Background: The aim of the study was to compare 7 available risk models in the prediction of 30-day mortality following transcatheter aortic valve implantation (TAVI). Heart team decision supported by different risk score calculations is advisable to estimate the individual procedural risk before TAVI.
Methods: One hundred and fifty-six consecutive patients (n = 156, 48% female, mean age 80.03 ± 8.18 years) who underwent TAVI between March 2010 and October 2014 were in-cluded in the study. Thirty-day follow-up was performed and available in each patient. Base-line risk was calculated according to EuroSCORE I, EuroSCORE II, STS, ACEF, Ambler's, OBSERVANT and SURTAVI scores.
Results: In receiver operating characteristics analysis, neither of the investigated scales was able to distinguish between patients with or without an endpoint with areas under the curve (AUC) not exceeding 0.6, as follows: EuroSCORE I, AUC 0.55; 95% confidence intervals (CI) 0.47-0.63, p = 0.59; EuroSCORE II, AUC 0.59; 95% CI 0.51-0.67, p = 0.23; STS, AUC 0.55; 95% CI 0.47-0.63, p = 0.52; ACEF, AUC 0.54; 95% CI 0.46-0.62, p = 0.69; Ambler's, AUC 0.54; 95% CI 0.46-0.62, p = 0.70; OBSERVANT, AUC 0.597; 95% CI 0.52-0.67, p = 0.21; SURTAVI, AUC 0.535; 95% CI 0.45-0.62, p = 0.65. SURTAVI model was calibrated best in high-risk patients showing coherence between expected and observed mortality (10.8% vs. 9.4%, p = 0.982). ACEF demonstrated best classification accuracy (17.5% vs. 6.9%, p = 0.053, observed mortality in high vs. non-high-risk cohort, respectively).
Conclusions: None of the investigated risk scales proved to be optimal in predicting 30-day mortality in unselected, real-life population with aortic stenosis referred to TAVI. This data supports primary role of heart team in decision process of selecting patients for TAVI.
Keywords: aortic stenosis; risk prediction; transcatheter aortic valve implantation.