Mortality prediction following transcatheter aortic valve replacement: A quantitative comparison of risk scores derived from populations treated with either surgical or percutaneous aortic valve replacement. The Israeli TAVR Registry Risk Model Accuracy A

BackgroundAccurate risk stratification is pivotal for appropriate selection of patients with severe symptomatic aortic stenosis for either surgical or transcatheter aortic valve replacement (TAVR). We sought to determine whether recent risk prediction models developed specifically in TAVR patients enhance prognostication in comparison with previous surgical scores used in clinical practice (EuroScore I, EuroScore II, STS).MethodsThe Israeli TAVR Registry Risk Model Accuracy Assessment (IRRMA) study utilized a multicenter prospective TAVR database (n = 1327) to perform a quantitative comparison between previous risk scores developed in either surgical or TAVR populations, with the present registry serving as an independent external validation set.ResultsIn the IRRMA population, 4 variables (NYHA functional class IV, chronic obstructive pulmonary disease, systolic pulmonary artery pressure ≥ 60 mm Hg, vascular access other than by the femoral route) identified by cross-validation and leave-one-out analyses provided the most discriminative model (C-statistic = 0.63) for predicting 30-day mortality. Previous scores developed in surgical (EuroScores I and II, STS), TAVR (FRANCE-2, OBSERVANT), or mixed (German AV score) populations were applied to the IRRMA cohort. Resultant C-statistics ranged between 0.52-0.71 (for the German AV and FRANCE-2 scores, respectively) and did not differ significantly (p = 0.07 for the comparison between the lowest and highest C-statistics). The observed C-statistic for 5 of these 6 scores was lower than originally reported when applied to the IRRMA population.ConclusionAvailable TAVR risk scores showed limited accuracy when applied to an independent validation set and did not enhance prognostication in comparison to previous surgical scores.