# Patient risk evaluation for transcatheter aortic valve replacement (PRE-TAVR) — identification of real-time predictors of short- and long-term mortality

## TL;DR

This study developed a new risk model (PRE-TAVR) to better predict short- and long-term mortality after TAVR procedures, outperforming existing scores.

## Contribution

A novel, more accurate risk prediction model for TAVR patients using real-time clinical variables.

## Key findings

- The PRE-TAVR model achieved an AUC of 0.770 for 1-year mortality, outperforming EuroSCORE II and STS/ACC TAVR scores.
- For 30-day mortality, the model showed an AUC of 0.699, slightly better than EuroSCORE II and comparable to the STS/ACC TAVR score.
- Key predictors included age, NYHA stage, COPD, atrial fibrillation, and biomarkers like C-reactive protein and platelet count.

## Abstract

The steadily increasing number of transcatheter aortic valve replacement (TAVR) procedures being performed on a heterogeneous patient population highlights the need for robust risk assessment. While EuroSCORE II is well established for surgical risks, it is less effective for TAVR, and the newer STS/ACC TAVR score has so far been validated mainly for in-hospital and 30-day mortality.

This study aims to improve risk stratification for TAVR patients by identifying real-time predictors of 30-day and 1-year mortality that incorporate comprehensive, procedure-specific factors.

Five-year data from 2256 transfemoral TAVR procedures performed at two German Heart Centers (2017–2022) were retrospectively analyzed. Predictors of 1-year and 30-day mortality were assessed using multivariable logistic and LASSO regression, considering a broad spectrum of patient demographics, comorbidities, and peri-procedural factors.

The analyses revealed a predictor model (PRE-TAVR predictors) for 1-year mortality (AUC 0.770; 95% CI 0.731–0.809), including age (> 81.5 years), NYHA stage IV, COPD (GOLD ≥ 2), atrial fibrillation, previous stroke or malignancy, elevated C-reactive protein (≥ 9.5 mg/L), aortic valve ΔP mean ≥ 48.5 mmHg, peripheral arterial disease (> stage 2) and low platelet count (≤ 228.5 g/L). The accuracy of the model exceeded the EuroSCORE II (AUC 0.645; 95% CI 0.599–0.691) and the STS/ACC TAVR score (AUC 0.714; 95% CI 0.670–0.758). For 30-day mortality, NYHA class IV was the only significant predictor in the bivariate analyses. However, additional LASSO analyses identified pre-existing renal insufficiency (KDIGO stage ≥ 3) and pre-TAVR sodium levels as further significant predictors. The AUC was 0.699 (95% CI 0.611–0.788) compared to an AUC of 0.680 (95% CI 0.604–0.756) for EuroSCORE II and 0.7129 (95% CI 0.633–0.793) for the STS/ACC TAVR score.

The PRE-TAVR study developed a robust model, particularly for predicting 1-year mortality. This model outperformed the EuroSCORE II and STS/ACC TAVR scores, despite requiring fewer variables. It provides a solid basis for future risk scores and enables more precise patient selection.

The online version contains supplementary material available at 10.1007/s00392-025-02704-6.