Gray Matter Volume Covariance Networks, Social Support, and Cognition in Older Adults

Abstract

Objective

We examined the neural substrates of social support in older adults. Social support is associated with better outcomes in many facets of aging—including cognitive and functional health—but the underlying neural substrates remain largely unexplored.

Methods

Voxel-based morphometry and multivariate statistics were used to identify gray matter volume covariance networks associated with social support in 112 older adults without dementia (M age = 74.6 years, 50% female), using the Medical Outcomes Study Social Support Survey.

Results

A gray matter network associated with overall social support was identified and included prefrontal, hippocampal, amygdala, cingulate, and thalamic regions. A gray matter network specifically associated with tangible social support (e.g., someone to help you if you were confined to bed) was also identified, included prefrontal, hippocampal, cingulate, insular, and thalamic regions, and correlated with memory and executive function.

Discussion

Gray matter networks associated with overall and tangible social support in this study were composed of regions previously associated with memory, executive function, aging, and dementia. Longitudinal research of the interrelationships between social support, brain structure, and cognition is needed, but strengthening social support may represent a new path toward improving cognition in aging that should be explored.

A broad social support system can promote healthy aging and as one grows older, this support can vary due to life events and declining health. Social support has a positive influence on a number of outcomes in aging (Seeman & Crimmins, 2001; Umberson & Montez, 2010), including cognitive (Barnes, Mendes de Leon, Wilson, Bienias, & Evans, 2004; Seeman, Lusignolo, Albert, & Berkman, 2001) and functional decline (Boult, Kane, Louis, Boult, & McCaffrey, 1994). The protective effects of social support on cognitive decline could be due to increased communication and interpersonal interactions that require more cognitive resources (Berkman, 2000). A broad social support system could also serve as a buffer between stress and health—a proposition referred to as the stress-buffering hypothesis (Cohen & Wills, 1985). Support from others could also encourage healthy behaviors such as exercise and proper nutrition (Emmons, Barbeau, Gutheil, Stryker, & Stoddard, 2007). Older adults with poor social support, however, display worse cognitive performance (Bassuk, Glass, & Berkman, 1999; Gow, Corley, Starr, & Deary, 2013) and an increased risk of dementia (Fratiglioni, Wang, Ericsson, Maytan, & Winblad, 2000; Kuiper et al., 2015).

Social support can be characterized in different ways, including social network size, social engagement and affiliation, and availability of social support. An important aspect of social support is the level of perceived social support—as positive perceptions of one’s amount of social support encourage greater health care utilization (Thoits, 2011) and could serve as a buffer to stressful life events (Cutrona, Russell, & Rose, 1986). The Medical Outcomes Study Social Support Survey (MOS-SSS) is a validated and reliable social support scale that distinguishes between emotional or informational support, tangible support, affectionate support, and positive social interactions—including such items as “someone to share your most private worries and fears with” (emotional/informational), “someone to help you if you were confined to bed” (tangible), “someone who hugs you” (affectionate), and “someone to do something enjoyable with” (positive social interaction) (Gjesfjeld, Greeno, & Kim, 2007; Sherbourne & Stewart, 1991).

The relationship between perceived social support and cognition in older adults, however, is not well understood. One study suggests that overall perceived social support is positively associated with global cognition in older adults (Pillemer & Holtzer, 2016). Of the four factors of the MOS-SSS, both emotional/informational support and positive social interaction were associated with better global cognition in this study. Another study has linked overall support, tangible support, affectionate support, and positive social interactions to global cognitive decline—particularly among older men (Pillemer, Ayers, & Holtzer, 2018). Other studies have observed a positive correlation between the number of individuals in a person’s social network and cognitive ability in older adults (Barnes et al., 2004; Gow et al., 2013). A social network is a social structure of interactions or relationships such as family members, friends, and colleagues that potentially can, but do not necessarily, provide social support. Social networks are pruned in old age, to prioritize closer or more meaningful relationships, and therefore, somewhat counterintuitively, older adults are more capable of regulating, remembering, and responding to their own and others’ emotions than younger adults. (Carstensen, Fung, & Charles, 2003). Thus, both perceived social support and social network size likely reflect the amount of support available to older adults.

The current literature suggests that the neural substrates of social interactions can be found in several regions throughout the brain, but few studies have examined the neural substrates of social support in older adults. Studies examining related measures such as social affiliation, social isolation, social evaluation, and social network size in a range of different age groups, however, have implicated a number of brain regions including amygdala and hippocampal regions (Bickart, Wright, Dautoff, Dickerson, & Barrett, 2011; Sherman, Cheng, Fingerman, & Schnyer, 2016) as well as precuneus (Beadle, Yoon, & Gutchess, 2012; Cassidy, Shih, & Gutchess, 2012), posterior cingulate (Cassidy et al., 2012; Che et al., 2014) and prefrontal regions—particularly medial prefrontal cortex (Beadle et al., 2012; Cassidy et al., 2012; Sherman et al., 2016). These regions have been previously linked to a variety of social, emotional, and cognitive processes—including social cognition, stress, episodic memory, and executive function. Yet, social support and stress has been shown to independently contribute to amygdala volume (Sherman et al., 2016). We have also linked social network size in older adults to functional connectivity in several resting-state functional networks, including sensory-motor, visual, vestibular-insular, and left fronto-parietal networks (Pillemer, Holtzer, & Blumen, 2017), as well as to a fairly distributed network of gray matter volume, including prefrontal (medial, lateral, and orbital), hippocampal, amygdala, precuneus, insular, and cingulate regions (Blumen & Verghese, 2018). Associating social networks and support to several different brain networks or brain regions is consistent with that social networks and support have been linked to several different social, cognitive, emotional, and functional outcomes in aging.

The present cross-sectional study aims to investigate the neural substrates of perceived social support in older adults and how they relate to different cognitive functions, including global cognition, processing speed, episodic memory, and executive function. Voxel-based morphometry methods and multivariate covariance-based statistics were used to identify gray matter covariance patterns or “networks” associated with informational support, tangible support, affectionate support, and positive social interactions as well as overall social support—using the MOS-SSS (Sherbourne & Stewart, 1991). Unlike traditional univariate approaches, which can be used to evaluate gray matter volumes on a voxel-by-voxel basis, these multivariate covariance-based analyses consider the covariance between gray matter volumes in different brain regions—and can therefore more easily be interpreted as neural networks (Habeck & Stern, 2010). Multivariate covariance-based analyses also avoid the multiple comparison problem of traditional univariate analyses and are reproducible across different data sets of older adults (Bergfield et al., 2010; Habeck et al., 2008). Although these analyses are data-driven, we expected perceived social support to be associated with distributed networks of brain regions that included prefrontal (particularly medial prefrontal), hippocampal, and amygdala regions. The extent to which older adults displayed these networks was then correlated with global cognition, processing speed, episodic memory, and executive function. Given the role that hippocampal and prefrontal cortex regions play in episodic memory and executive function, respectively, we expected the gray matter covariance networks associated with social support to be more strongly correlated with episodic memory and executive function, than with global cognition and processing speed. The cross-sectional design of this study, however, did not permit us to determine the directionality of these relationships.

Method

Participants

We examined gray matter covariance networks associated with perceived social support in 112 older adults without dementia (56 females; 56 males) from the ongoing Central Control of Mobility in Aging Study at Albert Einstein College of Medicine. The demographic characteristics of this sample are summarized in Table 1. Persons with dementia were excluded based on a telephone-based memory impairment screen of <5 (Lipton et al., 2003), Ascertain Dementia 8-item Informant Questionnaire score of >1 (Galvin et al., 2005) or the Diagnostic and Statistical Manual of Mental Disorders fourth edition (DSM-IV) criteria at consensus clinical case conferences (American Psychiatric Association, 2000). Additional exclusion criteria included serious chronic or acute illness (e.g., cancer), any medical illness or chronic medication use (e.g., neuroleptics) that could compromise safety or affect cognitive functions, terminal illness with life expectancy less than 12 months, progressive neurodegenerative disease (e.g., Parkinson’s disease), hospitalization for severe illness or surgery in the past 6 months, severe auditory or visual loss, active psychoses or psychiatric symptoms, and living in a nursing home. Older adults with MRI contraindications (e.g., pacemaker) were also excluded.

Table 1.

Demographics

Variable	N = 112	Range
Age (years), mean (SD)	74.6 (5.93)	65–91
Education (years), mean (SD)	15.7 (3.41)	6–28
Gender, n female (%)	56 (50%)
Race/Ethnicity, number (%)
Caucasian	78 (69.6%)
Black	27 (24.1%)
Hispanic, white	2 (1.79%)
Hispanic, black	0 (0.00%)
Asian	3 (2.68%)
Other/missing	2 (1.79%)
Total Intracranial Volume (liters), mean (SD)	1.34 (0.14)	1.0–1.6
Global Health Status (0–9), mean (SD)	1.20 (1.08)	0–4
Diabetes, number (%)	21 (18.8%)
Hypertension, number (%)	46 (41.4%)
Myocardial Infarction, number (%)	1 (0.89%)
Chronic heart failure, number (%)	1 (0.089%)
Arthritis, number (%)	52 (46.4%)
Depression, number (%)	6 (5.36%)
Stroke, number (%)	3 (2.68%)
COPD, number (%)	2 (1.79%)
Mild Cognitive Impairment, number (%)	12(10.71%)
Time Difference (days), mean (SD)	61.9 (73.5)	−177–178
Marital Status: number (%)
Never Married	13 (11.6%)
Separated/Divorced	26 (23.2%)
Widowed	24 (21.4%)
Currently Married	49 (43.8%)
MOS-SSS Overall Support, mean (SD)	4.07 (0.82)	1.63–5
Emotional/Informational, mean (SD)	4.05 (0.90)	1.75–5
Tangible, mean (SD)	3.82 (1.29)	1–5
Affectionate, mean (SD)	4.33 (0.91)	1–5
Positive Social Interaction, mean (SD)	4.22 (0.85)	2–5
RBANS Total, mean (SD)	92.6 (11.9)	62–122
TMT: B-A (seconds), mean (SD)*	71.5 (49.5)	8.98–261.9
FCSRT Total Free, mean (SD)*	31.6 (6.69)	6–46

Note: COPD = Chronic obstructive pulmonary disease, time difference refers to the number of days in between MRI scan and behavioral assessment; FCSRT = Free and Cued Selective Reminding Test (higher scores indicate better performance); MOS-SSS = Medical Outcomes Study Social Support Survey; RBANS = Repeatable Battery for the Assessment of Neuropsychological Status (higher scores indicate better performance); TMT = Trail Making Test (lower scores indicate better performance).

*Significantly (p < .05) associated with MOS-SSS Score 2 (Tangible) SSF.

Measures and Covariates

The Medical Outcomes Study Social Support Survey (MOS-SSS [Sherbourne & Stewart, 1991]) was used to quantify perceived level of social support with a total of 19 items, subdivided into four previously validated dimensions (Gjesfjeld et al., 2007; Pillemer & Holtzer, 2016; Sherbourne & Stewart, 1991): (a) emotional/informational support, (b) tangible support, (c) affectionate support, and (d) positive social interaction. For each item, participants were instructed to indicate how often each of the types of support was available to them if they needed it, ranging from none of the time (1), a little of the time (2), some of the time (3), most of the time (4), to all of the time (5). Higher scores indicate a greater perceived level of social support. The MOS-SSS measures emotional/informational support with eight questions about the availability of support such as “Someone whose advice you really want” or “Someone who understands your problems”. Tangible support consists of four items such as “Someone to help you if you were confined to bed”. Affectionate support consists of three questions such as “Someone who shows you love and affection”. Finally, positive social interaction is measured with three items such as “Someone to have a good time with”.

Using the Social Network Index (SNI) (S. Cohen, Doyle, Skoner, Rabin, & Gwaltney, 1997), we also determined current marital status and divided into four categories: “Currently married and living together, or living with someone in a marital-like relationship”, “Never married and never lived with someone in a marital-like relationship”, “Separated/divorced, or formerly lived with someone in a marital-like relationship”, and “Widowed”. The SNI can be used to quantify the number of high-contact social roles (e.g., spouse, parent, child) an individual interacts with at least biweekly, as well as the total number of networks members an individual interacts with at least biweekly, and has been used in our previous studies linking social networks to functional connectivity and gray matter covariance network in older adults (Blumen & Verghese, 2018; Pillemer et al., 2017)

The Repeatable Battery for the Assessment of Neuropsychological Status (RBANS (Randolph, Tierney, Mohr, & Chase, 1998)) was used to assess global cognition (Total Index). Processing speed and executive function were assessed with the Trail Making Test: Time to complete Part A (TMT:A) and Time to complete Part B minus time to complete Part A (TMT:B-A), respectively (Reitan, 1978). Episodic memory was assessed with the Free and Cued Selective Reminding Test: Total Free Recall (FCSRT (Buschke, 1973)). Finally, a global health score was obtained from dichotomous ratings (presence or absence) of physician diagnosed diabetes, chronic heart failure, arthritis, hypertension, depression, stroke, chronic obstructive pulmonary disease, angina, and myocardial infarction.

MRI Data Acquisition

Images were acquired at the Gruss Magnetic Resonance Research Center at the Albert Einstein College of Medicine (Bronx, NY). Images were acquired with a Philips 3T MRI scanner (Achieva Quasar TX; Philips Medical Systems, Best, Netherlands). Standard three-dimensional T1-weighted images were obtained: TR/TE of 9.9/4.6 ms, 240 mm² FOV, 240 × 240 × 240 matrix and 1 mm voxel size. MRI scans were obtained within 6 months of MOS-SSS assessment, and we dichotomized the number of days in between the scan and assessment into two groups (less than 90 days = 0, greater than or equal to 90 days = 1) to include as a covariate in our analyses.

MRI Preprocessing

T1-weighted images were manually reoriented to the anterior commissure-posterior commissure line and then preprocessed using SPM12 (Wellcome Department of Cognitive Neurology) and MATLAB R2016b (Mathworks, Natick, MA). Voxel-Based Morphometry was used to segment each T1-weighted image into gray matter (GM), white matter (WM), and cerebrospinal fluid (CSF), using the unified segmentation procedure and Diffeomorphic Anatomical Registration Through Exponentiated Line Algebra (DARTEL; Ashburner, 2007; Ashburner & Friston, 2005). DARTEL ensures proper intersubject alignment by modeling the shape of the brain using three parameters for each voxel, and simultaneously align gray matter and white matter to produce a study-specific and increasingly crisp template to which the data are iteratively aligned. DARTEL produces GM, WM, and CSF probability maps in the same space as the original T1-weighted images. These probability maps were then spatially normalized into Montreal Neurologic Institute (MNI) space and smoothed with an isotropic Gaussian kernel, full-width-at half-maximum = 8 mm. Only GM probability maps were used in upcoming multivariate analyses.

Multivariate Covariance-Based and Correlational Analyses

The principal components analysis suite, http://www.nitrc.org/projects/gcva_pca (Habeck & Stern, 2007) was used to identify gray matter covariance networks or patterns associated with overall perceived social support, and each subfactor of perceived social support (emotional/informational, tangible, affectionate, and positive social interaction). Because social support scores were not normally distributed, square transformations of these scores were used. These multivariate analyses were also adjusted for age, education, GHS, gender, marital status, total intracranial volume, and dichotomized date (less than 90 days and greater than or equal to 90 days) difference between MRI scan and social support assessment. First, gray matter probability maps were masked with a gray matter mask supplied by SPM12 to only include voxels with > 20% probability of being gray matter. After participant means were subtracted from each voxel, a PCA was performed, to generate a set of principal components and their associated participant-specific (or pattern) expression scores. Participant-specific expression scores reflect the degree to which a participant displays a particular component or pattern. A gray matter volume covariance pattern associated with perceived social support was then computed by regressing the participant-specific factor scores from the best linear combination of principal components, selected using the Akaike information criteria, against perceived social support. The stability of the voxels in this GM volume covariance pattern was then tested using 1,000 bootstrap resamples. Voxels with bootstrap samples of [Z] > + 1.96 or < −1.96, p < .05 were considered statistically significant.

These multivariate analyses allowed us to identify key “nodes” in the gray matter volume covariance “networks” associated with perceived social support. Covariance patterns obtained from a multivariate analysis assigns positive and negative weightings (or loadings) to each voxel (or variable) included in the analysis. A voxel with a positive weighting has a relatively greater value within the respective network, while a voxel with a negative weighting have a relatively lower value within the respective network (Habeck et al., 2008). Within the context of the current analyses, positively weighted regions show relatively more volume as a function of increased social support while negatively weighted regions show relatively less volume as a function of increased social support. Statistically significant clusters were labeled using the Automated Anatomical Labeling Brain Atlas through MRIcron software, and confirmed with visual inspection. Clusters either outside the brain or in white matter regions, or those smaller than 10 voxels were excluded. Finally, the factor scores obtained from the multivariate analyses of the statistically significant covariance patterns were correlated with cognitive measures, including the RBANS, FCSRT, and TMT. These correlations were Bonferroni corrected to adjust for multiple comparisons.

Results

Sample characteristics are reported in Table 1. A total of 112 CCMA participants who completed the MOS-SSS and underwent MRI scanning were included in the current analysis. The mean age of participants was 74.6 years (±5.93). On average, they had 15.7 years (±3.41) of education and 50% were female. Participants were of average global cognition, with a mean RBANS total index score of 92.6 (±11.9). The mean global health status score was 1.20 (±1.08), indicative of good general health. Forty-nine (43.8%) participants were currently married, 26 (23.2%) separated/divorced, 24 (21.4%) widowed, and 13 (11.6%) never married. On average, participants reported an overall perceived social support score of 4.07 (±0.82), suggesting that perceived overall support was available “most of the time.” For tangible support, participants reported an average score of 3.82 (±1.29), suggesting that on average the respondents perceive tangible support to be available somewhat less than “most of the time.” The mean and standard deviations (SD) for other subscores are reported in Table 1. Multivariate covariance-based analyses revealed a gray matter covariance network associated with overall and tangible support, but other subscores were not statistically significant. These gray matter covariance networks were composed of both overlapping and unique brain regions (see Figures 1 and 2). We also found that greater expression of the covariance pattern linked to tangible support was associated with better memory on the FCSRT (Total Free Recall), and better executive functioning (Trail-Making Test B-A). Global cognition (RBANS Total Index) and processing speed (Trial-Making Test A) were not found to be significantly associated. These results are further discussed below.

Figure 1.

Gray matter covariance pattern associated with social support (MOS-SSS Overall Score) in 112 older adults without dementia.

Figure 2.

Gray matter covariance pattern associated with tangible support (MOS-SSS Tangible Score) in 112 older adults without dementia.

Gray Matter Covariance Network Associated with Perceived Overall Social Support

The gray matter volume network associated with overall support was constructed from three principal components and had an R² of .18 (see Figure 1 and Table 2). Positively weighted regions included superior and inferior frontal (including medial orbitofrontal) regions, middle temporal regions (including hippocampal and amygdala), as well as parietal (supramarginal gyrus), cingulate (anterior and posterior), and thalamic regions. Negatively weighted regions included superior, middle, and inferior frontal (including orbital, triangular, and opercular) regions, superior, middle, and inferior temporal (including inferior temporal pole), superior parietal regions, as well as precentral, postcentral, supplementary motor, and cuneus regions. The extent to which older adults displayed this network was not significantly associated with any measures of cognitive function (Total RBANS, r = −.05, p = 1.00, (unadjusted p = .61; TMT: A r = .19, p = .43 (unadjusted p = .04); TMT: B-A r = .18 p = .64, (unadjusted p = .06); FCSRT Total Free Recall, r = .19, p = .39 (unadjusted p = .04). Both Bonferroni-adjusted and unadjusted values are reported above for completeness. Our unadjusted p-value threshold was p < .0083, as we ran a total of six bivariate correlations with cognitive measures.

Table 2.

Regions in the Gray Matter Covariance Network Associated with MOS-SSS Overall Score

Brain regions	L/R	x	y	z	k	z-value
Positive
Supramarginal Gyrus	R	54	−44	42	158	2.3840
Anterior Cingulum	R	8	36	24	2,773	2.3488
Thalamus	R	8	−14	12	7,153	2.2893
Posterior Cingulum	R	3	−36	34	131	2.2360
Hippocampus (extending into the amygdala)	R	36	−15	−18	203	2.1481
Middle Temporal Gyrus	R	52	−33	−4	15	2.0457
Inferior Frontal Gyrus (medial, orbital)	R	3	46	−8	66	2.0453
Hippocampus (extending into the amygdala)	L	−27	−9	−22	32	2.0364
Superior Frontal Gyrus	R	28	54	9	13	2.0241
Negative
Cuneus	L	−16	−90	42	9,011	−2.4336
Inferior Frontal Gyrus (opercular)	R	54	26	36	503	−2.4018
Middle Temporal Gyrus	L	−69	−15	−2	239	−2.3696
Precentral Gyrus	R	33	−18	74	937	−2.3630
Postcentral Gyrus	L	−62	−4	38	549	−2.3336
Inferior Temporal Gyrus	L	−68	−26	−27	1,457	−2.3009
Superior Frontal Gyrus	L	−14	45	51	248	−2.2796
Temporal Pole (inferior)	L	−27	−6	−51	206	−2.2392
Precentral Gyrus	L	−22	−21	78	98	−2.2058
Supplementary Motor Area	L	−12	21	68	135	−2.2004
Postcentral Gyrus	L	−46	−28	66	112	−2.1998
Postcentral Gyrus	R	68	−3	16	196	−2.1812
Superior Frontal Gyrus	R	24	8	70	38	−2.1711
Superior Frontal Gyrus	L	−21	−2	74	69	−2.1676
Inferior Frontal Gyrus (orbital)	L	−12	42	−30	68	−2.1663
Superior Frontal Gyrus	L	−28	63	16	65	−2.1614
Inferior Frontal Gyrus (orbital)	R	6	60	−27	234	−2.1581
Superior Temporal Gyrus	R	69	−8	2	87	−2.1580
Superior Parietal Lobule	L	−50	−68	45	68	−2.1517
Middle Frontal Gyrus	L	−46	26	42	64	−2.1453
Superior Frontal Gyrus	L	−12	68	22	58	−2.1186
Inferior Frontal Gyrus (opercular)	R	60	18	14	47	−2.1083
Middle Frontal Gyrus	R	32	50	38	37	−2.1019
Middle Frontal Gyrus	R	27	63	22	36	−2.0593
Superior Frontal Gyrus	R	24	70	8	46	−2.0510
Inferior Frontal Gyrus (triangular)	L	−50	38	27	35	−2.0483
Superior Parietal Lobule	L	−68	−36	33	10	−2.0016

Gray Matter Covariance Network Associated with Specific Aspects of Social Support

A reliable gray matter network specifically associated with tangible support was also identified—but the gray matter volume covariance networks associated with emotional/information support, affectionate support, and positive social interaction were not statistically significant. The gray matter volume covariance pattern associated with tangible support was constructed from one principal component and had an R² of .26 (see Figure 2 and Table 3). Positively weighted regions included superior, middle, and inferior frontal (triangular) regions, superior and middle temporal (including hippocampal, parahippocampal, and fusiform gyrus) regions, as well as precentral, postcentral, thalamic, insular, calcarine sulcus, cuneus, precuneus, and caudate regions. Negatively weighted regions included superior, middle, and inferior frontal (medial, orbital, and opercular) regions, middle temporal regions, middle occipital regions, inferior parietal regions (including angular gyrus), as well as cingulate (anterior and posterior), supplementary motor areas, insular, precentral, thalamic, precuneus, and cerebellar (Crus I) regions. We also found that greater expression of this gray matter covariance pattern linked to tangible support was associated with better memory on the FCSRT Total Free Recall, r = .29, p = .02 (unadjusted p = .002), and greater executive functioning on the TMT: B-A, r = −.30, p = .02 (unadjusted p = .002), but not with processing speed (TMT: A r = .11, p = 1.00 (unadjusted p = .26) or global cognition (Total RBANS r = .09, p =1.00 (unadjusted p = .32).

Table 3.

Regions in the Gray Matter Covariance Network Associated with MOS-SSS Score 2

Brain regions	L/R	x	y	Z	k	z-value
Positive
Calcarine Sulcus (extending bilaterally into cuneus and precuneus)	R	15	−60	20	10,617	2.1211
Middle Temporal Gyrus	L	−50	−56	22	1,509	2.0987
Precentral Gyrus	L	−40	9	32	373	2.0920
Postcentral Gyrus	L	−44	−18	36	767	2.0871
Middle Temporal Gyrus	R	46	−62	14	567	2.0823
Insula	L	−42	−20	14	5,432	2.0819
Thalamus (extending into hippocampus)	R	9	−16	14	4,337	2.0743
Middle Temporal Gyrus	R	56	−28	−3	1,803	2.0708
Inferior Frontal Gyrus (triangular)	R	44	14	26	315	2.0646
Superior Temporal Gyrus (extending into insula)	R	44	−27	16	2,470	2.0591
Insula	R	48	−9	32	799	2.0437
Superior Frontal Gyrus (medial)	R	6	46	10	1,329	2.0376
Middle Frontal Gyrus	R	28	57	4	81	2.0156
Parahippocampal Gyrus	L	−27	−48	−8	27	1.9905
Middle Frontal Gyrus	L	−28	54	8	47	1.9882
Fusiform Gyrus	L	−22	−75	−14	13	1.9815
Insula	R	40	18	3	89	1.9766
Caudate	R	14	16	9	31	1.9684
Negative
Supplementary Motor Area (extending bilaterally into superior frontal gyrus)	R	8	20	70	67,626	−2.1252
Posterior Cingulum	R	3	−40	14	335	−2.1138
Superior Frontal Gyrus (orbital)	R	30	15	34	976	−2.1049
Precuneus	L	−16	−57	34	1,879	−2.1031
Angular Gyrus	R	39	−51	24	195	−2.0926
Posterior Cingulum	R	15	−39	32	619	−2.0762
Middle Occipital Gyrus	L	−33	−70	−2	241	−2.0687
Inferior Frontal Gyrus (orbital)	R	30	28	10	74	−2.0600
Inferior Parietal Lobule	L	−39	−48	32	118	−2.0547
Insula	L	−28	33	4	213	−2.0492
Inferior Frontal gyrus (medial, orbital)	R	12	33	−6	64	−2.0402
Middle Temporal Gyrus	L	−45	−14	−18	139	−2.0392
Middle Temporal Gyrus	R	39	−51	−6	30	−2.0360
Middle Occipital Gyrus	R	18	−104	14	10	−2.0309
Middle Occipital Gyrus	R	36	−72	0	29	−2.0286
Cerebellum (Crus I, extending bilaterally)	L	0	−100	−14	17	−2.0254
Inferior Frontal Gyrus (medial, orbital)	L	−12	32	−8	54	−2.0218
Inferior Frontal Gyrus (opercular)	L	−66	−3	27	21	−2.0177
Middle Temporal Gyrus	R	48	−44	−6	89	−2.0171
Middle Frontal Gyrus	R	32	34	18	12	−2.0074
Precentral Gyrus	R	22	−26	78	11	−2.0033
Anterior Cingulum	L	−3	30	4	18	−1.9996
Thalamus	R	15	−15	−6	29	−1.9937
Superior Frontal Gyrus (orbital)	L	−18	26	−12	13	−1.9920
Middle Temporal Gyrus	R	42	−4	−27	13	−1.9895

Discussion

The current study aimed to examine the interrelationship between perceived social support, gray matter volume and cognitive functions in a sample of older adults without dementia. We identified gray matter volume covariance networks associated with both overall social support and tangible social support (see Figures 1 and 2). Statistically reliable gray matter volume covariance networks associated with emotional/information support, affectionate support, or positive social interaction, however, were not observed. We also found that expression of the gray matter covariance pattern linked to tangible support was associated with episodic memory and executive function, but that the gray matter covariance pattern linked to overall social support was not associated with cognitive function. We discuss the implications of these findings next.

Gray Matter Volume Network Associated with Overall Social Support

Several brain regions contributed to the gray matter volume network associated with overall perceived social support in this study of older without dementia. Key brain areas of this pattern include frontal (including prefrontal and supplementary motor area), middle and inferior temporal (notably the temporal pole, hippocampus, and amygdala), parietal (superior parietal lobule and postcentral gyrus), occipital (cuneus), and cingulate regions (posterior and anterior). Prefrontal, temporal, parietal, and cingulate regions, as well as the hippocampus, amygdala, fusiform gyrus, cuneus, and postcentral gyrus regions have been previously associated with social support in young adults (Beadle et al., 2012; Bickart et al., 2011; Cassidy et al., 2012; Che et al., 2014) and social networks in older adults (Blumen & Verghese, 2018). Many of these regions have also been linked to different cognitive functions. Prefrontal and cingulate cortex regions, for example, have previously been associated with executive functioning, episodic memory as well as social cognition and emotional processing (Botvinick, Cohen, & Carter, 2004; Bush, Luu, & Posner, 2000; Shackman et al., 2011; Spaniol et al., 2009). Temporal and hippocampal regions are some of the earliest affected regions in Alzheimer’s disease (Thompson et al., 2003), and the amygdala has been independently linked to both stress and social support in older adults (Sherman et al., 2016). Taken together, these results suggest that the regions associated with an overall measure of perceived social support in older adults without dementia are those same regions that undergo early AD-related change, have been linked to social cognition in younger and older adults, and are commonly associated with different measures of cognition, particularly those related to learning and memory.

Because the overall social support measure represents the cumulative effect of many different types of support (informational advice/emotional support, tangible support, affectionate support, and positive social interaction), we were interested in determining if there was one particular aspect that was driving this relationship. There were no significant patterns found among informational/emotional support, affectionate support, or positive social interaction; however, we identified a gray matter volume pattern that was significantly associated with tangible support.

Gray Matter Volume Network Associated with Tangible Support

Several brain regions contributed to the gray matter volume network associated with perceived tangible social support in this study of older nondemented adults. Some key brain areas in this pattern include frontal (prefrontal and supplementary motor area), temporal (middle, hippocampus, and fusiform gyrus), occipital (cuneus), cingulate (anterior and posterior), and thalamic regions. Other notable areas in this network include superior temporal, parahippocampus, angular gyrus, middle occipital, precuneus, inferior parietal, insular, and cerebellar regions. The angular gyrus, middle occipital, precuneus, and cerebellar regions have previously been linked to social support in younger adults (Beadle et al., 2012; Cassidy et al., 2012). Like other temporal regions, the parahippocampal gyrus has been associated with age-related decline and preclinical Alzheimer’s disease (Pantel, Kratz, Essig, & Schröder, 2003; Thompson et al., 2003). The insula is important in executive functioning and memory awareness (Cosentino et al., 2015; Menon & Uddin, 2010), as well emotional awareness and understanding (Lamm & Singer, 2010). Thus, these findings suggest that regions associated with early Alzheimer’s disease are associated with tangible support in older adults without dementia. These same regions have also been associated with a number of cognitive and emotional processes, and are part of the mesocorticolimbic dopamine system, consisting of areas such as the hippocampus, medial prefrontal, and anterior cingulate regions. These conclusions are further supported by the observed correlations between the expression of these networks and cognitive function, particularly episodic memory performance.

Gray Matter Networks Associated with Tangible Support Correlate with Episodic Memory and Executive Function

Greater expression of the gray matter covariance pattern associated with tangible support was associated with better performance on measures of episodic memory and executive functioning in this study of older adults without dementia. Many key nodes of this pattern (hippocampus, prefrontal, cuneus) have been linked to these cognitive measures in the past. Pillemer and Holtzer (2016) also examined the relationship between the MOS-SSS and cognition at cross-section and found that emotional/information support and positive social interaction, but not tangible support, were related to general cognitive function as measured by the RBANS. In a more recent longitudinal study, however, overall support, tangible support, affectionate support, and positive social interactions was associated with global cognitive decline—particularly among older men (Pillemer et al., 2018) We have also linked episodic memory (but not executive function or processing speed) to gray matter volume covariance networks associated with high-contact social relationships and total number of network member in older adults, which were also largely composed of prefrontal and hippocampal regions (Blumen & Verghese, 2018).

Although we found significant correlations of tangible support to cognition in the present study, none of the cognitive measures reliably correlated with the gray matter volume network associated with overall perceived social support. This finding tells us that there is a stronger interrelationship between tangible social support, brain structure and cognition than overall social support, or that some of the gray matter covariance network associated with tangible support is influenced by different levels of cognition. Indeed, follow-up analyses (see Supplementary Materials) that adjusted for global cognition, indicated that global cognitive abilities did not exert much influence on the gray matter network associated with overall support, while the gray matter network associated with tangible support was more restricted to hippocampal, prefrontal and thalamic regions after adjusting for global cognition. These follow-up analyses are consistent with our initial correlational analyses, and suggest that while the brain network associated with overall support may not differ as a function of different levels of cognition, the brain network associated with tangible support, particularly the insular, precuneus and posterior cingulate components, differ as a function of global cognitive abilities. These cross-sectional findings, however, do not speak to the directionality of these interrelationships. Nevertheless, the findings of the current study support the notion that social interventions, particularly those relating to areas of tangible support, could potentially improve memory performance and executive functioning in both healthy aging and dementia; however, further research is needed to determine the most pertinent areas of social support as well as the clinical efficacy of such interventions.

Limitations and Future Directions

The first limitation of the current study is that our sample of relatively healthy older adults without dementia may not generalize to older adults with dementia or other severe cognitive impairments. Research in participants with dementia and other cognitive impairments is necessary to broaden these findings to the general population and discern the feasibility of interventions at various stages of cognitive impairment. Second, a ceiling effect in the MOS-SSS is possible in our sample. A more refined scale would help further delineate the relationship between gray matter volume and social support. Third, other studies have found distinct results in both utilization of social support and its relation to cognition between men and women, as well as between married and non-married participants (Pillemer et al., 2018; Seeman & Crimmins, 2001). Though we controlled for both factors, our sample suggests that 71% of our male participants were currently married, while only 16% of our female participants were currently married. A closer look at the underlying neural substrates of the relationship between cognition and social support in males and females could answer more questions about the varying effects of social support in older adults. Fourth, although all scans were completed within six months of their social support assessment, 43% of participant completed their assessment more than 90 days from their MRI date. Although we adjusted for this factor in our analyses, structural changes that occurred during this time could have influenced our results. Finally, the cross-sectional design of this study makes it difficult to ascertain the directionality of the relationship, namely whether decreased social support causes gray matter volume changes and cognitive decline, or whether changes in gray matter volume and decreased cognitive function lead to impaired social relations. This is particularly difficult when conceptualizing tangible support, as the need for greater tangible support encompasses a wide variety of functional abilities compared with other aspects of social support. Tangible support consists of such items as “someone to take you to the doctor if you needed it,” which could imply difficulty in understanding medical instructions or difficulty in driving oneself due to cognitive decline or physical problems, as well as items that imply more significant cognitive or physical impairment such as “someone to prepare meals if you were unable to do it yourself.” Other potential consequences or causes of reduced tangible support that was not directly addressed in the present study include stress, depression, and vascular brain pathologies such as white matter lesions (Cohen & Wills, 1985; Flatt et al., 2015; Vu & Aizenstein, 2013). Longitudinal studies are needed to answer questions related to the direction of causality, and larger samples with more diverse levels of social support and cognition, coupled with additional measures such as stress and white matter lesions, are needed to address questions related to such individual differences. Yet, this cross-sectional study provides initial motivation for addressing such questions in the future.

Funding

This work is supported by National Institute on Aging (Grant/Award Number: 1R01 AG044007-01A1; 1RO1AG036921; 1K01AG049829-01A1).

Conflict of Interest

None reported.