

, Albert

(2010) Bilateral brain regions associated with naming in older adults. Brain Lang 113, 113–123.

80.

Armstrong

, An

, Shin

, Williams

, Doshi

, Erus

, Davatzikos

, Ferrucci

, Beason-Held

, Resnick

(2020) Associations between cognitive and brain volume changes in cognitively normal older adults. Neuroimage 223, 117289.

81.

Milham

, Erickson

, Banich

, Kramer

, Webb

, Wszalek

, Cohen

(2002) Attentional control in the aging brain: Insights from an fMRI study of the stroop task. Brain Cogn 49, 277–296.

82.

Mitolo

, Gardini

, Fasano

, Crisi

, Pelosi

, Pazzaglia

, Caffarra

(2013) Visuospatial memory and neuroimaging correlates in mild cognitive impairment. J Alzheimers Dis 35, 75–90.

83.

Montejo

, Montenegro

, Reinoso

, de Andrés Montes

, Claver Martín

(1999) Estudio de la eficacia de un programa de entrenamiento multicéntrico para mayores de 60 años. Rev Esp Geriatr Gerontol 34, 199–208.

84.

López-Higes

, Martín-Aragoneses

, Rubio-Valdehita

, Delgado-Losada

, Montejo

, Montenegro

, Prados

, de Frutos-Lucas

, López-Sanz

(2018) Efficacy of cognitive training in older adults with and without subjective cognitive decline is associated with inhibition efficiency and working memory span, not with cognitive reserve. Front Aging Neurosci 10, 23.

85.

, Gibson

, Bunarjo

, Buschkuehl

, Jaeggi

(2020) Quantifying the difference between active and passive control groups in cognitive interventions using two meta-analytical approaches. J Cogn Enhanc 4, 192–210.