Models of prognostic risk prediction have been widely used in the cardiovascular field, and several risk scores have been developed to predict the risk of short-term mortality associated with cardiac surgery on the basis of patients’ preoperative characteristics (table 1).1–8 Historically, the Society of Thoracic Surgeons Predicted Risk of Mortality (STS-PROM) score and the logistic European System for Cardiac Operative Risk Evaluation (EuroSCORE, now replaced by the EuroSCORE II) have been the most commonly used. These scores are simple to apply and have been widely adopted in the decision-making process of patients being evaluated for cardiac surgery.

Transcatheter aortic valve implantation (TAVI) has emerged as a less invasive treatment for patients with severe symptomatic aortic stenosis, and surgical risk scores have been widely used to identify those patients at high or prohibitive surgical risk who may benefit from this procedure.9 Therefore, the inclusion of patients in registries and randomised trials in the TAVI field has been mainly based on such surgical risk scores (particularly, the EuroSCORE and STS-PROM score), in addition to a thorough evaluation of the patient by the heart team. Furthermore, several studies have determined the potential usefulness of using such risk scores to predict clinical outcomes following TAVI, but they have systematically shown a relatively low accuracy, particularly the logistic EuroSCORE and the STS-PROM score, in the prediction of acute and midterm mortality (table 2).10–13 Also, no data exist on the prospective validation of surgical risk scores in the TAVI population. In fact, many studies found that some single risk factors were superior to the multifactorial surgical risk scores for the prediction of poor outcomes among TAVI candidates.9 The inclusion of a high proportion of octogenarians in TAVI studies, who were highly underrepresented in the cohorts used for the development of such surgical risk scores, and the limited data on the discriminatory and calibration power within this zone of (higher) risk have been argued as important reasons for the low prediction capacity of such scores.10–13

The development of a global TAVI risk score is therefore of major clinical importance in improving the patient selection process in patients considered otherwise at high or prohibitive surgical risk. Iung et al14 have evaluated the predictive factors of 30-day mortality following TAVI in 3833 consecutive patients from the French Aortic National CoreValve and Edwards (FRANCE 2) registry,15 and developed and validated a multifactorial predictive risk model for such patients. The authors identified up to nine pre-procedural factors associated with early mortality: age ≥90 years, body mass index <30 kg/m², four factors related to an impaired cardiovascular status (New York Heart Association class IV, pulmonary hypertension, critical haemodynamic state, ≥2 pulmonary oedemas during the last year), two non-cardiac co-morbidities (respiratory insufficiency, dialysis), and one factor related to the procedural strategy (non-transfemoral approach). Each factor was weighted and integrated into a 21-point risk score, with a mortality rate <5% and >40% among patients within the lowest and highest risk scores, respectively. While most of these factors had already been individually associated with a higher risk of early mortality following TAVI in previous studies,9 this important study represents one of the first attempts to combine them into a global risk score.

The performance of a risk score is usually determined by the c-index, which evaluates the discriminatory power of the final multivariate model. The c-index obtained in the work by Iung et al was 0.67 in the development sample and 0.59 in the validation cohort for the prediction of 30-day mortality. While such a discrimination level can be considered only moderate, it seems to be somewhat more accurate than the prediction performance obtained by classical surgical risk scores applied to the TAVI field, which indeed lack validation (table 2).10–13 The reasons for such a moderate prediction performance are probably multifactorial. First, although the study included a large cohort of patients, the number of patients was much lower than that used for most surgical scores. Second, the characteristics of the TAVI population, including a high number of very old patients with a high co-morbidity burden, may have a negative influence on the discriminatory capacity of any risk score. In this respect, the lack of functional as well as frailty parameters in Iung et al's work may indeed have reduced the predictive power of the model. Interestingly, the recently published ‘AHA/ACC Guidelines for the Management of Patients with Valvular Heart Disease’ have included frailty indices in addition to surgical risk scores in order to improve risk prediction with regard to valve surgery procedures.16 Third, unlike surgical series, a significant number of patients who survive TAVI die within the first months following the procedure: in FRANCE-2,15 the mortality rate at 1-year follow-up was 24%; ∼10% of patients died within the first 30 days, and up to 14% died within the months following the procedure. This can indeed be considered in part a failure in patient selection.

One may wonder whether a predictive model taking into account those deaths occurring within the few months following the procedure might be more appropriate and associated with a higher discriminatory capacity for the current TAVI population. Also, it has been shown that a significant number of TAVI patients fail to experience any improvement in quality of life and/or functional status following TAVI,17 ,18 so that the lack of improvement in quality of life at follow-up in addition to death as markers of poor outcomes may be further explored in future studies for the development of a risk score in TAVI candidates. Finally, while the increase in the experience of centres/operators and improvements in technology have been associated with a major reduction in procedural complications, TAVI is still associated with a relatively high rate of peri-procedural complications (vascular complications, life threatening bleeding, stroke, etc) that have been associated with a higher early and midterm mortality rate. These complications, which are only partially determined by baseline characteristics, play a role in the 30-day mortality associated with this procedure and may further reduce the accuracy of a predictive model based on pre-procedural factors.

The TAVI field has evolved in the treatment of elderly patients with a high risk profile, classically considered to be poor candidates for standard aortic valve surgery, which undoubtedly makes patient selection one of the most challenging aspects in the current TAVI era. The work of Iung et al14 represents an important step forward towards better and more objective risk estimation of such patients, and the implementation of the multiparametric FRANCE-2 risk score in clinical practice should help heart teams in the (commonly not easy) clinical decision-making process regarding the selection of TAVI candidates. In the near future, the addition of functional/frailty parameters, the extension of the period at risk (6–12 months rather than 30 days) and the addition of futility (lack of improvement in functional status/quality of life) as a marker of poor outcomes may provide a better estimation of those patients who would not benefit from these procedures. All this should translate into improved patient selection for TAVI, which in turn should be associated with both improved outcomes and a better cost–effectiveness ratio.