Abstract TP186: Heterogeneity Of Endovascular Treatment Effect: A Comparison Of Subgroup Identification Methods In Acute Stroke Trials

Background: Trials’ data are increasingly re-analyzed to identify treatment effect heterogeneity: that is, subgroups of patients who have either enhanced or adverse effects in a trial. This study investigates the robustness of subgroup identification methods in an acute stroke trial. Methods and Analysis: The Model-based recursive partitioning (MOB), Stochastic Subgroup Identification based on Differential Effects Search (Stochastic SIDEScreen), and Virtual Twin (VT) methods would be used to detect heterogeneity in Endovascular Treatment for Small Core and Anterior Circulation Proximal Occlusion with Emphasis on Minimizing CT to Recanalization Times (ESCAPE) trial. Results: In the ESCAPE trial, patients in the intervention group had a higher rate of functional independence (90-day mRS 0-2) than those in the control group (OR=2.6; p<0.001, and 95% CI=1.7–3.8). The three methods identified patients with differential treatment effects. The MOB identified 2-terminal subgroups, with the NIHSS > 11 group showing a significant treatment effect (OR=3.67; p<0.001 and 95% CI=2.11–6.40), while the subgroup of with a maximum NIHSS score of 11 did not (OR=1.63; p=0.463 and 95% CI=0.44–6.05). The stochastic SIDEScreen identified 4-terminal subgroups, but the group of patients with NIHSS greater than 9 and older than 54 years had a significant treatment effect (OR=4.92; p<0.001, and 95% CI= 2.66–9.10). Other three subgroups, like patients with a maximum NIHSS score of 9 and older than 54 years (OR=2.17, p=0.34, and 95% CI=0.44–10.65), did not have a significant treatment effect. VT identified 6-terminal subgroups; the subgroup consisting of patients older than 56 years and NIHSS > 11 had significant treat effect (OR=5.11; p<0.001 and 95% CI=2.68–9.73). As other renaming 4 subgroups, the subgroup consisting of younger patients and with a maximum NIHSS score of 11 did not show a treatment effect (OR=1.60, p=0.64, and 95% CI=0.39–6.30). Conclusion: Data-driven subgroup identification methods provide insight into the heterogeneity of treatment effects in acute stroke trials. Information about the identified subgroups might inform the development of clinical practice guidelines for acute stroke management.