To derive stroke-predicting serum immune biomarker signatures that predict incident stroke and assess their prognostic value beyond the Global Vascular Risk Score (GVRS).

Chronic immune activation contributes to vascular injury and stroke, yet prospective integrative biomarker studies remain limited. In the Northern Manhattan Study (NOMAS), circulating immune molecules have been linked to cerebral small/large vessel disease, poorer cognition, and incident cognitive decline. They may also relate to incident stroke risk.

Stroke-free NOMAS participants (age≥50) with serum concentrations of 60 immune-related proteins measured by customized Luminex assay were followed longitudinally. Least Absolute Shrinkage and Selection Operator (LASSO) proportional hazards Cox models with 5-fold cross-validation were applied to standardized markers to select predictors of time from blood draw to incident stroke, censored at last follow-up or death. Associations of the LASSO-selected markers with stroke hazard were assessed using GVRS-adjusted Cox models; GVRS is a validated stroke-predicting NOMAS score incorporating demographics and vascular burden.

Among 1,175 participants (mean age 70±9 years; 60% female; 68% Hispanic), 130 incident strokes occurred over a median 14-year (IQR 6.5) follow-up. LASSO prioritized C-X-C motif chemokine ligand 9 (CXCL9) and 10 (CXCL10). In GVRS-adjusted models, each 1-SD higher serum CXCL10 was associated with 20% greater stroke risk (HR=1.20, 95%CI 1.05–1.37, p=0.007). CXCL9 showed a similar direction of association that did not meet statistical significance (HR=1.10, 95%CI 0.96–1.27, p=0.157). In a model including both chemokines, only CXCL10 remained associated with stroke (HR=1.18, 95%CI 1.02–1.37, p=0.025).

In this large urban cohort followed for over 14 years, IFN-γ–inducible chemokines emerged as candidate predictors of incident stroke. The association of CXCL10 (interferon gamma–induced protein 10 [IP-10]) with stroke risk was beyond participants’ vascular risk burden and CXCL9 (monokine induced by IFN-γ [MIG]) concentration, supporting its biological plausibility as a potentially modifiable immune marker for stroke prevention.