Anxiety changes how inflammation marker relates to acute stroke severity

Among adults with mild-to-moderate ischemic stroke, the relationship between the blood inflammatory marker macrophage inflammatory protein-1α, or MIP-1α, and acute neurological impairment depended on the severity of post-stroke anxiety symptoms. A cross-sectional study of 263 patients found that MIP-1α was inversely associated with National Institutes of Health Stroke Scale, or NIHSS, scores among people with lower anxiety symptoms, while the analysis involving anxiety and MIP-1α together showed a significant statistical interaction. The result suggests that measuring an inflammatory marker alone may miss clinically relevant variation in how that marker relates to stroke severity.

The study, reported in Frontiers in Psychiatry, enrolled consecutive patients admitted to Northern Hospital of Sinopharm Group in China from June 2020 through December 2021. Participants were 18 years or older but were required to be between 45 and 80 for inclusion, and each had a clinical diagnosis of acute ischemic stroke supported by neurological symptoms, physical examination and neuroimaging. Only mild-to-moderate strokes were included, defined by an admission NIHSS score of no more than 15. Patients with severe stroke, impaired consciousness, pre-existing anxiety disorders, several major vascular or psychiatric conditions, substance dependence, or medications that could exacerbate anxiety were excluded.

Neurological severity was measured with the 15-item NIHSS, which ranges from 0 to 42 and evaluates domains including consciousness, eye movements, visual fields, facial movement, limb strength, sensation, coordination, speech and language. Higher scores represent greater impairment. Anxiety was assessed with the 7-item Generalized Anxiety Disorder scale, or GAD-7, covering symptoms during the preceding two weeks. The researchers divided participants into a lower-anxiety group with a score of 4 or less and a higher-anxiety group with a score above 4, equivalent to 5 or more. GAD-7 is a symptom-severity screen rather than a stand-alone diagnostic interview for an anxiety disorder.

Peripheral blood was collected on hospital admission, and serum was separated and stored at minus 80 degrees Celsius until analysis. Using enzyme-linked immunosorbent assays, investigators measured three CC chemokines: monocyte chemoattractant protein-1, MCP-1; MIP-1α; and macrophage inflammatory protein-1β, MIP-1β. Laboratory personnel were blinded to participants’ clinical and demographic characteristics. MIP-1α was selected for focused moderation analysis because initial correlation screening showed that its association with NIHSS scores differed markedly between anxiety groups, making the subsequent analysis exploratory and hypothesis-generating rather than a prespecified primary test.

Descriptive comparisons divided the cohort into four subgroups based on anxiety and MIP-1α concentration, with MIP-1α separated at the median within each anxiety group. NIHSS scores differed across these groups, with the lower-anxiety, higher-MIP-1α subgroup having a mean score of 0.35, compared with 1.32 in the lower-anxiety, lower-MIP-1α group; means were 1.24 and 1.54 in the higher-anxiety groups with higher and lower MIP-1α, respectively. The lower-anxiety, higher-MIP-1α group also had significantly lower scores than each of the other three groups after Holm-adjusted post-hoc comparisons. Sex distribution differed across the groups, but age, smoking, alcohol consumption, hypertension and diabetes did not differ significantly.

Spearman correlations showed that MIP-1α had opposite directional relationships with NIHSS scores in the two anxiety strata: the association was significantly negative among participants with GAD-7 scores of 4 or less but became positive in those scoring above 4. By contrast, MCP-1 and MIP-1β were positively correlated with neurological impairment in both anxiety groups. MIP-1α was also strongly positively correlated with MCP-1 and MIP-1β in both strata, all with p values below 0.001. These patterns linked the MIP-1α result to a broader chemokine profile, although the study did not directly test how the three markers mediate one another.

Hierarchical linear regression was used to test whether anxiety group moderated the MIP-1α–NIHSS association. Model 1 adjusted for age, sex, smoking, alcohol consumption, hypertension and diabetes. Model 2 added MIP-1α and anxiety group, and Model 3 added the MIP-1α-by-anxiety interaction. The interaction was significant, with a standardized beta of 0.16 and p = 0.006, while MIP-1α alone remained inversely associated with NIHSS scores at beta = minus 0.19 and p = 0.001. The adjusted R-squared increased from 0.10 in Model 2 to 0.13 in Model 3. Simple-slope analysis found a significant inverse MIP-1α association in the lower-anxiety group, with beta = minus 0.32 and p below 0.001, but no association in the higher-anxiety group, where beta was minus 0.003 and p was 0.969. Thus, the study’s detailed results do not support every directional claim in its abstract: the highest-anxiety stratum did not show a significant positive simple slope, even though the formal interaction was significant.

Receiver operating characteristic analysis assessed how well the models distinguished patients with no NIHSS-detectable deficit from those with at least one point. The area under the ROC curve was 0.67 for Model 1, 0.74 for Model 2 and 0.76 for Model 3. Model 2 significantly outperformed Model 1, and Model 3 also significantly outperformed Model 1, with p below 0.001 for the latter comparison. However, Model 3 did not significantly outperform Model 2: the AUC difference was 0.02, and DeLong’s test returned p = 0.14. The interaction therefore strengthened the model’s fit, but it did not add a statistically demonstrable improvement in discrimination beyond the separate contributions of anxiety and MIP-1α.

Several design features limit interpretation. The cross-sectional assessment cannot establish whether inflammation preceded anxiety, whether anxiety amplified inflammation, whether severe stroke produced both, or whether another factor influenced all three. It also cannot demonstrate that managing anxiety would change MIP-1α signaling or neurological recovery. The single-center sample was restricted to mild-to-moderate ischemic stroke and had a median NIHSS score of 1, limiting applicability to severe or hemorrhagic stroke. Anxiety was self-reported without a structured diagnostic interview, MIP-1α was measured only once, and important clinical variables—including onset-to-assessment interval, reperfusion treatment, infarct location and volume, stroke subtype, previous stroke, medication use and social support—were not recorded. The inverse association between age and NIHSS may likewise reflect selection or restricted score range rather than a protective effect of older age.

Taken together, the study provides cross-sectional clinical evidence that post-stroke anxiety symptom severity can alter the statistical relationship between MIP-1α and acute neurological impairment. Its strongest support is for a statistically detectable interaction, not for a causal pathway or a clinically validated prediction tool. The AUC result reinforces that caution: adding the interaction improved model fit, but did not significantly improve discrimination over a model already containing anxiety group and MIP-1α. Prospective studies with serial symptom and biomarker measurements, larger and more diverse stroke populations, and direct testing of the proposed stress and blood–brain-barrier mechanisms would be required to determine whether the relationship holds over time and whether it has practical value outside this selected cohort.

Subject of Research: Cross-sectional study of post-stroke anxiety, MIP-1α and acute ischemic stroke severity

Article Title: Post-stroke anxiety symptoms moderate the predictive efficacy of MIP1-α on acute ischemic stroke severity

Article References: Li, M., Chen, T., Wang, Y., Zhou, M., Zhou, X., Xing, L., Huang, H., Zhou, P., Liao, Z., Chang, Y., Hu, X., Luo, X., Chen, Y.-H., Chen, L., Liu, Y., Wang, W., Wang, F., & Jiang, H. (2026). Post-stroke anxiety symptoms moderate the predictive efficacy of MIP1-α on acute ischemic stroke severity. Frontiers in Psychiatry, 17, Article 1785971. https://doi.org/10.3389/fpsyt.2026.1785971

Image Credits: AI Generated

DOI: 10.3389/fpsyt.2026.1785971

Keywords: ischemic stroke, post-stroke anxiety, MIP-1α, neuroinflammation, NIHSS, GAD-7, chemokines, blood biomarkers, psychoneuroimmunology, Post-stroke, anxiety, symptoms