Abstract W P19: Propensity Scores Facilitate Unbiased Comparison of Endovascular Devices in IMS III
Introduction:The IMS III Trial randomized acute ischemic stroke patients to endovascular (IV tPA + intra-arterial) or IV tPA -only therapy to compare efficacy. However, endovascular subjects were not randomized to a particular approach; interventionalists chose the method. Hence, any statistical comparison of devices must account for potential selection bias.
Methods: To compare the efficacy and safety of various endovascular treatment modalities, the multiple propensity score (PS) method was implemented to account for potential bias and achieve balance between groups. The analysis is restricted to subjects with an ICA or M1 vessel treated using a standard microcatheter, the Merci retriever, or the Penumbra system. Propensity scores were estimated via a multinomial regression model that included demographic and clinical variables for which 1) an imbalance between devices was demonstrated (α=0.10) or 2) an association with outcome was established (α=0.10). PS tertiles were included as covariates in a logistic regression model relating outcome to device.
Results: The analysis included 167 subjects treated with a standard microcatheter (n=51), the Merci retriever (n=77) or the Penumbra system (n=39). Unadjusted comparison indicated imbalance between devices in the presence of atrial fibrillation, age, systolic blood pressure, onset to IV tPA start time, and IV tPA start to groin puncture time. Additional covariates included in the PS model varied by outcome. After adjustment for PS tertiles, the initial difference in baseline variables between devices was no longer statistically significant and there was insufficient evidence to conclude that the devices differed for any of the outcomes considered.
Conclusions: Selection bias was a concern in comparing endovascular approaches in IMS III. The PS method successfully balanced baseline characteristics and facilitated a valid conclusion: major efficacy and safety outcomes did not significantly differ between subjects treated with a standard microcatheter, the Merci retriever, and the Penumbra system. Limitations include small sample size and the possibility of differences in unmeasured baseline characteristics.
Author Disclosures: L.D. Foster: Research Grant; Significant; IMS III. S.D. Yeatts: Research Grant; Significant; IMS III Co-investigator. Consultant/Advisory Board; Modest; Genentech: PRISMS Steering Committee Member. T.A. Tomsick: Research Grant; Significant; IMS III, Covidien Ltd. J.P. Broderick: Other Research Support; Significant; Genentech-study medication for NINDS-funded IMS III and CLEARER trials, Genentech -PRISMS Trial consultant on steering committee, Penumbra-THERAPY Trial co-inv (paid to Dept). Y.Y. Palesch: Research Grant; Significant; IMS III. Consultant/Advisory Board; Modest; consulting fees from BrainsGate, consulting fees from Edge Therapeutics.
- © 2014 by American Heart Association, Inc.