In order to bring you the best possible user experience, this site uses Javascript. If you are seeing this message, it is likely that the Javascript option in your browser is disabled. For optimal viewing of this site, please ensure that Javascript is enabled for your browser.
Did you know that your browser is out of date? To get the best experience using our website we recommend that you upgrade to a newer version. Learn more.

We use cookies to optimise the design of this website and make continuous improvement. By continuing your visit, you consent to the use of cookies. Learn more

SOURCE REGISTRY TAVI: Development of a risk score for transcatheter aortic valve implantation: 1-year outcomes from over 1,000 patients in the source registry

Valvular Heart Diseases

Olaf Wendler
Presenter | see Discussant report Webcasts become available 24h after the presentation
Wendler, Olaf
(United Kingdom)
Open presentation slides

List of Authors:
Olaf Wendler, Gerhard Schymik, Thomas Walther, Dominique Himbert, Thierry Lefèvre, Hendrik Treede, Holger Eggebrecht, Paolo Rubino, Iassen Michev, Martyn Thomas


The SOURCE-Registry identified certain demographic variables as predictors of 30-day mortality after transcatheter aortic valve implantation (TAVI). However, preoperative variables to predict 1-year mortality are unknown at present. We analyze the 1 year mortality of patients enrolled in the SOURCE Registry in an attempt to develop a risk score which predicts mid-term outcome after TAVI.

The SOURCE-Registry collects data of patients treated at European centers following commercialization of the Edwards SAPIENTM bioprosthesis. Only data from centers that could provide all of their consecutively treated patients were included in the study. This univariate and multivariate analysis is based on Cohort I patients (n=1038), included from November 2007 to January 2009.

Out of the total of 1038 patients, 55.4% were treated using transapical (TA) TAVI and 44.6% with a transfemoral (TF) approach. One-year survival data are available on 98% of the patients.
The one-year Kaplan-Meier survival was 72.1% in TA and 81.1% in TF. Mortality between 30-days and 1-year were due in the majority to non-cardiac causes (49.2%) such as pulmonary, renal failure, cancer, stroke and gastrointestinal.
In proportional hazards multivariable analysis logistic EuroSCORE, liver disease, renal failure, smoking and “other cardiovascular” conditions were all predictors of inferior 1-year survival. The hazard ratio is 1.2 for a 10 point increase in EuroSCORE, 3.1 for patients with baseline liver disease, and 1.7 for patients with baseline renal failure.
Carotid artery stenosis, hypercholesterolemia and hypertension are also in the mathematical model, and work in a counterintuitive direction.

The 1-year survival after TAVI is convincing and in both implant approaches has improved compared to previously published early series. Although logistic EuroSCORE, liver disease and renal failure are strong predictors for 1-year mortality after TAVI, an appropriate risk score for TAVI is not available at present. Improvement on the data is being developed at present which hopefully will allow us to create a specific risk score for TAVI in the future.

Gerhard Schuler
Discussant | see Presenter abstract Webcasts become available 24h after the presentation
Schuler, Gerhard
Open presentation slides


Development of a Risk Score for TAVI

In the talk Olaf Wendler and colleagues presented the 1-year outcomes from over 1000 patients who had undergone percutaneous or transapical valve replacement for severe aortic stenosis using the Edwards Sapien Device. Data were collected in the SOURCE REGISTRY, which to date is the most comprehensive database on this topic. The aim of this analysis was to develop a dedicated risk score for predicting mortality for this unique group of patients.

A risk score is derived from a set of data by multivariate analysis by using C statistics, a measure of how well a clinical prediction can correctly rank-order pts by risk. A model that accurately discriminates pts 85% of the time has a C statistic of 0.85. A complete random statistic would be 0.50.

The third slide shows various ROC curves; the blue line has a C statistic of 0.5, i.e. it is no better than complete random order. Whereas the dark blue line has a C statistic of 0.85 and thus it is of clinical value to predict the patients' risk.
C statistics below 0.7 are only of limited or no value.

1-year Mortality by Euro-Score

Although 20 000 consecutive pts were recruited for the data base of the EuroScore with 97 risk factors assessed, this was not sufficient to predict 1-year mortality. For both transfemoral and transapical cases the 1-year prediction was not much better than random statistics.

Multivariate Analysis SOURCE REGISTRY

For the Source Registry, the C-statistics were slightly better, but not adequate for clinical purposes.

Risk Factors not assessed:
Frailty. Is there an explanation for the failure to create a clinical useful risk score from the data. There are several important risk factors which have not been included in the SOURCE registry. One of them is frailty which plays an important role in this particular patient population. Several frailty scores have been designed in recent years.

Causes of Death:
Another explanation for the failure of the C-statistics to reach a significant level is the great number of deaths that are not covered by appropriate variables, such as sudden death, cancer, stroke, and gastrointestinal.


709009 - 709010


Clinical Trial Update I
The content of this article reflects the personal opinion of the author/s and is not necessarily the official position of the European Society of Cardiology.