Methods
The study prospectively evaluates consecutive patients with AS referred to the HT meeting between June 2014 and February 2017 collated in a dedicated database. Patients were referred from the catchment area of our main hospital and from two other satellite hospitals with an established alliance. Our study was approved by the Institutional Review Board of our institution and all patients provided written informed consent for the procedures. Heart team meetings were weekly scheduled in the main hospital with the attendance of at least two cardiac surgeons, two interventional cardiologists, one imaging cardiologist and one clinical cardiologist. Physicians from other specialities, such as internal medicine, oncology or geriatric medicine, were invited to participate in the discussion when necessary and physicians from the satellite hospitals participated fully via video-link. By closely following previous Clinical Practice Guidelines (2,3), a local consensus document (Annex I in Supplementary data ) was developed and signed jointly by cardiac surgeons and cardiologists, with the aim of identifying potential candidates with AS for HT discussion. This document was distributed to all potential referral physicians within the three hospitals. All patients with AS originally referred for TAVR and those in whom the management was undecided were discussed by the HT and were included in this analysis. However, patients directly referred to SAVR who did not meet any of the criteria in the consensus document, were excluded from this analysis unless the cardiac surgeon deemed it necessary to discuss the case in the HT.
During the HT meeting, clinical data from each patient was summarized in a formal presentation, which also included a prospective evaluation of surgical risk using the logistic European System for Cardiac Operative Risk Evaluation logistic EuroSCORE and the Society of Thoracic Surgeons (STS) score. Each case presentation was followed by a discussion and assessment of the overall risk profile. The HT then decided to refer the patient to either medical treatment (MT), TAVR or SAVR. A prospective clinical follow-up at 1- 6- 12- and 24-month was carried out through clinical visits for all patients in the SAVR and TAVR groups, whereas clinical outcomes were analysed retrospectively in the MT group. The median follow-up time was 18 months [11-26] and only one patient was lost to follow-up. In-hospital and long-term outcomes were defined according to the Valve Academic Research Consortium 2 (VARC-2) criteria (8).
Quantitative continuous variables are expressed as mean (standard deviation) or median (interquartile range [IQR]) according to their distribution. Assessment of normality was performed using the Shapiro-Wilk test. Differences between treatment groups were evaluated using the non-parametric Kruskal-Wallis rank test and Wilcoxon rank test for continuous variables without normal distribution. Categorical variables were summarized as number (percentage) and comparisons were analysed by the chi-square or the Fisher´s exact test. Patient baseline characteristics were identified that significantly influenced decision making within the HT. Considering these factors, a decision tree to guide the decision-making process was built using CART (classification and regression tree) methodology (9). The CART method is used for constructing prediction models from data. The models are obtained by dividing the data and adjusting a simple prediction model within each partition. The programme determines cut-off points which best explain the categorical endpoint of the analysis (MT or TAVR or SAVR) and selects the predictor with the lowest p-value of a logistic regression to make a first division. The result is a decision tree. We used the registered baseline characteristics to reproduce the decision process by using the non-parametric CART methodology. Survival curves were calculated using the Kaplan-Meier method, and comparison was obtained with the log-rank test. All analyses were performed using Stata 14 (StataCorp, College Station, TX, USA) and RStudio Team (2018). RStudio: Integrated Development for R. RStudio, Inc., Boston, MA.