Associations between immunosenescence and domain-specific cognition in the Health and Retirement Study Harmonized Cognitive Assessment Protocol

Abstract

Background

Alzheimer’s disease and related dementias (ADRD) pose a critical challenge for the aging U.S. population, yet many biological pathways remain unclear. Recent work has focused on whether peripheral immunosenescence contributes to cognitive aging. Using the Health and Retirement Study (HRS), we examined associations between peripheral immunosenescence and domain-specific cognitive function.

Results

Immune function was associated with general and domain-specific cognition, with variation by sex and immune marker. The CD4 + EMRA:Naïve T Cell Ratio, CD4 + Naïve:CD8 + EMRA Ratio, and CMV IgG showed the most consistent associations, particularly with general and executive function. A one-SD increase in CMV IgG was associated with 0.06 SD lower executive function (95% CI: –0.07, –0.05), persisting after adjustment for confounders.

Conclusions

Associations with peripheral immune markers suggest a potential role for peripheral immunosenescence in cognitive aging. The immune system may be an early contributor to cognitive vulnerability, underscoring the need to integrate immune aging into ADRD research.

Supplementary Information

The online version contains supplementary material available at https://doi.org/10.1186/s12915-026-02679-4.

Background

Despite the critical challenge that Alzheimer’s disease and related dementias (ADRD) pose for the aging U.S. population, there remain many unanswered questions regarding the biological pathways leading to these conditions. Inflammation within the central nervous system (CNS) has been extensively referenced in studies of ADRD [1–3]. More recently attention has shifted toward the peripheral immune system and the potentially important role it might play by interacting with inflammatory processes in the brain which are hypothesized to be associated with cognitive changes observed in ADRD [4, 5].

Peripheral immunosenescence is characterized by alterations to both the innate and adaptive immune system. In the innate immune system, there is decreased leukocyte production and function, reduced phagocytic capability of neutrophils, monocytes, and macrophages, a functional shift towards a proinflammatory phenotype (i.e., ‘inflammaging’) in monocytes, and decreased cytotoxicity of natural killer cells [6]. In the adaptive immune compartment, antigenic stimulation across the life course drives expansion of memory and effector T cell subsets, increasing the number of pathogen-specific, highly differentiated T cells which also generate a greater inflammatory response [7, 8]. This leads to a decreasing availability of naïve T cells that can target new infections. Collectively, these processes coalesce to form a phenotype of immunosenescence in older age that is characterized by an increase of CD8 + cells relative to CD4 + cells, a decrease in naïve T cells, an accumulation of memory T cells-particularly late-stage differentiated CD8 + T cells with limited function, elevated levels of systemic inflammation, and elevated humoral antibody responses to viral infections such as CMV [9].

Multiple studies have documented associations between markers of peripheral inflammation and cognitive function. C-Reactive Protein (CRP),a peripheral immune inflammatory marker central to the innate immune response and capable of influencing adaptive immunity by stimulating T cell responses, has been well studied in relationship to Alzheimer’s Disease (AD) risk [10, 11]. Cross-sectional studies consistently report higher levels of CRP in the blood of individuals with dementia [12–18], and associations with lower cognitive function [18–21]. Several compelling longitudinal studies have corroborated these findings, demonstrating that higher CRP at baseline is with greater likelihood of subsequent cognitive decline [19–23] and AD [23–25].

Despite the compelling initial data pointing to a link between peripheral immune function and cognition, research on peripheral immune biomarkers beyond non-specific inflammatory biomarkers including CRP, IL-6 and other interleukins [26–29] has been far more limited. Emerging evidence from a few small but informative studies have found that individuals with ADRD have signs of advanced peripheral immunosenescence, relative to non-ADRD controls, characterized by decreases in naïve T cells and increases in memory T cells [30, 31]. A recent study by Gate et al. provided compelling evidence that adaptive immune cells are active within the central nervous system compartment; they identified clonally expanded CD8 + T effector cells in the cerebrospinal fluid (CSF) of individuals with AD compared to controls, suggesting antigen-specific T cell responses may play a role in neurodegeneration [30]. Despite these findings, there remains a critical gap in population-level research investigating how broader measures of peripheral immune function relate to specific domains of cognitive function.

There is also evidence suggesting that peripheral immunosenescence may be differentially associated with specific cognitive domains, such as processing speed, memory, and executive function [32, 33]. Given that cognitive function is a heterogenous and multi-dimensional process that does not follow a uniform trajectory, a more detailed examination of discrete cognitive domains may provide important insights into differential vulnerabilities associated with biological aging.

Using data from the U.S. Health and Retirement Study (HRS), we examined associations between multiple measures of peripheral immunosenescence and domain-specific cognitive function within a national, population-based sample of older adults. To our knowledge, few studies have applied a population health framework to investigate variation in peripheral immune aging markers in relation to specific cognitive domains. Rather than focusing solely on global screening measures or diagnostic categories related to cognitive function, we leveraged detailed domain-specific cognitive assessments to examine whether distinct immune phenotypes may be differentially associated with more specific areas of brain vulnerability. By integrating multidimensional immune markers with domain-specific cognitive outcomes, this study aims to generate biologically grounded hypotheses regarding how peripheral immune function may relate to heterogenous patterns of cognitive aging at the population level. Such findings may inform future research that focuses on improved early identification of AD and other neurodegenerative disorders characterized by region-specific changes.

Results

Sample description

The study sample included 2,215 individuals with a mean age of 74 years. The study sample was 56% female (Table 1). Twenty-seven percent of the sample had greater than a high school education. Females had higher mean scores than males on three of the five domain-specific cognitive measures. In addition, females had lower mean levels than males on four of the immune measures, consistent with a less aged immune system (e.g., CD8 + CD4 + T Cell Ratio, CD4 + terminally differentiated effector memory (EMRA): Naïve T Cell Ratio, CD8 + ERMA: Naïve T Cell Ratio, and IgD + Memory: Naïve B Cell Ratio). Females had a higher mean value than males of the Naïve CD4 + : EMRA CD8 + T Cell Ratio, which is also indicative of a less aged immune system. In contrast, females had a higher mean value than males of CMV IgG, suggestive of less immune control of the virus.

Table 1.

Weighted descriptive statistics for the sample, n = 2,215

	Total sample		Males		Females
Variable	N(%) or mean (SE)		N(%) or mean (SE)		N(%) or mean (SE)
Domain-specific cognition
General cognitive function	0.24	(0.03)	0.20	(0.04)	0.26	(0.04)
Memory	0.16	(0.02)	0.03	(0.04)	0.26	(0.03)
Executive function	0.23	(0.04)	0.27	(0.04)	0.20	(0.04)
Orientation	0.09	(0.02)	0.10	(0.02)	0.09	(0.02)
Language	0.15	(0.02)	0.15	(0.03)	0.16	(0.03)
Immune outcomes
CD8 + :CD4 + T Cell Ratio	0.44	(0.01)	0.48	(0.02)	0.42	(0.01)
CD4 + EMRA:Naïve T Cell Ratio	0.16	(0.01)	0.16	(0.01)	0.16	(0.01)
CD8 + EMRA:Naïve T Cell Ratio	7.78	(0.33)	8.58	(0.64)	7.15	(0.41)
Naïve CD4 + : EMRA CD8 + T Cell Ratio	1.38	(0.04)	1.28	(0.06)	1.45	(0.06)
IgD + Memory: Naïve B Cell Ratio	0.25	(0.01)	0.27	(0.01)	0.23	(0.01)
CMV IgG	313.39	(8.53)	249.69	(11.77)	363.62	(12.27)
Age (years)	74.16	(0.21)	73.68	(0.30)	74.54	(0.30)
Missing	0		0		0
Sex
Male	940	(44.09)
Female	1275	(55.91)
Missing	0
Education
Less than highschool	387	(14.62)	143	(12.45)	244	(16.33)
High school graduate or vocational training	1308	(57.98)	539	(55.44)	769	(59.99)
Greater than high school	520	(27.40)	258	(32.11)	262	(23.68)
Missing	0		0		0
Household wealth ($)	615,751.50	(44,925.31)	666,773.60	(70,532.39)	575,523.90	(50,766.28)
Missing	0		0		0
Current smoker
No	2016	(91.37)	859	(90.40)	1157	(92.13)
Yes	183	(8.63)	76	(9.60)	107	(7.87)
Missing	16		5		11
Depression score (CES-D score)	1.28	(0.04)	1.05	(0.06)	1.46	(0.07)
Missing	0		0		0
Comorbidity index	2.60	(0.03)	2.57	(0.06)	2.62	(0.05)
Missing	0		0		0

Associations between the CD4 + EMRA: Naïve T Cell Ratio and domain-specific cognition

Figure 1a and b illustrate the associations between the CD4 + EMRA: Naïve T Cell ratio and domain-specific function across sequential models, with negative values indicating poorer performance on the specific cognitive measure associated with more advanced immune aging. In Model 1, controlling for age, higher values of the CD4 + EMRA: Naïve T Cell Ratio (indicative of more advanced immunosenescence) were associated with lower language, orientation, and executive function scores, as well as lower general cognitive function scores, particularly in females (Fig. 1A, B, Additional file 1, Table S1). In Model 1, the confidence intervals did not include the null value, suggesting reasonably precise associations. For example, in females, in Model 1 a one standard deviation (SD) higher value of the CD4 + EMRA:Naïve T Cell Ratio was associated with −0.07 SD unit (95% CI: −0.1, −0.04) lower general cognitive function score, a −0.04 SD unit (95% CI: −0.07, −0.02) lower language ability score, a −0.03 SD unit (95% CI: −0.05, −0.01) lower orientation score, and a −0.09 SD unit (−0.12, −0.06) lower executive function score. However, in Model 2 controlling for age, education, household wealth, smoking status, chronic conditions, and depression, the estimates were attenuated, and the confidence intervals included the null (Fig. 1A, B, Additional file 1,Table S1). Notably, while we observed associations in the same directions for males, they were weaker in magnitude with confidence intervals crossing the null value.

Fig. 1.

Associations between each immune measure and domain-specific cognition function. Panels A, and B show results for the CD4 + EMRA:Naïve T cell ratio in which higher values are indicative of more advanced immunosenescence Panels C, and D show results for the CD8 + EMRA:Naïve T cell ratio in which higher values are indicative of more advanced immunosenescence. Panels E, F show results for the CD8 + :CD4 + T cell ratio in which higher values are indicative of more advanced immunosenescence. Panels G, and H show results for the CD4 + Naïve: CD8 + ERMA in which lower values are indicative of more advanced immunosenescence. Panels I, and J show results for the B cell Memory: Naïve ratio in which higher values are indicative of more advanced immunosenescence. Panels K, and L show results for CMV in which higher values are indicative of more advanced immunosenescence. Panel A. Female, CD4 +, EMRA: Naive Panel B. Male, CD4 +, EMRA: Naive. Panel C. Female, CD8 +, EMRA: Naive. Panel D. Male, CD8 +, EMRA: Naive. Part E. Female, CD8 + :CD4 +. Part F. Male, CD8 + :CD4 +. Part G. Female, CD4 + Naïve: CD8 + EMRA. Part H. Male, CD4 + Naïve: CD8 + EMRA. Panel I. Female, B Cell Memory: Naïve. Panel J. Male, B Cell Memory:Naïve. Panel K. Female standardized CMV IgG. Panel L. Male, standardized CMV IgG

Associations between the CD8 + EMRA: Naïve T Cell Ratio and domain-specific cognition

While we overall observed that higher values of the CD8 + EMRA: Naïve T Cell Ratio (indicative of more advanced immunosenescence) were associated with worse cognitive function across domains, the associations were weaker in magnitude than the associations observed in the CD4 + compartment and most of the confidence intervals included the null among both females and males (Fig. 1C, D, Additional file 1, Table S2). The exception was among females: a one SD higher value of the CD8 + EMRA: Naïve T Cell Ratio was associated with a −0.05 SD unit (95% CI: −0.09, −0.02) in executive function.

Associations between the CD8 + :CD4 + T Cell Ratio and domain-specific cognition

The findings were mixed as to the association between the CD8 + : CD4 + T cell ratio and domain-specific cognition (Fig. 1E, F, Additional file 1, Table S3). Higher values of this measure are indicative of more advanced immunosenescence. We observed associations among both females and males, where higher values of the CD8 + CD4 + ratio were associated with lower scores in multiple cognitive domains, particularly for executive function. Adjustment for additional covariates in Model 2 attenuated the observed associations and resulted in wider confidence intervals, indicating increased uncertainty around the estimates. Among females, for example, we observed an association that was relatively strong in magnitude between the CD8 + :CD4 + ratio and executive function, in which a one SD higher score was associated with a 0.14 lower SD unit (95% CI: −0.21, −0.08) executive function score in Model 1. However, in model 2, the association was attenuated and the confidence interval included the null (Fig. 1E, F, Additional file 1, Table S3).

Associations between the Naïve CD4 + : EMRA CD8 + T Cell Ratio and domain-specific cognition

We also interrogated a novel T cell ratio measure—the ratio of Naïve CD4 + T Cells to EMRA CD8 + T cells. A recent study by Xu et al., evaluated which T cell ratio measures explained the most variance across lymphocyte subsets in the HRS VBS. The ratio of naïve CD4 + : EMRA CD8 + T cells was one of the top ratio measures identified in this study; this ratio measure alone accounted for over 26% of the total variance in lymphocyte subsets [34].

Figure 1G and H display the associations between the ratio of Naïve CD4 + T Cells to EMRA CD8 + T cells and domain-specific cognitive performance across both models. In contrast to the other T cell ratio measures, for this measure a lower value is consistent with a more aged immune profile. We generally observed positive associations for both males and females, especially with regards to general cognitive function and executive function. For example, among both females and males a one SD higher score of the Naïve CD4 + :CD8 + EMRA T Cells was associated with a 0.17 SD unit higher executive function score (Female 95% CI: 0.11, 0.23; male 95% CI: 0.11, 0.25) in Model 1. While effect estimates were attenuated after covariate adjustment in Model 2, confidence intervals continued to exclude the null (Fig. 1G, H, Additional file 1, Table S4). Nonetheless, after correction for multiple comparisons, the FDR-adjusted p-value reached 1. Additional file 1, Table S4).

Associations between the Naïve: memory B cells and domain-specific cognition

Again, while we observed a modest signal suggesting that higher values of the Naïve: Memory B cell ratio were associated with worse function across cognitive domains, the associations were weak and imprecise among both females and males (Fig. 1I, J, Additional file 1, Table S5).

Associations between CMV IgG antibodies and domain-specific cognition

We also examined associations between standardized CMV IgG antibody levels and domain-specific cognition (Fig. 1K, L, Additional file 1, Table S6). Among females, we observed associations between CMV IgG and all cognitive domains except for orientation. Similar to the other measures, the strongest association was observed for executive function where a one SD higher level of CMV IgG was associated with a 0.06 SD unit lower executive function score (95% CI: −0.07, −0.05). The association was attenuated after covariate adjustment in Model 2 but the confidence intervals continued to exclude the null. Among males, we observed statistically significant association between CMV IgG and both general cognitive function and executive function with effect estimates of a similar magnitude to females. However, effect estimates were attenuated after covariate adjustment in Model 2 and while confidence intervals continued to exclude the null, after correction for multiple comparisons, the FDR-adjusted p-value reached 1.

Of note, not all HRS participants were CMV seropositive. Given the strong association between CMV and expansion of terminally differentiated effector memory T cells, we ran sensitivity analyses restricting the sample to only individuals that were CMV seropositive. The effect estimates were attenuated and estimated with more imprecision, but the overall inferences remained the same (see Additional file, Table S7).

Discussion

In a national population-based sample of older US adults, we observed associations between immunosenescence and general and domain-specific measures of cognitive function. However, the magnitude and precision of these associations depended on biologic sex and the specific immune marker analyzed. Specifically, the CD4 + EMRA: Naïve T Cell Ratio, the CD4 + Naïve: CD8 + EMRA Ratio, and CMV IgG exhibited the most consistent associations particularly with regards to general cognitive function and the executive function domain. We also broadly observed an attenuation of the associations after adjustment for socioeconomic and health factors, which is consistent with research showing that socioeconomic and health factors are correlated with both immune system aging and cognitive aging [35–37]. Notably, we observed fewer and weaker associations among males than females. These findings underscore the importance of examining sex-specific pathways in the relationship between immune aging and cognition.

Given the modest sample size and the novel population-health framing of this work, we interpret our findings with emphasis on the magnitude, direction, and coherence of associations across related immune markers and cognitive domains. Although several associations did not remain statistically significant after FDR correction, effect estimates were generally directionally consistent across markers and domains. We therefore do not interpret FDR-adjusted non-statistically significant findings as definitive evidence of an absence of association. Instead, we view these patterned and biologically aligned results as hypothesis-generating and warranting replication in larger samples with more granular immune characterization.There is a growing literature showing associations between markers of peripheral immunosenescence and general measures of brain health. Other studies have similarly found mixed or null findings with regards to the association between peripheral immune function and brain health [38–40]. Specifically, studies in the HRS using screening measures of global cognitive function such as the HRS-TICS have found null or mixed associations [41, 42]. These studies have been limited by the number and specificity of the immune markers available, focusing on, for example, single infections such as CMV or inflammatory cytokines. Another shortcoming is that cognitive outcomes are frequently limited to diagnostic categories (e.g., impaired or unimpaired or demented vs. nondemented) or rely on a single score based on a mental status screen such as the Mini Mental State Exam (MMSE) or TICS, rather than a comprehensive test battery assessing multiple domains such as the one used in HCAP [38, 42]. The current study underscores the importance of using a more detailed multi-domain cognitive evaluation in combination with a complex of immune markers sensitive to the aging process to characterize the complex relationships between cognition and peripheral immune function. We also recognize that, given some associations were null, it is possible that there are other salient indicators that were not part of our immunological panel that are important predictors of brain health.

Examining domain-specific cognition is crucial for advancing our understanding of the complex relationships between biological processes and cognitive function. Different cognitive domains, such as memory, executive function, language, and orientation are associated with different neural networks, which in turn may be differentially affected by various biological, environmental, and social factors [43, 44]. By focusing on specific cognitive domains, we can uncover more nuanced patterns of cognitive decline that may be obscured when focusing on more general measures of cognition [45] and mental health status screens. This approach can also identify early and domain-specific changes that may co-occur with broader changes in cognition, potentially informing future research on targeted interventions and biological mechanisms [46, 47].

Emerging evidence suggests several plausible biological pathways linking peripheral immunosenescence to age-related cognitive impairments, particularly with regard to executive function. For example, CD4 + EMRA cells, which represent terminally differentiated effector memory T cells, expand in response to persistent viral infections such as CMV and are a hallmark of immune aging [48, 49]. These cells secrete pro-inflammatory cytokines, exhibit limited replicative capacity, and contribute to systemic immune activation [50]. CMV IgG, a marker of lifelong antigenic stimulation, has been associated with elevated inflammatory profiles and skewed T cell phenotypes, including increased proportions of senescent and exhausted cells [51, 52]. Prior work has linked CMV antibody levels and inflammatory markers (e.g., CRP, IL-6, TNF-α, and IL-17A) to executive function performance, suggesting that inflammation may partially mediate associations between immunosenescence and cognition [41, 53, 54].

Importantly, studies in both human and non-human primates have shown that inflammation alters frontostriatal circuity critical for executive function. For example, experimental and observational studies demonstrate that elevated CRP is associated with reduced corticostriatal connectivity in prefrontal regions, implicating dopamine-dependent neural pathways central to executive functioning [55]. Consistent with this finding, higher peripheral inflammation has also been linked to reduced functional connectivity within central executive networks in humans [56]. Additionally, activated CD4 + T cells, including senescent subsets, have been shown to cross the blood–brain barrier under inflammatory conditions and localize in perivascular spaces, potentially altering neuronal signaling or contributing to neuroinflammation [57, 58]. Further, these processes have been shown to preferentially disrupt select regions of the brain such as the prefrontal cortex, an area integral to executive function and particularly vulnerable to inflammatory and vascular insults in later life [59, 60]. Together, these biological insights help contextualize our findings and underscore the need for future research on immune–CNS interactions in aging.

We also found that the observed associations were substantially attenuated in the model adjusting for the hypothesized confounders including chronic conditions, smoking status, education, wealth, and depression. However, these findings should be viewed with caution as some of these factors, particularly the health-related indicators, likely reflect shared determinants of both immune aging and cognitive function rather than purely confounding variables. Given the cross-sectional study design, we cannot determine the temporal order or clearly distinguish between confounding and pathway-related processes. Nonetheless, these findings offer useful hypothesis-generating evidence to inform future research on the complex social, economic, and health processes underlying these pathways.We hypothesized that we would observe strong sex differences based on research showing sex disparities in the relationship between immune function and aging-related outcomes [61–63]. We found that, while in males we observed associations in the same general direction as observed in females, the associations were often weaker and more imprecise. Future studies are needed that can better adjudicate sex differences in the processes. There are well-established biological differences in immune regulation by sex, which are shaped by myriad biological processes including genetic and endocrine mechanisms across the life course. Females generally mount more robust innate and adaptive immune responses, including stronger interferon signaling and humoral responses, which may enhance pathogen clearance but increase susceptibility to autoimmunity [64]. Sex hormones also influence adaptive immune composition; for example, estrogen’s effects on thymic output and peripheral differentiation are associated with higher numbers of functional CD4 + T cells among females, potentially delaying aspects of immunosenescence [65]. Importantly, immunosenescence unfolds with, and is sensitive to, gendered social contexts. Structural gender inequities may shape both differential exposure to trauma and stress and differential access to resources which may shape immune trajectories across the life course. Future studies that can distinguish between biological sex and gender are critical for fully understanding associations between immunosenescence and cognition. Our finding that the direction of the association was similar across sex suggests that the downstream cognitive consequences of peripheral immune aging may reflect broadly shared mechanisms across sexes in later life. Prior work has demonstrated sex-specific associations between immunosenescence and disability or mortality [61, 62], whereas findings in cognition have been mixed, with some studies reporting no differences and others observing higher mean cognitive performance among women [66–68]. Of note, the patterns described above largely reflect studies that rely on a binary classification of sex. Sex is not binary, however, and there is a dearth of studies that examine the unique immune dynamics individuals outside of this binary framework (e.g., transgender individuals, intersex individuals).

This study has several potential limitations that warrant consideration. First, despite leveraging a robust set of immune measures, we may still be missing key markers that could further our understanding of the relationship between the dynamic peripheral immune system and cognition such as those capturing compositional and functional heterogeneity. Furthermore, more studies are needed that can examine how and the extent to which the peripheral immune system interacts with the CNS immune system. Second, this study is cross-sectional, limiting our ability to infer temporal or causal relationships between immune function and cognition. Longitudinal data spanning the life course is essential to better understand how immune system changes over time are linked with domain specific cognitive trajectories and to identify potential critical periods for intervention. Future studies incorporating a life course framework and broader immune profiling will be critical for advancing this field. Third, the HRS-HCAP cohort may be subject to survival bias, as participants must have been healthy enough to survive to enrollment. Consequently, we may have missed individuals for whom the relationship between immune function and cognition is most pronounced. Relatedly, not all individuals participating in the 2016 VBS were selected for HCAP. We compared respondents included in both VBS and HCAP to those in the VBS only on key covariates and found that dual participants were older on average (74 years vs. 65 years) and exhibited higher mean values of immune markers indicative of more advanced immune age. These differences are consistent with the age-restricted design of HCAP, which was limited to adults aged 65 years and older. Although this dual selection shapes the composition of the overlap sample, it is unlikely to bias association estimates, as selection into HCAP was driven primarily by age rather than immune status.

Fourth, while we used latent factor scores intended to capture specific cognitive domains, the cognitive tests contributing to these domains do not map exclusively to discrete brain regions; rather, individual test performance scores likely reflect activity across multiple brain regions. Finally, we employed a modeling approach that adjusted for the number of chronic conditions, which may include variables on the causal pathway between immune aging and cognitive outcomes. As such, these models may underestimate the total effect of immunosenescence by partially adjusting for downstream mediators such as cardiovascular disease or cancer. Addressing these limitations in future research will be critical for advancing our understanding of the immune system's role in aging-related cognitive decline.

Conclusions

There is mounting evidence that aging-related changes in the peripheral immune system, including shifts in CD4 + T cell phenotypes and chronic antigenic stimulation by latent viruses such as CMV, play a fundamental role in the development of aging-related diseases. Our findings suggest that these immune alterations may also influence cognitive health, particularly executive function—a domain closely tied to functional independence and early cognitive decline. The consistent associations observed with CD4 + EMRA:Naïve ratios and CMV IgG levels, while modest in magnitude, provide biologically plausible signals that peripheral immunosenescence may contribute to neurocognitive aging. Notably, the strongest and most precise associations were observed among females, adding to a growing body of research suggesting there are sex differences in immunosenescence and its downstream consequences. These findings highlight the immune system as a potential early contributor to cognitive vulnerability and underscore the importance of integrating immunosenescence into the study of ADRD and cognitive decline.

Methods

Study sample

Study design and sample population

The study sample was derived from the U.S. HRS, a national population-based, multistage area probability household survey of Americans aged 51 years and older and their spouses. Study details have been previously described [69]. Sampled persons have been re-interviewed biannually since 1992, and younger cohorts have been added to the original sample every 6 years to maintain its representativeness of Americans over age 50 over time [70]. The HRS core public-use data are described in [84].

Our sample was based on a 2016 sub-study within the ongoing HRS, the Harmonized Cognitive Assessment Protocol (HCAP) [71]. The HCAP collected an in-depth battery of cognitive and neuropsychological assessments and informant reports to better characterize cognitive function in older adults. HCAP eligibility was determined based on the sample of HRS panel respondents born in or before 1952 (e.g., 65 years or older in 2016) who completed the 2016 core HRS interview. From those, one half of the uncoupled respondents were randomly selected and one respondent from each coupled household was randomly selected. The final HCAP sample included 3,496 cases (a 79% response rate among those invited to participate). The HCAP restricted dataset is available through the HRS [83].

From the HCAP sample, we selected individuals who also had valid immunological data from the 2016 HRS Venous Blood Sample (VBS). A total of 9,933 individuals participated in the 2016 VBS and had complete and valid test results. Of those, 2,423 were also included in the HCAP. We excluded 208 individuals who were missing data on key immune markers and one individual who was missing data on covariates of interest. Our final analytical sample size was 2,215 individuals (91% of those eligible). Immune marker data are from the 2016 HRS Venous Blood Study sensitive dataset [82].

Of note, the subset of respondents participating in both VBS and HCAP reflects additional age-based selection, as HCAP was restricted to individuals aged 65 years and older and sampled under a distinct framework. A comparison of individuals in both the HCAP and VBS to those only in the VBS is provided in Additional file, Table S8.

Measures

Primary exposures

Immunophenotyping was completed for the entire VBS sample. Details on the processing and laboratory assays, as well as the cell phenotyping protocols have been recently published [72–74]. Of note, the immunophenotyping was conducted by the HRS team. While the HRS provides derived immunophenotyping variables based on centralized, standardized protocols, detailed gating plots or full gating schematics are not publicly available.

We focused on six measures of immune function to conduct a more comprehensive assessment of peripheral immune function [75] [37, 62]. Following several recent studies, we constructed the following measures: (1) CD8 + T cells: CD4 + T cells; (2) EMRA CD4 + T cells: Naïve CD4 + T cells; (3) EMRA CD8 + T Cells: Naïve CD8 + T cells; (4) Naïve CD4 + T cells: CD8 + EMRA T cells; (5) IgD + Memory B cells: Naïve B Cells; and (6) continuous CMV IgG antibodies [34, 36, 37, 76, 77]. These ratio measures have been previously used in clinical and population-based research to examine patterns of immune aging [37, 61, 76].

Continuous CMV immunoglobulin (Ig) G antibody levels are often used as another indicator of overall immune fitness, where higher CMV IgG antibody levels reflect a waning ability of the immune system to keep the virus in a quiescent state and may correlate with frequency of reaction of the virus [78, 79]. CMV IgG levels were measured in serum using the Roche e411 immunoassay analyzer.

All immune markers used in the analyses were standardized. For full details on which variables were used to create each ratio measure and how each cell subset was determined see the Additional file 1, Table S9.

Primary outcomes

We utilized cognitive domain-specific factor scores estimated by Gross et al. to examine associations between immunosenescence and each of the five cognitive domains. For further details on how these scores were derived, see Gross et al., 2023. In brief, they assigned each cognitive test that was part of the HCAP battery to a particular domain according to both neuropsychological theory and psychometric testing showing which tests correspond best to each domain. Then, using a confirmatory factor analysis, they derived domain-specific factor scores for general cognitive function and specific domains of memory, executive function, orientation, and language, each derived from the covariance of cognitive test items corresponding with each domain [68]. The general and domain-specific latent factor scores are each standardized to have a mean of 0 and a standard deviation of 1, with higher scores representing better cognitive function. The individual tests that comprise each domain are listed in Additional file, Table S10.

Covariates

We constructed a directed acyclic graph (DAG) to select a control set of variables that represented confounders of the relationships of the peripheral immune markers and cognitive function (see Fig. 2). Based on the DAG, we included the following measures self-reported in the study interview: age (continuous, years), educational attainment (categorical), household net wealth as an indicator of financial resources (continuous), smoking status (ever, current), depressive symptom score (continuous, according to the eight-item Center for Epidemiologic Studies Depression Scale CES-D), and number of physician-diagnosed comorbidities (continuous, out of the eight conditions of high blood pressure, diabetes, cancer, lung disease, heart disease, stroke, psychiatric conditions, and arthritis).

Fig. 2.

Directed acyclic graph (DAG) depicting the hypothesized relationships between immunosenescence and domain-specific cognitive function, and potential confounders. Variables outlined in red show the hypothesized confounding variables that are part of the minimally sufficient adjustment set

Statistical analyses

Descriptive analyses were conducted using SAS survey procedures and were weighted using HRS VBS sampling weights to account for non-response and the complex survey design. Consistent with our prior work, [61–63] we stratified our statistical models by sex as there are known differences in immune function by biological sex [80, 81].

Associations between each immune marker and the general and domain-specific cognitive score were modeled using a series of linear regression models. We present point estimates and 95% confidence intervals (CI) to characterize both the magnitude and precision of the observed associations. All statistical tests were two-sided. False discovery rate (FDR) corrections were applied to control for multiple comparisons. However, interpretation emphasized the direction, size, and plausibility of the associations, with a focus on estimation rather than dichotomous significance thresholds.

We employed the following model building strategy: Model 1 was the null model in which we only controlled for chronological age. Model 2 included age in addition to key covariates that we hypothesized could be confounders of the association between immune function and cognition as directed by our DAG.

Analyses were carried out in SAS v9.4 and R v4.6 statistical software. This study was approved by the Institutional Review Board at the University of Michigan, HUM 00207292.

Abbreviations

AD

Alzheimer’s disease

ADRD

Alzheimer’s disease and related dementias

CES-D

Center for Epidemiologic Studies Depression Scale

CI

Confidence interval

CMV

Cytomegalovirus

CNS

Central nervous system

CRP

C-reactive protein

CSF

Cerebrospinal fluid

DAG

Directed acyclic graph

EMRA

Terminally differentiated effector memory cells

FDR

False discovery rate

HCAP

Harmonized Cognitive Assessment Protocol

HRS

Health and Retirement Study

IgD

Immunoglobulin D

IgG

Immunoglobulin G

IL

Interleukin

MMSE

Mini-Mental State Examination

SD

Standard deviation

TICS

Telephone Interview for Cognitive Status

TNF-α

Tumor necrosis factor alpha

VBS

Venous blood sample

Authors’ contributions

GAN, AEA, KL, LK, and RCS conceptualized the manuscript. KPZ carried out the analyses. GAN drafted the manuscript. GAN, RCS, KZ, LK, KL, BL, CL, and AEA contributed to the interpretation of findings. All authors assisted with editing and revising the manuscript. All authors read and approved the final manuscript.

Funding

Research reported in this publication was supported by the National Institute on Aging (NIA) of the National Institutes of Health (NIH) under awards R01 AG075719 (GAN and AEA) and R01 AG070953 (LCK, KZ, KML), U24 AG065182 (LCK, KML), and U01 AG009740 (LCK, KML). The content is solely the responsibility of the authors and does not necessarily represent the official views of the NIH.

Data availability

The data used in this study are from the Health and Retirement Study (HRS), a nationally representative longitudinal study of older adults in the United States. The datasets analyzed include sensitive, restricted-use files that are not publicly available due to privacy and confidentiality protections. These data are available to qualified researchers through application and approval of a Data Use Agreement (DUA) with the HRS (https://hrs.isr.umich.edu/data-products/restricted-data).

Declarations

Ethics approval and consent to participate

This study was approved by the Institutional Review Board at the University of Michigan, HUM 00207292.

Informed consent was obtained by the HRS study team for each participant.

Consent for publication

NA.

Competing interests

The authors declare no competing interests.