Serum IL-6 is independently associated with 90-day poor outcome, and incorporating cytokines into clinical models may improve prognostic estimation in EVT-treated acute ischemic stroke patients.
Key Findings
Results
The majority of EVT-treated large vessel occlusion patients experienced poor functional outcomes at 90 days.
139 consecutive patients were included from the prospective DETECT2-China registry between March 2022 and March 2025.
77 of 139 patients (55.4%) had 90-day poor functional outcomes, defined as modified Rankin Scale (mRS) 3-6.
Plasma cytokines were measured within 24 hours post-EVT.
The study population consisted of large vessel occlusion patients undergoing endovascular treatment.
Results
Serum IL-6 and IL-10 levels were significantly higher in patients with 90-day poor functional outcomes.
Both IL-6 and IL-10 were elevated in the poor outcome group (mRS 3-6) compared to the good outcome group.
Cytokine levels were analyzed on a log2-transformed scale to account for skewed distributions.
The finding was observed in univariable comparisons between outcome groups.
These associations prompted inclusion of both cytokines in multivariable regression modeling.
Results
Log2-transformed IL-6 was an independent predictor of 90-day poor functional outcome in multivariable logistic regression.
OR = 1.43 per doubling of IL-6, 95% CI: 1.14–1.79, p = 0.002.
This association was independent of conventional clinical predictors.
IL-6 was significant across multivariable models evaluated in the study.
The log2 transformation means each doubling of serum IL-6 was associated with a 43% increased odds of poor outcome.
Results
Log2-transformed IL-10 was a significant independent predictor of 90-day poor outcome in Adjusted Model 2.
OR = 1.86 per doubling of IL-10, 95% CI: 1.02–3.36, p = 0.041.
IL-10 significance was observed specifically in Adjusted Model 2, suggesting its effect may be model-dependent.
The confidence interval was wide (1.02–3.36), reflecting some statistical uncertainty.
Each doubling of serum IL-10 was associated with an 86% increased odds of poor outcome in this model.
Results
The combined model incorporating cytokines achieved superior discrimination compared to the clinical model alone.
Combined model AUC: 0.841 (95% CI: 0.770–0.905).
Clinical model AUC: 0.791 (95% CI: 0.717–0.865).
The difference in AUC between models was statistically significant (p = 0.015).
The combined model also demonstrated good calibration and clinical net benefit on decision curve analysis (DCA).
Results
Adding cytokines to the clinical model significantly improved patient reclassification.
Continuous Net Reclassification Improvement (NRI): 43.32%, p = 0.009.
Integrated Discrimination Improvement (IDI): 7.27%, p < 0.001.
Both reclassification metrics indicated meaningful incremental prognostic value of cytokines beyond conventional predictors.
These metrics were used alongside ROC analysis to comprehensively evaluate model improvement.
Conclusions
The authors caution that these findings are exploratory and require validation before clinical application.
The study was based on a single registry (DETECT2-China) with 139 patients, limiting generalizability.
The authors explicitly state these are 'exploratory findings' that 'require validation in independent cohorts before clinical application.'
The prospective registry design supports data quality but the sample size is relatively modest.
External validation in diverse populations was identified as a necessary next step.
What This Means
This research suggests that measuring certain inflammatory proteins in the blood shortly after a stroke procedure can help predict how well patients will recover. The study looked at 139 patients in China who had a stroke caused by a blocked large blood vessel and were treated with a procedure to remove the clot (endovascular treatment). More than half of these patients (55.4%) had poor recovery three months later, meaning they had significant disability or died. The researchers found that two immune signaling proteins — IL-6 and IL-10 — were notably higher in patients who did poorly, and that IL-6 in particular was independently linked to worse outcomes even after accounting for other known risk factors.
The study also tested whether adding these cytokine measurements to standard clinical information could improve doctors' ability to predict outcomes. The combined model (clinical information plus cytokines) was significantly better at distinguishing patients who would do well versus poorly compared to using clinical information alone, with the area under the ROC curve improving from 0.791 to 0.841. The model also correctly reclassified a meaningful proportion of patients into more accurate risk categories.
This research suggests that blood tests for inflammatory proteins taken within 24 hours of a stroke procedure could add useful prognostic information beyond what doctors currently use to estimate recovery. However, the authors emphasize that this was an exploratory study with a relatively small number of patients from a single registry in China, and the findings need to be confirmed in larger, independent studies before they could be considered for use in clinical practice.