Associations Between Mid-Life Psychosocial Measures and Estimated Late Life Amyloid Burden: The Atherosclerosis Risk in Communities (ARIC)-PET Study

Abstract

Background:

Psychosocial factors are modifiable risk factors for Alzheimer’s disease (AD). One mechanism linking psychosocial factors to AD risk may be through biological measures of brain amyloid; however, this association has not been widely studied.

Objective:

To determine if mid-life measures of social support and social isolation in the Atherosclerosis Risk in Communities (ARIC) Study cohort are associated with late life brain amyloid burden, measured using florbetapir positron emission tomography (PET).

Methods:

Measures of social support and social isolation were assessed in ARIC participants (visit 2: 1990–1992). Brain amyloid was evaluated with florbetapir PET standardized uptake value ratios (SUVRs; visit 5: 2012–2014).

Results:

Among 316 participants without dementia, participants with intermediate (odds ratio (OR), 0.47; 95% CI, 0.25–0.88), or low social support (OR, 0.43; 95% CI, 0.22–0.83) in mid-life were less likely to have elevated amyloid SUVRs, relative to participants with high social support. Participants with moderate risk for social isolation in mid-life (OR, 0.32; 95% CI, 0.14–0.74) were less likely to have elevated amyloid burden than participants at low risk for social isolation. These associations were not significantly modified by sex or race.

Conclusions:

Lower social support and moderate risk of social isolation in mid-life were associated with lower odds of elevated amyloid SUVR in late life, compared to participants with greater mid-life psychosocial measures. Future longitudinal studies evaluating mid-life psychosocial factors, in relation to brain amyloid as well as other health outcomes, will strengthen our understanding of the role of these factors throughout the lifetime.

Keywords

Alzheimer’s disease amyloid-beta dementia mid-life positron emission tomography psychosocial factors

INTRODUCTION

Researchers are actively studying what compensatory mechanisms allow some aging individuals to maintain normal cognitive function despite pathological disease burden [1 –3]. To understand these mechanisms, studying modifiable risk factors in the decades prior to the development of neuropathology and dementia has emerged as a primary focus [4]. This work aims to elucidate whether the influence of modifiable risk factors acts upon the biological components underlying Alzheimer’s disease (AD) and related dementias [2, 5] or if this influence acts indirectly by strengthening cognitive performance to the point that dementia is less likely to result [3, 6]. This latter suggestion is consistent with definitions of cognitive reserve [3].

Previous research in the Atherosclerosis Risk in Communities (ARIC) Study cohort has explored many risk factors in relation to cognition, dementia incidence, and AD pathology [7 –10]. Studies assessing education have shown that higher levels of education are associated with improved baseline cognitive performance, but that associations between education with longitudinal cognitive performance and amyloid-β (Aβ) are minimal [7 –9]. Beyond education, a previous study that assessed psychosocial measures in relation to cognition found minimal effects when examining long-term outcomes [10]. Studies in other aged cohorts have shown mixed results when examining relationships between psychosocial measures and structural brain measures such as volume and connectivity [11 –17].

Other ARIC studies have shown the strongest association between vascular risk factors measured in mid-life with dementia incidence and amyloid burden [18, 19], suggesting that mid-life may be a key window of exposure. We sought to examine this same time frame with a focus on the association between measures of social support and social isolation with late life amyloid burden. From this examination, we hoped to clarify whether psychosocial factors directly impact amyloid pathology or if their influence primarily affects baseline cognition, which ultimately delays cognitive decline and the diagnosis of dementia. Given previous findings which support the latter [7 , 20], we hypothesized that we would see no associations between mid-life psychosocial measures and brain amyloid in late life, measured using florbetapir positron emission tomography (PET). Due to observed disparities in brain amyloid and dementia prevalence [21], secondary analyses were performed to examine how these associations varied based on sex and race.

MATERIALS AND METHODS

Study population

The ARIC study is an ongoing community-based prospective cohort study that enrolled 15,792 adults aged 45 to 64 years from 4 US communities at baseline (visit 1: 1987–1989) [22]. Self-reported questionnaires regarding psychosocial measures were collected at visit 2 (1990–1992). At visit 5 (2011–2013), a subset of ARIC participants without contraindications for imaging underwent brain magnetic resonance imaging (MRI) [23, 24]. Among the subset who received brain MRI, 346 participants without dementia from 3 ARIC centers were recruited for florbetapir PET imaging (to occur within 1 year of cognitive testing) at visit 5 as part of the ARIC-PET ancillary study [21]. The study was approved by Institutional Review Boards at each study center.

For the present study, participants with race other than White or Black (n = 2), dementia at time of imaging (delayed dementia diagnosis; n = 1), missing APOE ɛ4 (n = 5), and/or missing or miscoded psychosocial/vascular risk factors measured at visit 2 (n = 22) were excluded. The analytic sample included 316 participants without dementia.

Psychosocial measures

Perceived social support was evaluated at visit 2 using the Interpersonal Support Evaluation List-Short Form (ISEL-SF). This 16-item scale was constructed by ARIC investigators from the original 40-item full scale and assesses perceived social support with four subscales including (a) appraisal support, (b) tangible assets support, (c) belonging support, and (d) self-esteem support. The total score is an equally weighted sum, with scores ranging from 0–48: the higher the score, the greater perceived social support [25, 26]. Total ISEL-SF score was categorized into distribution-based tertiles of the full ARIC sample assessed at visit 2 (high≥42, intermediate 36–41, and low≤35).

Social isolation was evaluated by the Lubben Social Network Scale (LSNS), also administered at visit 2. This 10-item scale assesses the size of the participant’s active social network and the perceived social support received by family, friends, and neighbors. The total score is an equally weighted sum, with scores ranging from 0–50; greater scores indicate lower risk of social isolation. Scores were distributed in the following categories: ≤20 = isolated; 21–25 = high risk for isolation; 26–30 = moderate risk for isolation;≥31 = low risk for isolation [27, 28]. These categories were not evenly distributed in this analytic sample. Because only 20 participants were categorized as “high risk for social isolation” or “isolated,” we further examined the association between social isolation and elevated global SUVR after consolidating the high risk for social isolation and isolated participants into one group. Models using this three-level categorization of social isolation were used in primary analyses.

Brain MRI and PET

Brain MRI scans, obtained as research studies at a 3T MRI facility near each field center, were read and preprocessed centrally at the Mayo Clinic [24, 29]. PET images were co-registered using magnetization prepared rapid acquisition gradient-echo sequences (MPRAGE). The details of PET image processing and co-registration with MRI, carried out at the Johns Hopkins University reading center, were described previously [21]. Florbetapir PET scans were obtained within 1 year of MRI scans (ideally within 6 months). Images were reviewed and quantified for standardized uptake value ratios (SUVRs). A global cortical measure of florbetapir uptake was used as a weighted average (based on region of interest size) of the orbitofrontal, prefrontal, and superior frontal cortices; the lateral temporal, parietal, and occipital lobes; the precuneus, the anterior cingulate, and the posterior cingulate. An automated region for cerebellum gray was used as reference. Because continuous florbetapir SUVRs are highly skewed, SUVRs were dichotomized at the sample median (SUVR > 1.2) to indicate elevated brain amyloid burden. This value corresponds with prior ARIC-PET studies [21]. Florbetapir uptake in the anterior cingulate, a region involved with feelings surrounding social interactions [30, 31], was examined in exploratory analyses.

Covariates

Age at visit 2 (when psychosocial measures were evaluated) was calculated from the date of birth reported at visit 1. Sex and race were self-reported at visit 1. Education level was categorized into basic (less than high school); high school or equivalent (GED/vocational school); or at least some college (more than high school). The apolipoprotein E (APOE) genotype (TaqMan assay; Applied Biosystems, Foster City, CA) was categorized as ɛ4 carrier or non-carrier based upon presence of the ɛ4 allele. Occupational and marital status were self-reported at visit 2 and categorized as employed or unemployed (which included being retired or a homemaker) and married or not married (divorced, separated, widowed, or never married), respectively. The Maastricht Vital Exhaustion Questionnaire was conducted at visit 2 to measure symptoms of depression and fatigue [32 –34]. Questionnaire scores were dichotomized at ≥14 to indicate depressive symptomology [32]. Vascular risk factors, including hypertension (systolic blood pressure > 140 mm Hg, diastolic blood pressure > 90 mm Hg, or use of antihypertensive medications), diabetes (fasting glucose≥126 mg/dL, non-fasting glucose≥200 mg/dL, HbA1c≥6.5, self-report of physician-diagnosed diabetes, or use of oral diabetes medications or insulin), smoking/drinking (self-report, binarized into current versus non), obesity (BMI≥30 kg/m²), and elevated total cholesterol (≥200 mg/dL) were collected at visit 2.

Statistical analysis

Stata SE, version 17 for Macintosh (Stata Corp, Austin, TX) was used for all analyses. Participant characteristics were assessed in the ARIC-PET analytic sample (Table 1, Supplementary Table 1) and the full ARIC visit 2 sample (Supplementary Tables 2 and 3). Logistic regression models were used to evaluate the independent relationships between social support and social isolation with dichotomous (SUVR > 1.2) florbetapir uptake in the global cortex. Tertiles for ISEL-SF and three categories for LSNS were evaluated with high social support and low risk for social isolation as the referent categories.

Table 1

Characteristics of the ARIC-PET analytic sample at visit 2, stratified by levels of social support (n = 316)

	Full sample	High	Intermediate	Low	p
	(n = 316)	(n = 103)	(n = 109)	(n = 104)
Age, y, mean (SD)	55.1 (5.2)	55.0 (5.3)	54.8 (5.1)	55.6 (5.1)	0.57
Race, No. (%)					0.34
Black	128 (40.5)	47 (45.6)	39 (35.8)	42 (40.4)
White	188 (59.5)	56 (54.4)	70 (64.2)	62 (59.6)
Sex, No. (%)					0.04
Female	177 (56.0)	65 (63.1)	64 (58.7)	48 (46.2)
Male	139 (44.0)	38 (36.9)	45 (41.3)	56 (53.8)
Field Center, No. (%)					0.46
Forsyth	71 (22.5)	25 (24.3)	25 (22.9)	21 (20.2)
Jackson	122 (38.6)	45 (43.7)	37 (33.9)	40 (38.5)
Minneapolis^a	–	–	–	–
Washington	123 (38.9)	33 (32.0)	47 (43.1)	43 (41.3)
APOEɛ4 Status, No (%)					0.89
Non-Carrier	217 (68.7)	72 (69.9)	73 (67.0)	72 (69.2)
Carrier	99 (31.3)	31 (30.1)	36 (33.0)	32 (30.8)
Education Level, No. (%)					<0.001
Less than high school	45 (14.2)	9 (8.7)	17 (15.6)	19 (18.3)
High school or equivalent	139 (44.0)	34 (33.0)	52 (47.7)	53 (51.0)
More than high school	132 (41.8)	60 (58.3)	40 (36.7)	32 (30.8)
Elevated Florbetapir Global Cortex SUVR (>1.2) at Visit 5, No. (%)	155 (49.1)	63 (61.2)	48 (44.0)	44 (42.3)	0.01
Occupational Status, No. (%)					0.87
Employed	257 (81.3)	85 (82.5)	87 (79.8)	85 (81.7)
Unemployed^b	59 (18.7)	18 (17.5)	22 (20.2)	19 (18.3)
Marital Status, No. (%)					0.25
Married	259 (82.0)	86 (83.5)	93 (85.3)	80 (76.9)
Not Married^c	57 (18.0)	17 (16.5)	16 (14.7)	24 (23.1)
Depressive symptoms, ^d No. (%)	93 (29.4)	11 (10.7)	36 (33.0)	46 (44.2)	<0.001
Hypertension, No. (%)	91 (28.8_	30 (29.1)	30 (27.5)	31 (29.8)	0.93
Diabetes, No. (%)	34 (10.8)	7 (6.8)	11 (10.1)	16 (15.4)	0.13
Body Mass Index (≥30 kg/m²), No. (%)	91 (28.8)	30 (29.1)	28 (25.7)	33 (31.7)	0.62
Total Cholesterol (≥200 mg/dl), No. (%)	144 (45.6)	44 (42.7)	59 (54.1)	41 (39.4)	0.08
Current smoker, No. (%)	48 (15.2)	17 (16.5)	13 (11.9)	18 (17.3)	0.5
Current drinker, No. (%)	168 (53.2)	63 (61.2)	52 (47.7)	53 (51.0)	0.13

Baseline differences assessed with analysis of variance (ANOVA) for continuous variables and χ² tests for categorical values. p values reflect between group (high, intermediate, and low social support) differences. ^aNo PET scans conducted at Minneapolis ARIC Center at Visit 5. ^bUnemployed including participants who identified as retired and no longer working for pay, homemaker or unemployed at visit 2. ^cNot married including participants who identified as being divorced, separated, widowed, or never married at visit 2. ^dDepressive symptoms scored using Maastricht Vital Exhaustion Questionnaire. APOE, apolipoprotein E; SUVR, standardized uptake value ratio.

Three logistic regression models were fit for each psychosocial measure: model 1 adjusted for demographics and APOE ɛ4 (carrier versus non-carrier); model 2 additionally adjusted for occupational status, marital status and (having) depressive symptoms at visit 2; and model 3 further adjusted for vascular risk factors measured at visit 2. Because of the possibility that vascular risk factors are on the causal pathway between psychosocial measures and amyloid burden (and thus are not clearly potential confounders), our primary analyses and inferences were based on model 2. Secondary analyses explored the relationship between each psychosocial measure and elevated global SUVR with formal tests for interaction (Wald test) by sex and race. The sample was further stratified by both measures. Because ARIC-PET participants represent only a portion of all participants evaluated for psychosocial measures at visit 2, additional analyses assessed the relationship between both psychosocial measures with incidence of dementia and/or death prior to visit 5 (n = 13,175; Supplementary Table 4). Both criteria preclude enrollment in the ARIC-PET study, and thus could provide further insight as to how selection bias shaped the formation of the analytic sample.

Sensitivity analyses were conducted in participants using the four-category Lubben distribution criteria and in those with normal cognition (excluding those with mild cognitive impairment (MCI)), no APOE ɛ4 alleles, or who were employed. The purpose of excluding unemployed participants was to ensure findings were not biased by the larger number of female participants who were retired or homemakers at visit 2. Additional analyses were repeated using other established florbetapir SUVR cut-points (1.11 and 1.10) and the ISEL-SF and LSNS as continuous measures. Lastly, because a greater proportion of females were categorized as having high or intermediate social support when using ISEL-SF cut-points based on the full sample, we further conducted a sensitivity analysis with ISEL-SF categorization specific to the distribution of scores within sex-stratified groups. To conduct this analysis, we re-categorized levels of social support within females and males then consolidated across (e.g., females with high social support and males with high social support were grouped together to create the “high social support” group, females with intermediate social support and males with intermediate social support became the “intermediate social support” group, and so on).

RESULTS

Demographic characteristics of the analytic sample (n = 316) are summarized in Table 1. Since the SUVR threshold was based upon the sample median, not surprisingly, nearly half of the sample (n = 155; 49.1%) had elevated global SUVRs. The mean (SD) age of participants at visit 2 when psychosocial measures were evaluated was 55.1 (5.2) years and the mean (SD) age at visit 5 when PET imaging occurred was 77.8 (5.3) years, median follow-up: 22.3 years (interquartile cut-points: 21.7 and 23.0 years).

The distribution of participants in social support tertiles was nearly even whereas over 80% (n = 257) of participants scored as low risk for social isolation. A greater proportion of females were categorized as having high or intermediate social support and low risk of social isolation compared to males. The distribution of participants in social support tertiles and social isolation categories did not significantly differ by race (Table 1, Supplementary Table 1). Analyses of the full ARIC sample provided evidence that participants with low social support or who were categorized as moderate or high risk of isolation/isolated at visit 2 had increased likelihood of dementia and/or death prior to visit 5 (Supplementary Table 4).

Evaluation of psychosocial measures and amyloid burden

In models adjusted for demographics, APOE ɛ4, occupation/marital status, and depressive symptoms (model 2), participants with intermediate (odds ratio (OR): 0.47, 95% CI: 0.25–0.88) or low social support (OR: 0.43, 95% CI: 0.22–0.83) in mid-life were less likely to have elevated global SUVRs than participants with high social support in mid-life (Table 2). Likewise, model 2 examining levels of social isolation with amyloid burden showed participants with moderate risk for social isolation (OR: 0.32, 95% CI: 0.14–0.74) in mid-life were less likely to have elevated global SUVRs, compared to participants with low risk for social isolation in mid-life (Table 3; for model findings that follow the four-category Lubben criteria, see Supplementary Table 5). This trend did not extend to the high risk for isolation/isolated group. Because the sample size was still small in the high risk/isolated group, we further included a sensitivity analysis where the sample was stratified at LSNS≥31 (all three subgroups with scores < 30 were consolidated; Supplementary Table 6). This model also showed that participants with greater risk of isolation were less likely to have elevated SUVRs compared to participants with low risk of isolation (LSNS≥31). Across models evaluating both psychosocial measures, we observed that the odds of elevated global SUVR were significantly increased in association with being older, Black, and having APOE ɛ4, which is consistent with published results [21]. Analyses including vascular risk factors (model 3) showed similar associations between both psychosocial measures and elevated global SUVR (Tables 2 and 3). We did not observe associations between either psychosocial measure and elevated anterior cingulate SUVR (Supplementary Table 7).

Table 2

Adjusted odds ratios for association of mid-life social support with florbetapir global cortex SUVR > 1.2 (n = 316)

	Model 1^a OR	Model 2^b OR	Model 3^c OR
	(95% CI)	(95% CI)	(95% CI)
Social Support (ISEL-SF)
High (Score≥42)	Reference	Reference	Reference
Intermediate (Score 36–41)	0.46 (0.25–0.83)^*	0.47 (0.25–0.88)^*	0.45 (0.23–0.86)^*
Low (Score≤35)	0.41 (0.22–0.77)^**	0.43 (0.22–0.83)^*	0.40 (0.20–0.80)^**
Age (per 1 year)	1.06 (1.01–1.12)^*	1.07 (1.02–1.13)^**	1.07 (1.02–1.13)^**
Race (Black)	2.25 (1.37–3.71)^**	2.24 (1.34–3.74)^**	2.12 (1.22–3.69)^**
Sex (Female)	1.46 (0.88–2.41)	1.57 (0.91–2.72)	1.33 (0.74–2.38)
APOEɛ4 Carrier	3.31 (1.94–5.67)^**	3.30 (1.92–5.65)^**	3.52 (1.98–6.26)^**
Education Level
Less than high school	1.61 (0.75–3.47)	1.72 (0.79–3.73)	1.81 (0.78–4.20)
High school or equivalent	1.22 (0.71–2.10)	1.27 (0.73–2.19)	1.19 (0.67–2.12)
More than high school	Reference	Reference	Reference
Occupational Status (Unemployed)	–	0.69 (0.36–1.35)	0.71 (0.35–1.41)
Marital Status (Not Married)	–	0.99 (0.51–1.92)	0.98 (0.49–1.96)
Depressive Symptoms	–	0.86 (0.47–1.56)	0.81 (0.43–1.53)
Hypertension	–	–	1.09 (0.60–2.00)
Diabetes	–	–	0.88 (0.38–2.05)
Body Mass Index (≥30 kg/m²)	–	–	2.82 (1.52–5.22)^**
Total Cholesterol (≥200 mg/dL)	–	–	1.85 (1.10–3.12)^*
Current smoker	–	–	0.99 (0.47–2.08)
Current drinker	–	–	1.36 (0.77–2.42)

^*p≤0.05; ^**p≤0.01. ^aModel 1 adjusted for age, race, sex, APOE ɛ4, education. ^bModel 2 adjusted for model 1 covariates, occupational/marital status and (having) depressive symptoms as measured at visit 2. ^cModel 3 adjusted for model 2 covariates and vascular risk factors as measured at visit 2. APOE, apolipoprotein E; ISEL-SF, Interpersonal Support Evaluation List – Short Form; OR, odds ratio; SUVR, standardized uptake value ratio.

Table 3

Adjusted odds ratios for association of mid-life social isolation with florbetapir global cortex SUVR > 1.2 (n = 316)

	Model 1^a OR	Model 2^b OR	Model 3^c OR
	(95% CI)	(95% CI)	(95% CI)
Social Isolation (LSNS)
Low risk (Score≥31)	Reference	Reference	Reference
Moderate risk (Score 26–30)	0.33 (0.14–0.74)^**	0.32 (0.14–0.74)^**	0.26 (0.10–0.63)^**
High risk/Isolated (Score≤25)	0.97 (0.36–2.63)	0.97 (0.35–2.68)	0.80 (0.28–2.27)
Age (per 1 year)	1.07 (1.02–1.12)^**	1.08 (1.02–1.13)^**	1.08 (1.03–1.14)^**
Race (Black)	2.35 (1.43–3.87)^**	2.31 (1.39–3.86)^**	2.18 (1.25–3.79)^**
Sex (Female)	1.57 (0.95–2.59)	1.76 (1.02–3.02)^*	1.48 (0.83–2.65)
APOEɛ4 Carrier	3.06 (1.80–5.22)^**	3.04 (1.78–5.20)^**	3.26 (1.84–5.76)^**
Education Level
Less than high school	1.17 (0.55–2.48)	1.33 (0.61–2.88)	1.33 (0.57–3.10)
High school or equivalent	1.01 (0.59–1.71)	1.09 (0.64–1.89)	1.03 (0.58–1.84)
More than high school	Reference	Reference	Reference
Occupational Status (Unemployed)	–	0.64 (0.33–1.24)	0.64 (0.32–1.27)
Marital Status (Not Married)	–	1.05 (0.53–2.06)	1.08 (0.53–2.20)
Depressive Symptoms	0.72 (0.41–1.27)	0.68 (0.37–1.25)
Hypertension	–	–	1.20 (0.65–2.22)
Diabetes	–	–	0.74 (0.32–1.75)
Body Mass Index (≥30 kg/m²)	–	–	3.07 (1.64–5.73)^**
Total Cholesterol (≥200 mg/dl)	–	–	1.72 (1.03–2.89)^*
Current smoker	–	–	1.10 (0.52–2.33)
Current drinker	–	–	1.42 (0.80–2.52)

^*P≤0.05; ^**P≤0.01. High risk for social isolation (n = 13) and isolated (n = 7) participants combined into one group in this analysis due to limited sample size. ^aModel 1 adjusted for age, race, sex, APOE ɛ4, education. ^bModel 2 adjusted for model 1 covariates, occupational/marital status and (having) depressive symptoms as measured at visit 2. ^cModel 3 adjusted for model 2 covariates and vascular risk factors as measured at visit 2. APOE, apolipoprotein E; LSNS, Lubben Social Network Scale; OR, odds ratio; SUVR, standardized uptake value ratio.

Stratified associations of psychosocial measures and amyloid burden

Although the relationship between mid-life psychosocial measures and amyloid did not differ significantly by sex or race (as indicated by non-statistically significant interaction terms; p-values between 0.31 and 0.92), stratified results and potential qualitative differences are noted. In males, there were no significant associations between either psychosocial measure and elevated global SUVR, whereas significant associations between both psychosocial measures and elevated global SUVR in females resembled those shown in the overall sample (Fig. 1, Supplementary Tables 8 and 9). In White participants, associations between both psychosocial measures with elevated global SUVR resembled those in the full sample whereas in Black participants, significant associations were mitigated although this may be due to the small sample size (Fig. 2, Supplementary Tables 10 and 11).

Fig. 1

Adjusted odds ratios for florbetapir global cortex SUVR > 1.2, stratified by sex. ^*p≤0.05; ^**p≤0.01; Adjusted odds ratios (with 95% CI as error bars) are shown for model 2 logistic regression models assessing the relationship of each psychosocial measure and elevated SUVR. Referent categories of high social support and low risk of social isolation shown.

Fig. 2

Adjusted odds ratios for florbetapir global cortex SUVR > 1.2, stratified by race. ^*p≤0.05; ^**p≤0.01; Adjusted odds ratios (with 95% CI as error bars) are shown for model 2 logistic regression models assessing the relationship of each psychosocial measure and elevated SUVR. Referent categories of high social support and low risk of social isolation shown.

Sensitivity analyses

When differences were analyzed in cognitive status subgroups, the associations between mid-life psychosocial measures with elevated global SUVR shown in the primary analysis were observed in participants with normal cognition. In the subgroup with MCI, the associations between psychosocial measures and elevated global SUVR were closer to null and non-significant (n = 86; Supplementary Tables 12 and 13). Results examining participants with no APOE ɛ4 alleles, limited to those who were employed at visit 2, or using different SUVR cut-points were similar to those shown in the full sample (Supplementary Tables 14–19). Associations between continuous ISEL-SF and LSNS scores with elevated global SUVR were non-significant (Supplementary Table 20). The association seen in participants with low social support and elevated global SUVR remained significant when using sex-stratified distributions for the ISEL-SF (Supplementary Table 21).

DISCUSSION

Contrary to our hypothesis, associations between mid-life psychosocial measures and late life amyloid burden in this sample of ARIC-PET participants were shown. Specifically, we found that intermediate/low levels of social support and a moderate risk of social isolation in mid-life were each associated with lower odds of elevated global SUVR, compared to participants with greater social support and a lower risk of social isolation in mid-life, respectively. Results remained robust to adjustments for demographics, APOE ɛ4, occupation/marital status, depressive symptoms, and vascular risk factors (as measured at visit 2).

Previous work in the ARIC Study cohort evaluating education and cognitive abilities in mid-life did not show an association with brain amyloid [8]. Other ARIC studies have shown a strong association between years of education and mid-life psychosocial measures with baseline cognitive performance but not with longitudinal decline [7 –10]. Our preliminary interpretation of these findings was that features of cognitive reserve, such as education and psychosocial factors, may not directly impact the disease process, but rather they may augment and help sustain cognitive performance for a longer period. On the contrary, a recent Rotterdam study showed beneficial associations between social support with brain volume and white matter microstructural integrity, albeit the follow-up period was shorter (median follow-up: 4.1 years) [17].

Surprisingly, our results reflected that many participants with high social support, and/or at a low risk for social isolation, also had high brain amyloid burden in late life. At least one other study examining social health has shown a positive correlation with Aβ in dementia-free older adults [35]. Although counterintuitive, this relationship illustrates how older adults who are socially engaged with others in mid-life may be able to remain dementia-free through late life despite having Aβ pathology. Such findings suggest that strong social health may promote cognitive reserve. Our secondary analysis assessing likelihood of dementia and/or death prior to visit 5 supports this possibility. These models show that participants with low social support or greater risk of social isolation in mid-life are less likely to be eligible for the ARIC-PET study due to higher likelihood of dementia and/or death prior to visit 5. Similarly, data assessing outcomes of the full ARIC visit 2 sample shows that despite potentially developing elevated brain amyloid, participants with greater social support and/or lower risk of social isolation in mid-life were more likely to be part of the analytic sample due to lower likelihood of attrition. Altogether, an interpretation of these findings could be that having greater social support in mid-life is “protective” against the cognitive consequences of elevated amyloid burden. This explanation is speculative and warrants further study to understand how psychosocial factors may affect survivorship.

Formal tests for interaction by sex and race revealed the associations between mid-life psychosocial measures and elevated global SUVR were not significantly modified by either demographic factor. Stratified analyses conducted to see how psychosocial measures influence amyloid burden in males and females and in different racial groups showed qualitative differences. Results indicated that associations between both psychosocial measures with elevated global SUVR observed in the whole sample persisted in females, but not in males. Similarly, associations seen in the whole sample were closely echoed in White participants, but not as closely in Black participants. Ultimately, because our results did not show evidence of effect modification by sex or race, these findings should be evaluated in larger samples prior to making any conclusions about true interactions between sex and race with psychosocial factors.

Naturally, the present study has limitations. Psychosocial measures were surveyed at visit 2 and amyloid was not measured until visit 5. While this can be viewed as a strength because it means there is a smaller chance that our results are reflecting reverse causation, we acknowledge that psychosocial health often changes as individuals age from their mid 50 s to late 70 s, and that psychosocial health in later life may have a stronger influence on amyloid burden. Additionally, we recognize that current measurement approaches assess multiple components of psychosocial health [36]. The ISEL-SF measures perceived social support and emphasizes support that participants feel they are receiving with less emphasis on reciprocated social support [37]. The LSNS captures many aspects of social relationships including social support, relationship closeness with friends and family, and aspects of loneliness; all of which may vary situationally and temporally [37]. Thus, it remains possible we are examining the impact of other features of psychosocial health (i.e., neediness, loneliness, self-worth) on amyloid burden in addition to or more closely than social support and social isolation.

We further acknowledge that psychosocial health is difficult to quantify [26, 36]. Like previous studies, we observed that the relationship between social behaviors and health outcomes is not linear, thus our primary interpretations were based off analyses using these variables categorically [10 , 39]. When we used numeric cut-offs previously established for the LSNS in this limited sample size, we saw an uneven distribution of scores, particularly in the high risk for social isolation and isolated groups, meaning inadequate power to detect all associations may be reflected in our findings. The small sample size of these groups may reflect survival bias and may have influenced our findings. However, having more participants who are less likely to have the risk (and/or given outcome) is somewhat innate to all longitudinal studies simply because participants with lower risk are more likely to stay alive and return for follow-up visits (as evidenced by our full ARIC visit 2 sample findings).

Lastly, we accounted for occupation/marital status, depressive symptoms, and vascular risk factors but the relationship between these factors with psychosocial measures varies. To account for potential mediating, confounding or causal effects, we created models with multiple adjustments to ensure our results remained robust to the influence of these factors. Across models, the observed associations between psychosocial measures and elevated global SUVR remained stable indicating that these covariates did not drive the effects nor were they possible mediators.

In conclusion, lower social support and a moderate risk of social isolation in mid-life were associated with lower odds of elevated global SUVR as measured using florbetapir PET. This may reflect that the psychosocial factors examined affected survivorship, ultimately shaping the formation of this analytic sample. Future studies should continue to evaluate mid-life psychosocial factors in relation to brain amyloid as well as other health outcomes. Such knowledge could strengthen our understanding as to the potential protective role these factors may exert throughout the lifespan.

CREDIT AUTHOR STATEMENT

Rebecca F. Gottesman (Conceptualization; Formal analysis; Funding acquisition; Investigation; Methodology; Writing – review & editing); Renee C. Groechel (Conceptualization; Formal analysis; Investigation; Writing – original draft); Albert C. Liu (Writing – review & editing); Silvia Koton (Writing – review & editing); Anna M. Kucharska-Newton (Investigation); Pamela L. Lutsey (Investigation; Writing – review & editing); Thomas H. Mosley, Jr. (Funding acquisition; Investigation; Writing – review & editing); Priya Palta (Investigation; Writing – review & editing); A Richey Sharrett (Investigation; Writing – review & editing); Keenan A. Walker (Investigation; Writing – review & editing); Dean F. Wong (Data curation; Investigation; Writing – review & editing).

Footnotes

ACKNOWLEDGMENTS

The authors thank the staff and participants of the ARIC study for their important contributions.

FUNDING

The Atherosclerosis Risk in Communities Study is carried out as a collaborative study supported by National Heart, Lung, and Blood Institute contracts (75N92022D00001, 75N92022D00002, 75N92022D00003, 75N92022D00004, 75N92022D00005). The ARIC Neurocognitive Study is supported by U01HL096812, U01HL096814, U01HL096899, U01HL096902, and U01HL096917 from the NIH (NHLBI, NINDS, NIA and NIDCD). The ARIC-PET study is funded by the National Institute on Aging (grant R01AG040282). Avid Radiopharmaceuticals provided the florbetapir isotope for the study. RCG and RFG were supported by the NINDS Intramural Research Program.

CONFLICT OF INTEREST

The authors have no conflicts of interest to report.

DATA AVAILABILITY

The data used in this study are available upon reasonable request and is available in repositories such as BioLINCC and dbGaP.

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

References

Katzman

(1993) Education and the prevalence of dementia and Alzheimer’s disease. Neurology 43, 13–20.

Soldan

, Pettigrew

, Albert

(2018) Evaluating cognitive reserve through the prism of preclinical Alzheimer’s disease. Psychiatr Clin N Am 41, 65–77.

Stern

(2012) Cognitive reserve in ageing and Alzheimer’s disease. Lancet Neurol 11, 1006–1012.

Chan

, Shafto

, Kievit

, Matthews

, Spink

, Valenzuela

, Cam-CAN, Henson

(2018) Lifestyle activities in mid-life contribute to cognitive reserve in late-life, independent of education, occupation, and late-life activities. Neurobiol Aging 70, 180–183.

Serra

, Cercignani

, Petrosini

, Basile

, Perri

, Fadda

, Spanò

, Marra

, Giubilei

, Carlesimo

, Caltagirone

, Bozzali

(2011) Neuroanatomical correlates of cognitive reserve in Alzheimer disease. Rejuvenation Res 14, 143–151.

Vemuri

, Lesnick

, Przybelski

, Knopman

, Prebsoke

, Karntarci

, Raman

, Machulda

, Mielke

, Lowe

, Senjem

, Gunter

, Rocca

, Roberts

, Petersen

, Jack

Jr (2015) Vascular and amyloid pathologies are independent predictors of cognitive decline in normal elderly. Brain 138, 761–771.

Gottesman

, Rawlings

, Sharrett

, Albert

, Alonso

, Bandeen-Roche

, Coker

, Coresh

, Couper

, Griswold

, Heiss

, Knopman

, Patel

, Penman

, Power MC Selnes

, Schneider

ALC

, Wagenknecht

, Windham

, Wruck

, Mosley

(2014) Impact of differential attrition on the association of education with cognitive change over 20 years of follow-up: The ARIC Neurocognitive Study. Am J Epidemiol 179, 956–966.

Rawlings

, Sharrett

, Mosley

, Wong

, Knopman

, Gottesman

(2019) Cognitive reserve in midlife is not associated with amyloid-B deposition late-life. J Alzheimers Dis 68, 517–521.

Schneider

ALC

, Sharrett

, Patel

, Alonso

, Coresh

, Mosley

, Selnes

, Selvin

, Gottesman

(2012) Education and cognitive change over 15 years: The Atherosclerosis Risk in Communities Study. J Am Geriatr Soc 60, 1847–1853.

10.

Kats

, Patel

, Palta

, Meyer

, Gross

, Whitsel

, Knopman

, Alonso

, Mosley

(2016) Social support and cognition in a community-based cohort: The Atherosclerosis Risk in Communities (ARIC) study. Age Ageing 45, 475–480.

11.

Anatürk

, Suri

, Smith

, Ebmeier

, Sexton

(2021) Leisure activities and their relationship with MRI measures of brain structure, functional connectivity, and cognition in the UK Biobank cohort. Front Aging Neurosci 13, 734866.

12.

Bygrave

, Gerassimakis

, Mwendwa

, Erus

, Davatzikos

, Wright

(2022) The role of race in relations of social support to hippocampal volumes among older adults. Res Aging 44, 205–214.

13.

Donovan

, Okereke

, Vannini

, Amariglio

, Rentz

, Marshall

, Johnson

, Sperling

(2016) Association of higher cortical amyloid burden with loneliness in cognitively normal older adults. JAMA Psychiatry 73, 1230–1237.

14.

Pettigrew

, Soldan

, Zhu

, Wang

M-C

, Brown

, Miller

, Albert

, the BIOCARD Research Team (2017) Cognitive reserve and cortical thickness in preclinical Alzheimer’s disease. Brain Imaging Behav 11, 357–367.

15.

Salinas

, O’Donnell

, Kojis

, Pase

, DeCarli

, Rentz

, Berkman

, Aeiser

, Seshadri

(2021) Association of social support with brain volume and cognition. JAMA Netw Open 4, e2121122.

16.

Treyer

, Meyer

, Buchmann

, Crameri

GAG

, Studer

, Saake

, Gruber

, Unschuld

, Nitsch

, Hock

, Gietl

(2021) Physical activity is associated with lower cerebral beta-amyloid and cognitive function benefits from lifetime experience – a study in exceptional aging. PLoS One 16, e0247225.

17.

van der Velpen

, Melis

RJF

, Perry

, Vernooij-Dassen

MJF

, Ikram

, Vernooij

(2022) Social health is associated with structural brain changes in older adults: The Rotterdam Study. Biol Psychiatry Cogn Neurosci Neuroimaging 7, 659–668.

18.

Gottesman

, Albert

, Alonso

, Coker

, Coresh

, Davis

, Deal

, McKhann

, Mosley

, Sharrett

, Schneider

ALC

, Windham

, Wruck

, Knopman

(2017) Associations between midlife vascular risk factors and 25-year incident dementia in the Atherosclerosis Risk in Communities (ARIC) Cohort. JAMA Neurol 74, 1246–1254.

19.

Gottesman

, Schneider

ALC

, Zhou

, Coresh

, Green

, Gupta

, Knopman

, Mintz

, Rahmim

, Sharrett

, Wagenknecht

, Wong

, Mosley

(2017) Associations between midlife vascular risk factors and estimated brain amyloid deposition. JAMA Neurol 317, 1443–1450.

20.

Soldan

, Pettigrew

, Zhu

, Wang

M-C

, Gottesman

, DeCarli

, Albert

, BIOCARD Research Team (2020) Cognitive reserve and midlife vascular risk: Cognitive and clinical outcomes. Ann Clin Trans Neurol 7, 1307–1317.

21.

Gottesman

, Schneider

ALC

, Zhou

, Chen

, Green

(2016) The ARIC-PET amyloid imaging study. Neurology 87, 473–480.

22.

Wright

, Folsom

, Coresh

, Sharrett

, Couper

, Wagenknecht

, Mosley

Jr , Ballantyne

, Boerwinkle

, Rosamond

, Heiss

(2021) The ARIC (Atherosclerosis Risk In Communities) Study: JACC Focus Seminar 3/8. J Am Coll Cardiol 77, 2939–2959.

23.

Knopman

, Gottesman

, Sharrett

, Wruck

, Windham

, Coker

, Schneider

ALC

, Hengrui

, Alonso

, Coresh

, Albert

, Mosley

Jr (2016) Mild cognitive impairment and dementia prevalence: The atherosclerosis risk in communities neurocognitive study. Alzheimers Dement (Amst) 2, 1–11.

24.

Knopman

, Griswold

, Lirette

, Gottesman

, Kantari

, Sharrett

, Jack

Jr , Graff-Radford

, Schneider

ALC

, Windham

, Coker

, Albert

, Mosley

Jr (2015) Vascular imaging abnormalities and cognition: Mediation by cortical volume in nondemented individuals: Atherosclerosis risk in communities-neurocognitive study. Stroke 46, 433–440.

25.

Cohen

, Hoberman

(1983) Positive events and social supports as buffers of life change stress. J Appl Soc Psychol 13, 99–125.

26.

Payne

, Andrew

, Butler

, Wyatt

, Dubbert

, Mosley

(2012) Psychometric evaluation of the interpersonal support evaluation list–short form in the ARIC Study Cohort. SAGE Open 2, 2158244012461923.

27.

Lubben

(1998) Assessing social networks among elderly populations. Fam Community Health 11, 42–52.

28.

Lubben

, Blozik

, Gillmann

, Iliffe

, von Renteln Kruse

, Beck

, Stuck

(2006) Performance of an abbreviated version of the lubben social network scale among three European community-dwelling older adult populations. Gerontologist 46, 503–513.

29.

Jack

Jr , Bernstein

, Borowski

, Gunter

, Fox

, Thompson

, Schuff

, Krueger

, Killiany

, Decarli

, Dale

, Carmichael

, Tosun

, Weiner

, Alzheimer’s Disease Neuroimaging Initiative (2010) Update on the magnetic resonance imaging core of the Alzheimer’s disease neuroimaging initiative. Alzheimer Dement 6, 212–220.

30.

Apps

, Rushworth

, Chang

(2016) The anterior cingulate gyrus and social cognition: Tracking the motivation of others. Neuron 90, 692–707.

31.

Rotge

, Lemogne

, Hinfray

, Huguet

, Grynszpan

, Tartour

, George

, Fossati

(2015) A meta-analysis of the anterior cingulate contribution to social pain. Soc Cogn Affect Neurosci 10, 19–27.

32.

Honda

, Mok

, Mathew

, Van’t Hof

, Daumit

, Kucharska-Newton

, Selvin

, Mosley

, Coresh

, Matsushita

(2021) Psychosocial factors and subsequent risk of hospitalizations with peripheral artery disease: The Atherosclerosis Risk in Communities (ARIC) Study. Atherosclerosis 329, 36–43.

33.

Appels

, Hoppener

, Mulder

(1987) A questionnaire to assess premonitory symptoms of myocardial infarction. Int J Cardiol 17, 15–24.

34.

Williams

, Mosley

Jr , Kop

, Couper

, Welch

, Rosamond

(2010) Vital exhaustion as a risk factor for adverse cardiac events (from the Atherosclerosis Risk in Communities [ARIC] study). Am J Cardiol 105, 1661–1665.

35.

Biddle

, Uquillas

FDO

, Jacobs

, Zide

, Kirn

, Rentz

, Johnson

, Sperling

, Donovan

(2019) Social engagement and amyloid-β-related cognitive decline in cognitively normal older adults. Am J Geriatr Psychiatry 27, 1247–1256.

36.

Holt-Lunstad

(2018) Why social relationships are important for physical health: A systems approach to understanding and modifying risk and protection. Annu Rev Psychol 69, 437–458.

37.

Sakr-Ashour

(2021) Concept Clarification and Assessment of Social Isolation and Social Support in Older Adults. Administration for Community Living Office of Performance and Evaluation. New Editions Consulting, Inc.

38.

Freak-Poli

, Ryan

, Neumann

, Tonkin

, Reid

, Woods

, Nelson

, Stocks

, Berk

, McNeil

, Britt

, Owen

(2021) Social isolation, social support and loneliness as predictors of cardiovascular disease incidence and mortality. BMC Geriatr 21, 711.

39.

Nagayoshi

, Everson-Rose

, Iso

, Mosley

, Rose

, Lutsey

(2014) Social network, social support, and risk of incident stroke. Stroke 45, 2868–2873.