Functional Connectivity and Response Inhibition: A Secondary Analysis of an 8-Week Randomized Controlled Trial of Computerized Cognitive Training

Abstract

Background:

Evidence suggests that computerized cognitive training (CCT) can improve cognitive function in older adults, particularly executive functions. However, the underlying mechanisms by which CCT may improve executive functions are not well established.

Objective:

To determine: 1) inter-network functional connectivity correlates of changes in executive functions; and 2) the effect of CCT on these functional connectivity correlates.

Methods:

This secondary analysis included a subset of 124 adults aged 65–85 years enrolled in an 8-week randomized controlled trial of CCT. Participants were randomized to either: 1) group-based CCT 3x/week for 1 hour plus 3x/week home-based training; 2) group-based CCT preceded by brisk walking (Ex+CCT) 3x/week for 1 hour plus 3x/week home-based training; or 3) group-based balanced and toned (BAT) classes 3x/week for 1 hour (control). At baseline and trial completion, 65 of the 124 participants completed resting-state functional magnetic resonance imaging and neuropsychological tests of executive functions, specifically the Stroop Colour-Word Test and Flanker Test.

Results:

Improved performance on the Stroop Colour-Word Test and Flanker Test were associated with decreased correlation between the default mode network (DMN) and the fronto-parietal network (FPN) (p < 0.05). Compared with BAT, CCT alone significantly decreased correlation between the left dorsolateral prefrontal cortex and both the left and right medial temporal gyrus (–0.143, 95%CI [–0.256,–0.030], p = 0.014, and –0.123, 95%CI [–0.242,–0.004], p = 0.043, respectively).

Conclusion:

Decreased correlation between DMN and FPN, indicating less connection between these networks, may be an underlying mechanism by which CCT improves executive functions. Future studies are needed to replicate this finding.

Keywords

Clinical trial cognitive aging executive function magnetic resonance imaging

INTRODUCTION

Aging negatively impacts multiple cognitive do-mains, such as memory, processing speed, and executive functions [1]. As the global population is aging, there are increasing efforts in finding strategies to combat cognitive decline [2]. Current evidence shows that exercise and cognitive training have the potential to promote cognitive outcomes in older adults [3 –5]. However, the underlying neural mechanisms are not well understood. A better understanding of underlying mechanisms will aid in the refinement of lifestyle strategies.

We previously demonstrated that in an 8-week randomized controlled trial (RCT) of CCT in com-munity-dwelling older adults, those assigned to CCT improved executive functions, specifically response inhibition, compared with those assigned to an active control group [6]. Executive functions (EF) are higher order cognitive processes involved in goal-directed behavior [7], and thus, are critical for one’s capacity to remain functionally independent. Specifically, response inhibition refers to the ability to resist acting on impulses and controlling interferences through selective attention and cognitive inhibition [7]. The underlying neural mechanisms of how CCT improves EF are not well examined [8]. We propose changes in inter-network functional connectivity as potential neural mechanisms.

The brain consists of functional networks that are interconnected; communication within and between these networks is crucial for cognitive performance [9]. Well-established neural networks include the default mode network (DMN), the fronto-parietal network (FPN), the salience network (SN), and the central executive network (CEN). These neural networks are classified as either task-positive or task-negative; where task-positive networks are involved in attention-demanding or goal-directed tasks, and task-negative networks are involved in involuntary actions or thoughts (e.g., mind wandering). The DMN is a network distinct from other brain systems because it is active in a task-negative state and inactive in task-positive states [10]. The network is involved in self-referential processes [11] and mind wandering [12]. The FPN is a task-positive network and is involved in attention and executive control, and is able to adjust and control processes based on changing demands [13]. The SN aims to identify relevant stimuli and helps guide behavior, and thus plays a role in processes of attention and cognitive control [14]. The main hub of the network, the anterior insula, is critical for the ability to switch between the DMN and the CEN by integrating information from multiple sources (e.g., emotional, sensory, and cognitive) [15]. The main function of the CEN is its involvement in EF, providing error feedback for top-down control, and help maintain associations between action versus outcome [14, 16].

Functional connectivity, both within and between these neural networks, is sensitive to aging effects [17]. Functional connectivity analysis examines the strength of the connections between brain regions that show temporally correlated activity and is mea-sured using resting-state fMRI (rs-fMRI). Regions of the neural networks can either correlate positively or negatively (i.e., anti-correlation) with each other. When focusing on inter-network functional connectivity, specifically between task-negative (i.e., DMN) and task-positive (i.e., FPN, CEN, and SN) networks; an anti-correlation is favorable for cognitive performance [18]. The degree of anti-correlation appears to vary across the lifespan; positive correlations are observed in early childhood, and evolve into anti-correlations in young adulthood [19]. With older age, the developed anti-correlation between neural networks tends to diminish [17 , 21]. For example, Geerligs and colleagues [22] showed that in adults aged 18 –26 years, the FPN and DMN acted as separate networks, while in older adults (aged 59 –74 years) the two networks act functionally more as one coherent network (i.e., reduced anti-correlation). These age-related decreases in anti-correlation between task-negative and task-positive networks have been linked to decreased network modularity (i.e., less segregation of networks [23]) and efficiency [22, 24], and are associated with reduced cognitive performance [25].

Therefore, the aim of this secondary analysis of an 8-week randomized controlled trial of CCT is two-fold: 1) To identify relevant changes in inter-network functional connectivity, specifically between task-positive (FPN, SN, and CEN) and task-negative (DMN) networks, that correlate with changes in the executive process of response inhibition; and 2) To examine the effects of an 8-week RCT of CCT on changes in regional inter-network functional connectivity (i.e., task-negative versus task-positive networks) compared with an active control group. We hypothesize that: 1) Improved response inhibition will be associated with decreased correlation between task-positive and task-negative networks; and 2) Compared with those assigned to the active control group, those assigned to the CCT groups will show decreased correlation between task-positive and task-negative networks on an overall-network level as well as a regional level.

METHODS

The protocol [26] and the primary findings [6] of the study have been published previously. A summary with key aspects of the protocol is described in the following sections; for more detailed information refer to the protocol paper [26].

Study design

This is a secondary analysis of a previously published 8-week, single-blinded, proof-of-concept RCT (ClinicalTrials.gov identifier: NCT02564809) at the University of British Columbia and Vancouver General Hospital campus with assessments at baseline and trial completion (i.e., 8 weeks). Secondary analyses can be seen as either supporting the primary analysis or as hypothesis-generating for future studies. MRI data were acquired at baseline and trial completion in a subset of eligible participants.

Participants

This planned secondary analysis included 68 community-dwelling older adults, a subset of 124 adults aged 65–85 years enrolled in an 8-week randomized controlled trial. Participants were recruited from metro Vancouver, British Columbia between September 2015 and April 2017 using advertisements in newspapers, flyers, and brochures in local community centers. We included community-dwelling older adults who: 1) were aged between 65 and 85 years; 2) completed high school education; 3) had preserved general cognitive function as indicated by a Mini-Mental State Examination (MMSE) [27] score ≥24/30; 4) scored ≥6/8 on the Lawton and Brody [28] Instrumental Activities of Daily Living Scale; 5) were not expected to start a new dose or those who remained stable on a fixed dose of anti-dementia medications (e.g., donepezil, galantamine, etc.) during the study period; and 6) were safe to engage in 15 min of brisk walking based on the Physical Activity Readiness Questionnaire [29]. We excluded individuals who: 1) were diagnosed with dementia of any type; 2) had a neurodegenerative disease as the cause of mild cognitive impairment (MCI) that is not AD, vascular dementia, or both (e.g., multiple sclerosis, Parkinson’s disease, etc.); 3) experienced clinically significant peripheral neuropathy or severe musculoskeletal or joint disease that impairs mobility, as determined by his/her family physician; 4) were taking medications that may negatively affect cognitive function (e.g., tranquilizers and anticonvulsants); and 5) were ineligible for MRI scanning.

Following screening over the phone, eligible participants came in for an information session to discuss additional study information as well as the consent form. Figure 1, the CONSORT (Consolidated Standards of Reporting Trials) flow chart, provides information about participant flow and distribution. Ethical approval was obtained from both the University of British Columbia Clinical Research Ethics Board as well as from the Vancouver Coastal Health Research Institute (VCHRI) ethics board.

Fig. 1

CONSORT flow diagram.

Descriptive variables

At baseline, general health, demographics, socioeconomic status, and education were ascertained by a questionnaire. Descriptive measures such as age in years, standing and sitting height in centimeters, mass in kilograms, and waist and hip circumference in centimeters were obtained. Global cognitive function was measured using both the MMSE [27] and the Montreal Cognitive Assessment (MoCA) [30].

Executive function: response inhibition

The Stroop Colour-Word Test [31] was used to assess response inhibition, by calculating incongruent condition completion time minus congruent condition completion time. In addition to standard paper and pen tests of response inhibition, we administered the Flanker Inhibitory Control and Attention Test from the cognition battery of the National Institute of Health (NIH) Toolbox [32].

Functional MRI acquisition

Participants completed an MRI scan at baseline and trial completion at the UBC MRI Research Centre on a 3.0 Tesla Intera Achieva MRI Scanner (Philip Medical Systems Canada, Markham, Ontario) using an 8-channel SENSE head coil. The 12-min resting-state fMRI scan consisted of 360 dynamic images of 36 slices (thickness of 3 mm) which were acquired using the following imaging parameters: repetition time (TR) of 2000 ms, echo time (TE) of 30 ms, flip angle (FA) of 90 degrees, field of view (FoV) of 240 mm, and an acquisition matrix of 80×80. During the rs-fMRI scan, no music was played, and participants were asked to keep their eyes open while looking at a fixed point without thinking of anything in particular. The anatomical T1-weighted images were acquired using the following imaging parameters: 170 slices (thickness of 1 mm), TR of 7.7 ms, TE of 3.6 ms, FA of 8 degrees, FoV of 256 mm, and an acquisition matrix of 256×200.

Randomization

Participants were randomly allocated to either Fit Brains^® Training (FBT), Exercise plus Fit Brains^® Training (Ex-FBT), or Balanced And Toned (BAT; i.e., control) with a ratio of 1:1:1 using the web application www.randomization.com. A research team member not involved with the study held this sequence at a remote location. Assessors were blinded to group allocation of the participants.

Sample size

Sample size calculations for the parent study [6] were based on predictions of changes in the Rey Auditory Verbal Learning Test (retention score), the primary outcome of the RCT, in the absence of previous trials testing the effects of FBT on memory and learning. Based on the work of Diamond and colleagues [33], and accounting for a 10%drop-out rate, 40 participants per group were needed for a power of 0.80, based on a two-group comparison (i.e., FBT versus BAT; Ex-FBT versus BAT). For the current secondary analysis of neuroimaging data, a convenience sample was used, where funding, eligibility, and willingness to participate determined the sample size. Our secondary analysis should be viewed as hypothesis-generating for future studies, as statistical power for this secondary analysis was not pre-specified.

Interventions

A succinct description of the intervention is described below. A more detailed description of the protocol is published elsewhere [26].

Fit Brains^® training

Participants randomized to Fit Brains^® Training (FBT) performed multi-domain computerized cognitive training 3x/week for 60 min at the research center, as well as 3x/week at home for 60 min. Games were performed on an iPad and consisted of 38 games targeting one of six domains: focus, speed, memory, visual, problem solving, and language. Games were individualized and adaptive throughout the 8-week program.

Exercise plus Fit Brains^® training

Participants randomized to the Exercise + Fit Brains^® Training (Ex-FBT) came to the research center 3x/week for 60 min, consisting of a 15 min brisk walk perceived as somewhat hard (i.e., up to 13-14 on the 6–20 Borg’s Rating of Perceived Exertion scale) [34] followed by a 45 min session of multi-domain computerized cognitive training. Additionally, they repeated the same 60 min training (i.e., 15 min walk + FBT training) 3x/week at home for 8 weeks. Compliance for both group-based training and home-based training was recorded via diaries and time stamped data recorded by the Fit Brains^® platform.

Balanced and toned

The Balanced and Toned (BAT) group attended three 60 min sessions/week at the research center for 8 weeks. Specifically, participants completed 8 h of sham cognitive training (e.g., word and drawing games, and creativity exercises), 8 h of sham exercise training (e.g., stretching, balancing, and core strength exercises), and 8 h of education regarding brain health (e.g., lectures on sleep, goal setting, mindfulness, and an educational project). Participants were asked to complete homework in order to complete their educational project.

Adverse events

Participants were asked about the presence of any adverse effects throughout the study, such as musculoskeletal pain or discomfort following the sham exercise portion (i.e., BAT) and the 15 min brisk walk (i.e., Ex-FBT). Participants were monitored for shortness of breath during the sham exercise and brisk walks.

Functional MRI data analysis

Preprocessing

Processing of the images was done using FEAT [35] (version 6.00), which is part of FSL (FMRIB Software Library; version 6.0) [36], MATLAB (Matrix Laboratory), and toolboxes from SPM (Statistical Parametric Mapping). Brain extraction in high resolution T1-weighted images was performed using optiBET [37] to remove unwanted structures (e.g., bones, skull). Manual checks were performed to ensure all brain tissue was included in the extraction; optiBET masks were edited where necessary by one individual (LFtB) to ensure rater consistency. Using the “fslmaths” function, final brain extraction was calculated by multiplying T1-weighted scans with the edited optiBET masks. FSL [38 –40] was used to create a study-specific template image (instead of MNI152 average brain) to ensure best template representation of the current study population. Registration with the study-specific average brain was checked manually by one individual for gross errors. Rigid body motion correction was completed using MCFLIRT [39]; participants were excluded when an absolute displacement of 2.0 mm or a relative displacement of 0.2 mm was exceeded. Spatial smoothing was carried out using a Gaussian kernel of 6.0 mm Full-Width-Half-Maximum (FWHM). A high-pass filter with a cut-off of 120 s was used for temporal filtering. Preprocessed functional data were registered to high-resolution T1-weighted anatomical images, which in turn were registered to the average study-specific space. Nuisance signals originated from the cerebrospinal fluid, white matter were extracted from respective masks and regressed out of the pre-processed time-series along with six motion-related regressors (one for each direction). The residual time-series data was used for subsequent functional connectivity analysis.

Functional connectivity analysis

The choice of resting-state networks and their corresponding regions of interest (ROI; i.e., seeds) are based on previous studies in aging [15 , 41], and consisted of the main hubs of each network. The networks and their corresponding key ROIs are displayed in Table 1. The DMN included the posterior cingulate cortex (PCC) and the bilateral middle temporal gyrus (i.e., RMTG and LMTG). The FPN included the right inferior parietal sulcus (RIPS) and the right and left dorsolateral prefrontal cortex (RdlPFC and LdlPFC, respectively). The CEN included the right and left anterolateral prefrontal cortex (RALPFC and LALPFC, respectively). The SN included the right ventral anterior insula (RVAI) and the dorsal anterior cingulate cortex (dACC). For each ROI, 5 mm radius spherical regions (i.e., diameter of 10 mm) were drawn on the study-specific average brain, from which preprocessed time-series data were extracted. A radius of 5 mm was chosen to avoid overlap of ROIs, which could lead to similar correlations between different ROIs. Subsequently, Fisher’s z transformed correlations were calculated between all ROIs.

Table 1

Brain network and included regions of interest

Network	ROI
Default Mode Network	PCC
	RMTG
	LMTG
Fronto-Parietal Network	RIPS
	RdlPFC
	LdlPFC
Central-Executive Network	RALPFC
	LALPFC
Salience Network	RVAI
	dACC

ROI, region of interest; PCC, posterior cingulate cortex; RMTG, right medial temporal gyrus; LMTG, left medial temporal gyrus; RIPS, right inferior parietal sulcus; RdlPFC, right dorsolateral prefrontal cortex; LdlPFC, left dorsolateral prefrontal cortex; RALPFC, right antero-lateral prefrontal cortex; LALPFC, left antero-lateral prefrontal cortex; RVAI, right ventral anterior insula; dACC, dorsal anterior cingulate cortex.

Guided by previous work [41], overall inter-net-work connectivity between the task-positive (FPN, SN, CEN) and task-negative (DMN) networks was calculated by categorically computing the average of all the pairwise ROI-ROI correlations with similar spatial designation to generate a network level correlation coefficient. Changes in inter-network connectivity coefficients were then calculated as trial completion connectivity coefficients minus baseline connectivity coefficients.

To examine inter-network (DMN–FPN) functional connectivity on the ROI level, we used a total of 9 Fisher’s z transformed correlations: PCC–RIPS, PCC–RdlPFC, PCC–LdlPFC, RMTG–RIPS, RMTG–RdlPFC, RMTG–LdlPFC, LMTG–RIPS, LMTG–RdlPFC,LMTG–LdlPFC.

Statistical analysis

Statistical analysis was performed using the statistical package SPSS 26.0 (IBM Corporation, Armonk, NY). To examine whether changes in response inhibition in the overall sample were associated with changes in the overall inter-network resting-state connectivity between task-positive and task-negative networks, we conducted a partial correlation analysis between change scores of response inhibition, as measured by the both the Stroop Colour-Word Test and the Flanker Inhibitory Control and Attention Test, and changes in overall inter-network functional connectivity between task-negative and task-positive (i.e., DMN –FPN; DMN–CEN;DMN–SN) networks. We controlled for baseline MoCA, baseline systolic blood pressure [42], and experimental group in these analyses.

After identifying the overall inter-network resting-state connectivity relevant to changes in response inhibition, we: 1) examined the effect of FBT and Ex-FBT, compared with BAT, on changes in overall inter-network connectivity (DMN–FPN), and 2) examined the effect of FBT and Ex-FBT, compared with BAT, on changes in the 9 ROI pairs that exist in the relevant overall inter-network connectivity of the DMN and FPN. We performed an ANCOVA, with baseline MoCA and baseline systolic blood pressure as covariates. The overall alpha was set at p < 0.05. No adjustment for multiple endpoints were made since in a proof-of-concept study, such as this study, a Type II error is of more concern than a Type I error [43].

RESULTS

Participants

Sixty-eight out of the 124 participants who consented and were randomized in the parent study underwent baseline MRI scanning. Three of the 68 MRI participants dropped out over the course of the study (1 = Ex-FBT, and 2 = BAT) and 65 participants completed a scan at trial completion. Scans of 10 participants (3 = FBT, 4 = Ex-FBT, and 3 = BAT) were excluded due to excessive motion (see Fig. 1). Baseline characteristics of the 55 participants are reported in Table 2. Compared with the parent sample of 124, this subset of participant had fewer females (p = 0.022), was taller (p = 0.039), and performed worse on the Stroop Colour-Word Test (p = 0.045).

Table 2

Participant characteristics at baseline (N = 55)

Variable	BAT (n = 21) Mean (SD)	FBT (n = 16) Mean (SD)	Ex-FBT (n = 18) Mean (SD)	Total (N = 55) Mean (SD)
Age (y)	71.43 (5.81)	70.75 (4.87)	72.44 (3.37)	71.56 (4.81)
Weight (kg)	72.18 (17.83)	69.26 (14.08)	73.46 (16.26)	71.75 (16.09)
Height (cm)	166.96 (11.16)	166.32 (12.36)	168.91 (10.94)	167.41 (11.29)
Sex (f, %)	10 (47.60)	10 (62.5)	7 (38.9)	27 (49.1)
Education (%)
High school certificate or diploma	3 (14.3)	1 (6.3)	1 (5.6)	5 (9.2)
Trades or professional certificate or diploma	3 (14.3)	–	4 (22.2)	7 (12.7)
University certificate or diploma	2 (9.5)	6 (37.5)	5 (27.8)	13 (23.6)
University degree	13 (61.9)	9 (56.3)	8 (44.4)	30 (54.5)
IADL (max. 8 pts)	7.81 (0.51)	7.94 (0.25)	7.94 (0.24)	7.89 (0.37)
Functional Comorbidity Index (max. 18 pts)	2.29 (1.45)	1.87 (1.36)	1.56 (1.25)	1.93 (1.37)
Physical Activity Scale for the Elderly	108.71 (46.08)	119.27 (40.21)	130.91 (46.26)	119.05 (44.70)
Systolic blood pressure (mmHg)	133.48 (19.39)	131.50 (16.70)	142.94 (25.02)	136.00 (20.94)
MoCA (max. 30 pts)	24.90 (3.75)	27.19 (2.76)	24.78 (3.87)	25.53 (3.64)
MMSE (max. 30 pts)	28.10 (1.79)	29.25 (0.86)	28.39 (1.50)	28.53 (1.53)
Stroop Incongruent –Congruent (s)^a	59.91 (32.33)	56.87 (41.34)	65.78 (37.16)	60.95 (36.19)
Flanker score	90.71 (11.25)	94.38 (7.10)	99.00 (9.96)	94.49 (10.22)

BAT, Balanced And Toned; FBT, Fit Brains Training; Ex-FBT, Exercise+Fit Brains Training; IADL, Instrumental Activities of Daily Living; MoCA, Montreal Cognitive Assessment; MMSE, Mini-Mental State Examination. ^aLower score reflect better performance.

Partial correlation: changes in response inhibition and changes in overall inter-network functional connectivity

Unadjusted functional connectivity coefficients for baseline and trial completion per group are displayed in Table 3. There was a statistically significant positive partial correlation between changes in Stroop Colour-Word Test performance and overall inter-network connectivity between DMN and FPN controlling for baseline MoCA, systolic blood pressure, and group (Pearson’s r = 0.358, p = 0.009; Table 4), such that improved behavioral performance over the course of the intervention was associated with decreased correlation between the DMN and FPN (Fig. 2A). In addition, a statistically significant negative partial correlation between Flanker test performance and overall inter-network connectivity between the DMN and FPN was observed (Pearson’s r = –0.275, p = 0.048; Table 4); improved performance on the Flanker test was associated with decreased correlation between the DMN and FPN (Fig. 2B). No other statistically significant associations were found (Table 4).

Table 3

Unadjusted baseline and final mean connectivity coefficients per intervention group

	Baseline						Final
ROI Pairs	BAT n = 21		FBT n = 16		Ex-FBT n = 18		BAT n = 21		FBT n = 16		Ex-FBT n = 18
	Mean	SE	Mean	SE	Mean	SE	Mean	SE	Mean	SE	Mean	SE
PCC –RIPS	0.148	0.056	0.127	0.052	0.045	0.036	0.109	0.040	0.125	0.046	0.092	0.034
PCC –RdlPFC	0.211	0.061	0.221	0.039	0.163	0.030	0.232	0.039	0.280	0.041	0.227	0.042
PCC –LdlPFC	0.219	0.056	0.378	0.033	0.229	0.039	0.286	0.037	0.320	0.046	0.241	0.054
RMTG –RIPS	0.093	0.040	0.051	0.033	0.028	0.042	0.094	0.037	0.069	0.049	0.047	0.028
RMTG –RdlPFC	0.181	0.040	0.121	0.042	0.203	0.048	0.232	0.042	0.158	0.046	0.203	0.045
RMTG –LdlPFC	0.119	0.026	0.238	0.042	0.152	0.048	0.185	0.039	0.179	0.053	0.170	0.035
LMTG –RIPS	0.088	0.038	0.059	0.035	–0.020	0.027	0.009	0.003	0.004	0.003	0.0005	0.03
LMTG –RdlPFC	0.166	0.036	0.073	0.036	0.125	0.033	0.207	0.035	0.127	0.041	0.140	0.039
LMTG –LdlPFC	0.197	0.038	0.209	0.043	0.193	0.043	0.234	0.038	0.132	0.063	0.162	0.038

DMN, default mode network; FPN, fronto-parietal network; PCC, posterior cingulate cortex, RIPS,right inferior parietal sulcus; RdlPFC, right dorsolateral prefrontal cortex; LdlPFC, left dorsolateral prefrontal cortex; RMTG, right medial temporal gyrus; LMTG, left medial temporal gyrus; BAT, Balanced and Toned (i.e., active control); FBT, Fit Brains Training; Ex-FBT, Exercise+Fit Brains Training.

Table 4

Partial correlations between change in response inhibition - change in functional connectivity

Variable ^†	Mean Change Mean (SD)	Δ DMN –FPN Pearson’s r	Δ DMN –CEN Pearson’s r	Δ DMN –SN Pearson’s r
Δ Stroop (s)	–17.230 (29.643)	0.358^**	0.189	0.079
Δ Flanker	9.200 (9.132)	–0.275^*	–0.193	–0.023

^†Stroop, Stroop 3 –Stroop 2 (seconds); DMN, default mode network; FPN, fronto-parietal network; CEN, central executive network; SN, salience network. Change calculated at trial completion minus baseline Correlations adjusted for: baseline Montreal Cognitive Assessment, baseline systolic blood pressure, and group. Δ Stroop interference: negative value represents improvement. Δ Flanker: positive value represents improvement. ^*p < 0.05; ^**p < 0.01.

Fig. 2

A) Partial Correlation of Stroop and Overall DMN-FPN Functional Connectivity; B) Partial Correlation of Flanker and Overall DMN-FPN Functional Connectivity. Correlations adjusted for: baseline Montreal Cognitive Assessment, baseline systolic blood pressure, and group. Increased correlation between DMN-FPN is associated with a decline in both Stroop and Flanker Test performance. Change in cognitive performance is measured by subtracting baseline from final. Lower change scores for the Stroop test reflects better performance. Higher change scores for the Flanker test reflects better performance. Negative change in functional connectivity reflects decreased correlation. Positive change in functional connectivity reflects increased correlation.

ANCOVA: Effect of FBT and Ex-FBT on overall and regional DMN-FPN connectivity

No statistically significant between-group differences (i.e., FBT versus BAT and Ex-FBT versus BAT) were found for overall inter-network DMN-FPN connectivity (p = 0.277 and p = 0.944, respectively) at trial completion. For the 9 ROI DMN-FPN pairs, there were statistically significant differences between FBT and BAT (Table 5). Specifically, there was a statistically significant difference in the ROI pair of RMTG and LdlPFC (contrast estimate, C =–0.143, 95%CI [–0.256, –0.030], p = 0.014; Table 5); where FBT demonstrated decreased correlation between RMTG and LdlPFC, whereas BAT showed increased correlation (Fig. 3A). In addition, there was a statistically significant difference in the ROI pair of LMTG and LdlPFC (contrast estimate, C =–0.123, 95%CI [–0.242, –0.004], p = 0.043; Table 5); where FBT decreased correlation compared with an increased correlation in BAT (Fig. 3B). No other statistically significant between-group differences were found in regional DMN-FPN connectivity (Table 5).

Table 5

Regional changes in inter-network functional connectivity (DMN –FPN)

ROI Pairs	BAT		FBT		Ex-FBT		p ^*
	Mean	SE	Mean	SE	Mean	SE	FBT versus BAT	Ex-FBT versus BAT
PCC –RIPS	–0.037	0.038	–0.028	0.045	0.066	0.042	0.880	0.078
PCC –RdlPFC	0.009	0.041	0.066	0.049	0.072	0.045	0.383	0.316
PCC –LdlPFC	0.066	0.046	–0.063	0.054	0.019	0.050	0.078	0.493
RMTG –RIPS	0.006	0.036	–0.006	0.042	0.035	0.039	0.835	0.585
RMTG –RdlPFC	0.043	0.040	0.030	0.046	0.014	0.043	0.834	0.624
RMTG –LdlPFC	0.077	0.036	–0.066	0.042	0.012	0.039	0.014^*	0.228
LMTG –RIPS	0.000	0.032	–0.042	0.038	0.044	0.035	0.405	0.367
LMTG –RdlPFC	0.032	0.038	0.052	0.044	0.028	0.041	0.745	0.940
LMTG –LdlPFC	0.040	0.038	–0.083	0.045	–0.029	0.042	0.043^*	0.226

DMN, default mode network; FPN, fronto-parietal network; PCC, posterior cingulate cortex; RIPS, right inferior parietal sulcus; RdlPFC, right dorsolateral prefrontal cortex; LdlPFC, left dorsolateral prefrontal cortex; RMTG, right medial temporal gyrus; LMTG, left medial temporal gyrus; BAT, Balanced and Toned (i.e., active control); FBT, Fit Brains Training; Ex-FBT, Exercise+Fit Brains Training. Negative change scores reflect more favorable connectivity (i.e., decreased correlation); positive change scores reflect less favorable connectivity (i.e., increased correlation). ^*p < 0.05; controlled for baseline Montreal Cognitive Assessment and systolic blood pressure.

Fig. 3

Between-Group Regional Differences in Inter-Network Functional Connectivity of DMN –FPN: A) RMTG –LdlPFC; B) LMTG–LdlPFC. A) ^*FBT versus BAT: –0.143, 95%CI [–0.256, –0.030], p = 0.014; B) ^*FBT versus BAT: –0.123, 95%CI [–0.242, –0.004], p = 0.043. Adjusted for baseline Montreal Cognitive Assessment, baseline systolic blood pressure, Negative change in functional connectivity reflects decreased correlation, and thus a favorable change in connectivity. Positive change in functional connectivity reflects increased correlation, and thus a non-favorable change in connectivity.

Adverse events

There was one adverse event over the course of the 8-week trial; a participant fell in the facility during the cognitive training classes which resulted in some bruising. This adverse event was not directly a result of the program (i.e., the participant fell while leaving the room); thus, no adjustments to the protocol were necessary to ensure participant safety. No participants reported musculoskeletal-related issues (e.g., muscle soreness or muscle strain) throughout the study.

DISCUSSION

This secondary analysis of an 8-week RCT of CCT in otherwise healthy older adults may suggest that: 1) improvements in executive functions, specifically response inhibition, were associated with decreased overall inter-network correlation between the DMN and FPN, and 2) compared with an 8-week sham exercise and cognitive training program, CCT alone decreased correlation between the left dorsolateral prefrontal cortex and bilateral medial temporal gyrus regions.

Importantly, our findings concur with prior cross-sectional evidence linking functional connectivity between the DMN and FPN to the efficacy of EF, specifically showing that greater segregation of the DMN and FPN (i.e., decreased correlation) is associated better performance on tasks of response inhibition [25]. We focused on inter-network connectivity between task-positive and task-negative networks as it is sensitive to aging effects and impacts cognitive performance [17, 22]. The FPN is a task-positive network involved in higher order cognitive functions [44] and a network of interest as the current analysis focused on changes in EF. We showed that a decrease in correlation of functional connectivity between the DMN-FPN, reflected by a negative connectivity change score, was associated with improved performance on tasks of response inhibition, a core aspect of EF [45]. Aging is associated with desegregation between the DMN and FPN [17]. The observed decrease in correlation between these two networks may help prevent the age-related desegregation of these networks, resulting in better independent network function, which in turn may benefit preservation of cognitive functions such as EF. Specifically, the consistent involvement of the left dlPFC versus the right dlPFC in CCT could be due to lateralized training-related effects in the Stroop Test. A review by Vanderhasselt & Baken stated a more direct impact of the left dlPFC in Stroop performance via mediation of cognitive control and prevention of conflict. This involvement is believed to increase during high-conflict trials (e.g., incongruent trails) [46]. Our current results extend these findings by demonstrating CCT may mitigate age-related changes in inter-network functional connectivity.

The current RCT extends the present literature with quality evidence by examining inter-network changes in resting-state functional connectivity, specifically functional connectivity between task-positive and task-negative neural networks. In a systematic review [8], we summarized the current evidence regarding the effect of CCT on neuroimaging outcomes, and the findings were equivocal and high-quality studies (i.e., RCTs) were needed. Five of the nine included studies, including 3 RCTs, examined both inter- and intra-network resting-state functional connectivity after CCT; results were very heterogeneous, with both increased and decreased levels of connectivity after the intervention, which were linked to maintained or improved cognitive performance. Two studies showed improved inter-network functional connectivity after CCT [47, 48]. Firstly, Suo and colleagues [48] showed that after 6 months of twice-weekly 90 min training sessions, those assigned to the groups that included multi-domain CCT (i.e., CCT + sham, CCT + progressive resistance training) increased functional connectivity between seeds (i.e., hippocampus –superior frontal cortex) of task positive networks compared to those groups that did not include CCT. Additionally, they showed that this increase in functional connectivity found in their MRI subsample of 79 older adults, was associated with improved overall memory performance after 6 months of training. Secondly, a 12-week trial of thrice-weekly 1 h multi-domain CCT sessions with assessments after both 3 and 12 weeks of training, showed increased anti-correlation between seeds of the DMN and executive control network after three weeks of training (i.e., 9 h of CCT) compared with the control. However, at trial completion (i.e., 12 weeks) this difference was no longer significant. At the three-week measure, changes in functional connectivity were not significantly associated with global cognition [47]. Thus, the current RCT extends the current literature by providing evidence of the effects of CCT on functional connectivity between task-positive and task-negative neural networks.

Contrary to our hypothesis, there was no effect of CCT, when combined with exercise (i.e., Ex-FBT group), on DMN-FPN functional connectivity. Per in-person session, the combined exercise and CCT group had 15 min less of CCT. Thus, over 8 weeks, the combined exercise plus CCT group received 36 h of CCT (i.e., 75%) versus 48 h for the CCT-only group. The lower dose of CCT could have contributed the null findings in inter-network functional connectivity. In addition, the absence of neural changes between the examined neural networks in the combined intervention group (i.e., exercise + CCT) indicates additional underlying neural mechanisms may be responsible for the previously reported benefits on EF in this group [6]. We encourage future research to look in other potential mechanisms of combined exercise and CCT on EF.

Current evidence suggests that neural changes evoked by exercise and cognitive training are different. Exercise stimulates the growth of new neurons (i.e., neurogenesis) [49, 50], while cognitive training promotes the functional neuronal structure via neuronal survival and differentiation (e.g., synaptogenesis, spine density, dendritic length) [49 , 52]. The duration of the current study intervention (i.e., 8 weeks) is likely insufficient to realize the neural benefits of exercise but sufficient for those associated with cognitive training, such as increased dendritic length [49]. In theory, these changes could improve the efficiency of inter-neuronal communication [49].

We note limitations of our study. The current findings are limited to an older population with access to and those comfortable with relevant technology (e.g., iPad, smartphone) necessary to access CCT platforms on a regular basis to complete training sessions, and thus the generalization of these findings to an older adult population is limited. To limit multiple comparisons, the current study confined regional analyses (i.e., ROI pairs between the DMN and FPN) based on the observed association between changes in response inhibition and overall inter-network connectivity of the DMN and FPN. However, 9 comparisons were still performed, making the results susceptible to Type-I error and potentially unreliable. Notably, the current results are no longer statistically significant when a Bonferroni correction is applied. These results are also from a secondary analysis where the statistical power was not pre-specified, and thus, are hypothesis-generating for future studies and must be interpreted with caution. Future studies with larger samples are needed to replicate and confirm these findings, as well as to allow for formal mediation analysis. Additionally, we would encourage future studies to include neuroimaging follow-up data, to examine whether maintenance or changes in functional connectivity persist or diminish over time.

In summary, results from this secondary analysis may suggest that improved response inhibition was related to decreased correlation between task-negative and task-positive networks in community-dwelling older adults, suggesting these networks are more functionally disconnected in those with improved EF. Moreover, evidence suggests that CCT alone may decrease regional correlation of these networks. This suggests that those who are able to better functionally segregate task-negative and task-positive networks with age, might better preserve performance on tasks requiring high-order cognitive functions and that CCT may be a promising strategy to promote the functional organization of the brain. However, we would encourage future studies with larger samples to replicate and extend these exploratory findings.

Footnotes

ACKNOWLEDGMENTS

We thank the participants for their time and dedication.

This work was supported by a grant provided by Rosetta Stone Canada to the University of British Columbia. They provided the Fit Brains® program and technical support for the program. Rosetta Stone Canada had no role in study design, study management, data collection, data analysis, data interpretation, and manuscript drafting, manuscript review for important intellectual content, or the decision to submit and publish the manuscript.

LTB is a Mitacs PhD trainee. TLA is a Canada Re-search Chair (Tier 2) in Physical Activity, Mobility, and Cognitive Neuroscience. CLH is a Canadian Institutes of Health Research Postdoctoral Fellow.

Authors’ disclosures available online ().

References

Deary

, Corley

, Gow

, Harris

, Houlihan

, Marioni

, Penke

, Rafnsson

, Starr

(2009) Age-associated cognitive decline. Br Med Bull 92, 135–152.

Polidori

, Nelles

, Pientka

(2010) Prevention of dementia: Focus on lifestyle. Int J Alzheimers Dis 2010, 393579.

Chiu

, Chu

, Tsai

, Liu

, Chen

, Yang

, Chou

(2017) The effect of cognitive-based training for the healthy older people: A meta-analysis of randomized controlled trials. PLoS One 12, e0176742.

Northey

, Cherbuin

, Pumpa

, Smee

, Rattray

(2018) Exercise interventions for cognitive function in adults older than 50: A systematic review with meta-analysis. Br J Sports Med 52, 154–160.

Rebok

, Ball

, Guey

, Jones

, Kim

, King

, Marsiske

, Morris

, Tennstedt

, Unverzagt

, Willis

, ACTIVE Study Group (2014) Ten-year effects of the advanced cognitive training for independent and vital elderly cognitive training trial on cognition and everyday functioning in older adults. J Am Geriatr Soc 62, 16–24.

Ten Brinke

, Best

, Chan

JLC

, Ghag

, Erickson

, Handy

, Liu-Ambrose

(2020) The effects of computerized cognitive training with and without physical exercise on cognitive function in older adults: An 8-week randomized controlled trial. J Gerontol A Biol Sci Med Sci 75, 755–763.

Diamond

(2013) Executive functions. Annu Rev Psychol 64, 135–168.

Ten Brinke

, Davis

, Barha

, Liu-Ambrose

(2017) Effects of computerized cognitive training on neuroimaging outcomes in older adults: A systematic review. BMC Geriatr 17, 139.

van den Heuvel

, Hulshoff Pol

(2010) Exploring the brain network: A review on resting-state fMRI functional connectivity. Eur Neuropsychopharmacol 20, 519–534.

10.

Greicius

, Srivastava

, Reiss

, Menon

(2004) Default-mode network activity distinguishes Alzheimer’s disease from healthy aging: Evidence from functional MRI. Proc Natl Acad Sci U S A 101, 4637–4642.

11.

Greicius

, Krasnow

, Reiss

, Menon

(2003) Functional connectivity in the resting brain: A network analysis of the default mode hypothesis. Proc Natl Acad Sci U S A 100, 253–258.

12.

Mason

, Norton

, Van Horn

, Wegner

, Grafton

, Macrae

(2007) Wandering minds: The default network and stimulus-independent thought. Science 315, 393–395.

13.

Marek

, Dosenbach

NUF

(2018) The frontoparietal network: Function, electrophysiology, and importance of individual precision mapping. Dialogues Clin Neurosci 20, 133–140.

14.

Seeley

, Menon

, Schatzberg

, Keller

, Glover

, Kenna

, Reiss

, Greicius

(2007) Dissociable intrinsic connectivity networks for salience processing and executive control. J Neurosci 27, 2349–2356.

15.

Menon

(2015) Salience network. In Brain Mapping: An Encyclopedic Reference, Toga AW, ed. Academic Press: Elsevier, pp. 597–611.

16.

Huang

, Fang

, Li

, Chen

(2016) Exercise-related changes of networks in aging and mild cognitive impairment brain. Front Aging Neurosci 8, 47.

17.

Andrews-Hanna

, Snyder

, Vincent

, Lustig

, Head

, Raichle

, Buckner

(2007) Disruption of large-scale brain systems in advanced aging. Neuron 56, 924–935.

18.

Fox

, Snyder

, Vincent

, Corbetta

, Van Essen

, Raichle

(2005) The human brain is intrinsically organized into dynamic, anticorrelated functional networks. Proc Natl Acad Sci U S A 102, 9673–9678.

19.

Chai

, Ofen

, Gabrieli

, Whitfield-Gabrieli

(2014) Selective development of anticorrelated networks in the intrinsic functional organization of the human brain. J Cogn Neurosci 26, 501–513.

20.

, Lo

, Lim

JKW

, Chee

MWL

, Zhou

(2016) Reduced functional segregation between the default mode network and the executive control network in healthy older adults: A longitudinal study. Neuroimage 133, 321–330.

21.

Zonneveld

, Pruim

, Bos

, Vrooman

, Muetzel

, Hofman

, Rombouts

, van der Lugt

, Niessen

, Ikram

, Vernooij

(2019) Patterns of functional connectivity in an aging population: The Rotterdam Study. Neuroimage 189, 432–444.

22.

Geerligs

, Renken

, Saliasi

, Maurits

, Lorist

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

23.

Sala-Llonch

, Bartres-Faz

, Junque

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

24.

Geerligs

, Maurits

, Renken

, Lorist

(2014) Reduced specificity of functional connectivity in the aging brain during task performance. Hum Brain Mapp 35, 319–330.

25.

Kelly

, Uddin

, Biswal

, Castellanos

, Milham

(2008) Competition between functional brain networks mediates behavioral variability. Neuroimage 39, 527–537.

26.

Ten Brinke

, Best

, Crockett

, Liu-Ambrose

(2018) The effects of an 8-week computerized cognitive training program in older adults: A study protocol for a randomized controlled trial. BMC Geriatr 18, 31.

27.

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.

28.

Lawton

, Brody

(1969) Assessment of older people: Self-maintaining and instrumental activities of daily living. Gerontologist 9, 179–186.

29.

Canadian Society for Exercise Physiology (1994) Par-Qand You. Canadian Society of Exercise Physiology, Gloucester, Ontario, Canada, pp. 1–2.

30.

Nasreddine

, Phillips

, Bedirian

, Charbonneau

, Whitehead

, Collin

, Cummings

, Chertkow

(2005) The Montreal Cognitive Assessment, MoCA: A brief screening tool for mild cognitive impairment. J Am Geriatr Soc 53, 695–699.

31.

Trenerry

, Crosson

, DeBoe

, Leber

(1988) Stroop Neuropsychological Screening Test. Psychological Assessment Resources.

32.

Mungas

, Heaton

, Tulsky

, Zelazo

, Slotkin

, Blitz

, Lai

, Gershon

(2014) Factor structure, convergent validity, and discriminant validity of the NIH Toolbox Cognitive Health Battery (NIHTB-CHB) in adults. J Int Neuropsychol Soc 20, 579–587.

33.

Diamond

, Mowszowski

, Cockayne

, Norrie

, Paradise

, Hermens

, Lewis

, Hickie

, Naismith

(2015) Randomized controlled trial of a healthy brain ageing cognitive training program: Effects on memory, mood, and sleep. J Alzheimers Dis 44, 1181–1191.

34.

Borg

(1982) Psychophysical bases of perceived exertion. Med Sci Sports Exerc 14, 377–381.

35.

Woolrich

, Ripley

, Brady

, Smith

(2001) Temporal autocorrelation in univariate linear modeling of FMRI data. Neuroimage 14, 1370–1386.

36.

Smith

, Jenkinson

, Woolrich

, Beckmann

, Behrens

, Johansen-Berg

, Bannister

, De Luca

, Drobnjak

, Flitney

, Niazy

, Saunders

, Vickers

, Zhang

, De Stefano

, Brady

, Matthews

(2004) Advances in functional and structural MR image analysis and implementation as FSL. Neuroimage 23 Suppl 1, S208–219.

37.

Lutkenhoff

, Rosenberg

, Chiang

, Zhang

, Pickard

, Owen

, Monti

(2014) Optimized brain extraction for pathological brains (optiBET). PLoS One 9, e115551.

38.

Greve

, Fischl

(2009) Accurate and robust brain image alignment using boundary-based registration. Neuroimage 48, 63–72.

39.

Jenkinson

, Bannister

, Brady

, Smith

(2002) Improved optimization for the robust and accurate linear registration and motion correction of brain images. Neuroimage 17, 825–841.

40.

Jenkinson

, Smith

(2001) A global optimisation method for robust affine registration of brain images. Med Image Anal 5, 143–156.

41.

Hsu

, Voss

, Handy

, Davis

, Nagamatsu

, Chan

, Bolandzadeh

, Liu-Ambrose

(2014) Disruptions in brain networks of older fallers are associated with subsequent cognitive decline: A 12-month prospective exploratory study. PLoS One 9, e93673.

42.

Son

, Kim

, Lee

, Park

, Namkoong

, Hong

, Ku

, Kim

, Oh

(2015) Effect of hypertension on the resting-state functional connectivity in patients with Alzheimer’s disease (AD). Arch Gerontol Geriatr 60, 210–216.

43.

Schoenfeld

(1980) Statistical considerations for pilot studies. Int J Radiat Oncol Biol Phys 6, 371–374.

44.

Reineberg

, Gustavson

, Benca

, Banich

, Friedman

(2018) The relationship between resting state network connectivity and individual differences in executive functions. Front Psychol 9, 1600.

45.

Miyake

, Friedman

, Emerson

, Witzki

, Howerter

, Wager

(2000) The unity and diversity of executive functions and their contributions to complex “Frontal Lobe” tasks: A latent variable analysis. Cogn Psychol 41, 49–100.

46.

Vanderhasselt

, De Raedt

, Baeken

(2009) Dorsolateral prefrontal cortex and Stroop performance: Tackling the lateralization. Psychon Bull Rev 16, 609–612.

47.

Lampit

, Hallock

, Suo

, Naismith

, Valenzuela

(2015) Cognitive training-induced short-term functional and long-term structural plastic change is related to gains in global cognition in healthy older adults: A pilot study. Front Aging Neurosci 7, 14.

48.

Suo

, Singh

, Gates

, Wen

, Sachdev

, Brodaty

, Saigal

, Wilson

, Meiklejohn

, Singh

, Baune

, Baker

, Foroughi

, Wang

, Mavros

, Lampit

, Leung

, Valenzuela

(2016) Therapeutically relevant structural and functional mechanisms triggered by physical and cognitive exercise. Mol Psychiatry 21, 1645.

49.

Gelfo

, Mandolesi

, Serra

, Sorrentino

, Caltagirone

(2018) The neuroprotective effects of experience on cognitive functions: Evidence from animal studies on the neurobiological bases of brain reserve. Neuroscience 370, 218–235.

50.

Lista

, Sorrentino

(2010) Biological mechanisms of physical activity in preventing cognitive decline. Cell Mol Neurobiol 30, 493–503.

51.

Shors

, Anderson

, Curlik

2nd , Nokia

(2012) Use it or lose it: How neurogenesis keeps the brain fit for learning. Behav Brain Res 227, 450–458.

52.

Gelfo

, De Bartolo

, Giovine

, Petrosini

, Leggio

(2009) Layer and regional effects of environmental enrichment on the pyramidal neuron morphology of the rat. Neurobiol Learn Mem 91, 353–365.

Functional Connectivity and Response Inhibition: A Secondary Analysis of an 8-Week Randomized Controlled Trial of Computerized Cognitive Training

Abstract

Background:

Objective:

Methods:

Results:

Conclusion:

Keywords

INTRODUCTION

METHODS

Study design

Participants

Descriptive variables

Executive function: response inhibition

Functional MRI acquisition

Randomization

Sample size

Interventions

Fit Brains® training

Exercise plus Fit Brains® training

Balanced and toned

Adverse events

Functional MRI data analysis

Preprocessing

Functional connectivity analysis

Statistical analysis

RESULTS

Participants

Partial correlation: changes in response inhibition and changes in overall inter-network functional connectivity

ANCOVA: Effect of FBT and Ex-FBT on overall and regional DMN-FPN connectivity

Adverse events

DISCUSSION

Footnotes

ACKNOWLEDGMENTS

References

Fit Brains^® training

Exercise plus Fit Brains^® training