Growing experience with transcatheter aortic valve implantation (TAVI) enables predictive factors of mid-term results to be identified.Between October 2006 and January 2010, 145 patients (pts) underwent TAVI because of symptomatic, severe aortic stenosis (mean valve area 0.70±0.18cm 2 ). TAVI used a transvascular (TV) approach in 96 cases and a transpical (TA) approach in 49. Prostheses were Edwards Sapien in 131 pts and Medtronic Corevalve in 14 pts. Follow-up was complete, median follow-up was 6 months. Predictive factors of 2-year mortality were selected among 16 variables using a univariate Cox model and then a Cox multivariate model including significant variables in univariate analysis and risk scores which were kept in the model. The effect of the learning curve was assessed by comparing the first 25 procedures with the subsequent 120 procedures.Mean age was 81±9 years, 77 pts (53%) were male, and 37 pts (26%) had had coronary artery bypass grafting (CABG). Mean Euroscore was 27±15% and mean Charlson comorbidity index was 3.7±2.5. Thirty-day survival was 88%. Two-year survival was 68±6%. The 3 significant predictive factors of 2-year mortality in multivariate analysis were TA approach, absence of prior CABG, and procedures within the first 25 cases (Table). Risk scores were not significantly associated with 2-year mortality.The identification of predictive factors of mid-term mortality after TAVI 1) Shows the impact of training 2) Suggests that patients with prior CABG and treated using the TV approach derive the greatest benefit 3) Further illustrates the limitations of conventional risk scores in predicting the results of TAVI.Table: Predictive factors of 2-year mortality in multivariate analysisHazard ratio95% CIpTA vs. TF approach3.6[1.6–7.9]0.002Early vs. late experience2.8[1.2–6.4]0.01Absence of prior CABG2.9[1.1–7.8]0.03Euroscore1.003[0.977–1.029]0.84Charlson comorbidity index1.03[0.88–1.19]0.72