Development and external validation of a preprocedural prediction model for 1-year mortality after transcatheter aortic valve implantation

Abstract

Abstract Background Transcatheter aortic valve implantation (TAVI) is a well-established treatment for severe aortic stenosis but not recommended for patients with a life-expectancy below 1 year. Nevertheless, 1 year mortality after TAVI is around 15% in current reports. Accurate prediction models are thus needed to prevent futile interventions. Methods Patients undergoing TAVI were prospectively recruited between 2008 and 2021 from a single-centre registry in Zurich, Switzerland and data split into derivation and validation datasets. Four expert cardiologists defined relevant candidate predictors. A prediction model was developed using multiple logistic regression after imputation of missing data. An independent cohort undergoing TAVI from 2008 to 2018 in London, UK served as an external validation dataset. Performance of the prediction model was assessed through internal and external validation. Receiver operating characteristic (ROC) curves and calibration plots were used for discrimination and calibration. Results A total of 1702 patients were included from the Swiss cohort. Among these 1278 were used for model development and 424 for validation. 12 predictors were incorporated including age, male sex, coronary artery disease, atrial fibrillation, chronic obstructive pulmonary disease, haemoglobin, creatinine, N-terminal pro-B-type natriuretic peptide, troponin, left ventricular ejection fraction, mean transaortic pressure gradient and severe mitral regurgitation. The prediction model achieved an area under the ROC curve (AUC) of 0.72 (95% CI 0.63 to 0.81) in internal validation and showed optimal sensitivity and specificity for 1-year mortality prediction at a probability threshold of 20%, outperforming European System for Cardiac Operative Risk Evaluation (EuroSCORE) II and Society of Thoracic Surgeons Predicted Risk of Mortality (STS-PROM) score. In the external validation dataset consisting of 851 patients from the UK cohort, the AUC was 0.64 (95% CI 0.58 to 0.70). Conclusion The TAVI-specific risk model demonstrated acceptable performance for prediction of 1-year mortality after TAVI across two contemporary real-world cohorts, highlighting its potential for clinical application instead of EuroSCORE II or STS-PROM.