Serum sCD163: a novel predictor of early neurological deterioration in non-reperfused acute ischemic stroke

Background and purpose Early neurological deterioration (END) is a common complication in non-reperfused acute ischemic stroke (AIS), yet reliable predictors are lacking. Soluble CD163 (sCD163), a monocyte/macrophage activation marker, is linked to neuroinflammation, but its role in non-reperfused END is unknown. We aimed to evaluate the serum sCD163–END association and externally validate a prediction model. Methods In this two-center prospective study, 664 non-reperfused AIS (training) and 300 (validation) patients were enrolled. Serum sCD163 was measured within 24 h. END was defined as an NIHSS increase ≥2 within 72 h. LASSO and multivariable logistic regression identified predictors and built a prediction model, with bootstrap internal and independent external validation assessing discrimination, calibration, and decision curve analysis. Results END occurred in 142 (21.4%) training and 65 (21.7%) validation patients. END patients had higher sCD163 (779 ± 110 vs. 690 ± 103 ng/mL training; 777 ± 110 vs. 685 ± 110 ng/mL validation, both P < 0.001). sCD163 independently predicted END (adjusted OR = 2.09 per 10 ng/mL, 95% CI: 1.72–2.54, P < 0.001). The prediction model showed good discrimination (bootstrap-corrected AUC = 0.887, 95% CI: 0.841–0.920; external AUC = 0.866, 95% CI: 0.819–0.903), calibration (Hosmer–Lemeshow P = 0.408 and 0.447), and positive net benefit in decision curve analysis. Adding sCD163 improved reclassification (NRI = 0.218, P = 0.003; IDI = 0.032, P = 0.001). Conclusions Serum sCD163 independently predicts END in non-reperfused AIS. The prediction model incorporating sCD163, baseline NIHSS, neutrophil percentage, and C-reactive protein offers a validated risk stratification tool.