Author + information
- Received September 26, 2017
- Revision received May 28, 2018
- Accepted June 26, 2018
- Published online September 3, 2018.
- Anthony W.A. Wassef, MDa,
- Josep Rodes-Cabau, MDb,
- Yaqing Liu, MSca,
- John G. Webb, MDc,
- Marco Barbanti, MDd,
- Antonio J. Muñoz-García, MD, PhDe,
- Corrado Tamburino, MD, PhDd,
- Antonio E. Dager, MDf,
- Vicenç Serra, MDg,
- Ignacio J. Amat-Santos, MD, PhDh,
- Juan H. Alonso Briales, MDe,
- Alberto San Roman, MDh,
- Marina Urena, MD, PhDi,
- Dominique Himbert, MDi,
- Lius Nombela-Franco, MD, PhDj,
- Alexandre Abizaid, MD, PhDk,
- Fabio S. de Brito Jr., MDl,
- Henrique B. Ribeiro, MD, PhDm,
- Marc Ruel, MDn,
- Valter C. Lima, MDo,
- Fabian Nietlispach, MDp and
- Asim N. Cheema, MD, PhDa,∗ ()
- aDivision of Cardiology, Department of Medicine, St. Michael’s Hospital, Toronto, Canada
- bQuebec Heart & Lung Institute, Laval University, Quebec City, Canada
- cDivision of Cardiology, Department of Medicine, St. Paul’s Hospital, University of British Columbia, Vancouver, Canada
- dDivision of Cardiology, Ferrarotto Hospital, University of Catania, Catania, Italy
- eDepartment of Cardiology, Hospital Universitario Virgen de la Victoria, Universidad de Málaga, Málaga, Spain
- fDepartment of Cardiology, Clínica de Occidente de Cali, Cali, Colombia
- gDepartment of Interventional Cardiology, Hospital Universitari Vall d’Hebron, Universitat Autònoma de Barcelona, Barcelona, Spain
- hCIBERCV, Hospital Clínico Universitario de Valladolid, Valladolid, Spain
- iDepartment of Cardiology, Bichat Hôpital, AP-HP, University Paris Diderot, Paris, France
- jInstituto Cardiovascular, Hospital Universitario Clínico San Carlos, Madrid, Spain
- kInstituto Dante Pazzanese de Cardiologia, São Paulo, Brazil
- lInterventional Cardiology Department, Hospital Israelita Albert Einstein, São Paulo, Brazil
- mHeart Institute (InCor), São Paulo, Brazil
- nDivision of Cardiac Surgery, Ottawa Heart Institute, Ottawa, Canada
- oHospital São Francisco-Santa Casa de Misericórdia de Porto Alegre, Porto Alegre, Brazil
- pUniversity Hospital Zürich, Zürich, Switzerland
- ↵∗Address for correspondence:
Dr. Asim N. Cheema, Division of Cardiology, St. Michael's Hospital, 30 Bond Street, Toronto, Ontario M5B 1W8, Canada.
Objectives The authors aimed to determine the procedural learning curve and minimum annual institutional volumes associated with optimum clinical outcomes for transcatheter aortic valve replacement (TAVR).
Background Transcatheter aortic valve replacement (TAVR) is a complex procedure requiring significant training and experience for successful outcome. Despite increasing use of TAVR across institutions, limited information is available for its learning curve characteristics and minimum annual volumes required to optimize clinical outcomes.
Methods The study collected data for patients at 16 centers participating in the international TAVR registry since initiation of the respective TAVR program. All cases were chronologically ordered into initial (1 to 75), early (76 to 150), intermediate (151 to 225), high (226 to 300), and very high (>300) experience operators for TAVR learning curve characterization. In addition, participating institutions were stratified by annual TAVR case volume into low-volume (<50), moderate-volume (50 to 100), and high-volume (>100) groups for comparative analysis. Procedural and 30-day clinical outcomes were collected and multivariate regression analysis performed for 30-day mortality and the early safety endpoint.
Results A total of 3,403 patients comprised the study population. On multivariate analysis, all-cause mortality was significantly higher for initial (odds ratio [OR]: 3.83; 95% confidence interval [CI]: 1.93 to 7.60), early (OR: 2.41; 95% CI: 1.51 to 5.03), and intermediate (OR: 2.53; 95% CI: 1.19 to 5.40) experience groups compared with the very high experience operators. In addition, the early safety endpoint was significantly worse for all experience groups compared with the very high experience operators. Low annual volume (<50) TAVR institutions had significantly higher all-cause 30-day mortality (OR: 2.70; 95% CI: 1.44 to 5.07) and worse early safety endpoint (OR: 1.60; 95% CI: 1.17 to 2.17) compared with the moderate- and high-volume groups. There was no difference in patient outcomes between intermediate and high annual volume groups.
Conclusions TAVR procedures display important learning curve characteristics with both greater procedural safety and a lower mortality when performed by experienced operators. In addition, TAVR performed at low annual volume (<50 procedures) institutions is associated with decreased procedural safety and higher patient mortality. These findings have important implications for operator training and patient care at centers performing TAVR.
Transcatheter aortic valve replacement (TAVR) has revolutionized the treatment of severe symptomatic aortic stenosis, with several randomized trials demonstrating equivalence or superiority to conventional surgical aortic valve replacement for inoperable (1), high-risk (2,3), and intermediate-risk patients (4,5). Concomitantly, there has been a large increase in the number of procedures being performed and the number of centers performing the procedure in North America as well as internationally (6,7). This trend is likely to continue, as there is expected to be a large increase in the number of elderly patients with aortic valve disease (8), as well as an increase in the utilization of this technology in lower-risk patients (9).
A learning curve phenomenon, defined as an improvement in outcome with increasing experience has been demonstrated for multiple cardiac (10,11) as well as noncardiac (12) procedures. In addition, maintaining minimal annual procedural volumes are associated with improved clinical outcomes (13). Although improved procedural results with greater TAVR experience have been described (14–17), adequate understanding of the TAVR learning curve and the minimal annual procedural volume required to achieve competence is lacking. In the present study, we characterize the TAVR procedure learning curve and investigate the relationship between annual institutional volume and clinical outcomes from the international multicenter TAVR registry.
All consecutive patients who underwent TAVR at 16 large, urban international academic teaching sites in North and South America and Europe since the initiation of the respective center’s TAVR program were included in this study. All centers employed a heart-team model of multidisciplinary decision making for patient selection, procedural planning and performance (6). Choice of TAVR device, valve size, approach, post-procedure care, and antithrombotic management were at the discretion of treating center’s heart team, and participation in clinical trials as well as investigational devices were included in this analysis. A total of 3,468 patients underwent TAVR during the study period with complete data were available for 3,403 (98.1%) patients comprising the study population. Patients with incomplete data, totaling <2% of the population were excluded. Baseline, procedural, echocardiographic, and outcome data were collected in a prospective manner at each center. Each center was responsible for collection of baseline demographic and procedural details, as well as relevant 30-day outcomes. Events were adjudicated at each site, with no central adjudication. The first patient included had the procedure performed January 6, 2005, and the last patient included had the procedure January 29, 2016.
Patients from each center were chronologically ordered from the initiation of the TAVR program for the learning curve analysis. To determine the procedural learning curve characteristics, TAVR cases from each center were chronologically grouped into initial (1 to 75), early (76 to 150), intermediate (151 to 225), high (226 to 300), and very high (>300) experience groups, to determine the effect of increasing procedural experience on procedural and clinical outcomes. To determine the minimum annual institutional TAVR volume for optimum clinical outcomes, each individual center’s TAVR volume per calendar year (January 1 to December 31) was determined, and grouped into low (1 to 49), intermediate (50 to 100) and high (>100) TAVR volume groups. Centers may contribute to different volume groups depending only on their annual volume per calendar year. For this analysis, all TAVR performed in the first and last calendar year from each center were excluded from analysis as the completeness of annual volume data could not be assured. Complete data available for 2,205 (64.8%) patients from 16 institutions.
All procedural and clinical outcomes were defined per Valve Academic Research Consortium-2 criteria (18) and determined at 30 days. Briefly, device success was defined as successful access, delivery, deployment in the correct position and anatomical location with aortic valve area >1.2 cm2 and mean aortic valve gradient <20 mm Hg or peak velocity <3 m/s, without moderate or severe aortic insufficiency. Early safety endpoint was defined as the composite of death, stroke, major bleeding, vascular complications, surgical conversion, and renal failure.
Categorical variables were reported as frequency and percentages and continuous variables as mean ± SD. For the learning curve analysis, the very high-experience group was used as a reference and baseline characteristics, procedural variables, and clinical outcomes for each experience group (initial, early, intermediate, and experienced) were compared with the very high-experience group using paired wise comparisons with a Student's t-test, Mann-Whitney U test, or a chi-square test as appropriate, with Dunnett’s test to control for multiple comparisons. A logistic regression was performed to determine association between experience and annual volume group allocation and all-cause mortality and composite early safety endpoint. Covariate adjustment with logistic regression has been demonstrated to compare well to propensity adjustment methods for clinical trials (19). Variables chosen for the model were included if there were significant baseline differences between the groups, and if they had previously been demonstrated to predict worse outcomes in TAVR (20). The variables included in the logistic regression for the learning curve analysis included baseline characteristics age, body surface area, sex, New York Heart Association (NYHA) functional class, prior coronary artery bypass grafting, prior coronary artery disease (CAD), chronic kidney disease, and continent of the participating institution (Europe, South America, North America), procedural variables (transfemoral, transapical, or other access) and prosthesis generation for the SAPIEN platform (SAPIEN, SAPIEN XT, SAPIEN 3) to account for the geographical differences in clinical practice and technological changes which may impact vascular access complications and paravalvular leak rates. The logistic regression analysis was performed first using procedural volume as a continuous variable and then repeated using the procedural volume categories as defined previously. Results of the multivariate analysis were expressed as odds ratio (OR) with 95% confidence interval (CI). All analyses were performed with SAS software version 9.4 (SAS Institute, Cary, North Carolina) and a p value < 0.05 defined statistical significance.
TAVR learning curve and clinical outcomes
Among the 3,403 patients enrolled at 16 international sites, 1,141 (33.5%) cases were performed by initial operators, 780 (22.9%) cases performed by early operators, 549 (16.1%) cases performed by intermediate-experience operators, 354 (10.4%) cases performed by high-experience operators, and 579 (17.0%) cases were performed by very high-experience operators. 16 centers contributed data to the initial cohort, 13 centers to early, 8 centers to intermediate, 6 centers to the high, and 4 centers to the very high cohort. Most participating centers showed gradually increasing TAVR volumes over the period of the study without a rapid ramp-up phenomenon from one year to the next (Figure 1, Online Figure). Baseline unadjusted clinical and procedural characteristics of the study population are shown in Table 1. The mean age of the study population was 82 ± 8 years, with 77% of patients undergoing transfemoral TAVR, and a balloon-expandable valve as used in 59% of patients. There were statistically significant intergroup differences in age (p < 0.001), mean aortic valve gradient (p < 0.001), body mass index (p < 0.001), Society of Thoracic Surgeons Predicted Risk of Mortality (p < 0.001), NYHA functional class IV (p < 0.001), CAD (p < 0.001), and the type of valve implanted (p < 0.001).
The unadjusted clinical and procedural outcomes for the different TAVR experience groups are shown in Table 2. There was a consistent decrease in all cause death with increasing TAVR experience and was significantly higher for the initial experience (9.6%; p < 0.001) and early experience (7.9%; p = 0.002) groups compared with the very high-experience (3.3%) operators (Figure 1A). Similarly, the composite early safety endpoint decreased with increasing procedural experience from 27.5% for the initial experience to 14.9% for very high-experience operators (Figure 1B), largely driven by reductions in major bleeding (Figure 1C) and major vascular complications (Figure 1D). The procedural time and length of stay post-procedure were only high in the initial experience with no significant difference between early, intermediate, experienced, and very high-experience groups (Table 2). The contrast volume administered was also high in the initial and early groups compared with very high-experience operators (p < 0.001). The rate of procedural success did not change with increasing experience, and the rates of stroke, myocardial infarction, new dialysis, severe aortic regurgitation, and surgical conversion were similar across different operator experience groups. Although, there were significant intergroup differences for procedure time and permanent pacemaker insertion, these did not demonstrate a stepwise learning curve phenomenon.
The results of multivariate logistic regression analysis showed that TAVR procedural volume when used as a continuous variable was independently associated with 30-day mortality (p < 0.0001) (Figure 2A), major bleeding (p < 0.0001), and major adverse cardiac events (p < 0.0001). In addition, the multivariate logistic regression analysis identified TAVR performed by the initial (OR: 3.83; 95% CI: 1.93 to 7.60; p < 0.001), early (OR: 2.41; 95% CI: 1.51 to 5.03; p = 0.020), and intermediate (OR: 2.53; 95% CI: 1.19 to 5.40; p = 0.016) experience groups were independently associated with higher mortality compared with the very high-experience operators (Figure 2B). In addition, the early safety endpoint was significantly higher for initial (OR: 2.02; 95% CI: 1.41 to 2.89; p < 0.001), early (OR: 1.74; 95% CI: 1.19 to 2.56; p = 0.005), intermediate (OR: 2.07; 95% CI: 1.40 to 3.07; p < 0.01), and high (OR1.70; 95% CI: 1.14 to 2.59; p = 0.009) experience groups compared with the very high-experience operators (Figure 2C). Higher Society of Thoracic Surgeons score (OR: 1.84; 95% CI: 1.20 to 2.82; p = 0.005) and transapical access (OR: 1.84; 95% CI: 1.20 to 2.82; p = 0.005) were also independent predictors of increased mortality, whereas left ventricular ejection fraction, sex, NYHA functional class, prior coronary artery bypass grafting, chronic kidney disease, chronic obstructive pulmonary disease status, and balloon-expandable versus self-expanding valve did not demonstrate a significant association with increased mortality or early safety endpoint (data not shown).
Annual volume and clinical outcomes
A total of 2,205 of 3,403 (64.8%) patients were included in this analysis, with 569 (25.8%) cases performed at low, 1,121 (50.8%) cases performed at intermediate, and 515 (23.3%) cases performed at high annual volume centers (Table 3). There were significant differences in the proportion of women (p = 0.018), mean body mass index (p = 0.001), left ventricular ejection fraction (p = 0.001), Society of Thoracic Surgeons Predicted Risk of Mortality (p = 0.001), CAD (p = 0.001), and chronic kidney disease (p = 0.001) among the 3 groups. The use of transfemoral approach (p < 0.001) and self-expanding valve (p < 0.001) was highest at low compared with high annual volume centers. The unadjusted clinical and procedural outcomes for the 3 annual TAVR volume groups are shown in Table 4. All-cause mortality was significantly higher in the low-volume centers compared with the high-volume centers (8.8% vs. 3.9%; p = 0.003), but was similar between intermediate- and high-volume centers (Figure 3A). The early safety endpoint (34.3% vs. 26.4%; p = 0.01), and major bleeding (11.1%, vs. 4.9%; p = 0.037) were also significantly higher for low-volume centers compared with high-volume centers, with no difference between intermediate- and high-volume centers (Figures 3B to 3D). The procedure time (184.3 ± 101.6 min vs. 84.8 ± 36.3 min; p < 0.001), and contrast volume (219.8 ± 92.4 ml vs. 74.8 ± 45.5 ml; p < 0.001) were higher in low-volume centers compared with high-volume centers. The device success rates and risk of post-procedural myocardial infarction, stroke, surgical conversion, new pacemaker, or aortic regurgitation were similar for the 3 volume groups. The results of logistic multivariate analysis showed that low annual volume group was independently associated with increased mortality (OR: 2.70; 95% CI: 1.44 to 5.07; p = 0.002) and high rate of composite early safety endpoint (OR: 1.60, 95% CI: 1.17 to 2.17; p = 0.003) (Figure 4).
The main findings of the present study are: 1) an important learning curve exists with incremental improvement in clinical outcomes for increasing TAVR experience; 2) a TAVR experience of at least 225 procedures is associated with reduced TAVR mortality while the early safety endpoint continues to improve beyond the 225-case volume; and 3) the institutional annual procedural volume affects TAVR clinical outcomes and low volume centers defined as an annual case volume <50 TAVR procedures are independently associated with higher mortality.
The number of TAVR procedures being performed has increased markedly as the indications have expanded from inoperable patients (1) to high-risk (2,3) and intermediate-risk patients (4). There is likely to be a further increase in TAVR utilization in view of demographic changes (21) and potential confirmation of safety and efficacy in lower risk patients (7). However, TAVR remains a technically challenging procedure requiring a high level of skill and prior experience with percutaneous and endovascular procedures. In addition to a comprehensive pre-procedural patient assessment by a dedicated heart team, the TAVR requires a unique skillset and expertise for implantation and expeditiously manage unexpected complications. Therefore, adequate training and experience play a critical role in improving procedural safety and clinical success. In addition, similar to other procedural skills, it is likely that a minimum annual procedural volume is required to provide optimum results.
TAVR learning curve
The 2 important components of any procedural learning curve include measures of clinical outcome including patient survival or procedural success and process efficiency including procedural time and resource utilization. For cardiovascular procedures, prior learning curve studies have demonstrated important relationship for increasing experience with both outcome measures as well as process efficiency (12,22). Four years’ experience post-training has shown to decrease mortality for coronary artery bypass surgery (23) and >150 cases required to optimize outcomes of percutaneous balloon mitral valvuloplasty (10). Similarly, minimum procedural volume criteria for optimizing procedural efficiency has been described for radial PCI (11,24).
In the present study, we examined 30-day mortality and composite early safety endpoint as measures of clinical outcome and procedural time and contrast media usage as measures of procedural efficiency. In addition, the very high-experience group was defined for operators with >300 case volume as the reference group to assess the learning curve of the all operators and effect on outcome measures with increasing procedural volumes. The patients undergoing TAVR by the very high-experience operators in the present study showed a mortality rate of 3.3%, which is lower than that reported for the STS/ACC TVT (Society of Thoracic Surgeons/American College of Cardiology Transcatheter Valve Therapy Registry (6) and similar to sites with the most TAVR experience in the United States (25). TAVR learning curve analysis from the PARTNER I (Placement of Aortic Transcatheter Valves) trial (14,15) showed an excess mortality and procedural failure only for initial TAVR experience but the findings are limited by a small sample size. These results of the present study are consistent with recently published data from the TVT registry, showing lower major adverse events with increasing TAVR experience (25). Carroll et al. (25) observed a reduction in vascular and bleeding complications most noted in the first 100 cases, and a reduction in mortality that became statistically insignificant after adjustment. When used as a continuous variable, the present study identified procedural volume to be independently associated with 30-day mortality (Figure 2A), major bleeding and major adverse cardiac events and the odds of mortality continued to decrease with increasing procedural experience. On comparing procedural experience as a categorical variable in multivariate regression analysis, the 30-day mortality showed a consistent decrease for up to a 225-case volume (Figure 2B). For the early safety endpoint up, we observed improvement beyond the 225-case volume (Figure 2C), a finding consistent with earlier reports (25). Similarly, the rates of stroke, paravalvular aortic regurgitation, and dialysis requirement did not change with increasing experience. The measures of procedural efficiency such as procedure time, contrast volume also improved with increasing experience but no significant difference was observed after initial and early experience when compared with the very high-experience operators.
Annual TAVR volumes and clinical outcomes
The present study showed that low-volume (<50 cases/year) centers experienced a 30-day mortality of 8.8%, which is significantly higher than 5.7% for intermediate-volume (51 to 100 cases/year) and 3.9% for high-volume (51 to 100 cases/year) centers (Figure 3). The mortality rates for high-volume centers in the present study are lower than those reported for contemporary TAVR experience (6,25,26) confirming that high annual volume is an important contributor to improved clinical outcomes. Similarly, the early safety endpoint was also significantly worse among the low-volume group and low-volume group was an independent predictor of worse 30-day mortality and the early safety endpoint (Figure 4). There is limited data for the relationship between annual institutional TAVR volume and clinical outcomes. A recent study by Khera et al. (27) showed that low-volume centers identified as performing <50 TAVR cases per year have a significantly higher 30-day readmission rates compared with high-volume centers performing >100 cases per year.
Despite the large increase in the number of TAVR procedures being performed at greater number of institutions, many hospitals perform a relatively small number of procedures. There are many low volume centers performing <50 procedures a year both in the United States and internationally. This trend is likely to continue with potential utilization of this technology in lower risk patients (9). The findings from the present study suggest a minimum annual volume threshold to provide the best clinical outcomes for patients undergoing TAVR procedures and can serve as a guide for optimal distribution of resources and technology.
First, there were significant differences in baseline demographic and procedural characteristics between the chronological and volume groups in the learning curve and annual volume analysis that may have confounded the outcomes of interest. Although, multivariate regression analysis was performed to adjust for these differences; other unmeasured confounders may still remain. Second, this study was only able to assess the center learning curve and center annual volume, not of individual operators. However, the current guidelines recommend a heart team approach to decision making and most centers require a 2-member TAVR team for procedures making individual operator data less relevant. Prior data from coronary interventions have shown center volume to be a more robust predictor of outcome compared with individual operator volume (13,28). However, the interaction between low-volume TAVR operators and high annual volume centers and vice versa could not be assessed in the present analysis as all operators performed TAVR procedures at a single institution. Our analysis included data from 16 centers only which may have influenced the findings due to a small sample size. Finally, many of the centers that were included in this study began their TAVR experience early; whether the learning curve may be abbreviated in the current environment with more centers performing the procedure, with newer technology and more educational opportunities is not known.
In this study examining the learning curve and annual procedural volumes of center’s performing TAVR using a large, international database, we found that at least 225 procedures are required to optimize mortality rates for TAVR and procedural complications continue to decrease beyond the 225 case volume. Annual institutional volume of <50 procedures per year was associated with worse clinical and procedural outcomes. These findings have important implications for operator training and patient care at centers performing TAVR. Further research is required to determine whether newer TAVR technology, focused training, and proctoring can abbreviate the TAVR learning curve.
WHAT IS KNOWN? The present study analyzed data from a large international multicenter registry and found an important TAVR learning curve with 30-day mortality that decreased with increasing procedural experience up to 225 cases.
WHAT IS NEW? The early safety endpoint, however, continued to improve beyond the 225-case volume. In addition institutions with an annual TAVR volume <50 cases were associated with worse 30-day mortality and early safety endpoint compared with institutions with an annual TAVR volume >100 cases.
WHAT IS NEXT? These findings have important implications for operator training and patient care at institutions performing TAVR.
Dr. Rodes-Cabau has received research grants from Edwards Lifesciences and Medtronic. Dr. Webb has received research grants from and served as a consultant for Edwards Lifesciences and Abbott. Dr. Barbanti has served as a consultant for Edwards Lifesciences. Dr. Ruel has received research grants from Medtronic and Edwards Lifesciences. Dr. Himbert has served as a proctor for Edwards Lifesciences and Medtronic. Dr. Nombela-Franco has served as a proctor for Abbott. Dr. Abizaid has served as a proctor for Edwards Lifesciences. Drs. de Brito and Nietlispach has served as a consultant for Edwards Lifesciences, Abbott, and Medtronic. All other authors have reported that they have no relationships relevant to the contents of this paper to disclose.
- Abbreviations and Acronyms
- coronary artery disease
- confidence interval
- New York Heart Association
- odds ratio
- transcatheter aortic valve replacement
- Received September 26, 2017.
- Revision received May 28, 2018.
- Accepted June 26, 2018.
- 2018 American College of Cardiology Foundation
- Reardon M.J.,
- Van Mieghem N.M.,
- Popma J.J.,
- et al.,
- SURTAVI Investigators
- Grover F.L.,
- Vemulapalli S.,
- Carroll J.D.,
- et al.,
- STS/ACC TVT Registry
- Mylotte D.,
- Osnabrugge R.L.,
- Windecker S.,
- et al.
- Thyregod H.G.,
- Steinbruchel D.A.,
- Ihlemann N.,
- et al.
- Ball W.T.,
- Sharieff W.,
- Jolly S.S.,
- et al.
- Hopper A.N.,
- Jamison M.H.,
- Lewis W.G.
- Suri R.M.,
- Minha S.,
- Alli O.,
- et al.
- Alli O.O.,
- Booker J.D.,
- Lennon R.J.,
- Greason K.L.,
- Rihal C.S.,
- Holmes D.R. Jr..
- Kappetein A.P.,
- Head S.J.,
- Genereux P.,
- et al.
- Elze M.C.,
- Gregson J.,
- Baber U.,
- et al.
- Holmes D.R. Jr..,
- Nishimura R.A.,
- Grover F.L.,
- et al.,
- STS/ACC TVT Registry
- Le Morvan P.,
- Stock B.
- Bridgewater B.,
- Grayson A.D.,
- Au J.,
- et al.
- Hess C.N.,
- Peterson E.D.,
- Neely M.L.,
- et al.
- Carroll J.D.,
- Vemulapalli S.,
- Dai D.,
- et al.
- Auffret V.,
- Lefevre T.,
- Van Belle E.,
- et al.,
- FRANCE TAVI Investigators
- Khera S.,
- Kolte D.,
- Gupta T.,
- et al.