Scholarlink

Background <p>The number of patients undergoing transcatheter aortic valve implantation (TAVI) is rising. While TAVI improves survival, symptoms, and quality of life, ESC guidelines highlight the lack of validated tools to predict futile outcomes. Around 20% of intermediate- and high-risk patients die within one-year post-TAVI, underscoring the need for robust risk stratification.</p> Objectives <p>To develop and validate a multivariable model predicting one-year mortality after TAVI, using real-world data from two Swiss centers.</p> Methods <p>We analysed 2,053 patients who underwent TAVI between 2014 and 2022 at two centers in the Swiss TAVI Registry. The primary endpoint was all-cause mortality at one-year. Candidate variables were selected based on a 2024 scoping review, clinical expertise, and data availability. Variables associated with mortality (<i>p</i> &lt; 0.2) in univariate Cox regression were included in a multivariate Cox model. Model performance was evaluated using ROC curves and Youden’s index.</p> Results <p>At 30 days, independent predictors of mortality were reduced left ventricular ejection fraction, COPD, peripheral artery disease, and age. At one-year, mortality was associated with atrial fibrillation, lower LVEF, mitral regurgitation grade, age, corticosteroid use, coronary artery disease, insulin-dependent diabetes, and reduced glomerular filtration rate. The model achieved an AUC of 0.814 for 30-day mortality and 0.763 at one-year, outperforming the STS score for short-term prediction (AUC 0.785).</p> Conclusion <p>This study demonstrates the feasibility of building a one-year mortality prediction model after TAVI using standard clinical variables. With 2,053 patients from two centers, the model is a promising step toward Swiss-specific risk tools. Incorporating frailty or functional status may further improve performance. This model may support heart teams in shared decision-making.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Predicting one-year outcomes after TAVI: evidence from a two-site Swiss registry

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2025年8月30日

查看全文
  • Charlie Ferry,
  • Dorian Garin,
  • Roberto Corti,
  • Aurélien Clerc,
  • Mario Togni,
  • Peter Wenaweser,
  • Stefan Stortecky,
  • Stephane Cook

摘要

Background

The number of patients undergoing transcatheter aortic valve implantation (TAVI) is rising. While TAVI improves survival, symptoms, and quality of life, ESC guidelines highlight the lack of validated tools to predict futile outcomes. Around 20% of intermediate- and high-risk patients die within one-year post-TAVI, underscoring the need for robust risk stratification.

Objectives

To develop and validate a multivariable model predicting one-year mortality after TAVI, using real-world data from two Swiss centers.

Methods

We analysed 2,053 patients who underwent TAVI between 2014 and 2022 at two centers in the Swiss TAVI Registry. The primary endpoint was all-cause mortality at one-year. Candidate variables were selected based on a 2024 scoping review, clinical expertise, and data availability. Variables associated with mortality (p < 0.2) in univariate Cox regression were included in a multivariate Cox model. Model performance was evaluated using ROC curves and Youden’s index.

Results

At 30 days, independent predictors of mortality were reduced left ventricular ejection fraction, COPD, peripheral artery disease, and age. At one-year, mortality was associated with atrial fibrillation, lower LVEF, mitral regurgitation grade, age, corticosteroid use, coronary artery disease, insulin-dependent diabetes, and reduced glomerular filtration rate. The model achieved an AUC of 0.814 for 30-day mortality and 0.763 at one-year, outperforming the STS score for short-term prediction (AUC 0.785).

Conclusion

This study demonstrates the feasibility of building a one-year mortality prediction model after TAVI using standard clinical variables. With 2,053 patients from two centers, the model is a promising step toward Swiss-specific risk tools. Incorporating frailty or functional status may further improve performance. This model may support heart teams in shared decision-making.