ARTICLE IN PRESS Transcatheter Aortic Valve Implantation Futility Risk Model Development and Validation Among Treated Patients With Aortic Stenosis Oren Zusman, MDa,b, Ran Kornowski, MDa,b,*, Guy Witberg, MDa,b, Adi Lador, MDa,b, Katia Orvin, MDa,b, Amos Levi, MDa,b, Abid Assali, MDa,b, Hana Vaknin-Assa, MDa,b, Ram Sharony, MDb,c, Yaron Shapira, MDa,b, Alexander Sagie, MDa,b, and Uri Landes, MDa,b Risk-benefit assessment for transcatheter aortic valve implantation (TAVI) is still evolving. A sizeable group of patients do not fully benefit from intervention despite a technically successful procedure. All patients who underwent TAVI with device success and with no Valve Academic Research Consortium (VARC)-2 defined complications were included. Various demographic data, clinical details, and echocardiographic findings were examined. The outcome was defined as 1-year composite of mortality, stroke, lack of functional-class improvement (by New York Heart Association class), and readmissions (?1 month after the procedure). Logistic regression was used to fit the prediction model. We used a 10-fold crossvalidation to validate our results. Of 543 patients, 435 met the inclusion criteria. The mean age was 82 (�5) years, 43% were men, and the mean Society of Thoracic Surgeons score was 6.6 (�7). At 1 year, 66 of 435 patients (15%) experienced the study end point. The final logistic regression model included diabetes, baseline New York Heart Association functional class, diastolic dysfunction, need for diuretics, mean gradient, hemoglobin level, and creatinine level. The area under the curve was 0.73 and was reduced to 0.71 after validation, with a 97% specificity using a single cutoff. Dividing to low-, medium-, and high-risk groups for futility produced a corresponding prevalence of 6%, 19%, and 59% futility. A web application for the prediction model was developed and provided. In conclusion, this prediction score may provide an important insight and may facilitate identification of patients who, despite a technically successful and uncomplicated procedure, have risk that may outweigh the benefit of a contemplated TAVI. � 2017 Elsevier Inc. All rights reserved. (Am J Cardiol 2017;??:?????) Transcatheter aortic valve implantation (TAVI) has been shown by to be a valid alternative to surgical aortic valve replacement1?4 and its use has grown exponentially.5 Although TAVI provides tremendous survival rate and symptom benefit for most, some patients die or gain little improvement in the quality of life or functional status after the procedure.6 Attention has previously been given to predicting periprocedural7,8 and 1-year9 mortality after TAVI, and to quality of life-related measures.10 In addition, there might be a substantial group of patients who would have a technically ?perfect? procedure and endure the periprocedural period, but would still not reap the expected benefits; that is, that procedure will be futile. The option to consider palliative care instead of an invasive procedure for these patients is increasingly recognized,11,12 and so is the need for an integrated benefit-risk assessment and shared decision making. We have therefore sought to identify and describe those who would a Department of Cardiology, Rabin Medical Center, Petah Tikva, Israel; b Sackler Faculty of Medicine, Tel Aviv University, Tel Aviv, Israel; and cDepartment of Cardiothoracic Surgery, Rabin Medical Center, Petah Tikva, Israel. Manuscript received July 27, 2017; revised manuscript received and accepted September 6, 2017. See page ?? for disclosure information. *Corresponding author: Tel: +972 39377107; fax: +972 39249850. E-mail address: firstname.lastname@example.org (R. Kornowski). 0002-9149/� 2017 Elsevier Inc. All rights reserved. https://doi.org/10.1016/j.amjcard.2017.09.007 have a ?futile? outcome, even though they had a perfect procedure, and develop a prediction model and calculator to try and identify these patients. Methods The TAVI screening policy at our medical center was described in detail previously.13 Briefly, each patient was initially evaluated in the clinic and then referred to a dedicated consultation ?heart team? forum that includes a multidisciplinary team. Severe aortic stenosis (AS) is defined as a valvular orifice area of <1.0 cm2 or <0.6 cm2/m2 and/or a mean pressure gradient of >40 mm Hg and/or a jet velocity of >4.0 m/s. Selected patients with low-flow, low-gradient severe AS are also considered for TAVI. All patients were assessed either by transesophageal echocardiography and electrocardiographygated cardiac computed tomography. Vascular access and coronary artery status were assessed by multislice computed tomography and peripheral and coronary angiography as needed. Both the logistic EuroScore and the Society of Thoracic Surgeons (STS) score were calculated. Risk factors not captured by traditional risk scores including frailty, porcelain aorta, severe pulmonary hypertension, hostile chest, severe liver disease or cirrhosis are also taken into account. The choice of therapy (i.e., surgical aortic valve replacement, TAVI, balloon aortic valvuloplasty, or palliative care) was tailored for each patient individually. Patients with an estimated life www.ajconline.org ARTICLE IN PRESS 2 The American Journal of Cardiology (www.ajconline.org) expectancy shorter than 1 year or a significant mental impairment were referred to palliative treatment and/or balloon aortic valvuloplasty (during our TAVI registry years, there were 38 such patients). The default access for TAVI is transfemoral and the valve choice is at the discretion of the heart team. After the index procedure, all patients are prospectively followed up in a dedicated TAVI clinic at 30 days, 6 and 12 months, and yearly thereafter. Clinical, procedural, and outcome information is prospectively collected for all patients. The institutional review board approved the prospective collection of our TAVI database (i.e., registry coronary cath Rabin database - ReCoRD TAVI institutional registry). For the present analysis, we included only patients with device success and without any major and/or debilitating complications as defined by the Valve Academic Research Consortium (VARC)-2,14 leaving only patients with an uncomplicated, optimal TAVI. The primary end point of the current study was procedural futility, defined as a composite of 1-year mortality, stroke, lack of functional-class improvement (by New York Heart Association [NYHA] class), and repeat (?2) admissions, both occurring more than 1 month after the procedure. A prediction model was then derived and validated. We planned to use the logistic model?s formula and to provide an accessible calculator (similar to the way the American College of Cardiology-STS TAVI In-Hospital Mortality Risk is calculated8) instead of a conversion to a point score (i.e. like CHA2DS2-VASc) to prevent loss of accuracy while maintaining ease of use. Continuous normally distributed variables are presented as means � standard deviations and compared using the Student t test. Ordinal and/or non-normally distributed continuous variables are presented by the median and the interquartile range and compared using the Wilcoxon rank-sum test. Normality was assessed using the Shapiro-Wilk test and a visual inspection of quantile-quantile plots. Categorical variables are compared using the chi-square test. The logistic regression model was developed by examining all possible variable subsets, including clinically plausible interactions. Continuous variables were assessed for nonlinearity and categorized accordingly. We used 10-fold crossvalidation repeated 5 times for the validation of the results: we avoided the use of the split-set approach as resampling such as cross-validation has been shown to be more appropriate for our expected sample size.15 Analysis was performed with R16 with several appropriate packages17,18 Results In November 2008 to November 2016, a total of 543 consecutive patients underwent TAVI in our center and were considered for inclusion. Of the 543 patients, we excluded patients with procedural-related complications as follows: 7 due to death and 6 due to stroke within 1 month, 29 due to postoperative renal failure, 13 due to significant (?2+) paravalvular leak, 13 due to major or life-threatening vascular complication, 25 patients in whom valve success was not achieved, and 15 patients with more than one of the previously mentioned exclusion criteria, leaving 435 patients (80%) who formed the study group. The patient characteristics of both groups (included and excluded) are presented in the supplementary. The patients? mean age was 83 (interquartile range 79 to 86) years, and 43% were men. The mean STS score was 6.58% (�74). At 1 year, 66 of the 435 patients (15%) experienced the adverse futility-related outcome. Of those, 26 (6%) were due to death, 9 (2%) were due to nonfatal stroke, 12 (3%) were due to NYHA class lack of improvement, and 20 (5%) were due to repeat hospitalizations (causes are detailed in Supplementary Figure S1). The characteristics of all patients and the outcomes are presented in Table 1. Using the STS score as a single predictor yielded a low area under the curve (AUC) (0.48) and was not significantly associated with futility (odds ratio 1.01, 95% confidence interval 0.81 to 1.41). The prediction models were developed as described. The distribution of the outcome did not differ significantly by the year TAVI was performed, and the year of TAVI was not significantly associated with the outcome in the regression model. Both the hemoglobin level and the preprocedural aortic valve (AV) mean gradient showed a nonlinear behavior; they were converted to appropriate categorical variables. The logistic regression final model included diabetes, diastolic dysfunction (any grade, per echo evaluation), the need for diuretics, the AV mean gradient from 35 to 50 mm Hg, the creatinine level, the baseline NYHA class of <3, and a hemoglobin level <10 and >15, all associated with increased risk. As the mean and the peak gradient were highly correlated, only the mean gradient was considered. The model is presented in Table 2, with corresponding points for each variable. The AUC for the model was 0.73, which was reduced to 0.71 after crossvalidation (Figure 1). The performance of the logistic regression model by the separation of raw probabilities to percentile bins is presented in Supplementary Figure S2, which shows appropriate response and higher rates of futility with a higher-risk percentage as produced by the model. Using a single cutoff of 45% of the probability, the model produced a sensitivity of 20% and a specificity of 97%. We then further divided the model into 3 groups with raw probabilities cut at 10% and 40%. This produced groups with 6%, 19%, and 59% futility corresponding with low, medium, and high risk of futility. The high-risk group consisted of 4% of the patients who underwent successful, uncomplicated TAVI, and the low-risk group consisted of 43% of these patients. We also applied the model on the original patient group (543 patients). The model AUC remained at 0.7, with a similar prevalence of futility in the group (6%, 21%, and 52%). For patients in the high-risk group, only 72% achieved a perfect procedure, whereas in the low-risk group, the prevalence was higher (84%). A web application or calculator utilizing the risk model was made available.19 Discussion In the present study, 15% of the patients experienced the composite outcome and are believed to represent the group of ?perfect but yet futile TAVI.? The model developed showed a fair performance with high specificity. Using 3-tier risk categories, the model assigns high risk to 4% of the patients (almost a third of ultimately futile patients) with at least 59% of futility. In contrast, 45% of the patients are identified as low risk (6%) of futility. One of the study?s strengths is in its attention to a clinical dilemma. As caregivers, the heart team confronts difficult ARTICLE IN PRESS Valvular Heart Disease/TAVI Futility Risk Model 3 Table 1 Patients characteristics Variable Age (year) Men Weight (kg) Height (m) BMI (kg/m2) Diabetes mellitus Obstructive lung disease Previous stroke Dyslipidemia* Hypertension? Pacemaker Atrial fibrillation Frailty Previous malignancy Smoker Need for diuretics Previous myocardial Infarction Recent hospitalization? Previous need for PCI Previous CABG Previous valvular surgery Pre-procedural PCI Base NYHA class 2 3 4 Syncope Dyspnea Chest pain STS score Mean aortic valve gradient (mmHg) Peak aortic valve gradient (mmHg) Aortic valve area (cm2) Aortic insufficiency grade None Mild or Moderate Severe Mitral insufficiency grade None Mild Moderate Moderate to severe Diastolic dysfunction None Grade 1 Grade 2?3 Systolic dysfunction None Mild, mild to moderate Moderate to severe, severe Left bundle branch block in baseline ECG Hemoglobin (mg/dl) Platelets (K/micl) Creatinine (mg/dl) Albumin (g/dl) Transfemoral access Overall Futile p No Yes (n = 435) (n = 369) (n = 66) 83 [79, 86] 186 (43%) 72.09 � 14.08 1.62 � 0.09 27.37 � 4.73 149 (34%) 85 (19%) 76 (17%) 371 (85%) 400 (92%) 33 (7%) 127 (29%) 68 (16%) 60 (14%) 44 (10%) 244 (56%) 32 (7%) 209 (48%) 127 (29%) 78 (18%) 40 (9%) 71 (16%) 83 [80, 86] 155 (42%) 72.01 � 14.12 1.62 � 0.09 27.37 � 4.77 116 (31%) 68 (18%) 61 (16%) 314 (85%) 338 (92%) 29 (8%) 108 (29%) 59 (16%) 48 (13%) 34 (9%) 200 (54%) 26 (7%) 174 (47%) 106 (29%) 70 (19%) 33 (9%) 59 (16%) 82 [77, 85] 31 (47%) 72.58 � 13.97 1.63 � 0.10 27.38 � 4.58 33 (50%) 17 (26%) 15 (23%) 57 (86%) 62 (94%) 4 (6%) 19 (29%) 9 (14%) 12 (18%) 10 (15%) 44 (67%) 6 (9%) 35 (53%) 21 (32%) 8 (12%) 7 (11%) 12 (18%) 43 (10%) 252 (58%) 140 (32%) 43 (10%) 402 (92%) 77 (18%) 6.6 � 4.74 49 [40, 59] 77 [65, 90] 0.60 [0.50, 0.75] 33 (9%) 219 (59%) 117 (32%) 38 (10%) 340 (92%) 65 (18%) 6.5 � 4.6 49 [40, 59] 77 [65, 91] 0.60 [0.50, 0.72] 10 (15%) 33 (50%) 23 (35%) 5 (8%) 62 (94%) 12 (18%) 6.8 � 5.6 46 [38, 56] 74 [66, 86] 0.70 [0.60, 0.80] 123 (28%) 260 (60%) 52 (12%) 104 (28%) 219 (59%) 46 (12%) 19 (28%) 41 (62%) 6 (9%) 76 (17%) 203 (47%) 146 (34%) 10 (2%) 63 (17%) 174 (47%) 123 (33%) 9 (2%) 13 (20%) 29 (44%) 23 (35%) 1 (1%) 170 (39%) 240 (55%) 25 (6%) 152 (41%) 198 (54%) 19 (5%) 18 (27%) 42 (64%) 6 (9%) 0.09 0.54 0.76 0.59 0.98 0.005 0.22 0.30 0.79 0.69 0.61 1 0.76 0.26 0.15 0.08 0.74 0.46 0.72 0.24 0.84 0.79 0.20 0.65 0.80 1 0.63 0.31 0.30 0.01 0.74 0.90 0.07 335 (77%) 60 (14%) 40 (9%) 53 (12%) 11.90 � 1.63 197 [158, 248] 0.99 [0.80, 1.30] 4.0 [3.7, 4.3] 406 (93%) 286 (77%) 52 (14%) 31 (8%) 43 (12%) 11.97 � 1.54 197 [159, 248] 0.97 [0.80, 1.29] 4.0 [3.7, 4.3] 346 (94%) 0.39 49 (74%) 8 (12%) 9 (14%) 10 (15%) 0.55 11.48 � 1.98 0.02 195 [148, 237] 0.34 1.10 [0.85, 1.50] 0.01 4.1 [3.7, 4.2] 0.87 60 (91%) 0.56 (continued on next page) ARTICLE IN PRESS 4 The American Journal of Cardiology (www.ajconline.org) Table 1 (continued) Variable Overall Urgent procedure Valve in valve� Porcelain aorta Futile p No Yes (n = 435) (n = 369) (n = 66) 31 (7.1) 29 (7%) 25 (6%) 23 (6.2) 25 (6%) 21 (6%) 8 (12%) 4 (6%) 4 (6%) 0.15 1 1 Values are mean � SD, median [IQR], or n (%). * Defined as elevated LDL levels or low HDL levels, or on statin treatment. ? Blood pressure over 140/90 or on hypertensive agents. ? 6 months prior to the procedure. � Defined as transcatheter heart valve implantation into a degenerative surgical bioprosthesis. Table 2 Logistic regression model variables Hemoglobin < 10 or > 15 (mg/dl) NYHA functional class < III Diabetes mellitus Diastolic Dysfunction Mean aortic valve gradient between 35 to 50 (mmHg) Need for Diuretics Creatinine (mg/dl) Odds ratio 95% CI SE p-value Points* 2.86 2.7 2.14 2.10 2.04 1.75 1.40 1.45?5.55 1.13?6.25 1.22?3.76 1.13?3.87 1.16?3.59 0.96?3.19 1.05?1.86 0.34 0.44 0.29 0.31 0.29 0.31 0.15 0.002 0.02 0.01 0.02 0.01 0.07 0.02 10 9 7 7 7 5 3 CI = confidence interval; SE = standard error; NYHA = New York Heart Association. * Sum of points 14 correspond to 10% probability, 21 points to 20% probability, 27 to 30%, 31 to 40%, 35 to 50%, 39 to 60%, and 43 points to 70%. Figure 1. Receiver operating curves of the original model (derivation) and after validation. ARTICLE IN PRESS Valvular Heart Disease/TAVI Futility Risk Model decision making. Especially tough is to deny a potentially therapeutic procedure from patients in need of remedy. The ultimate desire to ?do something? may bias the caregivers from making the best choice. Using an objective risk calculator that separates TAVI procedural success from subsequent patient?s outcomes may assist in the decision-making process by indicating that a particular patient may not fully benefit from TAVI, independent of how successful the procedure is. The model can also assist in clarifying the predicted gap between the patients? expectations and the projected clinical reality after TAVI. As the use of TAVI continues to spread, the need to define and articulate the anticipated patients? health benefit grows accordingly. The model can also assist in clarifying the predicted gap between patients? expectations and the projected clinical reality after TAVI. The study and model also stresses that a significant portion of patients with AS have co-morbidities and their symptomatic status is complex?a specific patient might not enjoy TAVI if he or she is anemic and transaortic gradients are not high. The study has high specificity, which is desired in this clinical situation (less false positives), but the trade-off is lower sensitivity. The study has several limitations. It is single centered, which might hinder generalizability. Although the resultant model parameters such as AUC were fair (0.71 after validation), a study with a larger sample size might have been able to include more variables and be more accurate. Naturally, the results of the present study should be externally validated. We chose a specific combination of outcomes; which outcomes are important when measuring the TAVI outcome is up for debate.20 A different choice or addition of outcome measures, such as quality of life assessment,10,21,22 might expose different variables and modify the defined risk. We chose to include the NYHA class, which is partially subjective, and rehospitalizations, which can mean a different burden to different patients. These variables should be carefully considered before applying this score for patients. In the present study, the portion of patients experiencing a perfect procedure outcome (80%) and a futile 1-year outcome (15%) might be an optimistic estimation, as adverse procedural outcomes from other reports ranged from a quarter to a third.1,6,23 In the assessment of poor predictors from the Placement of Aortic Transcatheter Valves (PARTNER) trial,10 poor outcome was identified in 33%. In the PARTNER 2A trial,3,4 the rate of death, stroke, or rehospitalization was 27%, whereas in the Surgical Replacement and Transcatheter Aortic Valve Implantation (SURTAVI) trial, the rate was 17%.4 In a metaanalysis of TAVI randomized controlled trials and matched trials, all-cause, 1-year mortality reached 13%.24 Higher prevalence would mean a higher positive predictive value for the model. The model components somewhat differ from TAVI risk factors previously described; they might stress the difference between procedural risk factors and factors associated with long-term risk-benefit unrelated to periprocedural events. For instance, our own institutional experience found an association with frailty13; others have reported that albumin is associated with an increased risk9,25; in FRANCE2, factors such as age, body mass index, and emergent procedure were included,7 and in Arnold et al,10 oxygen-dependent lung disease was included. These were examined but, because they did not contribute to the model, were excluded. Arnold et al also examined the performance on the mini-mental exam 5 and the 6-minute walk test, which were not captured in our patients. The issue of risk stratification in TAVI can be defined in 2 broad aspects.26,27 On one hand, we want to make sure that TAVI-eligible patients will not be denied the procedure due to perceived ?unfitness.? In contrast, some patients might undergo the procedure with its risks and costs but will not enjoy its benefits. The challenge is to define and then predict the two. The model described can aid in the heart-team decision process. For ease of use, we supply the prediction score through a web application.19 The approach of taking periprocedural complications out of the equation represents a novel one. The findings of the present study will need external validation. In summary, our prediction score could provide important insight and can aid in identifying patients who may experience futility after successful TAVI, having hazards that may outweigh the benefit of the procedure. Disclosures The authors have no conflicts of interest to disclose. Supplementary Data Supplementary data associated with this article can be found, in the online version, https://doi.org/10.1016/ j.amjcard.2017.09.007. 1. Leon MB, Smith CR, Mack M, Miller DC, Moses JW, Svensson LG, Tuzcu EM, Webb JG, Fontana GP, Makkar RR, Brown DL, Block PC, Guyton RA, Pichard AD, Bavaria JE, Herrmann HC, Douglas PS, Petersen JL, Akin JJ, Anderson WN, Wang D, Pocock S, PARTNER Trial Investigators. Transcatheter aortic-valve implantation for aortic stenosis in patients who cannot undergo surgery. N Engl J Med 2010;363:1597? 1607. 2. Adams DH, Popma JJ, Reardon MJ, Yakubov SJ, Coselli JS, Deeb GM, Gleason TG, Buchbinder M, Hermiller JJ, Kleiman NS, Chetcuti S, Heiser J, Merhi W, Zorn G, Tadros P, Robinson N, Petrossian G, Hughes GC, Harrison JK, Conte J, Maini B, Mumtaz M, Chenoweth S, Oh JK. Transcatheter aortic-valve replacement with a self-expanding prosthesis. N Engl J Med 2014;370:1790?1798. 3. Leon MB, Smith CR, Mack MJ, Makkar RR, Svensson LG, Kodali SK, Thourani VH, Tuzcu EM, Miller DC, Herrmann HC, Doshi D, Cohen DJ, Pichard AD, Kapadia S, Dewey T, Babaliaros V, Szeto WY, Williams MR, Kereiakes D, Zajarias A, Greason KL, Whisenant BK, Hodson RW, Moses JW, Trento A, Brown DL, Fearon WF, Pibarot P, Hahn RT, Jaber WA, Anderson WN, Alu MC, Webb JG. Transcatheter or surgical aortic-valve replacement in intermediate-risk patients. N Engl J Med 2016;374:1609?1620. 4. Reardon MJ, Van Mieghem NM, Popma JJ, Kleiman NS, S鴑dergaard L, Mumtaz M, Adams DH, Deeb GM, Maini B, Gada H, Chetcuti S, Gleason T, Heiser J, Lange R, Merhi W, Oh JK, Olsen PS, Piazza N, Williams M, Windecker S, Yakubov SJ, Grube E, Makkar R, Lee JS, Conte J, Vang E, Nguyen H, Chang Y, Mugglin AS, Serruys PWJC, Kappetein AP. Surgical or transcatheter aortic-valve replacement in intermediate-risk patients. N Engl J Med 2017;376:1321?1331. 5. Grover FL, Vemulapalli S, Carroll JD, Edwards FH, Mack MJ, Thourani VH, Brindis RG, Shahian DM, Ruiz CE, Jacobs JP, Hanzel G, Bavaria JE, Tuzcu EM, Peterson ED, Fitzgerald S, Kourtis M, Michaels J, Christensen B, Seward WF, Hewitt K, Holmes DR, Registry ST. 2016 Annual Report of The Society of Thoracic Surgeons/American College of Cardiology Transcatheter Valve Therapy Registry. J Am Coll Cardiol 2017;69:1215?1230. 6. Reynolds MR, Magnuson EA, Wang K, Thourani VH, Williams M, Zajarias A, Rihal CS, Brown DL, Smith CR, Leon MB, Cohen DJ, PARTNER Trial Investigators. Health-related quality of life after transcatheter or surgical aortic valve replacement in high-risk patients ARTICLE IN PRESS 6 The American Journal of Cardiology (www.ajconline.org) 7. 8. 9. 10. 11. 12. 13. 14. 15. with severe aortic stenosis: results from the PARTNER (Placement of AoRTic TraNscathetER Valve) Trial (Cohort A). J Am Coll Cardiol 2012;60:548?558. Iung B, Laou閚an C, Himbert D, Eltchaninoff H, Chevreul K, DonzeauGouge P, Fajadet J, Leprince P, Leguerrier A, Li鑦re M, Prat A, Teiger E, Laskar M, Vahanian A, Gilard M, FRANCE 2 Investigators. Predictive factors of early mortality after transcatheter aortic valve implantation: individual risk assessment using a simple score. Heart 2014;100:1016? 1023. Edwards FH, Cohen DJ, O?Brien SM, Peterson ED, Mack MJ, Shahian DM, Grover FL, Tuzcu EM, Thourani VH, Carroll J, Brennan JM, Brindis RG, Rumsfeld J, Holmes DR, 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. JAMA Cardiol 2016;1:46?52. Hermiller JB, Yakubov SJ, Reardon MJ, Deeb GM, Adams DH, Afilalo J, Huang J, Popma JJ, CoreValve United States Clinical Investigators. Predicting early and late mortality after transcatheter aortic valve replacement. J Am Coll Cardiol 2016;68:343?352. Arnold SV, Reynolds MR, Lei Y, Magnuson EA, Kirtane AJ, Kodali SK, Zajarias A, Thourani VH, Green P, Rod閟-Cabau J, Beohar N, Mack MJ, Leon MB, Cohen DJ, PARTNER Investigators. Predictors of poor outcomes after transcatheter aortic valve replacement: results from the PARTNER (Placement of Aortic Transcatheter Valve) trial. Circulation 2014;129:2682?2690. Otto CM, Kumbhani DJ, Alexander KP, Calhoon JH, Desai MY, Kaul S, Lee JC, Ruiz CE, Vassileva CM. 2017 ACC expert consensus decision pathway for transcatheter aortic valve replacement in the management of adults with aortic stenosis: a report of the American College of Cardiology Task Force on Clinical Expert Consensus Documents. J Am Coll Cardiol 2017;69:1313?1346. Kirkpatrick JN, Hauptman PJ, Swetz KM, Blume ED, Gauvreau K, Maurer M, Goodlin SJ. Palliative care for patients with end-stage cardiovascular disease and devices: a report from the Palliative Care Working Group of the Geriatrics Section of the American College of Cardiology. JAMA Intern Med 2016;176:1017?1019. Levi A, Landes U, Assali AR, Orvin K, Sharony R, Vaknin-Assa H, Hamdan A, Shapira Y, Schwartzenberg S, Codner P, Shaul AA, Vaturi M, Gutstein A, Sagie A, Kornowski R. Long-term outcomes of 560 consecutive patients treated with transcatheter aortic valve implantation and propensity score-matched analysis of early- versus new-generation valves. Am J Cardiol 2017;119:1821?1831. Kappetein AP, Head SJ, G閚閞eux P, Piazza N, Mieghem NM, van Blackstone EH, Brott TG, Cohen DJ, Cutlip DE, Es G-A, van Hahn RT, Kirtane AJ, Krucoff MW, Kodali S, Mack MJ, Mehran R, Rod閟Cabau J, Vranckx P, Webb JG, Windecker S, Serruys PW, Leon MB, Valve Academic Research Consortium (VARC)-2. Updated standardized endpoint definitions for transcatheter aortic valve implantation: the Valve Academic Research Consortium-2 consensus document (VARC2). Eur J Cardiothorac Surg 2012;42:S45?S60. Steyerberg EW, Harrell FE, Borsboom GJ, Eijkemans MJ, Vergouwe Y, Habbema JD. Internal validation of predictive models: efficiency of 16. 17. 18. 19. 20. 21. 22. 23. 24. 25. 26. 27. some procedures for logistic regression analysis. J Clin Epidemiol 2001;54:774?781. R Core Team. R: A Language and Environment for Statistical Computing. Vienna, Austria: R Foundation for Statistical Computing; 2015. Available at: https://www.R-project.org/. Accessed on March 1, 2017. Harrel FEJ. rms: regression modeling strategies; 2016. Available at: http:// CRAN.R-project.org/package=rms. Accessed on January 1, 2017. Kuhn M, Weston S, Williams A, Keefer C, Engelhardt A, Cooper T, Mayer Z, Kenkel B, Team the RC, Benesty M, Lescarbeau R, Ziem A, Scrucca L, Tang Y, Candan C, Hunt T, Wing J. caret: classification and regression training; 2016. Available at: https://CRAN.R-project.org/ package=caret. Accessed on January 1, 2017. Zusman O. TAVI futility risk model. Available at: https:// rabinmedicalcenter.shinyapps.io/TAVI_futility_score/. Accessed on July 20, 2017. Arnold SV, Spertus JA, Lei Y, Green P, Kirtane AJ, Kapadia S, Thourani VH, Herrmann HC, Beohar N, Zajarias A, Mack MJ, Leon MB, Cohen DJ. How to define a poor outcome after transcatheter aortic valve replacement: conceptual framework and empirical observations from the placement of aortic transcatheter valve (PARTNER) trial. Circ Cardiovasc Qual Outcomes 2013;6:591?597. Arnold SV, Reynolds MR, Wang K, Magnuson EA, Baron SJ, Chinnakondepalli KM, Reardon MJ, Tadros PN, Zorn GL, Maini B, Mumtaz MA, Brown JM, Kipperman RM, Adams DH, Popma JJ, Cohen DJ, CoreValve US Pivotal Trial Investigators. Health status after transcatheter or surgical aortic valve replacement in patients with severe aortic stenosis at increased surgical risk: results from the CoreValve US pivotal trial. JACC Cardiovasc Interv 2015;8:1207?1217. Green CP, Porter CB, Bresnahan DR, Spertus JA. Development and evaluation of the Kansas City Cardiomyopathy Questionnaire: a new health status measure for heart failure. J Am Coll Cardiol 2000;35:1245? 1255. Smith CR, Leon MB, Mack MJ, Miller DC, Moses JW, Svensson LG, Tuzcu EM, Webb JG, Fontana GP, Makkar RR, Williams M, Dewey T, Kapadia S, Babaliaros V, Thourani VH, Corso P, Pichard AD, Bavaria JE, Herrmann HC, Akin JJ, Anderson WN, Wang D, Pocock SJ, PARTNER Trial Investigators. Transcatheter versus surgical aorticvalve replacement in high-risk patients. N Engl J Med 2011;364:2187? 2198. Gargiulo G, Capodanno D, Tamburino C, Trimarco B, Esposito G. Transcatheter aortic valve implantation versus surgical aortic valve replacement. Ann Intern Med 2017;166:606. Yamamoto M, Shimura T, Kano S, Kagase A, Kodama A, Sago M, Tsunaki T, Koyama Y, Tada N, Yamanaka F, Naganuma T, Araki M, Shirai S, Watanabe Y, Hayashida K. Prognostic value of hypoalbuminemia after transcatheter aortic valve implantation (from the Japanese Multicenter OCEAN-TAVI Registry). Am J Cardiol 2017;119:770?777. Puri R, Iung B, Cohen DJ, Rod閟-Cabau J. TAVI or No TAVI: identifying patients unlikely to benefit from transcatheter aortic valve implantation. Eur Heart J 2016;37:2217?2225. Lindman BR, Alexander KP, O?Gara PT, Afilalo J. Futility, benefit, and transcatheter aortic valve replacement. JACC Cardiovasc Interv 2014;7:707?716.