Mild cognitive impairment impacts association of functional brain connectivity with cognition in older adults with and without multiple sclerosis

Abstract

Cognitive decline is common in multiple sclerosis (MS), although neural mechanisms are not fully understood. The objective was to investigate the impact of mild cognitive impairment (MCI) on the relationship between resting state functional connectivity (RSFC) and cognitive function in older adults with multiple sclerosis (OAMS) and age matched healthy controls. Participants underwent magnetic resonance imaging (MRI) scans and cognitive assessments. The Oral Symbol Digit Modalities Test (SDMT) was the outcome measure. Multiple linear regressions examined the relationship between RSFC networks and SDMT performance across the cohort and within group-specific analyses. These analyses were repeated with MCI as a moderator. Adjusted analyses showed that better SDMT performance correlated significantly with higher RSFC in sensorimotor (p = 0.01), left frontoparietal (p = 0.027), and salience (p = 0.047) networks. Stratified analyses showed significant positive associations for the OAMS group in medial visual (p = 0.007) and sensorimotor (p = 0.018) networks. Among OAMS, MCI moderated the association between RSFC and SDMT performance in the posterior default mode (p = 0.016) and subcortical (p = 0.042) networks, while no significant moderation effects were observed for the control group. Higher RSFC in brain networks correlates with better cognition in aging, but this relationship is modified in the presence of MS and MCI.

Supplementary Information

The online version contains supplementary material available at 10.1007/s11357-025-01854-9.

Introduction

Multiple sclerosis (MS), an autoimmune-mediated neurological disease, can be understood as a disorder of brain networks [1] composed of structurally and functionally interconnected gray matter regions [2]. Focusing on network-level changes, as opposed to regional damage, has provided groundbreaking insights into the neural mechanisms implicated in cognitive and clinical manifestations of MS [1]. Peak MS prevalence has shifted toward older age groups, suggesting a potential increase in the overlap between age-related and MS-specific declines in both cognitive outcomes and their related networks. While older adults with MS (OAMS) often display cognitive impairments [3], the brain mechanisms underlying such declines are poorly understood.

A systematic review [4] of resting-state functional MRI connectivity (RSFC) changes and cognitive impairment found that MS patients experience substantial reductions in RSFC across various brain networks, including the default mode network (DMN) and frontoparietal networks, two of the networks also affected in healthy aging [5]. Altered RSFC plays a role in the cognitive decline observed in MS, affecting patients’ capacity to maintain attention and processing speed [6]. However, previous studies reported conflicting findings, where both increases and decreases in RSFC within the DMN and frontoparietal networks were related to cognitive impairment in MS [4]. In aging, the presence of mild cognitive impairment (MCI) [7] has been associated with lower RSFC in brain networks [8]. However, the effects of RSFC, MCI, and their possible interaction on cognition among older adults with MS have not been reported.

The current study was designed to address this important gap in knowledge. Specifically, we aimed to determine whether MS, MCI, and their interaction influenced the strength and association of brain network RSFC with cognitive performance. Participants were OAMS and healthy controls who underwent cognitive testing and magnetic resonance imaging (MRI) of the brain. The oral version of the symbol digit modalities test (SDMT) [9] served as the cognitive outcome in this investigation.

Methods

Participants

Data were collected as part of a study titled “Brain Predictors of Mobility and Falls in Older Adults with Multiple Sclerosis.” Participants aged 60 years or older with complete cognitive and brain imaging data were included. MS participants were recruited from regional treatment centers and patient registry lists. MS diagnoses were physician-confirmed by medical record review using the revised McDonald criteria [10]. MS participants were stable on a disease-modifying therapy (DMT) for at least 6 months prior to enrollment. Control participants were recruited from population lists, by mail, and follow-up phone call. Participants underwent two in-person visits; the first included neuropsychological tests, mobility protocols, and questionnaires, while the second included an MRI and additional questionnaires. Cognitive status was determined via established clinical case conference procedures [11]. Exclusion criteria included diagnosis of any major neurological, psychiatric, or medical disease (excluding MS); inability to move independently; contraindication to MRI; and vision or hearing impairment. The study was approved by the IRB of Albert Einstein College of Medicine. Written informed consent was obtained from all participants at the first visit. The work described herein was executed in adherence with The Code of Ethics of the World Medical Association (Declaration of Helsinki).

Neuropsychological evaluation and determination of MCI status

Participants underwent a structured neuropsychological evaluation at the first visit to assess cognitive domains including premorbid ability, current cognitive status, processing speed, attention, visual-spatial processing, verbal and visual memory, and executive functions. Consensus diagnostic procedures based on published guidelines [12] determined cognitive status (normal, MCI, dementia). MCI was defined by performance at or below 1.5 SD of age and education-adjusted norms in at least two tests, preserved instrumental activities of daily living, and no dementia. Cognitive complaints were evaluated through clinical interviews and questionnaires [13].

Study outcome

Symbol digit modalities test (SDMT)

The oral version of the SDMT is a measure of speed of processing and working memory, requiring participants to pair symbols and digits based on a key, within a 90-s time interval; lower scores indicate worse performance. The SDMT is used as a primary cognitive outcome measure in MS clinical trials [9]. We used the total number of correct responses as the cognitive outcome measure.

Brain magnetic resonance imaging (MRI)

Image acquisition

Resting-state functional scans were acquired with eyes open on a 3 T Philips scanner in a 32-channel head coil (Elition 3.0TX, Philips Medical Systems, Best, The Netherlands) at the Gruss MRRC, Albert Einstein College of Medicine. Resting-state blood oxygen level-dependent (BOLD) images used single shot echo planar imaging, with 256-mm field of view (FOV) and 2.5 mm isotropic; TE/TR = 28/1000 ms, flip angle = 70°, 60 trans-axial slices, and multiband 4. T1-weighted structural images were acquired using axial 3D-MPRAGE with 240 mm FOV and 1.0 mm isotropic, TE/TR/TI = 4.6 ms/9.9/900 ms, SENSE acceleration 2 × 1.3.

Resting-state fMRI pre-processing

Resting-state BOLD images were preprocessed using FSL (Version 4.1), FMRIB’s Software Library (http://fsl.fmrib.ox.ac.uk/fsl) [14]. Preprocessing steps were similar to those in our previously published article [15] consisting of non-brain removal using BET, motion correction with MCFLIRT, slice-timing correction for interleaved acquisitions using Fourier-space time-series phase shifting, and high-pass temporal filtering using Gaussian-weighted least-squares straight line fitting (r = 100 s), and spatial smoothing using a Gaussian kernel with full-width half-maximum 5 mm, co-registration to high-resolution T1-weighted images, and normalization to standard space (Montreal Neurological Institute atlas, 2-mm resolution) using combined affine and nonlinear registration with Advanced Normalization Tools (ANTs) [16]. Denoising steps included removing motion artifacts and high-pass filtering to highlight slow-frequency fluctuations while reducing physiological and other noise sources.

Group-level independent component analysis (GICA)

Independent component analysis (ICA), a data-driven approach, was used to decompose resting-state data into a limited number of functionally related brain networks. Functional networks from multiple subjects were extracted using Group-level ICA (GICA) [17]. This was implemented using the Group ICA of fMRI Toolbox (GIFT) [21], which runs on MATLAB (R2023b, MathWorks, Natick, MA). Similar to dual regression, GIFT performs group ICA on the entire cohort, and subject-level networks are then estimated by back-reconstruction. In Line with previous literature examining major brain networks, the number of components was limited to 30 [18]. We manually identified and retained 12 out of the 30 components, which showed overlap with major functional networks commonly identified in the resting-state literature [19]. The brain networks identified included the medial visual, salience, anterior default-mode network (aDMN), posterior DMN (pDMN), sensorimotor network (SMN), left and right frontoparietal networks (LFPN/RFPN), and subcortical network, consistent with prior research [20, 21]. Network masks were generated using t ≥ 4 (i.e., p < 0.0001) from group-level t-stats images for each network, similar to prior group-ICA literature [22]. Subject-level RSFC brain scores were calculated by averaging all voxels within each group-level brain network mask. Brain networks were visualized with BrainNet Viewer (http://www.nitrc.org/projects/bnv/) [23].

Covariates

Age, sex, education, and global health status (GHS) were included as covariates. The GHS measures comorbidity with score ranges from 0 to 10 [11], indicating presence or absence of several clinical conditions, such as diabetes, chronic heart failure, and hypertension.

Additional measures used to characterize study participants

Two other measures, namely, the Repeatable Battery for the Assessment of Neuropsychological Status (RBANS) and the Patient Determined Disease Steps (PDDS), were collected as part of the study to characterize study participants but were not used in any analyses. RBANS has been thoroughly researched and validated for use in aging and individuals with MS [24]. Its effectiveness as a comprehensive measure of cognitive function has been well-established. PDDS is a valid commonly used self-report ordinal scale describing increasing degrees of MS-related disability and mobility limitations, and was recently validated in OAMS [25].

Statistical analysis

Descriptive statistics were used to summarize all study measures (mean ± SD for continuous measures, count and percent for categorical variables) and tabulated per group (MS vs. control). Independent sample t-tests were used to compare differences among the brain networks between MS and control groups. Average RSFC derived from each brain network served as the predictor and SDMT as the primary outcome for all analyses. The first set of analyses used multiple Linear regression analyses to test associations between brain network RSFC and SDMT across the whole sample, adjusted for covariates. Stratified Linear regressions examined whether these associations differed in OAMS compared to controls. To evaluate the moderating effect of MCI on RSFC, we performed a second set of analyses with an added MCI-by-RSFC interaction term using Hayes macro model 1 [26] in SPSS. Significant interactions were followed by conditional effects analyses. Furthermore, stratified linear regressions were conducted to determine whether the RSFC-SDMT associations differed in OAMS compared to controls in interaction models. Covariates included age, gender, education, and global health status (GHS). Statistical analyses were done in SPSS (version 29; SPSS, Inc., Chicago, IL), tests were two-sided, and p < 0.05 was considered statistically significant.

Results

Participants

Of 181 participants with resting-state functional scans, 167 participants were included in all analyses (exclusions: n = 8 due to dementia diagnosis, n = 6 due to motion artifacts). Participant characteristics are summarized per group (Table 1). There were more females in both groups (OAMS 68.3%, control 68.2%, p = 0.994). OAMS participants were on average younger than controls (64.4 ± 4.1 vs. 68.6 ± 7.1 years, p < 0.001) and had less education (14.82 ± 2.36 vs. 16.10 ± 2.40 years, p < 0.001). SDMT (47.51 ± 9.38 vs. 47.99 ± 10.09, p = 0.752) and total RBANS scores (91.20 ± 11.98 vs. 93.11 ± 11.97, p = 0.304) were comparable. There was a significantly greater number of MCI participants in OAMS than controls (OAMS 40.2%, control 15.3%, p < 0.001). SDMT scores were lower in MCI as compared to non-MCI participants (MCI 41.67 ± 9.26; non-MCI 50.07 ± 8.89; p < 0.001) in the full cohort, in OAMS (MCI 42.42 ± 9.39; non-MCI 50.94 ± 7.74; p < 0.001), and in controls (MCI 39.77 ± 8.97; non-MCI 49.47 ± 9.61; p < 0.001). Most OAMS participants were of relapsing remitting subtype (65.85%), and the median PDDS score of 2 indicates relatively low MS-related disability.

Table 1.

Sample characteristics. Values are quoted as mean ± SD or % for sex

Variables	Group
	MS (N = 82)	Controls (N = 85)	p-value
Age (years)	64.4 ± 4.1	68.6 ± 7.0	< 0.001
Sex (M/F, %F)	26/56 (68.3%)	27/58 (68.2%)	0.994
Education (years)	14.82 ± 2.36	16.11 ± 2.47	< 0.001
GHS (0–10)	1.37 ± 1.29	1.24 ± 0.91	0.452
SDMT	47.51 ± 9.38	47.99 ± 10.09	0.752
MCI (n, %)	33, 40.2%	13, 15.3%	< 0.001
RBANS index score	91.20 ± 11.98	93.11 ± 11.97	0.304
PDDSa	2 (0–5)
MS Subtypesb
Relapsing remitting	54 (65.85%)
Secondary progressive	20 (24.39%)
Primary progressive	3 (3.66%)

GHS Global Health Status; SDMT Symbol Digit Modalities Test; MCI Mild Cognitive Impairment; RBANS Repeatable Battery for the Assessment of Neuropsychological Status; PDDS Patient Determined Disease Steps

The entries in bold represent significant entries

^aMedian values used for PDDS

^bMS subtype was undetermined for 5 MS participants

Group-level ICA-derived brain networks

Of the eight brain networks identified by GICA at the group level, we present representative components portraying well-known brain networks in Fig. 1 (sensorimotor, salience, and left frontoparietal networks).

Fig. 1.

Resting-state brain networks obtained after running group ICA are shown on an MNI 152 surface template. The networks shown are a sensorimotor network (SMN), b salience network, and c left frontoparietal network (LFPN)

Cohort-wide associations between SDMT and brain networks

Results indicated that higher RSFC in SMN (t = 2.613, p = 0.01), LFPN (t = 2.231, p = 0.027), and RFPN (t = 2.003, p = 0.047) were significantly associated with better SDMT performance in the overall sample, controlling for covariates (see Table 2). The complete list of associations between SDMT and all brain networks is shown in Supplemental Table S1.

Table 2.

Associations of SDMT with brain networks in the overall sample of OAMS and healthy controls (N = 167)

Group	Network	Unstandardized		Standardized Beta	95% CI		T-value	p-value
		B	SE
Overall	SMN	14.379	5.503	0.195	3.511	25.246	2.613	0.01
	LFPN	10.883	4.879	0.165	1.248	20.519	2.231	0.027
	Salience	11.761	5.872	0.145	0.166	23.356	2.003	0.047
	RFPN	6.295	5	0.092	− 3.579	16.17	1.259	0.21

SMN Sensorimotor network; LFPN Left Fronto-Parietal Network; RFPN Right Fronto-Parietal Network

The entries in bold represent significant entries

Group-stratified associations between SDMT and brain networks

Within OAMS, higher RSFC in the SMN (t = 2.411, p = 0.018) and medial visual (t = 2.783, p = 0.007) networks were associated with better SDMT performance. No significant associations were found in controls (see Table 3). Supplemental Table S2 shows the complete list of group-stratified SDMT-RSFC associations for all networks.

Table 3.

Associations of SDMT with brain networks stratified by group status

Group	Network	Unstandardized		Standardized	95% CI		T-value	p-value
		B	SE	Beta
MS	Medial visual	11.651	4.187	0.295	3.312	19.991	2.783	0.007
	SMN	20.94	8.687	0.263	3.639	38.241	2.411	0.018
Control	Medial visual	− 3.839	4.349	− 0.095	− 12.495	4.817	− 0.883	0.38
	SMN	10.998	7.084	0.155	− 3.102	25.097	1.553	0.125

SMN Sensorimotor network

The entries in bold represent significant entries

Cohort-wide associations between SDMT and brain networks: main effects and interaction with MCI status

We observed significant associations of the main effect of RSFC with SDMT in LFPN (t = 2.431, p = 0.016), RFPN (t = 2.442, p = 0.016), SMN (t = 2.191, p = 0.03), and salience (t = 2.029, p = 0.044) networks (see Table 4). In addition, there was significant MCI-by-RSFC interaction in the subcortical network (t = − 2.105, p = 0.037). Subsequent conditional effects analysis showed that associations between subcortical RSFC and SDMT were in opposite directions in the MCI (t = − 1.712, p = 0.088) and non-MCI (t = 1.233, p = 0.219) groups, although not statistically significant. The complete list of cohort-wide associations between SDMT and all brain networks and interaction with MCI is shown in Supplemental Table S3.

Table 4.

Associations of SDMT with RSFC in brain networks and its interaction with MCI status in the overall sample of OAMS and healthy controls (N = 167)

Network	Measure	Unstandardized		Standardized Beta	95% CI		T-stats	p-value
		B	SE
Subcortical	RSFC	8.862	7.19	0.013	− 5.338	23.062	1.233	0.22
	RSFC × MCI	− 27.717	13.169	− 0.131	− 53.726	− 1.707	− 2.105	0.037
LFPN	RSFC	12.231	5.031	0.137	2.294	22.167	2.431	0.016
	RSFC × MCI	− 11.471	9.578	− 0.078	− 30.386	7.445	− 1.198	0.233
RFPN	RSFC	12.399	5.077	0.113	2.372	22.427	2.442	0.016
	RSFC × MCI	− 17.069	9.629	− 0.112	− 36.087	1.949	− 1.773	0.078
SMN	RSFC	12.513	5.711	0.14	1.233	23.792	2.191	0.03
	RSFC × MCI	− 8.143	11.27	− 0.049	− 30.402	14.116	− 0.723	0.471
Salience	RSFC	12.967	6.39	0.1	0.346	25.588	2.029	0.044
	RSFC × MCI	− 17.548	10.964	− 0.097	− 39.203	4.107	− 1.6	0.111

LFPN Left Fronto-Parietal Network; RFPN Right Fronto-Parietal Network; SMN Sensorimotor Network; RSFC Resting-state functional connectivity; MCI Mild Cognitive Impairment

The entries in bold represent significant entries

Group-stratified associations between SDMT and brain networks: main effects and interaction with MCI status

Within OAMS, higher pDMN RSFC was associated with better SDMT performance (t = 2.229, p = 0.029). In addition, there were significant MCI-by-RSFC interactions in pDMN (t = − 2.454, p = 0.016) and subcortical (t = − 2.065, p = 0.042) networks (see Table 5). Conditional effect analysis revealed that the association between pDMN and SDMT was negative but not significant in the MCI group (t = − 1.291, p = 0.201), whereas it was positive and significant in the non-MCI group (t = 2.229, p = 0.028). The associations between RSFC and SDMT in the MCI (t = − 1.863, p = 0.066) and non-MCI (t = 1.01, p = 0.316) groups were also in opposite directions for the subcortical network but not statistically significant. Within the control group, we observed higher RSFC in salience (t = 2.383, p = 0.02), LFPN (t = 2.143, p = 0.035), and RFPN (t = 2.109, p = 0.038) networks were associated with better SDMT but no significant MCI-by-RSFC interactions for any of the brain networks (see Table 4). The complete list of group-stratified associations between SDMT and all networks, as well as interactions with MCI, is shown in Supplemental Table S4.

Table 5.

Associations of SDMT with RSFC in brain networks and its interaction with MCI status stratified by group status

Group	Network	Measure	Unstandardized		Standardized Beta	95% CI		T-stats	p-value
			B	SE
MS	Subcortical	RSFC	11.027	10.915	− 0.033	− 10.72	32.775	1.01	0.316
		RSFC × MCI	− 34.858	16.882	− 0.188	− 68.496	− 1.219	− 2.065	0.042
	pDMN	RSFC	24.257	10.88	0.097	2.578	45.936	2.229	0.029
		RSFC × MCI	− 39.126	15.944	− 0.22	− 70.896	− 7.357	− 2.454	0.016
Controls	Salience	RSFC	22.08	9.267	0.187	3.627	40.534	2.383	0.02
		RSFC × MCI	− 34.914	18.561	− 0.141	− 71.873	2.045	− 1.881	0.064
	LFPN	RSFC	14.448	6.742	0.18	1.022	27.874	2.143	0.035
		RSFC × MCI	− 16.317	17.808	− 0.088	− 51.776	19.143	− 0.916	0.362
	RFPN	RSFC	14.516	6.883	0.174	0.811	28.221	2.109	0.038
		RSFC × MCI	− 15.908	18.762	− 0.082	− 53.268	21.451	− 0.848	0.399
	SMN	RSFC	14.323	7.289	0.178	− 0.191	28.838	1.965	0.053
		RSFC × MCI	− 11.243	17.49	− 0.057	− 46.07	23.583	− 0.643	0.522

pDMN posterior Default Mode Network; LFPN Left Fronto-Parietal Network; RFPN Right Fronto-Parietal Network; SMN Sensorimotor Network; RSFC Resting-state functional connectivity; MCI Mild Cognitive Impairment

The entries in bold represent significant entries

Discussion

This study investigated the relationship between cognition and brain integrity, based on functional MRI network analysis, in older adults with multiple sclerosis and healthy controls. Specifically, we examined whether RSFC in empirically derived functional brain networks was associated with SDMT performance and if these associations differed between OAMS and controls. Given that MCI is common in both aging and OAMS and has been shown to influence RSFC, we further evaluated whether the presence of MCI modified these associations in the entire cohort and within each group. We found that higher RSFC in specific networks was related to better SDMT performance, but the presence of MS and MCI modified these associations. These findings are discussed in detail below.

Overall, better performance on the SDMT, a measure of processing speed and executive functions, was related to higher connectivity in sensorimotor, left frontoparietal, and salience networks in older adults with or without MS. The association between RSFC in brain networks and SDMT reflects the multi-dimensionality of the SDMT as a measure of cognition, assessing multiple cognitive functions such as processing speed, working memory, and visual scanning [27]. However, stratified analyses showed that this association was influenced by the presence of MS (see below). Prior studies reported that better SDMT performance correlated with higher RSFC in the motor network among older healthy adults [28] and younger adults with MS [29]. A recent study in older adults found that functional connectivity (FC) across various neural networks—sensorimotor, subcortical, frontoparietal, medial frontal, default-mode, and visual—correlated with processing speed [28]. Previous studies on younger adults with MS have associated processing speed with multiple brain networks, including the default-mode [21], frontoparietal [20, 30], and visual [31] networks, aligning with current findings. The salience network, which involves regions for higher-order cognitive integration [42], mediates the interaction between the default-mode and frontoparietal networks during task-based processing. In the current study, higher RSFC in the salience network was associated with better SDMT performance suggesting greater reliance on cognitive processes in older adults with and without MS. The salience network is important in older adults, as its interplay with other networks is crucial in the transition to cognitive impairment during normal aging [32] and in MS [33].

Among OAMS, better SDMT performance correlated with higher RSFC in sensorimotor and medial visual networks. Two networks, aDMN and LFPN, showed trends toward significance in the OAMS group (see Supplementary Table S3). No significant SDMT-RSFC associations were found in the control group. A prior study noted a positive correlation between sensorimotor network FC and processing speed in younger individuals with RRMS [20], while another found a negative correlation between PASAT scores and medial visual network FC [34]. Differences in SDMT and visual network associations across studies may stem from variations in disease duration, phenotype, and cognitive impairment levels. Notably, network integration fluctuates with age [28]; thus, our results in older adults may represent associations between RSFC and cognition that are distinct from those reported in previous studies with younger adults with MS. Consistent with our findings, DMN structural integrity and dynamic FC were both positively related to SDMT scores indicating its crucial role in information processing [21].

Adding MCI as a moderator clarified associations between RSFC and cognitive performance. These analyses showed that higher SDMT scores correlated with greater connectivity in the SMN, LFPN, RFPN, and salience networks. While previous studies have examined cognitive impairment in MS through cognitive scores or RSFC robustness [35], fewer have explored the effects of cognitive impairment on cognitive measures specifically through brain network associations [30, 36, 37]. Two [36, 37] of the three aforementioned studies used SDMT to stratify participants into cognitively impaired and cognitively preserved groups, limiting comparisons with our current study. One study found significant associations between SDMT scores and higher RSFC in LFPN and RFPN [30], mirroring our results. Regarding the somewhat counterintuitive findings of negative associations of RSFC with SDMT, a previous study reported that FC in DMN and cognition were negatively correlated in MCI participants [38], consistent with our findings. Additionally, the conditional effects in the subcortical network observed in the current study suggest network inefficiency in MCI, as highlighted by the inverse relationship between RSFC and SDMT.

In models using MCI as a moderator, better SDMT performance correlated with higher RSFC in brain networks, primarily within the control group for LFPN, RFPN, and salience networks, with SMN trending toward significance (p < 0.053; see Supplementary Table S5). Among OAMS, a significant association was found between higher SDMT scores and FC of the pDMN. Consistent with the literature, the current study revealed a higher percentage of MS participants with MCI compared to controls. Group stratified analyses indicated that differing associations between the SDMT and brain networks were influenced by cognitive impairment status, confirming the effect of MCI on the relationship between brain networks and cognition. The conditional effects in the pDMN and subcortical networks for OAMS support the idea of network inefficiency due to cognitive impairment; specifically, associations between RSFC and SDMT performance were negative in those with MCI but positive in individuals with normal cognition. Subcortical structural changes correlate with cognitive changes in MS [39], and our FC results support these findings. The limited number of MCI participants among controls likely restricted our power to detect interaction effects. Nonetheless, incorporating MCI clarified the main effects in the control group that might have been overlooked [40]. The results underscore the importance of cognitive impairment in the RSFC–cognition relationship. Previous studies suggest that MS is characterized by unstable RSFC [41], worsening at higher disability levels, and potentially leading to cognitive impairment [4]. Our findings indicate that cognitive impairments in older adults with MS may lead to functional network changes despite no measurable difference in cognitive scores as measured by both SDMT and RBANS.

Our data-driven methodology for RSFC calculation in cerebral networks is robust and effectively captures extensive neuronal activity [42, 43]. We have a few limitations to mention. Although our functional data were not collected during active cognitive tasks, some literature addresses functional brain correlates of SDMT using task-based functional MRI [44]; thus, our data complement these studies. The unweighted summation of ten diseases used for the GHS variable limited our adjustment for comorbidities. This was the most parsimonious approach, as we lacked a priori hypotheses for specific comorbidities to determine weightings. Analyses controlled for chronological age indicated that healthy controls were older than OAMS participants, potentially impacting findings. However, this difference, if anything, may have attenuated the demonstrated group differences. Furthermore, significant results in the stratified analysis for the MS group, but not for controls, suggest age did not drive the significant associations. Various methodological challenges in RSFC estimations can influence study outcomes [42]. Comparing seed-based, voxel-based, and data-driven approaches to calculate RSFC networks was beyond the scope of this investigation. Future studies could assess different RSFC methods and link them to cognition. Lastly, our cohort included individuals with relatively low levels of disability and cognitive impairments, so results may not generalize to a broader MS spectrum with more severe deficits.

Conclusions

In conclusion, the current study presents evidence that in older adults the presence of both MS and MCI influences brain support, measured through RSFC of empirically derived networks, of cognitive task performance.

Supplementary Information

Below is the link to the electronic supplementary material.

Abbreviations

GICA

Group-level independent component analysis

GHS

Global health status

aDMN

Anterior default mode network

LFPN

Left frontoparietal network

MCI

Mild cognitive impairment

MS

Multiple sclerosis

OAMS

Older adults with MS

pDMN

Posterior default mode network

RFPN

Right frontoparietal network

RSFC

Resting-state functional MRI connectivity

SDMT

Symbol digit modalities test

SMN

Sensorimotor network

Author contribution

Siddharth Nayak: drafting/revision of the manuscript for content, including medical writing for content; study concept or design; analysis or interpretation of data. Roee Holtzer: drafting/revision of the manuscript for content, including medical writing for content; major role in the acquisition of data; study concept or design; analysis or interpretation of data. Mark Wagshul: drafting/revision of the manuscript for content, including medical writing for content; analysis or interpretation of data.

Funding

This work was supported by funding from NINDS: R01NS109023.

Data availability

The corresponding author has full access to all study data and takes responsibility for data integrity and analysis. Anonymized data are available on reasonable request.

Declarations

Ethics approval

The study was approved by the IRB of Albert Einstein College of Medicine (Protocol #2019–10049). Written informed consent was obtained from all participants at the first visit. The work described herein was executed in adherence with The Code of Ethics of the World Medical Association (Declaration of Helsinki).

Consent for publication

All authors have consented to the publication of this study and its findings as an original research article in the journal GeroScience.

Competing interests

The authors declare no competing interests.

Disclaimer

The funders played no role in the study’s design, the collection, analysis, or interpretation of data, the writing of the manuscript, or the decision to publish the results.