Abstract 20166: Development of a Risk Prediction Model for 1-Year Mortality After Surgical vs. Transcatheter Aortic Valve Replacement in Patients With Severe Aortic Stenosis
Introduction: Treatment options for patients (pts) with severe aortic stenosis (AS) include surgical aortic valve replacement (SAVR) and transcatheter aortic valve replacement (TAVR). Although randomized trials have shown similar long-term survival with TAVR and SAVR in selected pts, individual risks may vary. Presently, no tool exists to predict 1 year mortality in AS pts treated with SAVR or TAVR.
Methods: Using Medicare claims-linked databases of pts treated with TAVR (Transcatheter Valve Therapy (TVT) registry) or SAVR (Society of Thoracic Surgeons (STS) Registry), we developed a unified risk model to predict 1 year mortality after AVR. Pts aged 65 to 90 yrs, who underwent TAVR (2011-2015) or isolated SAVR or SAVR/CABG (2011-2013) were included in the analysis. Exclusion criteria included inoperability, pure aortic insufficiency, and emergent procedures. Since falsification endpoints indicated substantial unmeasured confounding in non-high risk pts, the model was restricted to pts with STS predicted mortality risk ≥ 8%. The associations between mortality, baseline covariates, AVR type and selected interaction terms were estimated using weighted Cox proportional hazards models.
Results: The analytic cohort included 7230 TAVR pts and 4632 SAVR pts (median age 83 yrs; 47% male; mean STS score 13.2%). The final model consisted of 33 covariates including 13 interaction terms (c statistic 0.62). Home O2 use, immunosuppression, reduced LVEF and prior cardiac surgery were associated with increased mortality with SAVR while peripheral vascular disease and atrial fibrillation were associated with increased mortality with TAVR. An example of differential treatment effects according to specific risk factor combinations is shown in the figure.
Conclusion: Data from the TVT and STS registries have been used to develop a predictive model of 1 year mortality for AS pts treated with TAVR or SAVR. This model is a valuable tool for shared decision making by pts and providers.
Author Disclosures: S.J. Baron: Consultant/Advisory Board; Modest; Edwards LifeSciences, St. Jude Medical Inc. D.J. Cohen: Research Grant; Significant; Eli Lilly, Astra Zeneca. Consultant/Advisory Board; Significant; Eli Lilly, Astra Zeneca. S. Suchindran: None. M. Pencina: None. D. Dai: None. R. Matsouaka: None. J.M. Brennan: None.
- © 2016 by American Heart Association, Inc.