Peripheral Inflammation and Executive Function among Community Samples Across the Lifespan: A Systematic Review and Meta-Analysis

Search Strategy

This review was registered in PROSPERO (CRD420250655231) prior to data extraction and followed PRISMA 2020 guidelines. Electronic searches in Web of Science, PubMed (including MEDLINE), and PsycINFO were conducted on February 4, 2025, for studies that examined the link between inflammation and executive functioning (see full list of search terms in Supplementary Information). Searches also included filters for studies that included human subjects and were published and/or presented in the English language. To mitigate the file-drawer problem, we included unpublished dissertations via the PsycINFO database.

Selection Criteria

Figure 1 provides a detailed overview of the systematic search and the number of studies excluded at each review stage. Following the removal of duplicates, four trained coders (AS, JS, FP, OB) independently reviewed the titles and abstracts of all records, with the first author (AS) reviewing all records. To meet inclusion criteria, the study needed to (a) report a concurrent association between a measure of peripheral inflammation and a self-report or behavioral measure of executive function, (b) examine community-based samples representative of the population, (c) be available in the English language, and (d) be empirical in nature. We defined “community” samples as non-medical and non-psychiatric samples that were generally representative of the specific community under investigation, even if not recruited randomly (e.g., convenience sample using undergraduates, population-based sample that over-sampled a specific racial minority group, risk-enriched sample for memory decline but otherwise representative). We excluded studies that (a) recruited individuals with a medical diagnosis, psychiatric diagnosis, or who were part of a specific population (e.g., postpartum women, following a cardiac episode), (b) based on a clinical trial and/or acute intervention (e.g., vaccination, immune challenge, psychosocial stress test), (c) did not include a validated self-report or behavioral measure of executive function (see Table 1 for a list of included measures of executive function¹), (d) did not collect estimates of inflammatory markers and/or did not measure inflammatory proteins in the periphery (e.g., lumbar puncture), (e) measures of inflammation and executive function were not given at the same time point, (f) did not use human subjects, and (g) were not empirical studies (e.g., review, case report). Further, we excluded studies when outcome data still were not available following two attempts to contact corresponding author(s) over a one-month period, or data only were provided in conference abstracts, editorials, and/or letters, but full methodological information could not be extracted. When manuscripts employed data from the same cohort, the manuscript with the largest sample size was included.²

During the initial calibration phase, the first author (AS) independently coded the same set of 50 records as each coder (JS, FP, OB) to determine inter-rater reliability. Coders demonstrated substantial agreement during this phase, with Cohen’s κs ranging from 0.64–0.79. Articles not excluded during this phase were subject to full-text review by the first author (AS) and one other coder (JS, FP, OB) to determine whether the article met inclusion criteria. Cohen’s κs during the full-text review stage also demonstrated strong agreement (0.71–0.75). All discrepancies were discussed and resolved through weekly group consensus. Study quality and risk of bias were determined for all included studies using a modified version of the Newcastle Ottawa scale for cross-sectional studies (Peterson et al., 2011, see Table 1 in Supplementary Information for assessment protocol) and modeled off previous meta-analytic reviews (Mac Giollabhui, Ng, et al., 2021; Chat et al., 2024).

Data Extraction

All coding and data extraction were performed in Covidence (Covidence systematic review software, 2025). For each study, two authors (AS, JS, FP, OB) independently extracted study information (i.e., author, year of publication, author email address, country, analytic sample size, inclusion and exclusion criteria) and methodological information, including (1) study design (i.e., cross-sectional or longitudinal design with cross-sectional assessments), (2) demographic characteristics of the sample cohort (age, sex, race, socioeconomic status, average body mass index [BMI]), (3) executive function measure, and (4) inflammatory marker assayed (e.g., IL-6, CRP) and information concerning inflammatory data collection and analysis (e.g., drawn from plasma/serum/saliva; kit used; time of day/fasting; how authors handled outliers).

Additionally, the following statistical parameters were collected: (1) method of analysis (i.e., bivariate correlation, partial correlation, linear regression), (2) covariates included (e.g., medication, BMI), and (3) effect size. The outcomes of interest were unadjusted (i.e., Pearson’s correlation coefficients) and adjusted linear associations (i.e., partial correlation and standardized regression coefficients) between inflammatory biomarkers and executive function. If the appropriate effects were not reported but could be extrapolated using other information presented in the manuscript (e.g., standardizing regression coefficients), the first author completed this task. When possible, we extracted both the unadjusted model of the association between inflammation and executive function and the model with the most complete set of covariates. For manuscripts that presented multiple effect sizes due to multiple measures of inflammation and/or executive function, unique effect sizes were extracted to represent each inflammation-executive function relationship. For manuscripts that presented insufficient data, corresponding authors were contacted twice to obtain relevant data. If the corresponding author did not respond via email within four weeks of the second contact attempt, the study was excluded from analysis. However, if authors did not respond to correspondence and appropriate statistics were not reported due to the creation of composite scores that included non-relevant indices (i.e., composite scores that combined inflammatory and non-inflammatory biomarkers and/or executive function measures with other measures of general cognition) or due to presenting associations among only a subset of the sample, coders qualitatively described results to be included for discussion. Of note, all articles meeting inclusion criteria were reviewed by the first author (AS) to determine whether there was sufficient data for the meta-analysis.

Analytic Approach

All analyses were conducted in R (version 2024.21.1) using the “metafor” package (Viechtbauer, 2010). Random-effects meta-analyses were conducted to examine the unadjusted and adjusted associations between inflammatory biomarkers and executive functioning, as these models do not assume a common effect size among samples (Borenstein et al., 2022). Given that 76% of studies included in the present meta-analysis reported more than one inflammation-executive function association, we conducted analyses for inflammatory markers most represented in the literature (i.e., CRP and IL-6). If studies provided multiple effect sizes with CRP or IL-6, effect sizes that represented the most general/non-specific facet of executive functioning were selected. Next, we conducted meta-regressions to examine whether the strength of these associations differed when accounting for demographic information from the selected sample. Age and BMI were centered on the mean at baseline to facilitate easier interpretation.

Finally, to better understand whether the magnitude of effect sizes differed by executive functioning domain, secondary analyses using executive function domains were conducted. Measures of executive function were classified into one of four groups: (1) general/non-specific, (2) updating/working memory, (3) inhibition, and (4) cognitive flexibility (see Table 1 for classification information). Studies that created composite scores were classified as general/non-specific, unless the measures all assessed the same executive function domain. If a study reported effect sizes for the same measure of executive function with more than one inflammatory biomarker, effect sizes using the most commonly reported inflammatory biomarker (i.e., CRP) were selected to maximize consistency.

Prior to analyses, all correlation coefficients were transformed to Fisher’s Z; following meta-analysis, Fisher’s Z values were transformed back to correlation coefficients to aid in interpretation. Heterogeneity between studies was assessed using Cochrane Q and the inconsistency index (I²). Funnel plots were visually analyzed, and Egger’s regression tests were conducted to determine the presence of publication bias. Finally, sensitivity analyses were conducted to determine whether (1) results differed when verbal fluency measures were removed and (2) study quality or mean level of inflammation moderated the inflammation-executive function relationship.

Overall Association

Secondary Analyses

Sensitivity Analyses

Our results largely did not differ when analyses were replicated among high-quality studies (i.e., studies with quality scores greater than or equal to 4). However, the adjusted association between inflammation and updating performance was rendered non-significant once low-quality studies were removed (r = −0.022, SE_z = 0.013, p = 0.085). Supplementary Table 2 contains full model information.

Despite dispute over the nature of verbal fluency tasks as a measure of executive functioning or language ability (Aita et al., 2019; Whiteside et al., 2016), verbal fluency was included in the present meta-analysis as a proxy for cognitive flexibility. To assess verbal fluency, participants typically are asked to provide as many responses as possible that begin with a certain letter (i.e., phonemic verbal fluency) or fit a certain category (i.e., semantic verbal fluency) within a specified timeframe, indicating that verbal fluency requires both language (e.g., lexical access ability) and executive function (e.g., switching) skills. To examine whether the associations between inflammatory proteins and executive functioning differ when tasks with a strong language component were removed, sensitivity analyses were conducted that removed verbal fluency measures. As shown in Supplementary Table 3, adjusted and unadjusted associations between inflammatory proteins and executive functioning, including cognitive flexibility, remained significant once measures of verbal fluency were removed. Neither mean age, BMI, nor percentage of the sample identifying as female significantly moderated inflammation-executive function relationships.

Finally, to determine if the magnitude of the association between inflammatory biomarkers and executive function performance differed depending on the mean level of inflammation, post-hoc analyses that explored mean level of inflammation as a moderator were conducted. As shown in Supplementary Table 4, results did not differ when the mean level of CRP or IL-6 for the overall sample was entered as a moderator.

Are markers of peripheral inflammation associated with general executive function skills?

Based on a meta-analysis of 41 studies, surveying 255,539 community individuals across the lifespan, CRP and IL-6 showed small, but reliable, associations with executive function skills. That is, higher levels of CRP and IL-6 were associated with worse executive functioning among community participants. For both inflammatory markers, these associations were consistent across models that accounted for potential covariates. As few studies examined other biomarkers of peripheral inflammation from the innate immune system (e.g., TNF-α, IL-8, IL-1β), quantifying these associations via a meta-analytic investigation was not feasible.

The present review supports the notion that higher levels of pro-inflammatory proteins in the periphery, namely CRP and IL-6, are associated with worse executive function performance, even among individuals not recruited based on a specific psychiatric or medical condition. Our results align with previous reviews and meta-analyses that document inflammation-cognition relationships among older individuals (Fard et al., 2022) and following stimulated inflammatory response (Mekhora et al., 2024). However, the present study expands upon this work by focusing on measures of executive function, specifically, to suggest that heightened levels of peripheral inflammatory biomarkers have important implications for executive functioning skills. For example, given that these proteins can communicate across the blood-brain barrier, it is possible that markers of peripheral inflammation influence the connectivity in brain regions essential for this higher-order cognitive process (e.g., dorsolateral prefrontal cortex to posterior parietal cortex connectivity; Nusslock et al., 2019). This interpretation is aligned with evidence from preclinical and human neuroimaging studies that pro-inflammatory biomarkers circulating in the periphery may influence activity in brain networks thought to underpin executive functioning (e.g., McAfoose & Baune, 2009; Mekhora et al., 2024; Nusslock et al., 2019; Walker et al., 2020).

Importantly, it should be noted that the present review focused on concurrent relationships, so the role of inflammatory markers as predictive or causal mechanisms underlying executive dysfunction cannot be determined. That is, although the present study supports the notion that heightened inflammation and executive function difficulties are correlated, the precise nature of this association is not clear. It may be possible that low-grade inflammation underpins executive dysfunction, possibly indicating a causal role. Support for this possibility stems from longitudinal studies of inflammation and general cognitive decline (Feng et al., 2023; Long et al., 2023; Mac Giollabhui et al., 2024; J. Yang et al., 2015). However, it also may be possible that poor executive functioning enhances stress reactivity that induces an elevated immune response (Slavich & Irwin, 2014; Williams et al., 2009), or a third, underlying factor contributes to both heightened inflammation and executive function difficulties (e.g., obesity, sleep deprivation; Miller & Spencer, 2014; Yuan & Wang, 2025).

The current study is among the first to consider both unadjusted and adjusted associations between inflammation and cognitive ability, documenting consistent relationships between specific markers of peripheral inflammation (i.e., CRP, IL-6) and executive functioning when considering potential confounding factors. Of note, we detected substantial between-study variability in effect sizes, even among adjusted associations, which may be due to differences in sample characteristics. As shown in Table 2, studies included in the present meta-analysis not only varied widely in the number and type of covariates considered, but also in the definitions of “community” samples. To consider generally representative individuals of all ages, we included studies that recruited based on convenience (e.g., civil service members; Calice-Silva et al., 2021), at-risk status (e.g., genetic predispositions for Alzheimer’s disease; Low et al., 2021), and/or other demographic factors (e.g., race; Arce Renteria et al., 2020). Albeit inclusive, this approach may have resulted in sample compositions that substantially differed on important factors known to influence both inflammation and executive functions (e.g., socioeconomic status; Lawson et al., 2018; Muscatell et al., 2020). Although our findings provide support for the notion that higher levels of CRP and IL-6 in the periphery are reliably related to poor executive functioning among non-medical and non-psychiatric samples, it is likely that the magnitude of these relations differs by context and specific sampling conditions. As such, caution should be exercised when interpreting the overall inflammation-executive function effect size.

Is the relationship between inflammation and executive function dependent on age, adiposity, or sex?

Despite significant heterogeneity detected in inflammation-executive function associations, we did not find evidence for the moderating roles of age, adiposity, or sex assigned at birth. That is, contrary to our hypotheses, we did not find support for the moderating role of age in either adjusted or unadjusted models of CRP or IL-6 and executive functions. As normative aging is associated with general increases of pro-inflammatory markers in the periphery and central nervous system – a process known as inflammaging (Fard et al., 2022; Fulop et al., 2017) – coupled with evidence that older individuals perform worse on measures of executive function compared to younger individuals (e.g., Kirova et al., 2015), it was surprising that we did not find evidence that the inflammation-executive function relationship strengthens as individuals age. However, it should be noted that the average age was over 60 years old for approximately 40% of included studies, whereas only four studies included individuals aged 18 or younger (i.e., Caldú et al., 2023; Mac Giollabhui et al., 2020; Prunell-Castañé et al., 2024; Smith et al., 2021). Additional work is needed to determine whether these results would differ with the inclusion of more samples with younger participants.

Adiposity, proxied by BMI, also did not strengthen the association between higher circulating CRP or IL-6 and general executive functioning skills. While it may be possible that adiposity holds a mediating (rather than a moderating) role in this relationship, it also may be possible that we did not find support for adiposity as a significant moderator due to a limited variability of BMI, given that the present review explicitly excluded studies that recruited based on a medical diagnosis that could co-occur with obesity (e.g., hypertension, type 2 diabetes). To explore whether these relations are specific to samples where individuals have higher levels of CRP and IL-6, on average, we conducted post-hoc analyses in which the mean level of inflammatory biomarker was included as the moderating variable. These meta-regressions revealed that higher average circulating levels of inflammatory biomarkers did not significantly strengthen the relationship between inflammation and executive functioning (see Supplementary Table 4). Of note, approximately 25% of included studies that collected measures of CRP excluded observations with CRP values greater than or equal to 10 mg/L, as these individuals may be experiencing acute infection or illness. As noted by Mac Giollabhui, Ellman, and colleagues (2020), this approach may be problematic because (1) many factors, particularly adiposity and smoking, are associated with circulating levels of CRP that are >10 mg/L and not related to acute illness (e.g., diet, sleep, and socioeconomic status), and (2) individuals of true interest to researchers investigating human health may be excluded. Importantly, average reported levels of CRP may be unduly influenced by this cutoff – suggesting that additional work is needed to determine whether these relations only hold among samples with higher levels of peripheral inflammation.

Finally, when considering the role of sex in the relationship between inflammation and general executive functions, we did not find evidence that sex significantly influences the inflammation-executive function association. Our findings are not surprising given the inconsistent literature on sex-specific differences in the association between inflammation and cognitive functioning. For example, a previous review documented an association between CRP and poorer general cognition (measured by the Mini-Mental State Examination) among elderly men, but not women (Hedges et al., 2012), whereas a more recent study found that the association between CRP and cognition was stronger among female community participants compared to male participants (B.-A. Chen et al., 2023). Moreover, West and colleagues (2020) found no evidence of sex-specific differences in the association between inflammatory markers and a decline in executive functioning, specifically. Although we did not find support for the moderating roles of age, BMI, mean level of inflammation, or sex in the present review, future researchers should carefully consider these factors and clearly define and present the characteristics of the sample under study.

Methodological considerations in the assessment of executive functions

Secondary analyses determined that higher levels of inflammatory markers were negatively associated with scores on all subdomains of executive function. That is, our results suggest that higher circulating inflammatory biomarkers are related to worse performance on tasks where individuals must (1) hold and manipulate information in one’s mind (i.e., working memory/updating), (2) switch between tasks (i.e., cognitive flexibility), and (3) deliberately control and suppress a dominant, automatic response (i.e., inhibition). These results were consistent when possible confounding factors were considered (e.g., race, education level, history of traumatic brain injury), indicating that each subdomain of executive function may be associated with, or potentially influenced by, heightened inflammation. Furthermore, the association between peripheral inflammation and cognitive flexibility did not seem to be driven by measures that required a strong language component, as our results remained significant even when verbal fluency measures were not considered. Given that executive functions are thought to consist of distinct functions that share a commonality because of coordination of the prefrontal cortex – a pattern known as “unity and diversity” (Friedman & Miyake, 2017) – it is unsurprising that we found similar sized associations between inflammation and each executive functioning domain (unadjusted rs range from −0.065 to −0.087; adjusted rs range from −0.017 to −0.025). However, as only a handful of studies (ks = 7–12) explicitly quantified the concurrent relationship between each domain of executive function and inflammation, coupled with our findings that the relationship between inflammation and updating attenuated when adjusting for study quality, additional research is needed to determine whether all facets of executive function are associated with inflammation.

Taken together, our review highlights the robust variability in the measurement and classification of executive functions when assessed alongside inflammation. Although our meta-analysis included studies in which composite cognitive variables were created from validated executive function assessments, studies that employed composite measures derived from behavioral assessments of executive function and other cognitive domains (e.g., long-term memory) were excluded. Table 3 summarizes the findings from studies that used composite measures that tapped into executive functioning in addition to other cognitive domains. Of the studies that were excluded from our meta-analysis due to the creation of cognitive composite measures, seven reported non-significant relationships between markers of peripheral inflammation and composite cognitive measures, all of which contained at least one measure of executive functioning (Botoseneanu et al., 2023; Estrella et al., 2024; Hajjar et al., 2018; Hu et al., 2020; Koblinsky et al., 2023; Noble et al., 2010; Wennberg et al., 2019). Only two studies documented an association between inflammation and cognitive composites (Heringa et al., 2014; Tegeler et al., 2016), whereas three others demonstrated mixed support (Baune et al., 2008; Boots et al., 2022; Karlamangla et al., 2014). Taken together, these studies indicate that heterogeneity in study design may contribute to the appearance of weak or mixed associations between inflammation and executive functions. Although future researchers may be inclined to create composite scores and/or latent variables derived from computational models (e.g., factor analysis), our findings strongly encourage researchers to consider which cognitive function is of interest and to use putative theory to guide decision-making. If composite measures are employed that tap into multiple cognitive domains, researchers also should report associations with individual tasks to provide a clearer picture.

Strengths and Limitations

Surveying 255,539 individuals recruited from broadly representative communities, the present review serves as the first systematic picture of the association between markers of peripheral inflammation and executive function performance across the lifespan. However, interpretations drawn from this large, meta-analytic review are not without limitation. First, this review only included cross-sectional associations, so causal inferences were not possible. Given the complex nature of the innate immune system and neurobiology underpinning executive functioning, future research should employ rigorous, prospective designs to better elucidate this relationship. Second, we were underpowered to conduct meta-regressions using inflammatory markers other than CRP and IL-6; thus, interpretations are limited to these biomarkers. Finally, studies varied widely in criteria for what constitutes a “community” sample, in the measurement and collection of inflammatory biomarkers (e.g., fasting conditions, kit used), in other potential moderating variables and/or sources of inflammation (e.g., amount of sleep, current level of stress, history of mental health concerns), and in the covariates considered in adjusted analyses; all of which contribute to increased heterogeneity among effect sizes that influence the strength of our results.

Conclusion

In line with the growing literature on the association between peripheral inflammation and cognition, this meta-analytic review bolsters support for the idea that markers of low-grade inflammation and executive function performance are concurrently related across the lifespan. Even among individuals recruited without major medical, psychiatric, or neuropsychological disorders, higher levels of CRP and IL-6 were associated with worse executive function performance across a variety of tasks. To draw further conclusions as to the exact nature of this relationship, additional cross-sectional work is needed because heterogeneous study designs and methodologies limited our ability to draw definitive conclusions as to whether this relationship is context dependent and/or specific to a domain of executive functioning. Longitudinal designs that assess community individuals across development also are needed to enhance our understanding of the temporal nature of the inflammation-executive function relationship. Further research into this association not only may inform current cognitive interventions but also advance detection and prevention efforts.