[Skip to Navigation]
Sign In
Views 2,946
Citations 0
Invited Commentary
April 2016

Predicting Outcomes in Individual Patients After Transcatheter Aortic Valve Replacement: Small Steps on the Path to Improved Decision Making

Author Affiliations
  • 1Cardiovascular Division, Department of Medicine, Brigham and Women’s Hospital, Harvard Medical School, Boston, Massachusetts
JAMA Cardiol. 2016;1(1):53-54. doi:10.1001/jamacardio.2016.0006

The application of risk scores to predict outcomes following medical and surgical interventions has become an increasingly common step in the process of clinical decision making. Ideally, risk scores can be used to inform patients and clinicians about expected outcomes in an individualized fashion. Tools can be developed to facilitate complicated and contextualized conversations with patients, and both hospitals and heart care teams can obtain risk-adjusted patient outcomes as a means for ongoing care improvement. A premium is to be placed on any tool that enhances the intertwined processes of shared decision making and informed consent, while providing a benchmark for quality improvement. However, implicit in the appropriate use of a prediction tool for these purposes is the confidence that it is valid, accurate, and robust; that it incorporates the experience of a seasoned clinician while benefiting from current information; and that it ensures the predicted outcomes are relevant.

Professional surgical and cardiology societies in both the United States and Europe have spearheaded the use of risk scores to predict outcomes following cardiac operations and percutaneous coronary interventions for many years. Indeed, the Society of Thoracic Surgeons Predicted Risk of Mortality (STS-PROM) and European System for Cardiac Operative Risk Evaluation II models are routinely used in preoperative patient assessment and planning for valvular heart surgery.1,2 These scores constitute an important aspect of the evaluation of patients for transcatheter aortic valve replacement (TAVR), a transformative and progressively improving technology initially approved for commercial use in the United States in 2011 and 2012 for the management of patients with prohibitive (STS-PROM score >15%) or high (STS-PROM score, 8%-15%) surgical risk, respectively. Additional factors that are unmeasured by the score also play an important role in determining therapy, as evidenced by a mean STS-PROM score of only 8.34% reported to the STS–American College of Cardiology Transcatheter Valve Therapy Registry (TVTR) for the more than 26 000 patients commercially treated in the United States between 2012 and 2014.3

Not unexpectedly, models based solely on prior surgical experience and data have underperformed when applied to a different patient phenotype (older and sicker) undergoing TAVR. Models combining data across TAVR and surgical aortic valve replacement (SAVR) have uncertain value, as it may not be valid to combine different techniques and divergent patient populations into a single model.4 A TAVR-specific risk model derived from a sample of 2552 patients treated from 2010 through 2011 was reported in 2014 from the French Aortic National CoreValve and Edwards-2 (FRANCE-2) registry.5 These investigators identified 9 factors predictive of 30-day or in-hospital mortality and reported similar discrimination to models cited in previous studies using SAVR risk scores applied to TAVR populations (C statistics of 0.67 for derivation set and 0.59 for validation).6

In the current issue of JAMA Cardiology, Edwards and colleagues7 report a first attempt using the TVTR to develop a model to predict in-hospital mortality for patients undergoing TAVR in the United States. The model was developed with data from 13 718 consecutive patients undergoing TAVR between November 2011 and February 2014; validation was performed among 6868 patients treated between March and October 2014. From an initial list of 39 candidate patient variables, 14 variables were selected via expert consensus, and, ultimately, 9 of these covariates were retained after statistical selection. With variables similar to those used by the FRANCE-2 investigators (age, renal function, dialysis, New York Heart Association status, chronic lung disease, no femoral access site, and procedural acuity), the model provided modest discrimination (C statistic of 0.67 for derivation set and 0.66 for validation).

The investigators emphasize several limitations to this first iteration that help place it in sharper perspective: missing data, an incomplete audit process, lack of frailty and quality-of-life indicators, and focus on an in-hospital rather than a 30-day or 1-year mortality end point. To this list should be added the lack of clarity regarding the process by which expert opinion (rather than incorporation of new findings from data analysis) drove development of the model and the absence of information regarding major in-hospital cardiac (need for permanent pacemaker or implantable defibrillator, atrial fibrillation, and cardiac arrest) and noncardiac (stroke and vascular access) complications, as well as patient-oriented outcomes (functional status and quality of life). It is encouraging to note the TVTR investigators plan to refine their model further, with the ultimate goal of creating a tool that provides a fuller picture of anticipated survival and functional outcomes for the TAVR population, the demographics of which may change considerably in the years ahead.

In most instances, the first decision facing heart care teams has to do with a patient’s candidacy for SAVR, which is usually addressed in a multidisciplinary manner, taking into account the STS-PROM (or European System for Cardiac Operative Risk Evaluation II) score plus information regarding vital organ system function, frailty, and procedure-specific obstacles. In its present iteration, the TVTR risk score provides a tool to identify individual patients with elevated in-hospital mortality. To improve its accuracy and relevance, future iterations of the risk score should stem from more complete ascertainment of patient and site factors and incorporate data-driven findings. In addition, development of models to forecast postdischarge survival and its quality would be most empowering for the patient and the heart care team.

The TVTR risk model cannot predict treatment benefit and risk between alternative strategies for management of aortic stenosis (eg, TAVR vs SAVR or TAVR vs conservative therapy). While it may be tempting to compare risks estimated from 2 models (eg, TVTR risk and STS-PROM) to help decide which type of treatment to pursue, an undertaking of this nature is likely to be fraught with limitations. This first-generation TVTR risk score was developed from patients rarely (or not at all) represented in the surgically treated population from which the STS-PROM score was derived. Inadequate overlap in the patient populations represented, as well as differences in the execution of data definitions and collection, signify that even large sample sizes and statistical methods to adjust for bias would be inadequate to allow valid comparison of treatment effects across these disparate groups. For estimation of treatment differences, one must refer primarily to randomized evidence.

The pendulum has clearly swung in favor of TAVR for high surgical risk patients. Experience in Europe demonstrates ongoing dissemination of TAVR to intermediate surgical risk patients and a similar trend is widely expected to occur in the United States following publication of completed randomized clinical trials. Trials to evaluate the benefits and risks of TAVR in low surgical risk patients are anticipated. The TVTR offers a rich data set from which to characterize TAVR patients treated in practice, whose attributes may differ from those included in randomized clinical trials. The TVTR sites can already compare their TAVR mortality outcomes against the benchmark of the combined performance of all participating institutions. A reliable risk score might in the future provide sites a method to compare their patients and outcomes more accurately to provide the local heart care team a tool for continuous quality improvement. While the currently reported risk score can be improved, Edwards and colleagues7 have marked another small yet important step on the pathway to making better decisions together with our patients.

Back to top
Article Information

Corresponding Author: Patrick T. O’Gara, MD, Cardiovascular Division, Department of Medicine, Brigham and Women’s Hospital, Harvard Medical School, 75 Francis St, Boston, MA 02115 (pogara@partners.org).

Published Online: March 9, 2016. doi:10.1001/jamacardio.2016.0006.

Conflict of Interest Disclosures: All authors have completed and submitted the ICMJE Form for Disclosure of Potential Conflicts of Interest. Dr Mauri reports grants to her institution from Abbott, Boston Scientific, and Medtronic. No other disclosures were reported.

O’Brien  SM, Shahian  DM, Filardo  G,  et al; Society of Thoracic Surgeons Quality Measurement Task Force.  The Society of Thoracic Surgeons 2008 cardiac surgery risk models, part 2: isolated valve surgery.  Ann Thorac Surg. 2009;88(1)(suppl):S23-S42.PubMedGoogle ScholarCrossref
Nashef  SA, Roques  F, Sharples  LD,  et al.  EuroSCORE II.  Eur J Cardiothorac Surg. 2012;41(4):734-744.PubMedGoogle ScholarCrossref
Holmes  DR  Jr, Nishimura  RA, Grover  FL,  et al; STS/ACC TVT Registry.  Annual outcomes with transcatheter valve therapy: from the STS/ACC TVT registry.  J Am Coll Cardiol. 2015;66(25):2813-2823.PubMedGoogle ScholarCrossref
Kötting  J, Schiller  W, Beckmann  A,  et al.  German Aortic Valve Score: a new scoring system for prediction of mortality related to aortic valve procedures in adults.  Eur J Cardiothorac Surg. 2013;43(5):971-977.PubMedGoogle ScholarCrossref
Iung  B, Laouénan  C, Himbert  D,  et al; FRANCE 2 Investigators.  Predictive factors of early mortality after transcatheter aortic valve implantation: individual risk assessment using a simple score.  Heart. 2014;100(13):1016-1023.PubMedGoogle ScholarCrossref
Ribeiro  HB, Rodés-Cabau  J.  The multiparametric FRANCE-2 risk score: one step further in improving the clinical decision-making process in transcatheter aortic valve implantation.  Heart. 2014;100(13):993-995.PubMedGoogle ScholarCrossref
Edwards  FH, Cohen  DJ, O’Brien  SM,  et al; Steering Committee of the Society of Thoracic Surgeons/American College of Cardiology Transcatheter Valve Therapy Registry.  Development and validation of a risk prediction model for in-hospital mortality after transcatheter aortic valve replacement [published online March 9, 2016].  JAMA Cardiol. doi:10.1001/jamacardio.2015.0326.Google Scholar