Elevated C-Reactive Protein in Alzheimer’s Disease Without Depression in Older Adults: Findings From the Health and Retirement Study

Abstract

We examined the association between differential diagnoses of major stroke and probable Alzheimer’s disease (AD) and mixed AD on C-reactive protein (CRP) in older adults with and without depression. Secondary data analyses examined associations between blood-based measures of probable peripheral inflammation using CRP collected from dried blood spots in the Health and Retirement Study, a nationally representative sample of individuals aged 50 and older. A validated pattern-recognition algorithm was utilized to identify cognitive decline indicative of probable AD, mixed AD, and major stroke. Negative binomial regressions were utilized to model concentrations of serologic CRP. On average, participants (N = 4 601) were 70 years old, female, and non-Hispanic White. Mixed AD participants had a 0.26 mg/dL increase in CRP compared to unimpaired participants, controlling for demographics, health behaviors, and comorbidities. Those with mixed AD had 2.14 times increased odds of having high CRP (odds ratio = 2.14 [1.19–3.85]). In analyses stratified by depression, adults with mixed AD and without depression had an additional 0.37 mg/dL increase in CRP (SE = 0.06; p < .001) compared to unimpaired adults. Those with AD without depression had a 0.20 mg/dL increase in CRP (SE = 0.07; p < .01). Age was not associated with increased CRP in nondepressed older adults. Depressed adults with major stroke had a −0.26 mg/dL decrease in CRP (SE = 0.11; p = .02), controlling for hypertension, alcoholic drinks/beverages per week, and smoking status. Concentration modeling revealed that participants with major stroke, probable AD, and probable mixed AD without depression had significantly higher CRP concentrations when compared to unimpaired older adults.

Both Alzheimer’s disease (AD) and major stroke are significant sources of disability among older adults and are both contributors to the top 10 causes of death in the United States (1,2). Inflammation has long been thought to play a vital role in the pathophysiology of AD in older adults, whereby inflammation is related to further neurodegeneration and cognitive decline (3–6). This proinflammatory environment is associated with a deteriorating central nervous system (CNS) (7). There is increasing recognition of the critical role that the breakdown of the blood–brain barrier (BBB) plays in AD pathogenesis (8).

C-reactive protein (CRP) is an acute-phase protein that increases in distribution volume in the blood when there is a condition, infectious or systemic, causing a neuroinflammatory response (9). While CRP does not normally cross the BBB, it has the potential to modulate BBB permeability and appears to interact with the CNS as evidenced by, for example, a high degree of correspondence between CRP levels in plasma and in the cerebrospinal fluid (r > 0.85) (10). Seeking to better understand this correspondence, researchers have highlighted the potential for CRP to interact with the CNS in a number of ways, including, for example, by modulating BBB permeability, during regular maintenance or response operations, helping to relay cytokine signals to the brain, initiating a neuroinflammatory response when the body is injured, or passing accidentally through leaky regions of the BBB (11). For example, the interactions between CRP and the BBB follow a biphasic process with increased intracellular permeability at a high dose that enables its entry into the CNS and serves to activate glia (eg, reactive gliosis) and impair CNS function (12). Additionally, elevated CRP increases the permeability of the BBB by binding to the Fcγ receptor, resulting in the activation of microglia in the brain (13). Reactive gliosis and the complement-dependent pathway can overprune functional synapses and lead to neurodegeneration and cognitive impairment in AD (14). Once engaged, neuroinflammation may accelerate AD pathogenesis (15) by inciting tau propagation in the cerebrum of those who have existing tauopathy (16,17). CRP is also known to be elevated in individuals with risk factors common to major stroke, including diabetes (18), obesity (19), and smoking (20).

CRP is a putative bioindicator of interest in predicting cognitive aging (21) and may be indicative of increased risk of both major stroke and AD (22,23). Inflammation has long been thought to play a vital role in the pathophysiology of AD and prestroke risk and poststroke cognitive impairment in older adults, whereby inflammation in general, but CRP specifically, is related to further neurodegeneration and cognitive decline (24). CRP is associated with an overall adverse risk of primary major stroke (25) and increased major stroke severity (26). Studies suggest that CRP deposition in the postmortem brain was highly concentrated in brain regions infarcted by stroke, with distinct and separate interactions and co-localization of CRP with β-amyloid and phosphorylated tau (27). In mouse models, the injection of CRP into the hippocampus caused cognitive decline and induced phosphorylation of tau like that observed in AD mouse models (27). CRP is also “dramatically overexpressed” in the brain’s extracellular matrix in postmortem brain tissues following an acute ischemic stroke (28). However, CRP does not affect all brain regions and tissues equally. It is unclear why CRP affects the BBB and no other cerebrovascular beds following stroke (29). Still, CRP has been shown to be useful in independently predicting subtypes of stroke, and studies have found an association between elevated CRP and ischemic stroke, but not hemorrhagic stroke (30). Elevated CRP affects cognitive decline poststroke as well as carries a higher risk of intracranial arterial stenosis (31).

CRP has also been identified in frontal cortex regions of those diagnosed with AD and unimpaired controls, thus suggesting a plausible pathological role in the development and/or progression of AD pathology (32). CRP may also reflect later-stage changes in AD that occur with the integration of β-amyloid-40 in as neuritic plaques mature (33). CRP expression was increased in brain regions that exhibited AD pathologies of amyloid deposits and neurofibrillary tangles, yet CRP was absent from unimpaired regions of brain tissue without signs of injury, insult, or neurodegeneration (34). CRP might be indicative of a pathological cascade if β-amyloid deposition increases in damaged areas as a result of chronic inflammation after major stroke (35) or AD (36,37) associated with chronic cerebrovascular dysfunction (38).

Despite evidence that CRP may be present in the brain tissues of individuals with major stroke or AD, the mechanisms relating peripheral CRP in blood-based plasma studies and AD and/or major stroke are unclear as it is unknown whether CRP elevation precipitates β-amyloid deposition and phosphorylation of tau, or whether the presence of β-amyloid and total-tau incites CRP elevation. However, CRP is produced and laid down in large quantities within the brain following stroke, other brain injuries, or conditions linked with neuroinflammation. There is, therefore, a need to translate postmortem and mouse model studies of CRP to human populations using peripheral blood-based analyses of CRP using a large cohort of aging individuals in the United States.

Together, the above results suggest that CRP may be indicative of the presence of neurodegeneration and increased risk for associated neurocognitive dysfunction, but also highlight the fact that relationships between CRP and AD, aging, and/or major stroke remain unclear. For example, a recent meta-analysis that evaluated whether CRP concentrations were significantly higher in older adults with AD using a pooled sample of 1 645 AD and 14 363 healthy age-matched controls reported that no significant differences between those with AD when compared with healthy controls (p = .07) (7). This result was confirmed by a second meta-analysis, which reported no significant difference in CRP concentrations between having AD and healthy controls in a meta-analysis (N = 2 093) (9).

One potential source of variability is in the background levels of inflammation among those with psychiatric comorbidities including, especially, major depression. Indeed, inflammation has long been thought to play a vital role in the pathophysiology of depressive disorders in older adults (39,40). Recent studies support this conclusion showing, for example, that elevated CRP and other peripheral biomarkers of inflammation are associated with the incidence of depression in individuals with lung cancer (41) and the risk of treatment-resistant depression (42). Recent work has even clarified that inflammation is coincident with neuroinflammation as measured by the presence of translocator protein (43). Concurrent evidence of higher CRP concentrations would support the psycho-neuroinflammatory theory as it contributes to depressive symptoms and AD in older adults (40). Thus, we expect older adults with depression and cognitive impairment to have an increased immune response due to depression (44), but the remaining question then is whether CRP is increased in people with AD or major stroke apart from a neuropsychiatric disease burden of depression (45).

While AD and major stroke represent distinct pathologies, the 2 conditions are commonly comorbid and share risk factors (46). Both AD and major stroke are common neurological problems that tend to be underdiagnosed, often cause cognitive decline, and differ in prognosis highlighting the need to find new ways of objectively determining risk and differentiating disease processes and, at the end of life, most individuals with dementia have neuropathological indications consistent with both AD and major stroke commonly referenced as mixed AD (47), thereby highlighting the need to better identify the timing of onset in AD and major stroke. To that end, recent work has proposed that AD and major stroke are longitudinally differentiable in terms of objectively observed patterns of decline in episodic memory over time (48). These efforts have thus far noted high levels of sensitivity and offer a possible way to objectively determine the risk of AD and of major stroke among individuals who may otherwise be asymptomatic. Thus, it may be possible to improve understanding of the underlying biological changes longitudinally years before a clinical diagnosis when relying on algorithms targeted at detecting and characterizing different types of cognitive decline. In addition, to the best of our knowledge no study has specifically compared the peripheral levels of CRP between older adults with and without depression and major stroke, AD, mixed AD, and cognitively unimpaired comparison groups in a large, nationally representative sample. We propose that longitudinal patterns of cognitive change characteristic of progression of preclinical decline in AD and related dementias and of major stroke will be associated with increased CRP in groups with and without depression. The goal of this study was therefore to rely on a recently validated pattern-recognition protocol to improve our understanding of the temporality of probable systemic inflammation and probable AD and major stroke using a pattern-recognition algorithm and blood-based CRP.

Method

Data from the 2014 wave of the Health and Retirement Study (HRS), the largest prospective study of cognitive aging in the United States, were retrieved for secondary data analysis. The HRS has a good response rate across entry cohorts (81.5%) (49) and includes an intentional oversampling of minority populations to maintain representativity. The HRS sample has been built over time, with the initial cohort enrolled in 1992 and followed up every 2 years after that for a total of 13 waves, at the time this article was written. The HRS is a rich source of longitudinal and cross-sectional data for researchers and policymakers who study aging. The HRS is open to enrollment at subsequent waves, and data are publicly available online (http://hrsonline.isr.umich.edu). The overall completion rate was 90.3%.

Outcome: CRP to Measure Probable Peripheral Inflammation

Details regarding blood collection, CRP assay procedures, and quality control are previously described (50). The CRP assay used by the HRS is a sandwich enzyme-linked immunosorbent assay applied to a punch from a dried blood spot (DBS) card containing either a CRP assay standard, a quality control (QC) sample, or a patient sample eluted in a buffer solution. DBSs are regarded as a commonly applicable method for collecting, storing, transporting, and analyzing a variety of human body fluids (51). Acceptability of the assay was determined by comparing the CRP concentrations of the QC samples with established values (50). Analyses excluded CRP values that were more than 30 mg/dL, which are likely indicative of acute infections when compared with chronic disease burden. DBS samples were collected in 2006, 2008, 2010, and 2014. The equivalent values make the CRP assay levels for the HRS data based on DBS similar to the level in the National Health and Nutrition Survey (NHANES), where values were modified to match averages in NHANES while retaining variability evident in the HRS sample (50). The CRP within-assay imprecision (CV) was 8.1% and between-assay imprecision was 11.0%.

Independent Variables

Inferential identification of AD and major stroke was determined by a pattern-recognition algorithm that detected probable major stroke and probable AD based upon patterned changes in episodic memory, using 5 or more serial cognitive assessments. Cognitive assessments included 10-item word recall (episodic memory), where participants were asked to correctly recall to the best of their ability with each correct answer scoring one point. The Total Episode Memory Index was the summation of both the immediate and delayed word recall out of 20 points. The first 2 waves were excluded because they used a 20-item scale, as compared to the 10-item one used here. While not used here, orientation was measured using the degree of confusion over everyday facts, including what the president’s name is, or the current day of the month, day, and year. Prior efforts have clarified that changes seen in episodic memory are highly associated with transitions into disorientation both for major stroke (48) and AD-like outcomes (52). The pattern-recognition program is known to be particularly sensitive to random variation in the first and/or last waves, so onsets of major stroke or AD that occurred before the second or after the penultimate wave were generally ignored (48). The algorithm assigned each participant a score based on longitudinal assessments, and each exclusive diagnostic category had a cutoff to determine cognitive states. To clarify these groups in the present analyses, participants were divided into 4 exclusive diagnoses: cognitively unimpaired, probable AD, probable major stroke, and probable mixed AD.

Race/ethnicity was categorized into 4 groups (non-Hispanic White, non-Hispanic Black, other race, and Hispanic ethnicity). White was the reference group in the analyses. Age was a continuous variable. Sex was a dichotomous variable (male/female) with males as the reference group. Alcohol consumption was measured continuously in alcoholic drinks/beverages. History of cardiovascular disease (CVD) was recorded as yes/no. Smoking, former asked if participants previously smoked cigarettes ever in one’s life was recorded as yes/no. Smoking, current asked if the participant currently smokes cigarettes as yes/no. History of high blood pressure of more than 130 systolic mmHg or more than 80 diastolic mmHg. Educational attainment was categorized into 4 categories: less than high school, GED or high school graduate, some college, and college degree or above. Obesity was determined by a cutoff of a body mass index of 30 or higher.

Possible depression was assessed using the validated Center for Epidemiologic Studies—Depression scale (CES-D). The CES-D-8 is an 8-item scale with a score of 3 or greater indicating depression. CES-D-8 was measured at each wave beginning at Wave 2. Longitudinal assessments of 6 or more waves of the CES-D were used for all models. Respondents were asked to rate the frequency of 8 symptoms of psychological distress as yes/no responses (eg, Much of the time during the past week, you felt depressed. Would you say yes or no?).

Statistical Analyses

Descriptive analyses reported percentages (%) or means and standard deviations (mean; SD). We conducted a Spearman’s rho correlation matrix of all key variables to show significant relationships between AD and CRP. Negative binomial regression models (53) appropriately estimate counts or distribution volume (numbers of proteins as expressed by weight of proteins in a standardized amount of blood) that follow a Poisson process when the data are overdispersed (54). When modeling concentration based on counting the number of, or tallying the weight of, proteins in the blood, these efforts assume that analytes in the blood are not modified by blood volume in retrieved specimens or in the body as a whole. Negative binomial regression is like regular multiple regression in that it allows researchers to examine associations between outcomes and variables of interest when adjusting for possible confounding, here indicators of inflammation, and indicators including disease of interest, age, race/ethnicity, and sex. The log of the outcome is predicted with a linear combination of the predictors. Five models were fit using maximum likelihood estimation. Three models use the whole sample of older adults and examine the relationship between inferential diagnosis and CRP concentrations, accounting for demographics, comorbidities, and lifestyle choices. Two other regression models were run using a sample stratified between depressed and nondepressed older adults. Missing data were carefully considered, and sensitivity analysis was run on missing data, and CRP values were above the 30 mg/dL cutoff for acute infection. A 2-tailed α = 0.05 was used to determine statistical significance.

Sensitivity Analyses

We performed additional sensitivity analyses to determine whether the high CRP values were associated with a cognitive diagnosis or varied by age, sex, and race/ethnicity. Results show no significant increased odds of having a CRP value of more than 30 mg/dL for males or females, older adults, Black, Hispanics, or others.

Results

Sample Characteristics

In 2014, there were 4 801 participants with an inferential diagnosis available, age, a CES-D-8 score, and CRP data. Those with high CRP concentrations exceeding 30 mg/dL were excluded due to possible acute infection (N = 200). In total, 4 601 participants were included. Finally, having an elevated CRP above 30 mg/dL was not associated with increased odds of having missing depression scores (p = .812).

On average, participants in this study (N = 4 601) were 65.8 ± 8.8 years old for participants without an inferential diagnosis, 72.2 ± 10.0 years old for major stroke, 72.8 ± 8.0 years old for probable AD, and 78.5 ± 9.4 years old for mixed AD groups (Table 1). The distribution of CRP represented blood volume concentrations that were highly skewed (p < .0001). For example, mean CRP levels were only 4.0 mg/L, and yet 19.8% of the total sample had elevated levels of CRP above 5.0 mg/L. All covariates varied significantly between diagnostic categories (Table 1). Approximately 32% of those with probable AD also had a diagnosis of heart disease (32%) and likely smoked cigarettes (59%) and drank more than 2 alcoholic drinks/beverages per week (46%). About one third of those with AD were obese, compared to 35% of unimpaired respondents.

Table 1.

Sample Characteristics, Health and Retirement Study (N = 7 222), Kruskal–Wallis Test

	Normal		Major Stroke		ADRD		Mixed		p
	M	SD	M	SD	M	SD	M	SD
Age, years	65.8	8.8	72.2	10.0	72.8	8.0	78.5	9.4	<.001
CRP, mg/dL	3.4	7.0	3.6	7.1	4.0	8.4	4.5	11.0	<.001
Alcoholic drinks/beverages per week	1.19	2.1	1.19	2.06	1.04	2.0	0.96	2.0	.030
	%		%		%		%
Male	39.3		37.70		39.0		34.0		<.001
Female	60.7		62.30		61.0		66.0
White	75.7		77.20		74.9		80.3		<.001
Black	13.7		12.70		14.3		11.8
Other	2.0		2.00		2.2		1.6
Hispanic	8.5		8.00		8.6		6.3
Less than high school	17.0		15.0		21.0		18.0		.010
High school/GED	35.0		37.0		35.0		36.0
Some college	23.0		23.0		22.0		27.0
College and above	25.0		26.0		22.0		20.0
Diabetes	17.0		22.0		27.0		24.0		<.001
Cardiovascular disease	21.0		28.0		32.0		36.0		<.001
Smoking, former	57.3		56.0		59.3		54.8		<.001
Smoking, current	10.6		10.2		9.4		8.8		.507
Hypertension	53.0		63.9		68.0		69.0		<.001
Depression	17.73		20.22		23.61		25.03		<.001
Obesity	35.31		33.81		33.07		28.21		<.001

Note: SD = standard deviation; ADRD = Alzheimer’s disease and related dementia; CRP = C-reactive protein; GED = general education development test.

Table 2 presents the Spearman’s rho correlation matrix for all variables. Having an inferential diagnosis was associated with increased CRP. Age was not associated with CRP but was correlated with AD (r = 0.29; p < .001), major stroke (r = 0.21; p < .001), and mixed AD (r = 0.27; p < .001). Increased intake of alcoholic drinks/beverages per week consumed was negatively correlated with CRP concentrations (r = −0.09; p < .001). Those with mixed AD alone had 2.14 times increased odds of having high CRP (odds ratio [OR] = 2.14 [1.19–3.85]). We also analyzed the relationship between clinical depression and CRP, finding that depressed participants were not at increased odds of having CRP values (p = .743). Further analyses reveal that females had increased odds of clinical depression (OR = 1.73 [1.67–1.79]). Blacks (OR = 1.65 [1.58–1.72]) and Hispanics (OR = 2.50 [2.39–2.61]) were at increased odds of clinical depression. Those with cognitive impairment had increased odds of clinical depression, including, major stroke (OR = 1.12 [1.08–1.17]), AD (OR = 1.25 [1.17–1.34]), and mixed AD (OR = 1.33 [1.26–1.40]). Older age was protective against odds of depression (OR = 0.997 [0.995–0.998]) after adjusting for inferential diagnostic status.

Table 2.

Spearman’s Rho Correlation Matrix Showing Associations Between Alzheimer’s Disease and Mixed Alzheimer’s Pathology Determined Using Pattern-Recognition Profiling and C-Reactive Protein Concentration, Health and Retirement Study

	CRP	AD vs Unimpaired	Stroke vs Unimpaired	Mixed AD vs Unimpaired	Diabetes	Alcoholic Drinks/Beverages per Week	Cardiovascular Disease	Smoke, Former	Smoke, Current	Hypertension	Obesity	Depression
CRP	1.0000
		1.0000
AD vs Unimpaired	0.2910
	<0.001		1.0000
Major stroke vs Unimpaired	0.2120
	<0.001
Mixed AD vs Unimpaired	0.2720			1.0000
	<0.001
Diabetes	0.0829	0.0625	0.0552	0.0560	1.0000
	<0.001	<0.001	<0.001	<0.001
Alcoholic drinks/beverages per week	−0.0944	−0.0295	−0.0307	−0.0713	−0.1357	1.0000
	<0.001	<0.001	<0.001	<0.001	<0.001
Cardiovascular disease	0.0443	0.0705	0.0765	0.1149	0.7646	−0.0958	1.0000
	<0.001	<0.001	<0.001	<0.001	<0.001	<0.001
Smoking, former	0.0525	0.0086	−0.0067	−0.0137	0.5284	0.1016	0.5262	1.0000
	<0.001	0.0061	0.0203	<0.001	<0.001	<0.001	<0.001
Smoking, current	0.0414	−0.0219	−0.0325	−0.0425	0.7815	−0.0266	0.7483	0.6629	1.0000
	<0.001	<0.001	<0.001	<0.001	<0.001	<0.001	<0.001	<0.001
Hypertension	0.1021	0.0688	0.0879	0.0924	0.6142	−0.1077	0.5852	0.3629	0.5260	1.0000
	<0.001	<0.001	<0.001	<0.001	<0.001	<0.001	<0.001	<0.001	<0.001
Obesity	0.3053	−0.0085	−0.0118	−0.0493	0.2904	−0.0674	0.2440	0.1664	0.2450	0.2290	1.0000
	<0.001	0.0065	<0.001	<0.001	<0.001	<0.001	<0.001	<0.001	<0.001	<0.001
Depression	0.0622	0.0240	0.0239	0.0408	0.6837	−0.1149	0.6632	0.4710	0.7050	0.4950	0.2928	1.0000
	<0.001	<0.001	<0.001	<0.001	<0.001	<0.001	<0.001	<0.001	<0.001	<0.001	<0.001

Notes: AD = Alzheimer’s disease; CRP = C-reactive protein. Estimates are statistically significant (p < .05).

Univariate regression modeling showed that both AD and mixed AD were associated with increased CRP (Model 1; Table 3; p < .01). Model 2 builds on the previous model and adds covariates of age, sex, race/ethnicity, and education. The relationship between AD, mixed AD, and CRP remained significant and positive. Model 3 in Table 3 revealed a relationship between mixed AD and CRP, such that mixed AD participants have a 0.26 mg/dL increase in CRP compared to other participants. Additionally, in both Models 1 and 2, Blacks had increased CRP compared to non-Hispanic Whites (p < .001). Covariates in Model 3 associated with increased CRP include having diabetes, being a current smoker, and having hypertension and being obese (p < .001).

Table 3.

Associations Between Alzheimer’s Disease and Mixed Alzheimer’s Pathology Determined Using Pattern-Recognition Profiling and C-Reactive Protein Concentration, Health and Retirement Study

	Model 1			Model 2			Model 3
	Pattern-Recognition Diagnostic			Demographic-Adjusted Model			Fully Adjusted
	B	SE	p	B	SE	p	B	SE	p
Algorithmic diagnosis
Normal	–	–	–	–	–	–	–	–	–
Stroke	0.07	0.05	.158	0.07	0.05	.160	0.04	0.05	.423
AD	0.15	0.06	.010	0.14	0.06	.021	0.11	0.06	.069
Mixed AD	0.29	0.06	<.001	0.29	0.06	<.001	0.26	0.06	<.001
Age				0.00	0.00	.317	0.01	0.00	.005
Race
White				–	–	–
Black				0.36	0.06	<.001	0.20	0.06	.000
Other				0.02	0.13	.873	−0.02	0.13	.867
Hispanic				−0.01	0.06	.954	0.02	0.07	.810
Female				−0.02	0.07	.757	−0.01	0.04	.761
Education (years)				−0.78	0.02	<.001	−0.03	0.02	.081
Diabetes							0.09	0.05	.037
Alcoholic drinks/beverages per week							−0.05	0.01	<.001
Cardiovascular disease							0.14	0.04	<.001
Smoking, former							−0.04	0.04	.386
Smoking, current							0.31	0.07	<.001
Hypertension							0.18	0.04	<.001
Obesity							0.41	0.04	<.001

Notes: AD = Alzheimer’s disease; Normal = no stroke or AD decline; CRP = C-reactive protein. Bolded estimates are statistically significant (p < .05).

Multivariate regression modeling revealed that participants with major stroke, probable AD, and probable mixed AD without depression had significantly higher CRP concentrations when compared to individuals who were not determined to have a cognitive pathology (Table 4). Mixed AD was associated with the largest additional increase of CRP at 0.37 mg/dL, followed by AD at 0.21 mg/dL, then major stroke at 0.14 mg/dL. After adjusting for age, race/ethnicity, education and sex, diabetes, hypertension, alcohol consumption, smoking, and CVD, those without depression and probable AD had higher CRP when compared with other participants. In contrast, participants without depression and classified as mixed AD had an increase of 0.37 mg/dL in CRP concentrations (p < .001). Black adults without depression had higher CRP (B = 0.29; SE = 0.06; p < .001) compared to Whites. Current smokers, and those with diabetes, hypertension and obesity without depression had associated increased CRP concentrations (p < .001). Increased educational attainment and alcohol consumption in older adults without depression were associated with decreased CRP concentrations. Aging was not associated with increased CRP in depressed older adults. Model 3 also shows sex differences in elevated CRP among clinically depressed females, such that women had decreased CRP compared to males with depression.

Table 4.

Associations Between Alzheimer’s Disease and Mixed Alzheimer’s Pathology Determined Using Pattern-Recognition Profiling and C-Reactive Protein Concentration by Depression, Health and Retirement Study

	Not Depressed			Depressed
	B	SE	p	B	SE	p
Algorithmic diagnosis
Normal	Reference			Reference
Stroke	0.14	0.05	.006	−0.25	0.11	.024
AD	0.21	0.07	.001	−0.09	0.13	.485
Mixed AD	0.37	0.06	<.001	−0.05	0.12	.716
Age	0.01	2.60E−03	.003	5.14E−04	4.90E−03	.916
Race
White	Reference			Reference
Black	0.28	0.06	<.001	−0.08	0.13	.526
Other	−0.01	0.14	.970	−0.11	0.30	.721
Hispanic	−0.09	0.08	.291	0.05	0.13	.705
Female	0.05	0.05	.271	−0.25	0.10	.012
Education (years)	−0.03	0.02	.071	0.01	0.04	.881
Diabetes	0.18	0.05	<.001	−0.15	0.10	.12
Alcoholic drinks/beverages per week	−0.04	0.01	<.001	−0.07	0.03	.006
Cardiovascular disease	0.13	0.05	.007	0.05	0.09	.609
Smoking, former	−0.03	0.05	.574	−0.03	0.10	.759
Smoking, current	0.30	0.08	<.001	0.19	0.14	.156
Hypertension	0.16	0.05	<.001	0.19	0.12	.107
Obesity	0.36	0.05	<.001	0.57	0.10	<.001

Notes: AD = Alzheimer’s disease; Normal = no stroke or AD decline; CRP = C-reactive protein. Bolded estimates are statistically significant (p < .05).

For those with depressive symptoms, CRP was negatively associated with major stroke (B = −0.25; SE = 0.11; p = .024). Female sex, alcohol consumption, age, and major stroke were associated with lower CRP, while hypertension was associated with 0.27 mg/dL (p < .001) increased CRP concentrations in those with depression. Education, diabetes, and currently smoking cigarettes were not significant for those with depression, unlike those without depression. Older age was associated with increased CRP concentrations for those with without depression (B = 0.01; p = .003). Blacks with depression did not show an increase in CRP. Females with depression had associated decreased CRP (B = −0.21; SE = 0.10; p = .012). Notably, those with obesity and depression had increased CRP concentrations compared to those with depression and body mass indicative of normal or overweight status (Table 4).

Discussion

This study relied on a recently validated pattern-recognition protocol to improve our understanding of the role of probable AD and major stroke as well as depression as correlates of systemic inflammation measured using a validated blood-based CRP indicator. We relied on a large nationally representative study of older adults in the United States to examine the relationship between probable AD, mixed AD, and major stroke with increased serological indicators of inflammation in adults with and without depression. We found that older adults with probable mixed AD had significantly higher CRP concentrations when compared to individuals who were determined to lack cognitive pathology. The relationship between CRP and major stroke, AD, and mixed AD persisted after controlling for hypertension, alcohol consumption, smoking status, and CVD. Furthermore, elevated CRP was associated with aging in the nondepressed group but was not associated with aging in the depressed group of older adults. In summary, our results showed a clear relationship between high concentrations of CRP and major stroke, probable AD, and mixed AD for those without depression, yet an inverse relationship between major stroke and CRP in depressed older adults. CRP was not related to AD or mixed AD in depressed older adults. The present study has the strength of combining concentration modeling with a novel diagnostic measure that relies on pattern recognition to algorithmically define diagnostic categories following existing clinical criteria.

The emerging research on probable systemic inflammation indicated that older adults without dementia with higher levels of inflammatory markers perform worse on verbal memory tests at baseline (22) and are at risk for future cognitive decline (21). However, the relationship between inflammation and memory is not consistent in the literature, as additional cross-sectional (23) and longitudinal studies (55) have failed to find an appreciable connection between CRP, AD, and memory function. We aimed to address these issues by using a pattern-recognition algorithm that objectively characterizes possible disease states prior to clinical presentation. We related individual pattern changes in episodic memory across 12 waves of data to increased systemic inflammation.

Our results contrast with previous studies that show decreased CRP concentrations associated with AD at an older age, or no association between AD and CRP concentrations (7,9,56,57). A recent systematic review and meta-analysis of 1 645 AD participants and 14 363 controls without depression or AD showed no differences in CRP, IL-6, and TNF-α between AD participants and controls (p = .07) (7). Similarly, a 9-year follow-up study on cognitively unimpaired community-dwelling older women found no difference across CRP tertiles in individuals with faster rates of cognitive decline (58). However, in a prospective cohort study, serum CRP level was inversely associated with decreased executive function and processing speed within the normal range (59). Our results build on prior work showing that elevated CRP concentrations in adults with AD are associated with reduced cognitive function and survival (56,60). We hypothesize that some patients with both elevated CRP levels and AD may have more severe dementia indicating that the presence of an inflammatory state worsens cognitive function in adults with AD.

Our results also show higher CRP concentrations in AD and mixed AD, and this relationship remains significant after accounting for age, demographics, health behaviors like smoking and consuming alcohol, and comorbid diabetes, CVD, smoking status, and hypertension. These results are suggestive of a possible proinflammatory endophenotype in AD but only in those without depression. Our results do not support the psycho-neuroinflammatory theory as depressed older adults in the present study with AD or mixed AD did not have a significant relationship with CRP.

Depression, Major Stroke, and CRP

Furthermore, our results show an interesting relationship between major stroke and CRP in those with depression compared to those without depression. For major stroke survivors, they had decreased CRP compared to cognitively unimpaired older adults. However, prior studies show that postmajor stroke depression at 6 months is strongly associated with increased odds of elevated CRP (61,62). Other studies have shown higher CRP levels in those with greater depressive symptomology, with symptom-specific associations for women (44). However, the women in the present study experienced decreased CRP levels when depressed compared to men. Our results do not support the “sickness syndrome” theory (44), suggesting that chronic low-grade inflammation may be associated with a subtype of depression in women. Sex differences in neuropsychopathology may be explained by biological differences in depression and AD physiology for men versus women, who differ in the distribution of circulating levels of protein-tau even at midlife (33). Clinical studies should investigate sex-specific treatments guided by both depressive severity and CRP.

Lifestyle Correlates of CRP

We also found a negative association with more alcoholic drinks/beverages per week and decreased CRP in those with and without depression, which is consistent with previous studies showing lower CRP in stable moderate consumption of alcoholic drinks/beverages per week across 10 years of observation (63). This effect was not present in older adults with depression. One explanation is that moderate alcohol consumption has a protective cardiometabolic effect compared to abstinence, resulting in lower CRP levels (64). Although studies using repeat measures of alcohol consumption are rare, moderately drinking wine was shown to be associated with lower inflammatory risk factors across 8 years (65). Chronic probable systemic inflammation has been hypothesized to contribute to the development of type II diabetes as well as its complications (66). One study showed increased mean CRP in diabetics versus healthy controls (67). Our findings also confirm that participants with diabetes have associations with increased CRP in nondepressed adults.

We also found the relationship between CRP and CVD became nonsignificant after controlling for current smokers in those with and without depression. The relationship between CVD, hypertension, and CRP is well known (68) however, our results suggest that in people without depression there was a robust association between inflammatory biomarkers and major stroke, AD, and mixed AD after adjusting for current and former smoking status. In prior studies, CRP has been found to be elevated in current smokers when compared to never smokers but those studies did not account for depression. Like the present study, which reported no differences in CRP between never and former smokers, smoking cessation is thought to cause inflammatory markers to revert to levels seen in never smokers (69), though there remains some debate as to the consistency of reversion results (70).

Strengths and Limitations

The present study used a stable measure of systematic inflammation to examine the relationship between long-term patterns of decline in episodic memory and elevated blood-based biomarkers of probable systemic and neuroinflammation. The present study had the additional strength of using pattern-recognition algorithms to estimate neurological diagnosis over 12 waves of nationally representative data in the HRS, making the study one of the first and largest estimations of cognitive function and probable systemic inflammation in a nationally representative sample of Americans. Respondents in the current study included a wide array of American residents from disparate social and racial backgrounds.

This study sought for the first time to validate diagnostic criteria relying on longitudinal pattern-recognition profiling using blood-based biomarkers and specifically CRP; however, this study lacks information about the degree and/or type of neuropathology that may be present. Abnormally high concentrations of CRP may also indicate acute infections, rather than underlying probable systemic inflammation implicated in AD and major stroke. To that end, in this study, we excluded observations among individuals with CRP values more than 30 mg/dL as an indicator of acute infection. Our sensitivity analyses showed that those with mixed AD were at 2.14 times increased odds of abnormally high CRP values. This was surprising, and little is known about the integrity of the BBB in adults with mixed AD. Abnormally high CRP values in mixed AD adults might signify advanced neuropathology and the subsequent degradation of the BBB, which could further incite systemic inflammatory response in the body and the brain.

Additional sensitivity analyses showed that Blacks, Hispanics, and others are not at increased odds of abnormally high CRP values. One explanation is that chronic anti-Black discrimination directed toward African Americans contributes toward a chronic stress response that moderately elevates CRP concentrations. Experiencing chronic stressors, poverty, and discrimination may lead to higher rates of depression in Black communities as well (71). While the experience of being Black in America varies tremendously, there are shared cultural factors that play a role in helping to define mental health, well-being, and resiliency. Racism, discrimination, and income or educational inequity significantly affect older Black Americans’ mental health. Being treated or perceived as “less than” because of the color of one’s skin can be stressful and even traumatizing (72,73). Additionally, members of the Black community face structural challenges accessing the medical care and treatment they need. Black adults living below the poverty line are more than twice as likely to report “serious psychological distress” than those with more financial security (74).

Conclusions

Early identification of elevated serologic inflammatory biomarkers before the onset of clinical AD or mixed AD is crucial, and our results demonstrate that depression needs to be considered. These results were promising and highlighted a potentially large burden of heretofore unobserved inflammatory processes in older Americans with symptoms consistent with major stroke, AD, and with mixed AD. One clinical use of these data will be to corroborate the objectively identified history-based diagnosis in the clinic. Given the consistent effects of probable and mixed AD on elevated CRP, the present study suggests that more work is warranted to replicate, extend, and validate current clinical AD diagnoses with CRP using data with measures of neuropathology.

Funding

The work was supported by the National Institutes of Health/National Institute on Aging (NIH/NIA RF1 AG058595). Funding for the Health and Retirement Study is provided by the NIA at NIH (U01 AG009740), with supplemental support from the Social Security Administration.

Conflict of Interest

None declared.

Author Contributions

G.N. conceptualized the research design and methods and performed analyses and data analysis. G.N., S.A.P.C., and D.M.S. prepared and revised the manuscript for publication. S.A.P.C. verified statistical analyses and data interpretation.