The Society of Thoracic Surgery Risk Score as a Predictor of 30-Day Mortality in Transcatheter vs Surgical Aortic Valve Replacement: A Single-Center Experience and its Implications for the Development of a TAVR Risk-Prediction Model

Read more about risk assessment and the Society of Thoracic Surgery (STS) Scoring

• Secondary Femoral Access Hemostasis During Transcatheter Aortic Valve Replacement: Impact of Vascular Closure Devices

• Patient Risk Assessment for Transcatheter Aortic Valve Replacement at Veterans Health Administration Hospitals

• Comparison of 26-mm Evolut and 23-mm Sapien 3 Valves in TAVR for Small Aortic Annulus

Abstract

Background. The Society of Thoracic Surgery (STS) risk score serves as an important determinant of eligibility for transcatheter aortic valve replacement (TAVR). The STS score’s validity for predicting TAVR mortality, however, is incompletely understood. This study compares the STS score’s discriminatory power for TAVR mortality as compared with surgical aortic valve replacement (SAVR) mortality. Methods. A retrospective analysis of STS score and 30-day mortality for TAVR patients (n = 426) and SAVR patients (n = 297) at a single institution was performed. The performance of the STS score was evaluated from the standpoint of discriminatory power. The predictive ability of STS for 30-day mortality was detected by generation of receiver operator characteristic (ROC) curves. Results. The STS score possesses predictive ability for 30-day SAVR mortality with an area under the ROC curve of 0.791 (95% confidence interval [CI], 0.690-0.893). The STS score also possesses predictive ability for 30-day TAVR mortality with an area under the ROC curve of 0.674 (95% CI, 0.541-0.807). When stratifying TAVR by access route, the STS score for transfemoral TAVR provides an area under the ROC curve of 0.789 (95% CI, 0.569-1.000). There is not a statistically significant difference in predictive ability between SAVR and TAVR. Conclusion. The STS score possesses predictive value for 30-day mortality in both SAVR and TAVR. Although not designed for TAVR, the STS score may provide some insight into TAVR mortality, and therefore serves as an appropriate model for efforts to develop a TAVR-specific risk prediction instrument. 

J INVASIVE CARDIOL 2017;29(3):109-114

Key words: transcatheter aortic valve replacement, surgical aortic valve replacement, risk prediction

Transcatheter aortic valve replacement (TAVR) has become an accepted alternative therapy for patients with severe aortic stenosis deemed at elevated risk for surgical aortic valve replacement (SAVR). Recently, the Sapien  3 prosthesis (Edwards Lifesciences) has been approved for intermediate or higher-risk patients. The preprocedure assessment of surgical risk is made by a “heart team” using the Society of Thoracic Surgery (STS) score as a guide. The STS score is a validated risk-prediction model for open surgery based on data from the STS National Adult Cardiac Surgery Database.¹ In general, an STS predicted risk of surgical mortality of 4%-8% is considered intermediate risk and 8% or greater is considered high risk. The predictive value of the STS score in the setting of TAVR has previously been examined, but its utility in contemporary practice has not been clearly defined.

Analysis of the PARTNER and CoreValve US Pivotal trial data identifies an STS score of 15% or more as a threshold beyond which the mortality benefit of TAVR is impaired.^2,3 Yet studies of the predictive power of the STS score in the setting of TAVR have yielded discrepant results.^4,5 Given this ambiguity in the value of the STS score, it is necessary to more closely analyze the value of the STS score in the setting of TAVR. While the STS score should not substitute for the development of a TAVR-specific risk predictor, understanding the predictive value of the STS score may help the TAVR community use the knowledge gained from the STS process to inform the ongoing development of a TAVR risk prediction model. The goal of this study, therefore, was to compare the STS score’s predictive ability in the setting of TAVR as compared with its predictive ability in the setting of SAVR by retrospectively analyzing both TAVR and SAVR outcomes by STS score at a single large-volume valvular and structural heart disease center. 

Methods

STS scores and 30-day mortality were gathered from 426 consecutive patients who underwent TAVR at a single center from November 2011 to December 2014 using the balloon-expandable Edwards Sapien line of bioprostheses (Edwards Lifesciences). STS scores and 30-day mortality data were also obtained from 297 patients who underwent SAVR at the same institution from 2004-2014. The primary endpoint for all TAVR and SAVR patients was 30-day mortality. We obtained institutional review board (IRB) approval for retrospective analysis of the above-mentioned data. 

The STS score was taken from source documentation at the time of procedure planning so as to accurately capture the risk determination upon which decisions were made at the time of AVR. The current STS score is based on 24 variables derived from voluntarily reported surgical registry data in the United States.^1,5 Over the period from 2004-2014, the STS scoring process did change, meaning that the STS scores for the surgical cohort are not uniform in character. The STS scores for the TAVR cohort are based on a single STS model and are thus uniform in character. Due to limitations in the availability of source documentation, it was not possible to retrospectively recalculate STS scores for all patients in the SAVR cohort using a single uniform model. While this is an admitted limitation of the present study, the authors felt it was important to capture the STS risk estimation made by the operators at the time of work-up for AVR, as it was this number that was used by the operators to make decisions at the time of valve replacement. It was thus decided to take all STS scores at face value despite the fact that different iterations of the STS model may have been used for different patients in the SAVR cohort. Among the TAVR cohort, although an STS score of 8% or more in general correlated with high surgical risk, some patients with lower STS scores were deemed high-risk surgical candidates due to anatomic features such as midline left internal mammary artery (LIMA) graft, prior chest radiation, porcelain aorta, or frailty. An integrated multidisciplinary heart-valve team led by a group of cardiothoracic surgeons and interventional cardiologists made all decisions regarding triage for TAVR vs SAVR.

The performance of the STS score was evaluated in terms of discrimination. Discriminatory power was evaluated using the C-index, area under (AUC) the receiver-operating characteristic (ROC) curve with its 95% confidence interval (CI). A C-index of 0.5 indicates the absence of predictive ability, while a C-index of 1.0 represents perfect discriminatory ability. U-statistics were used to investigate differences in C-statistics across different scores. All data were analyzed using SPSS software, version 23 (IBM, Inc). 

Results    

Of the 426 patients who underwent TAVR, the mean STS score measured was 9.6% and of the 297 patients who underwent SAVR the mean STS score was 4.4%. The observed 30-day mortality was 4.2% (18 patients) in the TAVR group and 4.7% (14 patients) in the SAVR group (Table 1). As expected, the mean STS score was relatively high in the TAVR group, whereas the mean STS score for the SAVR group fell in the intermediate range. The two cohorts were statistically significantly different from one another (P<.001) in terms of STS score. However, the observed 30-day mortality rates in the two groups were not statistically significantly different from one another.

With respect to discriminatory power, the STS score demonstrated reasonable predictive ability in both the SAVR and TAVR populations. As expected, the STS score possesses strong predictive ability for 30-day SAVR mortality, with an area under the ROC curve of 0.791 (95% CI, 0.690-0.893; P<.001). The STS score also demonstrated reasonable predictive ability for 30-day TAVR mortality, with an area under the ROC curve of 0.674 (95% CI, 0.54-0.807; P=.01). When stratifying TAVR by access route, the STS score again demonstrated reasonable predictive ability for 30-day transfemoral (TF)-TAVR mortality, with an area under the ROC curve of 0.789 (95% CI, 0.569-1.000; P=.047). With respect to transapical (TA)-TAVR, however, the STS score displayed poor discriminatory ability, with an area under the ROC curve of 0.583 (95% CI. 0.316-0.851; P=.46) (Table 2). Figure 1 graphically represents these data. Comparison of the AUC values for TAVR (0.674) and SAVR (0.791) yielded a P=.17, indicating that there was not a statistically significant difference in the STS score’s predictive ability in the setting of SAVR vs TAVR. Furthermore, comparison of the AUC values for TF-TAVR (0.789) and SAVR (0.791) yielded a P=.99, again indicating that there was not a statistically significant difference in the STS score’s predictive ability in the setting of SAVR vs TF-TAVR. Post-TAVR mortality does increase with increasing STS score when stratifying the TAVR population into low-to-intermediate risk, high-risk, and extreme-risk cohorts (Table 3). 

Discussion

The predictive ability of the STS score for TAVR was recognized as early as 2012 with analysis of data from cohort B of the original PARTNER trial, which randomized inoperable patients to TF-TAVR vs standard (medical) therapy.⁶ In reporting the 2-year outcomes from PARTNER cohort B, Makkar et al noted that although the mortality benefit of TAVR as compared with standard therapy remained substantial at 2 years, the survival benefit of TAVR diminished with increasing STS score.² Stratification by STS score (<5%, 5%-14.9%, ≥15%) revealed that TAVR patients with STS scores <15% derived a substantial survival benefit relative to those treated with standard therapy, but those with STS scores ≥15% saw no mortality benefit when compared with those treated with standard therapy.² Similarly, multivariate analysis of the 2-year outcomes from the CoreValve US Pivotal trial extreme-risk cohort revealed that an STS score ≥15% was predictive of 2-year all-cause mortality.³ It is therefore well understood that the STS score has some value in the setting of TAVR.

Further exploration of the STS score applied to TAVR, however, apparently revealed poor calibration and imperfect discriminatory ability with respect to TAVR mortality. Beohar et al evaluated the predictive ability of the STS score using data from cohort A of the PARTNER trial. They compared the predicted vs observed 30-day and 1-year mortality in the TAVR and SAVR patients treated in PARTNER. Given an average STS score of 11.4% for TAVR patients and 11.7% for SAVR patients, they observed a 30-day mortality rate of 6.5% in the TAVR arm and 10.5% in the SAVR arm. This produced an observed-to-predicted ratio of 0.57 in the TAVR arm and 0.89 in the SAVR arm. Based on these findings, the authors concluded that the STS score overestimated 30-day mortality in TAVR patients and thus the STS score is poorly calibrated for TAVR. Furthermore, using ROC curves with associated areas AUC, these authors found the STS score to be a poor discriminator for mortality both post TAVR (AUC, 0.60) and post SAVR (AUC, 0.58).⁴ These authors therefore urged the development of a TAVR-specific risk prediction instrument.

The data on the calibration and discriminatory power of the STS score for TAVR risk prediction are, however, discrepant. An analogous but smaller European study examining the predictive power of the STS score in a series of TAVR patients at a single center found good calibration but only moderate discriminatory ability. Durand et al looked at 250 consecutive patients undergoing TAVR in Rouen, France and found an average STS score of 7.3% with an observed mortality rate of 7.6%, thereby indicating good calibration. When examining discriminatory ability, however, the ROC curve showed an AUC of only 0.58, indicating that the STS score possesses unimpressive discriminatory power. These authors concluded that “Additional efforts are therefore required to establish a new ‘TAVR score’ to identify patients who will benefit from TAVR and to better predict mortality in individual patients.”⁵

Unlike the studies by Beohar et al and Durand et al, however, the present analysis demonstrates that the STS score does possess discriminatory power with respect to 30-day post-TAVR mortality. Rather than dispensing with the STS score then, it may be necessary to return to the STS score and reexamine its value in the setting of TAVR. The population studied here consists of two cohorts, TAVR patients and SAVR patients. Chronologically, the SAVR cohort extended from 2004 to 2014, thereby spanning both the pre-TAVR and post-TAVR eras. The mean STS score of 4.4% for the SAVR cohort likely reflects that in the pre-TAVR era, some high-risk patients with high STS scores may not have been offered surgery. It is well documented that anywhere from 40%-60% of patients with severe aortic stenosis were not offered SAVR due to excessive surgical risk.^7,8 It is precisely this unmet need that led to the development of TAVR. Following the advent of TAVR, few high STS score patients went for SAVR as the high-risk population was preferentially treated with TAVR. There is no question that the SAVR and TAVR cohorts studied here represent statistically significantly different risk categories. This is to be expected, given that TAVR and SAVR were intended to serve different patient populations for the time period retrospectively reviewed in this study. 

From the standpoint of discriminatory power, however, the STS score’s predictive ability in the setting of TAVR is not statistically significantly different from that in SAVR (AUC, 0.674 vs 0.791, respectively; P=.17). In other words, the STS score does possess some predictive power for TAVR. Interestingly, the STS score proved a better mortality predictor for TF-TAVR (AUC, 0.789) than TA-TAVR (AUC, 0.583). Excluding TA-TAVR, the AUC values for TF-TAVR and SAVR (0.789 vs 0.791; P=.99) are statistically indistinguishable from one another. These findings are different from what was found in prior studies examining the value of the STS score for TAVR risk assessment.^4,5 Unlike the prior studies, data from this single-center experience suggest that the STS score has value in TAVR risk assessment, particularly from the standpoint of discrimination. 

To understand why this may be the case, it is useful to look back at the original study examining the relationship between the STS score and TAVR mortality. From the PARTNER cohort B 2-year data, it was clear that an STS score ≥15% portended a worse outcome and loss of the survival benefit than was seen with lower-risk patients.² In other words, the STS score appeared sensitive to risk in an extremely high-risk cohort. Examination of our data stratifying by STS score <8%, between 8%-15%, and ≥15% raises similar concerns. Table 3 summarizes these findings and confirms a significantly higher 30-day post-TAVR mortality rate in those with STS scores ≥15%. 

These data indicate that patients at very high risk for SAVR also possess sufficient co-morbidities that they may be higher-risk TAVR candidates. While it is true such patients will do better with TAVR because of the less-invasive nature of the procedure, there remains a corresponding level of increase in risk with TAVR. While the risk with TAVR is less than with SAVR, as we rise through the continuum of risk, the STS score appears sensitive to that increasing risk in the setting of TAVR. 

However, the STS score should not be used as a simple surrogate for a TAVR risk prediction model. The STS score is based on registry data from SAVR and is designed to predict risk in the setting of SAVR. Rather, as the TAVR community develops a TAVR-specific risk-assessment tool, the elements of the STS score that provide value in the setting of all forms of AVR should be retained and examined in the setting of TAVR registry data.

The difference in the predictive ability of the STS score for TF-TAVR as compared with TA-TAVR also deserves some attention. The STS score possesses poor discriminatory power with respect to TA-TAVR (AUC, 0.583). The TA-TAVR cohort represents those patients with insufficient ileo-femoral access to undergo TF-TAVR. The insensitivity of the STS score for risk prediction in this cohort may reflect the possibility that this cohort represents a different category of risk. It is well documented that when comparing TAVR outcomes between evenly matched TA and TF patients, TAVR outcomes are poorer among TA patients.⁹ The STS score is based on surgical registry data from patients who underwent SAVR. The patients triaged to TA-TAVR, such as vasculopaths with advanced peripheral arterial disease, may not be well represented in the STS registry framework. If this supposition is accurate, it lends credence to the notion that a TAVR risk-prediction instrument should be based on registry data derived from patients undergoing TAVR. 

The TAVR literature has identified numerous individual risk factors associated with impaired outcomes.^10-17 Furthermore, a number of groups have attempted to develop TAVR risk-prediction scores.^18-21 Most recently, the Steering Committee of the Society of Thoracic Surgery/American College of Cardiology (STS/ACC) Transcatheter Valve Therapy (TVT) registry published a risk-prediction model for post-TAVR in-hospital mortality.²² Employing the TVT registry data gathered from all commercial TAVRs performed in the United States, this group identified a series of variables – each independently predictive of in-hospital mortality after TAVR. These variables include age, glomerular filtration rate (GFR), hemodialysis, New York Heart Association functional class, chronic lung disease, non-femoral access, and procedural acuity. Procedural acuity was assessed based on the presence of cardiogenic shock, need for inotropes, need of a mechanical assist device, or recent cardiac arrest. This TVT score constitutes the first rigorous statistical risk-prediction model based on data from a very large population of patients who have undergone TAVR.²² Interestingly, most of the elements of the TVT score are also elements of the STS score. Furthermore, the TVT score includes the absence of femoral access as a factor increasing TAVR risk. It is therefore not surprising that the STS score possesses predictive power with respect to mortality post TAVR, particularly post TF-TAVR.

While it may appear obvious that the STS score should possess predictive power for post-TAVR mortality, the present study is the only analysis that has been able to demonstrate this correlation.^4,5 Furthermore, stratification by route of access makes it clear that the correlation between STS score and post-TAVR mortality is not a simple one. Given that the STS score possesses predictive power in the setting of TF-TAVR but not TA-TAVR, the STS score may inform “heart team” decisions regarding TF patients, but may not be as useful in those patients lacking femoral access. The STS score thus possesses some value in TAVR risk estimation, but should not substitute for the development of a more comprehensive TAVR risk-prediction instrument. 

Conclusion

The present analysis demonstrates that the STS score possesses discriminatory power for 30-day post-TAVR mortality. When stratifying TAVR by route of access (TF vs TA), it becomes clear that the predictive power of the STS score is limited to 30-day mortality post TF-TAVR. The STS score may therefore be used as a rough guidepost of TF-TAVR risk. Given the lack of predictive power for TA-TAVR, however, the STS score cannot be used as a simple surrogate for a TAVR risk-prediction model. The STS score is based on registry data from patients undergoing SAVR and is designed to predict risk in the setting of SAVR. Furthermore, the STS score is based on surgical registry data in which the bulk of the patients possessed low-to-intermediate risk scores. Since fewer high-risk patients underwent SAVR, the STS score inherently under-represents the higher-risk population treated by TAVR. Building a TAVR-specific risk calculator will therefore require incorporating many of the valuable aspects of the STS score while also including those elements not captured by the STS score. Only analysis of granular TAVR data with appropriate weighting of the individualized risk factors will yield a highly calibrated and discriminative TAVR risk-prediction tool. 

While the present study offers some insight into the meaning of the STS score in the setting of TAVR, it has a number of limitations. First, this is a single-center, non-randomized, retrospective analysis. Furthermore, the STS scores compared between the SAVR and TAVR cohorts were not recalculated based on a uniform version of the STS model due to limitations in the availability of historical data. The TAVR cohort also saw rapid progression in technology over the time period analyzed, meaning that the TAVR cohort was treated with a series of different devices with later platforms potentially exposing patients to lesser risk. Despite these limitations, the present analysis does provide some insight into the value of an existing risk-prediction model applied to a new cohort of patients. 

With the exponential growth in the volume of TAVR and continued improvement in outcomes, there is no question that the data will exist to build a robust TAVR risk-prediction model.²³ The recently developed TVT in-hospital mortality risk-assessment tool is a useful first step.²² The present study highlights the value of retaining those elements of the STS score that have predictive value for TAVR in the development of a TAVR risk-prediction model. While the present study possesses a number of limitations, we believe these data offer some insight into the meaning of the STS score in the setting of TAVR. Furthermore, finding predictive value in the application of the STS score to TAVR only encourages the cardiovascular community to focus on similar risk factors and employ similar methods in developing a TAVR risk-prediction model. Just as the cardiovascular community realized that some patients posed excessive risks for SAVR, so too must they recognize that some patients, burdened by numerous co-morbidities, pose excessive risks for TAVR.^24,25 While the STS score provides a glimpse into who some of those patients might be, the current effort to build a TAVR-specific risk-prediction model based on TVT registry data in a manner analogous to the STS score will significantly improve our clinical judgment. This will allow us to optimize patient selection for TAVR and thus optimize TAVR outcomes.

Related Reading

• Reframing Optimal Implantation of the Sapien 3 Transcatheter Heart Valve
• Transcatheter Aortic Valve Replacement for Mixed Aortic Valve Disease: A Propensity Score-Adjusted Analysis From the RISPEVA Registry
• Comparison of 26-mm Evolut and 23-mm Sapien 3 Valves in TAVR for Small Aortic Annulus
• Outcomes of Radial Versus Femoral Access in Patients With Severe Aortic Stenosis Undergoing Percutaneous Coronary Intervention Prior to Transcatheter Aortic Valve Replacement

References

 1.    O’Brien SM, Shahian DM, Filardo G, et al. The Society of Thoracic Surgeons 2008 cardiac surgery risk models: part 2 — isolated valve surgery. Ann Thorac Surg. 2009;88(1 Suppl):S23-S42. 

2.    Makkar RR, Fontana GP, Jilaihawi H, et al. Transcatheter aortic-valve replacement for inoperable severe aortic stenosis. N Engl J Med. 2012;366:1696-1704. 

3.    Yakubov SJ, Adams DH, Watson DR, et al. 2-year outcomes after iliofemoral self-expanding transcatheter aortic valve replacement in patients with severe aortic stenosis deemed extreme risk for surgery. J Am Coll Cardiol. 2015;66:1327-1334. 

4.    Beohar N, Whisenant B, Kirtane AJ, et al. The relative performance characteristics of the logistic European System for Cardiac Operative Risk Evaluation score and the Society of Thoracic Surgeons score in the placement of aortic transcatheter valves trial. J Thorac Cardiovasc Surg. 2014;148:2830-2837. 

5.    Durand E, Borz B, Godin M, et al. Performance analysis of EuroSCORE II compared to the original logistic EuroSCORE and STS scores for predicting 30-day mortality after transcather aortic valve replacement. Am J Cardiol. 2013;111:891-897.

6.    Leon MB, Smith CR, Mack M, et al. Transcatheter aortic valve implantation for aortic stenosis in patients who cannot undergo surgery. N Engl J Med. 2010;363:1597-1607. 

7.    Pellika PA, Sarano ME, Nishimura RA, et al. Outcome of 622 adults with asymptomatic, hemodynamically significant aortic stenosis during prolonged follow-up. Circulation. 2005;111:3290-3295.

8.    Freed BH, Sugeng L, Furlong K, et al. Reasons for non-adherence to guidelines for aortic valve replacement in patients with severe aortic stenosis and potential solutions. Am J Cardiol. 92010;105:1339-1342.

9.    Blackston EH, Suri RM, Rajeswaran J, et al. Propensity-matched comparison of clinical outcomes after transapical or transfemoral transcatheter aortic valve replacement: a placement of aortic transcatheter valves (PARTNER)-I trial substudy. Circulation. 2015;131:1989-2000. 

10.    Rodes-Cabau J, Webb JG, Cheung A, et al. Transcatheter aortic valve implantation for the treatment of severe symptomatic aortic stenosis in patients at very high or prohibitive surgical risk: acute and late outcomes of the multicenter Canadian experience. J Am Coll Cardiol. 2010;55:1080-1090. 

11.    Moat NE, Ludman P, de Belder MA, et al. Long-term outcomes after transcatheter aortic valve implantation in high-risk patients with severe aortic stenosis: the UK TAVI (United Kingdom Transcatheter Aortic Valve Implantation) registry. J Am Coll Cardiol. 2011;58:2130-2138. 

12.    Tamburino C, Capodanno D, Ramondo A, et al. Incidence and predictors of early and late mortality after transcatheter aortic valve implantation in 663 patients with severe aortic stenosis. Circulation. 2011;123:299-308. 

13.    Thomas M, Schymik G, Walther T, et al. One-year outcomes of cohort 1 in the Edwards Sapien Aortic Bioprosthesis European Outcome (SOURCE) registry: the European registry of transcatheter aortic valve implantation using the Edwards Sapien valve. Circulation. 2011;124:425-433. 

14.    Green P, Woglom AE, Généreux P, et al. The impact of frailty status on survival after transcatheter aortic valve replacement in older adults with severe aortic stenosis: a single-center experience. JACC Cardiovasc Interv. 2012;5:974-981. 

15.    Pilgrim T, Kalesan B, Wenaweser P, et al. Predictors of clinical outcomes in patients with severe aortic stenosis undergoing TAVI: a multistate analysis. Circ Cardiovasc Interv. 2012;5:856-861. 

16.    Dewey TM, Brown DL, Herbert MA, et al. Effect of concomitant coronary artery disease on procedural and late outcomes of transcatheter aortic valve implantation. Ann Thorac Surg. 2010;89:758-767; discussion 767. 

17.    Bedogni F, Latib A, De Marco F, et al. Interplay between mitral regurgitation and transcatheter aortic valve replacement with the CoreValve Revalving System: a multicenter registry. Circulation. 2013;128:2145-2153. 

18.    Van Mieghan NM, Head SJ, van der Boon RM, et al. The SURTAVI model: proposal for a pragmatic risk stratification for patients with severe aortic stenosis. EuroIntervention. 2012;8:258-266

19.    Capodanno D, Barbanti M, Tamburino C, et al. A simple risk tool (the OBSERVANT score) for prediction of 30-day mortality after transcatheter aortic valve replacement. Am J Cardiol. 2014;1:1851-1858.

20.    D’Ascenzo F, Capodanno D, Tarantini G, et al. Usefulness and validation of the survival post TAVI score for survival after transcatheter aortic valve implantation for aortic stenosis. Am J Cardiol. 2014;114:1867-1874.

21.    Debonnaire P, Fusini L, Wolterbeck R, et al. Value of the “TAVI2-score” versus surgical risk scores for prediction of one-year mortality in 511 patients who underwent transcatheter aortic valve implantation. Am J Cardiol. 2015;115:234-242. 

22.    Edwards FH, Cohen DJ, O’Brien SM, et al. Development and validation of a risk prediction model for in-hospital mortality after transcatheter aortic valve replacement. JAMA Cardiol. 2016;1:46-52.

23.    Holmes DR, Nishimura RA, Grover FL, et al. Annual outcomes with transcatheter valve therapy. J Am Coll Cardiol. 2015;66:2813-2823.

24.    Herrmann HC, Han Y. Identifying patients who do not benefit from transcatheter aortic valve replacement. Circ Cardiovasc Interv. 2014;7:136-138. 

25.    Mack MJ, Holmes DR, Webb JG, et al. Patient selection for transcatheter aortic valve replacement. J Am Coll Cardiol. 2013;62:1-10.

From the ¹Department of Internal Medicine, Division of Cardiology; and the ²Department of Cardiothoracic Surgery, The McGovern Medical School at The University of Texas Health Science Center Houston, Memorial Hermann Texas Medical Center Heart & Vascular Institute, Houston, Texas.

Disclosure: The authors have completed and returned the ICMJE Form for Disclosure of Potential Conflicts of Interest. Dr Nguyen reports consulting/proctor fees from Edwards LifeSciences and St. Jude Medical. The remaining authors report no conflicts of interest regarding the content herein.

Manuscript submitted October 2, 2016, provisional acceptance given December 5, 2016, final version accepted January 11, 2017.

Address for correspondence: Prakash Balan, MD, JD, 6431 Fannin Street, MSB 1.257, Houston, TX 77030. Email: prakash.balan@uth.tmc.edu