Acute Versus Chronic Inflammatory Markers and Cognition in Older Black Adults: Results from the Minority Aging Research Study

Abstract

Peripheral inflammation is elevated in older Black adults, an elevation which prior work has suggested may be due to chronic stress associated with systemic racism and related adverse cardiovascular health conditions. Inflammation is also involved in the pathogenic processes of dementia; however, limited (and mixed) results exist concerning inflammation and cognitive decline in Black adults. We characterized patterns of inflammation and their role in cognitive decline in 280 older Black adults (age=72.99±6.00 years; 69.6% female) from the Minority Aging Research Study (MARS) who were without dementia at baseline and followed between 2 and 15 years (mean=9 years). Participants completed a blood draw at baseline and annual cognitive evaluations. Serum was assayed for 9 peripheral inflammatory markers; 19 neuropsychological test scores were used to create indices of global cognition and five cognitive domains. Principal component analysis with varimax rotation characterized patterns of inflammation with factor loadings >0.6 per component contributing to two composite scores representing acute/upstream and chronic/downstream inflammation. These composites were used as separate predictors in linear mixed regression models to determine associations with level and change in cognition adjusting for relevant covariates. Higher baseline upstream/acute inflammation associated with lower baseline semantic memory (p=.040) and perceptual speed (p=.046); it was not related to cognitive decline. By contrast, higher baseline downstream/chronic inflammation associated with faster declines in global cognition (p=.010), episodic (p=.027) and working memory (p=.006); it was not related to baseline cognition. For older Black adults, chronic, but not acute, inflammation may be a risk factor for changes in cognition.

1. Introduction

Numerous studies report that exposure to chronic psychosocial stressors, such as persistent poverty, neighborhood and housing concerns, poor health, and other race-driven structural determinants are prevalent in minoritized populations (Brown, Mitchell, & Ailshire, 2020; Zuelsdorff et al., 2020) and may result in systemic inflammation (Noren Hooten, Pacheco, Smith, & Evans, 2022). For example, discrimination, an important psychosocial stressor reported by Black adults, has been associated with higher levels of peripheral inflammation in many, but not all, studies of middle-aged and older persons (Cuevas et al., 2020; Stepanikova, Bateman, & Oates, 2017). It has been posited that exposure to race-related stressors, like discrimination, can result in upregulation of the stress response in the body, which includes increases in peripheral inflammation as part of the immune response, along with changes to neuroendocrine and cardiovascular function (Clark, Anderson, Clark, & Williams, 1999; Anthony D. Ong, Deshpande, & Williams, 2017; Pascoe & Smart Richman, 2009). Over time, this chronic upregulation of the stress response can lead to greater allostatic load across physiological systems (A. D. Ong, Williams, Nwizu, & Gruenewald, 2017) and result in greater risk for chronic disease including cardiovascular disease and neurodegenerative diseases (Bagby, Martin, Chung, & Rajapakse, 2019; Bisht, Sharma, & Tremblay, 2018).

In fact, inflammation is thought to be a mediator and common pathway between psychosocial stressors and neurodegenerative diseases such as Alzheimer’s disease and related dementias (ADRD; Liu, Wang, & Jiang, 2017), and accumulating evidence suggests that greater levels of peripheral inflammation are implicated in the pathophysiological cascade of ADRD (Heneka et al., 2015; Holmes, 2013; Walker, Ficek, & Westbrook, 2019). To date, several markers of peripheral inflammation have been linked to cognitive dysfunction (Bettcher et al., 2012; Gross et al., 2019; Tampubolon, 2016; Tegeler et al., 2016; Wichmann et al., 2014). However, the vast majority of these studies have been conducted in predominately White cohorts, and there has been limited work investigating associations between inflammation and cognitive outcomes in older Black adults. This is despite the fact that Black adults have elevated risk for inflammation (Farmer et al., 2020) as well as greater risk for cognitive impairment and ADRD (Mayeda, Glymour, Quesenberry, & Whitmer, 2016). As such, improved understanding of the relationship between peripheral inflammation and the clinical correlates of ADRD, especially cognitive dysfunction in older Black adults, has become increasingly relevant.

Of the relatively few studies that have examined links between peripheral inflammation and cognition in older Black adults, most have been cross-sectional studies. Generally, these studies have reported an association between particular inflammatory markers and worse memory, executive function, and processing speed (Boots et al., 2020; El Husseini et al., 2020; Goldstein, Zhao, Steenland, & Levey, 2015; Windham et al., 2014); however, null findings between inflammatory markers and cognition have also been reported within these same studies. Longitudinal findings have been mixed as well, with studies reporting inflammation linked to faster rates of decline (Walker, Gottesman, et al., 2019; West, Kullo, Morris, & Mosley, 2020), improvements in cognition (Beydoun et al., 2018), or no specific associations with cognition over time (Arce Renteria et al., 2020; Yaffe et al., 2003). Overall, there has been a lack of consistency in findings across cross-sectional versus longitudinal approaches as well as across inflammatory markers chosen and how they have been used in analyses. Specifically, in the above-mentioned studies, age at time of study has varied from mid-forties (Beydoun et al., 2018) to early seventies (Yaffe et al., 2003) and range in length of follow up has varied between two and twenty years, which could influence both levels of inflammation over time as well as cognitive function outcomes, e.g., less likelihood of significant cognitive change in younger groups. Additionally, the above studies used between one and five inflammatory markers; all were investigated separately, and two studies created composite measures (Beydoun et al., 2018; Walker, Gottesman, et al., 2019). Resultingly, cognitive outcomes varied by which inflammatory marker was being investigated with inconsistencies in results across studies, e.g., CRP was associated with cognitive decline in some studies (Walker, Gottesman, et al., 2019; West et al., 2020), but not others (Arce Renteria et al., 2020; Yaffe et al., 2003). Given these variations in study design, a statistically based approach to analyzing inflammatory markers with a more comprehensive set of markers could provide clarity to these mixed findings, in addition to examining these questions in an annually-followed cohort of older Black adults – a group thought to be at high risk for cognitive dysfunction.

To our knowledge, we are aware of only two studies that have used a data-driven, statistical approach to incorporate inflammatory markers into analyses of cognitive function in older adults. One study employed a confirmatory factor analysis to estimate a latent factor of neutrophil activation from five inflammatory markers that was associated with decline in executive function over one year in a majority White cohort of adults with ADRD (Bawa et al., 2020). Another study used a principal component analysis (PCA) that combined four inflammatory markers into one component, which was associated with faster longitudinal decline in visual memory in older men within a biracial cohort of middle- to older-aged adults (Beydoun et al., 2018). These types of approaches to systematically characterizing several inflammatory markers may aid in clarifying mixed results seen previously in the literature. To date, however, this approach has never been taken in a solely older Black adult population. In prior work using data from this study, we investigated the link between only one inflammatory marker – CRP – and discrimination (Lewis, Aiello, Leurgans, Kelly, & Barnes, 2010) but did not examine cognition in that study. Given increased risk for elevated inflammation due to race-related stressors and resulting risk for cognitive decline, a comprehensive grouping of a wider range of inflammatory markers investigated over time as related to cognition in older Black adults is warranted.

The aim of our study was to statistically categorize patterns of inflammation using principal component analysis (PCA) for nine peripheral inflammatory markers and employ resulting categorizations to investigate cross-sectional and longitudinal associations with global cognition and five cognitive domains in a cohort of older Black adults. For our study, we hypothesized that statistical categorizations of peripheral inflammation using PCA would be differentially associated with both baseline level and longitudinal change in cognition. By utilizing a statistical approach to inflammatory marker analysis with a comprehensive number of inflammatory markers that have various functions in the inflammatory cascade, this study will build upon prior research that has used single inflammatory markers or a priori composite measures to characterize inflammation in a data-driven and more comprehensive format. Additionally, by examining both cross-sectional and longitudinal associations of inflammation and cognition annually over a nine-year period on average in a well-characterized cohort of older Black adults, this study will advance understanding of patterns of inflammation specific to the older Black population, a population that is understudied yet at greater risk of inflammation and cognitive impairment due to discrimination, stress, and health inequities.

2. Materials & Methods

2.1. Participants

Study participants were enrolled in the Minority Aging Research Study (MARS), an ongoing, longitudinal epidemiological cohort study of community-dwelling older Black adults that began in 2004 (Barnes, Shah, Aggarwal, Bennett, & Schneider, 2012). The study was approved by the Rush University Medical Center Institutional Review Board and conducted in accordance with the Declaration of Helsinki where written informed consent was obtained from all participants. As has been previously described (Barnes, Shah, et al., 2012), participants were recruited from churches, senior buildings, and community-based organizations in the Chicagoland area, as well as through the Clinical Core of the Rush Alzheimer’s Disease Center. Eligible participants included individuals who self-identified as Black, were 65 years of age or older, and were without known dementia at study entry.

All participants agree to complete annual in-home clinical evaluations. The evaluations include a physical examination, medical history, comprehensive neuropsychological evaluation, and blood specimen collection. Participants are also asked to answer questions about various risk factors and present prescription and non-prescription medications for visual inspection and recording using the Medi-Span Drug Database^®.

There were 350 participants originally enrolled in the MARS study, and 307 of these individuals agreed to annual blood specimen collection and had stored blood specimens from the baseline visit. These 307 individuals were included in an ancillary study for inflammatory marker assay data collection from baseline blood specimens. Given the aims of this study, we additionally excluded 27 participants who had not completed at least two in-home clinical evaluations, were missing relevant covariates, or had serum concentrations below the limit of detection for the assays. Therefore, our analytic sample consisted of 280 individuals, with a mean follow up time of nine years (range 2–15 years).

2.2. Inflammatory Markers

As detailed previously (Lewis et al., 2010), phlebotomists skilled in venipuncture collected blood samples from study participants in 2-ml EDTA tubes. The samples were subsequently spun for ten minutes, aliquoted into 1.8 cc Nunc vials, placed on dry ice, and transported to freezers in a storage facility at the Rush Alzheimer’s Disease Center where they were stored at −80°C. As part of the ancillary study, blood serum was quantified for nine inflammatory markers by Endogen Searchlight technologies (Billerica, MA) using highly sensitive multiplexed sandwich enzyme-linked immunosorbent assay (ELISA) arrays. The nine inflammatory markers that were chosen have been commonly reported in the cognitive aging literature and represent a range of pro-inflammatory, anti-inflammatory, and vascular inflammation processes. These nine inflammatory markers included the following: interleukin-6 (IL-6), interleukin-10 (IL-10), interleukin-1 beta (IL-1β), tumor necrosis factor alpha (TNFα), c-reactive protein (CRP), interleukin 1 receptor antagonist (IL-1ra), vascular cell adhesion molecule (VCAM), interleukin-6 receptor (IL-6r), and matrix metallopeptidase-9 (MMP9). Given our statistical categorization approach, we included all nine inflammatory markers for the current study. The ELISA arrays were tested for spotting consistency and specificity to rule out the presence of cross-reactivity or nonspecific binding resulting from multiple antibody combinations, and the mean intra-assay coefficient of variability was <15% across all markers and deemed acceptable as outlined elsewhere (Lewis et al., 2010). Markers below the lower limit of quantification were excluded from analysis as noted above. All inflammatory markers were log-transformed for data normalization purposes.

2.3. Cognition

All participants underwent annual comprehensive neuropsychological evaluation to characterize cognitive functioning. As described previously (Wilson et al., 2002), raw scores from neuropsychological tests were converted to z-scores using the baseline mean and standard deviation and averaged to create a global cognitive functioning composite score as well as five cognitive domains. The five cognitive domains, and their constituent tests, are as follows: Episodic Memory – Word List Memory, Word List Recall, and Word List Recognition established by the Consortium to Establish a Registry for Alzheimer’s Disease (CERAD; Morris et al., 1989), immediate and delayed recall from Story A of the Logical Memory subtest of the Wechsler Memory Scale-Revised (WMS-R; Wechsler, 1987), and immediate and delayed recall of the East Boston Story (Albert et al., 1991); Semantic Memory – Semantic Verbal Fluency (animals and fruits/vegetables) from CERAD (Morris et al., 1989), the 15-item version of the Boston Naming Test (Kaplan, Goodglass, & Weintraub, 1983), and the 15-item reading test from the Wide Range Achievement Test – Third Edition (Wilkinson, 1993); Working Memory – Digit Span Forward and Backward of the WMS-R (Wechsler, 1987) and Digit Ordering (Cooper & Sagar, 1993); Visuospatial Ability – 15-item version of Judgment of Line Orientation (Benton, Sivan, & Hamsher, 1994) and the 16-item Standard Progressive Matrices (Raven, Court, & Raven, 1992); Perceptual Speed – Oral Symbol Digit Modalities Test (Smith, 1982), Number Comparison (Ekstrom, French, & Harman, 1976), and two indices from the Stroop Neuropsychological Screening Test including color naming minus errors and color reading minus errors (30 seconds each; Trenerry, Crosson, & DeBoe, 1989).

2.4. Covariates

In addition to demographic variables including age, sex, and years of education, we also included cardiovascular risk factor burden, body mass index (BMI), statin medication use, and analgesic medication use as covariates in all analyses given their relevance to both inflammatory marker levels and cognitive function (Bourassa & Sbarra, 2017; Golia et al., 2014). Specifically, cardiovascular risk factor burden was quantified by medication inspection and responses to self-report questions for hypertension, diabetes mellitus, and smoking (never versus current/former) resulting in a score of 0 to 3 for each participant. BMI was calculated as weight in kilograms divided by height in meters squared. Statin and analgesic medication use were determined by visual inspection of medications taken with two weeks of the study visit and coded using the Medi-Span Drug Database^®. Analgesic medications included non-steroidal anti-inflammatories along with acetaminophen, aspirin, glucocorticoids, antimalarials, and opioids.

2.5. Statistical Analyses

Analyses were programmed using SAS Software Version 9.4 with p ≤ 0.05 for significance. Descriptive statistics characterized demographic, health, and medication variables. PCA with varimax rotation was used to statistically characterize patterns of inflammation utilizing all nine inflammatory markers. Then, individual inflammatory markers with factor loadings >0.6 per rotated component were z-scored and averaged (as relevant) for all participants to compute composite scores.

The PCA-derived composite scores of inflammation were used as continuous predictors in separate linear mixed regression models that examined their associations with baseline level of and longitudinal change in global cognition and the five cognitive domains (six individual and continuous outcomes). Models were fit using restricted maximum likelihood estimation with random intercept and slope for time (measured in years from baseline visit). All models included the terms for time, age (mean centered), sex, years of education (mean centered), cardiovascular risk factor burden, BMI, statin and analgesic medication use, and interaction terms of each of these variables with time.

3. Results

3.1. Participant Characteristics

For the 280 older Black adults included in this study, the average baseline age was 72.99 years and 69.7% of the sample was female. Participants were well-educated, with an average of 14.98 years of education. Out of a possible three cardiovascular risk factors, participants had a mean of 1.43 while approximately 35.4% of participants were taking a statin and 72.1% of participants were taking analgesic medications. Average BMI fell within the overweight range (see Table 1 for these and other participant characteristics).

Table 1.

Participant Characteristics at Baseline, N=280

Characteristic	Mean (Standard Deviation)*
Age, years	72.99 (6.00)
Female, %	69.7
Education, years	14.98 (3.59)
Cardiovascular risk factors, count (0–3)	1.43 (0.86)
Hypertension, %	71.1
Type II Diabetes Mellitus, %	21.8
Current or Former Smoker, %	50.0
Statin medication use, %	35.4
Analgesic medication use, %	72.1
Body mass index, kg/m2, n=276	29.42 (5.82)
Length of follow-up, years	9.20 (4.11)

^*Values are Mean (Standard Deviation) unless otherwise noted.

Note: The cardiovascular risk factors variable is the number of three risk factors present, including hypertension (via self-report and medication review), type 2 diabetes mellitus (via self-report and medication review), and current/former smoking (self-report).

3.2. Principal Component Analysis

The PCA with varimax rotation resulted in a two-factor solution that accounted for 60.3% of the variance (see Table 2). Composite 1 comprised 38.2% of the variance and consisted of the following peripheral inflammatory markers with factor loadings >0.6: IL-6, IL-10, IL-1β, TNFα, and IL-1ra. Composite 2 comprised 22.1% of the variance and consisted of the following peripheral inflammatory markers with factor loadings >0.6: MMP9, VCAM, and IL-6r. CRP did not load on either factor at our threshold for composite inclusion, i.e., a factor loading >0.6, nor did it load onto an isolated factor of its own; thus, CRP was not included in either composite score or further analysis.

Table 2.

Principal Component Analysis with Varimax Rotation for Nine Peripheral Inflammatory Markers

Inflammatory Marker	Factor 1	Factor 2
Interleukin-6 (IL-6)	.754	−.263
Interleukin-10 (IL-10)	.792	−.182
Interleukin-1β (IL-1β)	.826	−.294
Tumor Necrosis Factor-α (TNF-α)	.761	−.056
Interleukin-1 Receptor Agonist (IL-1ra)	.814	−.144
Interleukin-6 Receptor (IL-6r)	.344	.757
Matrix Metallopeptidase 9 (MMP9)	.330	.689
Vascular Cell Adhesion Molecule (VCAM)	.287	.803
C Reactive Protein (CRP)	.097	.288

Note: Inflammatory markers with factor loadings greater than 0.6 per rotated component (indicated by bolding) contributed to z-scored and averaged composite inflammatory factors. Composite 1 was characterized as the “Acute/Upstream” factor, and Composite 2 was characterized as the “Chronic/Downstream” factor.

3.2.1. Principal Component Analysis Composite Labeling

Given the solution from the PCA revealed two factors with differential loadings of inflammatory markers, we labeled the two resulting composites based on the commonalities of inflammatory markers included within each composite. Specifically, we labeled Composite 1, which consisted of the inflammatory markers IL-6, IL-10, IL-1β, TNFα, and IL-1ra, as representative of acute/upstream inflammation. We labeled Composite 2, which consisted of the inflammatory markers MMP9, VCAM, and IL-6r, as representative of chronic/downstream inflammation. We labeled the acute/upstream inflammation composite as we did given the tendency for the inflammatory markers included in the acute/upstream composite to show increased expression early in the immune response. Specifically, IL-6, IL-1β, and TNFα are primary proinflammatory cytokines that activate acute pathways including the nuclear factor kappa-B (NF-κB) and Janus kinase-signal transducer and activator of transcription (JAK-STAT) pathways (Chen et al., 2018). While IL-10 and IL-1ra are anti-inflammatory in nature, they are also active in the acute inflammatory response (Couper, Blount, & Riley, 2008; Palomo, Dietrich, Martin, Palmer, & Gabay, 2015). In contrast, the inflammatory markers included in the chronic/downstream composite tend to be more distal mediators of the immune response. For example, VCAM is activated in a downstream pathway of TNFα, where it is expressed in endothelial cells (Kong, Kim, Kim, Jang, & Lee, 2018). MMP9 is upregulated downstream of IL-1β and TNFα where it degrades extracellular matrix proteins, and IL-6r is part of the IL-6 anti-inflammatory classic signaling response that is downstream of the initial immune response (Wolf, Rose-John, & Garbers, 2014). Thus, while each composite included both pro- and anti-inflammatory markers, our statistical characterization resulted in factor loadings that we felt mimicked the immune response and labeled each composite accordingly.

3.3. Acute/Upstream Inflammation and Cognition

Linear mixed regression models adjusted for age, sex, education, cardiovascular risk factor burden, BMI, statin medication use, and analgesic medication use resulted in significant associations between the acute/upstream inflammation composite and baseline levels of select cognitive domains (see Table 3). Specifically, higher baseline levels of acute/upstream inflammation were associated with lower baseline semantic memory (p=.040) and perceptual speed (p=.046). Our acute/upstream inflammation composite was not associated with longitudinal change for global cognition or any cognitive domain.

Table 3.

Association Between Acute/Upstream Inflammation (Composite 1) and Level and Change in Cognitive Function

	Global Cognition	Episodic Memory	Semantic Memory	Working Memory	Perceptual Speed	Visuospatial Ability
Model Term	β ± SE	β ± SE	β ± SE	β ± SE	β ± SE	β ± SE
Time	−0.144 ± 0.017***	−0.124 ± 0.024***	−0.171 ± 0.024***	−0.119 ± 0.022***	−0.122 ± 0.018***	−0.053 ± 0.020**
Acute/Upstream	−0.068 ± 0.037	−.072 ± 0.047	−0.092 ± 0.045*	−0.010 ± 0.052	−0.096 ± 0.048*	−0.044 ± 0.049
Acute/Upstream x Time	−0.003 ± 0.007	−0.003 ± 0.009	−0.0004 ± 0.009	−0.004 ± 0.006	−0.002 ± 0.005	−0.003 ± 0.005

^*Denotes significant result at p < 0.05

^**Denotes significant result at p < 0.025

^***Denotes significant result at p < 0.0001.

Note: Results from six separate linear mixed models are shown, where each model additionally adjusted for age, sex, education, vascular risk factors, body mass index, statin medications, and analgesic medications.

3.4. Chronic/Downstream Inflammation and Cognition

Linear mixed regression models adjusted for age, sex, education, cardiovascular risk factor burden, BMI, statin medication use, and analgesic medication use did not show any significant associations between the chronic/downstream inflammation composite and baseline cognitive function (see Table 4). Chronic/downstream inflammation was associated with longitudinal change in global cognition (p=.010), episodic memory (p=.027), and working memory (p=.006; see Table 4). Specifically, higher baseline levels of chronic/downstream inflammation were related to faster rates of decline across global cognition and both cognitive domains (see Figure 1).

Table 4.

Association Between Chronic/Downstream Inflammation (Composite 2) and Level and Change in Cognitive Function

	Global Cognition	Episodic Memory	Semantic Memory	Working Memory	Perceptual Speed	Visuospatial Ability
Model Term	β ± SE	β ± SE	β ± SE	β ± SE	β ± SE	β ± SE
Time	−0.144 ± 0.017***	−0.126 ± 0.024***	−0.172 ± 0.024***	−0.118 ± 0.021***	−0.122 ± 0.018***	−0.052 ± 0.019**
Chronic/Downstream	−0.029 ± 0.036	−0.008 ± 0.046	−0.023 ± 0.044	−0.066 ± 0.050	−0.040 ± 0.047	−0.042 ± 0.048
Chronic/Downstream x Time	−0.018 ± 0.007**	−0.020 ± 0.009*	−0.016 ± 0.009	−0.017 ± 0.006**	−0.008 ± 0.005	−0.004 ± 0.005

^*Denotes significant result at p < 0.05

^**Denotes significant result at p < 0.025

^***Denotes significant result at p < 0.0001

Note: Results from six separate linear mixed models are shown, where each model additionally adjusted for age, sex, education, vascular risk factors, body mass index, statin medications, and analgesic medications.

Figure 1.

Figures depict baseline level and rate of change in cognition over time for participants with low (25^th percentile in blue) versus high (75^th percentile in green) levels of the chronic/downstream inflammation factor while adjusting for age, sex, education, cardiovascular risk factors, body mass index, statin medications, and analgesic medications. Specifically, participants with higher levels of the chronic/downstream inflammation factor had steeper rate of decline for global cognition (Panel A), episodic memory (Panel B), and working memory (Panel C) compared to participants with lower levels of the chronic/downstream inflammation factor. No significant differences in baseline level of cognition were observed.

4. Discussion

In this study of over 250 older Black adults who were cognitively normal at baseline, we statistically summarized nine peripheral inflammatory markers, which resulted in a two-component solution of acute/upstream and chronic/downstream inflammatory marker loadings. Greater levels of acute/upstream inflammation were associated with lower baseline levels of semantic memory and perceptual speed but not cognitive decline. By contrast, greater levels of chronic/downstream inflammation were associated with faster declines in global cognition and episodic and working memory, but not with baseline cognition. Together, these findings suggest a dissociation between associations of acute/upstream versus chronic/downstream inflammatory markers with cognitive aging among older Black adults.

Previous studies of inflammation and cognitive aging in older Black adults have yielded mixed results, even within the same study. For example, significant cross-sectional associations have been noted between higher levels of inflammatory markers including CRP, IL-6, IL-8, soluble TNF receptors 1 and 2 (sTNFR1, sTNFR2), TNFα, VCAM, erythrocyte sedimentation rate, and albumin, and poorer performance in memory, executive function, and processing speed (Boots et al., 2020; El Husseini et al., 2020; Goldstein et al., 2015; Windham et al., 2014). Within these same studies, however, null associations were reported between cognition and CRP, IL-6, IL-8, IL-10, TNFα, IL-1β, IL-1ra, interferon gamma, or thrombin-anti-thrombin. We are aware of three longitudinal studies that showed significant associations between inflammatory markers and cognitive decline in Black adults. In one longitudinal study, greater sTNFr1 was associated with memory decline and greater CRP and IL-6 were associated with faster declines in global cognition and processing speed in Black men only (West et al., 2020), In another study, higher levels of CRP were associated with faster cognitive decline in Black adults who were late-middle-aged at baseline, but an inflammation composite score was not associated with cognitive change in this sample (Walker, Gottesman, et al., 2019). The other study found that CRP was associated with improved attention over time across both Black men and women (Beydoun et al., 2018). Still, other longitudinal studies have not noted any associations between inflammatory markers and cognitive decline (Arce Renteria et al., 2020; Yaffe et al., 2003). Reasons for the discrepant findings are not clear but may be due, at least in part, to differing methodological details (i.e., participant ages, cross-sectional versus longitudinal, length of follow-up) and variability in inflammatory marker methodology. For example, some studies employed single inflammatory markers (Arce Renteria et al., 2020; Goldstein et al., 2015; West et al., 2020; Windham et al., 2014; Yaffe et al., 2003) or a priori inflammatory composite measures (Walker, Gottesman, et al., 2019) in their investigations. By contrast, our study opted for a statistical approach for inclusion of inflammatory markers by utilizing a PCA – an approach that has, to the best of our knowledge, been utilized only once with four inflammatory markers (Beydoun et al., 2018). In the present study, the PCA of nine inflammatory markers revealed two factors, which we labeled as acute/upstream inflammation (comprised of IL-6, IL-10, IL-1β, TNFα, and IL-1ra) and chronic/downstream inflammation (comprised of MMP9, VCAM, and IL-6r); CRP did not load substantially onto either factor or onto its own. This statistical approach to grouping inflammatory markers, rather than using inflammatory markers individually or as part of composites made a priori, resulted in a statistical characterization of two factors that mimicked the immune response quite well, suggesting PCA may be a useful statistical approach for future studies.

Our results showed a dissociation between greater acute/upstream inflammation and lower baseline cognition compared with greater chronic/downstream inflammation and faster longitudinal cognitive decline, which may provide clarity to the disparate findings seen previously in the literature. Associations of greater acute/upstream inflammation with semantic memory and perceptual speed align well with prior cross-sectional findings linking IL-6 and TNFα receptors with processing speed, language, and executive function (Boots et al., 2020; Windham et al., 2014) and further extend these studies to suggest a grouping of inflammatory markers that may be driving the results. We did not find any longitudinal studies to date that have investigated the inflammatory markers included in the chronic/downstream component, which may explain why some prior longitudinal studies of other inflammatory markers did not find associations with cognitive decline in older Black adults (Arce Renteria et al., 2020). Cross-sectional studies of VCAM in patients with ADRD, however, have reported that higher VCAM is associated with lower global cognition and episodic memory (Huang et al., 2015; Tchalla et al., 2017) and our longitudinal findings extend this cross-sectional result. By contrast, one longitudinal study noted that greater levels of IL-6, an inflammatory marker in our acute/upstream component, were associated with steeper decline in global cognition and processing speed in older Black men, but not older Black women (West et al., 2020). Our sample of approximately 70% women could explain these differential findings. Finally, while CRP did not load preferentially onto either of our PCA-derived components, it is worth noting that higher CRP levels have been cross-sectionally associated with lower attention and processing speed performance (Beydoun et al., 2018; Windham et al., 2014) and with steeper longitudinal declines in global cognition and processing speed (West et al., 2020).

Prior investigations point to mediating factors that link acute/upstream and chronic/downstream inflammation to baseline and longitudinal cognitive dysfunction. Generally, peripheral inflammatory markers are known to influence neuroinflammation by crossing the blood-brain barrier (BBB), inducing BBB endothelial cell activation, and signaling to microglia across the BBB and via circumventricular organs (Holmes, 2013; Lampron, Elali, & Rivest, 2013; Perry, 2004). Acute/upstream inflammatory markers can instigate ‘sickness behavior’ in the brain during an acute inflammatory increase, resulting in slowed psychomotor processing speed (Holmes, 2013). By contrast, chronic/downstream inflammatory markers employed in our study, including VCAM and MMP9, are known to play a role in endothelial dysfunction which over time can contribute to hypertension, atherosclerosis, and cardiovascular disease (Tchalla et al., 2017; Yabluchanskiy, Ma, Iyer, Hall, & Lindsey, 2013); factors that are themselves inflammatory and are known to impact brain health and cognitive function (Boots et al., 2019). In fact, in addition to cognitive function, VCAM has been linked to greater white matter hyperintensities and poorer white matter microstructure (Huang et al., 2015) and has also been associated with ADRD biomarkers of amyloid beta, total tau, and phosphorylated tau in the cerebrospinal fluid (Janelidze et al., 2018). MMP9 is also associated with amyloid-beta and t-tau in ADRD patients (Gu et al., 2020) and is implicated in BBB breakdown in older adults leading to cognitive dysfunction (Montagne et al., 2020). These findings lend credence to these chronic inflammatory markers’ direct impacts on ADRD-related pathogenesis. Additional research is needed to further elucidate the roles of acute/upstream and chronic/downstream inflammation as related to the pathophysiological cascade of ADRD and incident dementia in older Black adults.

Although underlying biological mechanisms relating inflammatory markers to cognition are important to consider; there are also potent underlying societal and psychosocial factors that may explain elevated levels of inflammation that subsequently contribute to greater risk for cognitive decline in older Black adults. Although not examined in the current study, experiences of discrimination and mistreatment secondary to structural racism have been associated with higher levels of inflammation in Black adults (Cuevas et al., 2020; Friedman, Williams, Singer, & Ryff, 2009; Lewis et al., 2010; Pascoe & Smart Richman, 2009). Experiences of discrimination and mistreatment are thought to upregulate the stress response (Simons et al., 2018), triggering increases in peripheral inflammation along with other cardiovascular and neuroendocrine responses (Anthony D. Ong et al., 2017). Over time, these physiological responses can lead to chronic disease including cardiovascular diseases and ADRD – both of which disproportionately affect older Black adults (Bagby et al., 2019). Indeed, prior work has shown that inflammation moderates the association between discrimination and poor cognitive outcomes in older Black adults (Zahodne, Kraal, Sharifian, Zaheed, & Sol, 2018). Additional work investigating the systemic and psychosocial factors that lead to biological dysregulation, currently ongoing in our group, will be critical for targeted intervention to reduce these health disparities in aging Black adults.

This study has several strengths. First, we characterized the effect of peripheral inflammation on cognitive function over an average follow-up period of nine years in a sample of 280 older Black adults. Given the paucity of longitudinal data investigating these relationships in older Black adults – individuals not only underrepresented in ADRD research (Barnes & Bennett, 2014), but with greater levels of inflammation and risk for ADRD (Farmer et al., 2020; Mayeda et al., 2016) – our findings are an important addition to the literature. Additionally, we used a data-driven approach to statistically characterize these patterns, lending confidence to our results. We also characterized cognitive function utilizing a comprehensive battery of neuropsychological tests including global as well as domain specific scores for a nuanced perspective. Finally, we adjusted for several relevant covariates that are known to impact inflammation and cognition, including cardiovascular risk factor burden, BMI, statin medications, and analgesic medications.

This study is not without limitations. While utilizing several inflammatory markers for statistical characterization was a strength, there are additional inflammatory markers from both the innate and adaptive immune response that were not collected and may have led to a more refined characterization of inflammation. Further, given the complexities of the inflammatory markers used in our research, and the fact that both the acute and chronic markers are being expressed in participants at the same time over many years, our categorization based on acute/upstream versus chronic/downstream may not encompass the finer-grained differences inherent in these PCA-derived composite scores. Future work may aid in developing a more precise statistical characterization of these and other inflammatory markers. Additionally, while several relevant covariates were accounted for in analyses, kidney function as measured by estimated glomerular filtration rate (eGFR) was unable to be adjusted for in our models as available eGFR values in this sample were based on an outdated and methodologically flawed calculation (Delgado et al., 2022). Given known relationships between eGFR, inflammatory markers, and cognition (Fox et al., 2010; Murray et al., 2016), it is possible that our findings could be impacted by kidney function; future work using the updated eGFR calculation will be important to better define kidney-related contributions to these findings. Further, this study did not investigate psychosocial factors, such as stress (Turner, James, Capuano, Aggarwal, & Barnes, 2017) and discrimination (Barnes, Lewis, et al., 2012) that may contribute to or explain the associations noted between inflammatory markers and cognitive function over time. Although investigating both the individual and environmental/systemic factors that may be contributing to these relationships is imperative (Bagby et al., 2019), understanding how inflammatory markers associate with cognition is the first step in a line of ongoing research to address these issues. Finally, the study sample may not be representative of older Black adults living in other geographical areas throughout the U.S. in terms of education, psychosocial factors, and/or broader cultural and environmental contexts.

4.1. Conclusions

In sum, our study showed that acute/upstream peripheral inflammatory markers were associated with baseline semantic memory and perceptual speed, while chronic/downstream peripheral inflammatory markers were associated longitudinally with faster declines in global cognitive function as well as episodic and working memory in older Black adults. Our findings highlight the utility of statistical characterization of inflammatory markers as a means of providing a data-driven approach to incorporating several types of inflammatory markers in one study. It also emphasizes the importance of investigating cross-sectional as well as longitudinal associations between inflammatory markers and cognitive function given differential relationships emerged depending on both the type of inflammatory markers and the baseline versus longitudinal nature of the cognitive data. Together, this work provides increased clarity to previous mixed findings in studies of inflammation and brain-behavior in older Black adults, and may serve as a springboard for future work investigating the psychosocial, environmental, and systemic contexts that may influence these associations.

Highlights.

• Limited work has investigated inflammation and cognition in older Black adults

• Principal component analysis revealed acute and chronic inflammation factors

• Higher acute inflammation associated with lower level but not change in cognition

• Higher chronic inflammation associated with faster decline in cognition

• Chronic inflammation may increase risk for cognitive decline in older Black adults