Abstract WP251: Long Term Outcome Prediction After Ischemic Stroke Using Gene Expression

Objective: Prediction of the long-term outcome in Ischemic Stroke (IS) patients can have a significant impact on design of clinical trials and on patients’ care. We studied gene expression (GE) as a novel biomarker to provide an accurate prediction of 90-day outcome in IS patients. Methods: RNA from 72 samples from 2 peripheral blood draws (at ≤3 and 24 hrs post IS onset) was analyzed on Affymetrix U133 Plus 2.0 microarrays. These represented samples from 36 CLEAR trial IS patients that had blood drawn within 3 hrs of stroke onset and were then treated with tPA with or without eptifibatide. The samples were split into derivation (n=25) and validation (n=11) sets. We identified the differential GE in blood at 24 hrs and the difference in GE between 24 hrs and 3 hrs post IS that was associated with 90-day post stroke outcome using the model: GE = μ + NIHSS_24hr+mRS_90day+ ε. Good outcome was defined as mRS 0-2; Poor - as mRS 3-5. Logistic regression was used to derive a biomarker classifier. Results: Using 24 hrs GE, we identified 14 probesets (12 genes) with the highest discriminative power for predicting outcome. The model achieved recall (the probability of correctly identifying the patients with Good outcome) of 0.88 and specificity (the probability of correctly identifying the patients with Poor outcome) of 0.67 in the validation set (The AUC-ROC = 0.88). The biomarker genes were enriched in immune responses such as IL and cytokine signaling. Among the predictors were genes important for stroke and repair after stroke (e.g., MACC1 , GDF11 ). MACC1 has been considered as a potential treatment target for IS with a protective role in hypoxia-induced human brain microvascular endothelial cells. GDF11 plays a role in brain repair after IS. We also determined how the change of GE from 3 hrs to 24 hrs would predict the 90-day outcome. A panel of ten genes was able to predict outcome in the validation set (recall= 1, specificity = 0.67, AUC-ROC=0.88). These included AVPR1A , which mediates platelet aggregation and release of coagulation factors and exacerbates brain inflammatory response to injury. Conclusion: This pilot study suggests gene expression can be used to predict stroke outcome. Some of the genes may serve as potential therapeutic targets.