Results
A total of 286 consecutive patients with severe AS were included in the study. There was a progressive growth in the number of patients referred for HT discussion, with a 26% increase seen between the first and third time periods of the study. Of the 286 patients, 53 were referred for surgical therapy, 210 for TAVR (188 to transfemoral and 22 to non-transfemoral approach) and 23 were referred for MT (Figure 1 ). Reasons for choosing MT are summarized in Table 1.
Baseline characteristics according to the initial selected therapeutic strategy are shown in Table 2 . Analysing data on an intention-to-treat basis, patients referred for SAVR were younger (82 [78-84] year-old) than patients in the TAVR and MT group (85 [81-87] and 86 [83-90] year-old, respectively) (p<0.001). There were no significant differences in gender nor in cardiovascular risk factors. Patients in the SAVR group had lower risk scores than those assigned to TAVR or MT. Median left ventricular ejection fraction was lower in the MT group compared to TAVR and SAVR groups (55% [35-60] vs 60% [52-66] and 60% [51-64], respectively). Significant mitral regurgitation (MR≥grade III) was less frequent in the TAVR group (5.5%) compared to MT and SAVR groups (42.1% and 22.0%, respectively, p<0.001 for both comparison).
A total of 22 patients changed their therapy arm after being initially discussed in HT session due to various reasons (supplementary Table S1). Thus, considering the final definitive treatment, SAVR was performed in 50 patients, 195 patients underwent TAVR (176, 90.3% through transfemoral approach) and 41 patients received MT (Figure 1 ). Baseline characteristics according to definitive treatment group (as-treated group) are shown in Supplementary Table S2 .
Fifteen baseline clinical characteristics were included in the CART analysis to determine relevant variables in the decision process (supplementary Table S3 ). Finally, the HT decision algorithm was built with six of those variables [age, logistic EuroSCORE, significant MR, frailty, STS score and estimated glomerular filtration rate (eGFR)] (figure 2) . Age was the first split point identified with a cut-off value of 88 years old. Among patients 88-years and older, logistic EuroSCORE was the determinant parameter which assigned patients to TAVR or MT. Among patients<88 years old, significant MR was the next split variable. In patients with significant MR frailty was the conditioning variable to assign patients to SAVR or TAVR. In those without MR, STS score and eGFR were recognized as further relevant factors to decide between SAVR or TAVR.
Procedural characteristics and in-hospital outcomes among patients who underwent TAVR or SAVR are depicted in Tables 3 and 4 , respectively. Importantly, 20% in the SAVR group underwent a concomitant second valve intervention. In-hospital mortality in the TAVR group was 4.6% and 12.0% in SAVR group. Specifically, in patients undergoing isolated aortic valve intervention in-hospital mortality was 7.5% with SAVR, compared with 3.4% in the transfemoral TAVR cohort (p=0.447). While major vascular complications were more frequent in the TAVR group, acute kidney injury, significant bleeding, new onset atrial fibrillation and longer hospital stay occurred more frequently in the SAVR group. There were no differences between groups in terms of stroke or the need for permanent pacemaker implantation.
A total of 89 deaths were recorded during follow-up, 45 (53.6%) from a cardiac cause. In the as-treated analysis, all-cause and cardiovascular mortality at 1-year were 16.6% and 7.2% in the invasive groups (17.2% and 6.4% in the TAVR group, 14.0% and 10.2% in SAVR group), and 68.7% and 60.7% in the MT arm, respectively (figure 3).Survival rate according to access site (transfemoral versus non-transfemoral) and the type of surgery (single versus multiple valve intervention) are depicted in figure 4 . The survival analysis according to the intention to treat are shown in supplementary figure S1 and S2 . All cause rehospitalization rate and NYHA class in follow-up are shown in supplementary figure S3 .