Abstract W P20: A Model to Calculate the Expected Treatment Effect in Acute Endovascular Stroke Trials

Background: Recent trials have shown no benefit of endovascular therapy. This may, in part, be explained by inaccurate estimates of the treatment effect used in the sample size calculations of these trials. A predictive model which includes variables that modify the expected treatment effect might yield more accurate estimates, and could be valuable in the design of future acute stroke trials. Methods: We conducted a literature review to obtain estimates of parameters that are associated with good functional outcome (GFO) following recanalization. We developed a model to estimate the treatment effect in endovascular stroke trials and applied this model to two recently published endovascular stroke trials. Results: We estimated a 40% absolute difference in the proportion of GFO (mRS 0-2 at 90 days) associated with reperfusion in patients with ICA or M1 occlusions who have a substantial ischemic penumbra at baseline. To estimate the effect size in trials, this value was multiplied by: 1) the proportion of patients undergoing endovascular therapy in the active treatment arm; 2) the proportion of patients with occlusions of the ICA or MCA-M1; 3) the proportion of patients with a substantial penumbra and a DWI lesion <50mL; and 4) the absolute difference in the proportion of patients with reperfusion, defined as TICI 2B-3, between the endovascular treatment and control arms. Based on literature review we assumed a reperfusion rate of 20% in the control arms of IMS III and MR Rescue, a 50% prevalence of patients with substantial penumbra and DWI lesions<50 mL in IMS III, and a 75% prevalence in the penumbral arms of MR Rescue. Based on these model inputs, a 2.2% increase in GFO with endovascular therapy was expected in IMS III, which closely matches the observed 2.1% increase. For MR Rescue, the model predicted a 1.5% increase in GFO with endovascular therapy. Considering the small sample size, this equates to 0.5 additional patients with GFO which closely matches the observed result of 3 fewer patients with GFO. Conclusion: A simple model shows promise for estimating the treatment effect of endovascular stroke trials. It may be useful for the design of future trials and could lead to different inclusion criteria or larger sample sizes compared to the recently conducted studies.