Author + information
- Received September 1, 2015
- Revision received November 12, 2015
- Accepted December 17, 2015
- Published online March 28, 2016.
- Edward L. Hannan, PhDa,∗ (, )
- Zaza Samadashvili, MDa,
- Nicholas J. Stamato, MDb,
- Stephen J. Lahey, MDc,
- Andrew Wechsler, MDd,
- Desmond Jordan, MDe,
- Thoralf M. Sundt III, MDf,
- Jeffrey P. Gold, MDg,
- Carlos E. Ruiz, MDh,
- Mohammed H. Ashraf, MDi and
- Craig R. Smith, MDe
- aSchool of Public Health, University at Albany, State University of New York, Albany, New York
- bDepartment of Cardiology, Campbell County Memorial Hospital, Gillette, Wyoming
- cDivision of Cardiothoracic Surgery, University of Connecticut, Storrs, Connecticut
- dDepartment of Cardiothoracic Surgery, Drexel University College of Medicine, Philadelphia, Pennsylvania
- eDepartment of Surgery, Columbia-Presbyterian Medical Center, New York, New York
- fCardiac Surgical Division, Massachusetts General Hospital, Boston, Massachusetts
- gUniversity of Nebraska Medical Center, Omaha, Nebraska
- hDivision of Structural and Congenital Heart Disease, Lenox Hill Hospital, New York, New York
- iDepartment of Cardiothoracic Surgery, Kaleida Health, Buffalo, New York
- ↵∗Reprint requests and correspondence:
Dr. Edward L. Hannan, School of Public Health, State University of New York, University at Albany, One University Place, Rensselaer, New York 12144-3456.
Objectives The purpose of this study was to investigate changes in the use of transcatheter aortic valve replacement (TAVR) relative to surgical aortic valve replacement (SAVR) and to examine relative 1-year TAVR and SAVR outcomes in 2011 to 2012 in a population-based setting.
Background TAVR has become a popular option for patients with severe aortic stenosis, particularly for higher-risk patients.
Methods New York’s Cardiac Surgery Reporting System was used to identify TAVR and SAVR volumes and to propensity match TAVR and SAVR patients using numerous patient risk factors contained in the registry to compare 1-year mortality rates. Mortality rates were also compared for different levels of patient risk.
Results The total number of aortic valve replacement patients increased from 2,291 in 2011 to 2,899 in 2012, an increase of 27%. The volume of SAVR patients increased by 7.1% from 1,994 to 2,135 and the volume of TAVR patients increased 157% from 297 to 764. The percentage of SAVR patients that were at higher risk (≥3% New York State [NYS] score, equivalent to a Society of Thoracic Surgeons score of about 8%) decreased from 27% to 23%, and the percentage of TAVR patients that were at higher risk decreased from 83% to 76%. There was no significant difference in 1-year mortality between TAVR and SAVR patients (15.6% vs. 13.1%; hazard ratio [HR]: 1.30 [95% confidence interval (CI): 0.89 to 1.92]). There were no differences among patients with NYS score <3% (12.5% vs. 10.2%; HR: 1.42 [95% CI: 0.68 to 2.97]) or among patients with NYS score ≥3% (17.1% vs. 14.5%; HR: 1.27 [95% CI: 0.81 to 1.98]).
Conclusions TAVR has assumed a much larger share of all aortic valve replacements for severe aortic stenosis, and the average level of pre-procedural risk has decreased substantially. There are no differences between 1-year mortality rates for TAVR and SAVR patients.
Until recently, many patients with severe symptomatic aortic stenosis were not candidates for surgical intervention via surgical aortic valve replacement (SAVR) because their perioperative risk was thought to be too high. However, with the introduction of transcatheter aortic valve replacement (TAVR), many of these patients have been faring better than they would have with medical treatment alone. Also, TAVR has become an option for patients who are operable but have very high short-term risks with SAVR, and numerous randomized controlled trials and observational studies have shown that TAVR outcomes are frequently as good or better than SAVR outcomes (1–33). However, for the most part, these studies were limited to very high-risk patients or examined short-term outcomes (29,30).
TAVR is now used in some settings for moderate-risk as well as high-risk patients, and information about the use of TAVR in standard practice for lower-risk patients is limited and based on small samples. The purposes of this study were to compare 1-year mortality rates for TAVR and SAVR after having adjusted for differences in acuity among TAVR and SAVR patients in the observational database, and to use a 2010 to 2012 New York State (NYS) short-term mortality risk model for valve patients to examine relative outcomes of TAVR and SAVR across different ranges of pre-procedural risk.
The primary database used for the analyses is New York’s Cardiac Surgery Reporting System, which is a clinical registry developed in New York in 1989 that, since its inception, has been used by the New York State Department of Health to publicly release risk-adjusted mortality rates for coronary artery bypass graft surgery and cardiac valve surgery. These reports contain risk-adjusted mortality rates for hospitals and surgeons and are released on an annual basis. Data in the system include patient demographics (age, sex, race, and so on), numerous patient risk factors and comorbidities, patient disposition, complications of care, and hospital and surgeon identifiers.
New York State’s vital statistics data were matched to the Cardiac Surgery Reporting System using unique patient identifiers to capture deaths that occurred in New York State following discharge from the hospital. The study was limited to New York State residents to minimize the chance of deaths occurring outside of New York.
Patients and hospitals
There were a total of 5,190 patients with aortic stenosis who underwent valve replacement/implantation in New York between January 1, 2011, and December 31, 2012, 4,129 of whom underwent SAVR without coronary artery bypass graft surgery and 1,061 of whom underwent TAVR without percutaneous coronary intervention. For the portion of the study related to 1-year term outcomes, exclusions consisted of out-of-state patients (n = 697; 420 SAVR, 277 TAVR), patients in non-TAVR hospitals (n = 1,963, all SAVR), patients with invalid personal identifiers (n = 106; 97 SAVR, 9 TAVR), patients with shock or hemodynamic instability (n = 4; 4 SAVR, 0 TAVR), patients with acute myocardial infarction onset within 24 h prior to the procedure (n = 5; 4 SAVR, 1 TAVR), patients with emergency priority (n = 18; 14 SAVR, 4 TAVR), patients with active endocarditis (n = 10; 10 SAVR, 0 TAVR), patients <60 years of age (n = 547; 538 SAVR, 9 TAVR), and patients ≥95 years of age (n = 44; 4 SAVR, 40 TAVR). This left 2,292 patients, 1,545 of whom underwent SAVR and 747 of whom underwent TAVR. Patients in non-TAVR hospitals (hospitals with no TAVR procedures) were excluded to make the paired patients as similar as possible.
Hospitals in the study include all 17 hospitals in which TAVR procedures were performed during the time period of the study. The median hospital volume for all aortic valve replacements was 105, with an interquartile range of 53 to 226. Patients in the study were followed for 1 year.
The outcomes of interest are all-cause in-hospital/30-day mortality and 1-year mortality. Only patients who could be followed for an entire year were included in the 1-year mortality part of the study. We also examined changes in TAVR and SAVR volumes as a function of risk level for 2011 and 2012.
TAVR and SAVR patients were classified into short-term risk categories of <3% and ≥3% for 2011 and 2012 separately based on the NYS in-hospital/30-day mortality risk model for isolated valve surgery (34). Although other studies used the Society of Thoracic Surgeons (STS) predicted rate of mortality (PROM) risk score (35) to classify patients, we were unable to use that score because some data elements in it are not contained in the New York State registry. However, because 1 of the TAVR hospitals in New York was able to calculate both NYS and STS scores with its expanded database, we were able to correlate the 2 scores based on that hospital’s 2013 to 2014 data (R = 0.72) and translate the STS score into an NYS score based on a linear regression model. An STS score of 8, used in some studies to distinguish between medium- and high-risk patients, translated into an NYS score of 2.94, so we used an NYS score of 3 as the cutoff point. Changes in volumes of SAVR and TAVR patients between 2011 and 2012 were then examined by NYS risk group (<3%, ≥3%) to understand the penetration of TAVR over time.
Prevalences of all available patient risk factors were compared for SAVR and TAVR patients, and standardized differences were calculated. Although frailty was not contained in the registry, we captured frailty using a measure developed by the Centers for Medicare & Medicaid Services for their mortality reports (36) by matching registry data to New York’s administrative dataset, the Statewide Planning and Research Cooperative System, and then using decubitus or chronic skin ulcer, dementia, senility, metastatic cancer, acute leukemia, hemiplegia, paraplegia, paralysis, functional disability, or stroke coded as a secondary diagnosis as a measure of frailty.
Propensity score matching was used to identify a set of TAVR/SAVR pairs matched on all available characteristics so that the selection bias associated with our observational study could be minimized. The propensity score was derived by developing a logistic regression model that predicted the probability that a given patient would receive TAVR on the basis of all of the risk factors available in the registry. The propensity score was used to match patients on a 1-to-1 basis so as to minimize the overall distance in propensity scores between the groups (37–40). The TAVR and SAVR patients were matched unless their estimated log-odds from the logistic regression model were more than 0.2 SDs apart. Standardized differences in the prevalence of propensity model variables were then calculated (41).
The propensity matched pairs were then used to analyze differences in mortality rates between the 2 groups. To eliminate further differences within the matched pairs, a Cox proportional hazards model was used to calculate the hazard ratio for the 2 procedures after adjusting for all risk factors in the propensity score.
The mortality rates were also compared for patients in 2 ranges (<3% and ≥3%) of pre-procedural short-term risk, and for subsets of patients (age <80 and ≥80 years, nonfrail patients). The analyses were repeated with propensity matching that required each matched pair of patients to be from the same hospital to ensure that variable hospital quality did not affect the results. The findings were essentially unchanged and are not reported here. All tests were 2-sided and conducted at the p = 0.05 level, and all analyses were conducted in SAS version 9.2 (SAS Institute, Cary, North Carolina).
Table 1 presents the procedural volumes for TAVR and SAVR patients in New York for 2011 and 2012 separately, broken down by NYS predicted mortality ranges of <3% and ≥3%. The total number of aortic valve replacement patients increased from 2,291 in 2011 to 2,899 in 2012, an increase of 27%, and the increase in TAVR patients was from 297 to 764, an increase of 467 (157%). The percentage of TAVR patients that were at higher risk decreased from 83% to 76%.
After exclusions, 1,545 SAVR patients and 747 TAVR patients were candidates for propensity matching. As demonstrated, the standardized differences for most risk factors exceeded the recommended maximum value of 10%, with TAVR patients generally having higher prevalences (Table 2). Before propensity matching, the median NYS score for TAVR patients was 4.91% and the median score for SAVR patients was 2.13% (p < 0.001).
A total of 405 pairs resulted from the propensity match. The propensity model has a very good discrimination, with C statistic of 0.881 and acceptable calibration (p = 0.08). Also, standardized differences for all variables were lower than the typical cutoff of 10% for this measure (Table 3). The median NYS score for propensity-matched TAVR patients was 3.74, and the median score for SAVR patients was 3.93 (p = 0.48).
The respective in-hospital/30-day mortality results for TAVR and SAVR patients were 5.49% and 2.52% (p < 0.001) prior to propensity matching, and 4.69% and 4.69% (p = 1.00) after matching. The respective observed 1-year mortality rates for TAVR and SAVR prior to adjustment were 21.4% and 7.1%. There was no significant difference in 1-year all-cause mortality between TAVR and SAVR patients (15.6% vs. 13.1%; hazard ratio [HR]: 1.30 [95% confidence interval (CI): 0.89 to 1.92]) (Table 4). There were no significant differences between the procedures for the 2 ranges of NYS scores: 12.5% vs. 10.3%; HR: 1.42 (95% CI: 0.68 to 2.97) for NYS score <3% and 17.10% vs. 14.5%; HR: 1.27 (95% CI: 0.81 to 1.98) for patients with NYS score ≥3%. There were also no differences between the mortality rates for the 2 procedures for patients age <80 and ≥80 years or for patients without frailty.
In a relatively short period of time, TAVR has become a very popular alternative for treating high-risk patients with severe aortic stenosis. Since 2011, 4 randomized controlled trials that compared TAVR and SAVR have been published. In the PARTNER (Transcatheter Versus Surgical Aortic-Valve Replacement in High-Risk Patients) trial for patients deemed to be at high surgical risk (coexisting conditions associated with a risk of death of at least 15% within 30 days of the procedure), Smith et al. (1) randomized 348 patients to undergo TAVR and 351 to undergo SAVR in 25 centers. For the primary outcome of all-cause mortality at 1 year, the mortality rates were not significantly different (24.2% vs. 26.8%; risk difference −2.6%, 95% CI: −9.3% to 4.1%). Also, Kodali et al. (2) reported 2-year all-cause mortality in the PARTNER trial in 2012. The authors found that there was no significant difference between TAVR and SAVR (34% vs. 35%; p = 0.78).
In the STACCATO (A Prospective randomized Trial of Transapical Transcatheter Aortic Valve Replacement in Operable Elderly Patients with Aortic Stenosis) trial, Nielson et al. (3) randomized 70 patients of age 75 and older with operable severe aortic stenosis to undergo transapical TAVR or SAVR in 2 Denmark hospitals (3). Using a primary endpoint of 30-day all-cause mortality, major stroke, or renal failure requiring dialysis, the trial was ended prematurely by the data safety monitoring board because of an excess of endpoint events in the TAVR group (14.7% vs. 2.8%; p = 0.07).
In the U.S. CoreValve High Risk Study in 2014, Adams et al. (4) randomized 390 TAVR and 357 SAVR (both numbers after patient refusal following randomization) “high-risk” surgical patients to undergo TAVR or SAVR in 45 centers in the United States. With a primary endpoint of all-cause mortality at 1 year, the authors found a significantly lower rate for TAVR (14.2% vs. 19.1%; p = 0.04).
In the NOTION (Nordic Aortic Valve Intervention Trial) trial, Thyregod et al. (28) randomized a total of 280 patients (with a mean STS PROM score of 3.0) with severe aortic stenosis in 3 Nordic centers. Findings from the study were that there was no significant difference in the primary endpoint (death, stroke, or myocardial infarction) after 1 year (13.1% vs. 16.3%; p = 0.43 for superiority).
With regard to nonrandomized studies, in a meta-analysis of 11 nonrandomized trials, Cao et al. (5) found that there was no significant difference between TAVR and SAVR in periprocedural all-cause in-hospital/30-day mortality (8.4% vs. 7.2%; relative risk [RR]: 1.24; 95% CI: 0.92 to 1.62). Also, Holmes et al. (29) used data from the STS/American College of Cardiology Transcatheter Valve Therapies Registry linked with Centers for Medicare & Medicaid Services claims data to examine 1-year outcomes for TAVR patients at 299 U.S. hospitals. They found that the patients had a median STS PROM score of 7.1%, a 30-day mortality rate of 7.0%, and a 1-year mortality rate of 23.7%.
Our study yielded some interesting information on the evolution of TAVR in New York during 2011 to 2012. We found that there was a total increase of 27% in all aortic valve replacement procedures between 2011 and 2012, and that the increase was almost entirely attributable to the increase in TAVR procedures (from 297 to 764, an increase of 157%). Also, there was a decrease in the proportion of higher-risk patients undergoing aortic valve replacement (NYS score >3%, approximately equivalent to an STS score of >8%). The percentage of these patients undergoing SAVR decreased from 27% to 23% and the percentage undergoing TAVR decreased from 83% to 76%. Also, TAVR procedures with an NYS score of <3% increased from 17% of all TAVR procedures in 2011 to 24% of all TAVR procedures in 2012.
After propensity matching and using a Cox proportional hazards model to further reduce differences among TAVR and SAVR patients, we then compared 1-year mortality rates for TAVR and SAVR. We found no significant overall difference in mortality between TAVR and SAVR patients (15.6% vs. 13.1%; HR: 1.30 [95% CI: 0.89 to 1.92]). We also found no differences among patients with NYS score <3% (12.5% vs. 10.2%; HR: 1.42 [95% CI: 0.68 to 2.97]) or among patients with NYS score Holmes ≥3% (17.1% vs. 14.5%; HR: 1.27 [95% CI: 0.81 to 1.98]).
The findings of our study are in accord with the findings of the observational study by Piazza et al. (31), which found no differences at 1 year either for patients with 3% to 8% STS scores or for patients with <3% scores (note that the cutoff point of 3% NYS score is essentially the same as an 8% STS score). However, it is important to note that although the HRs in our study were not significant, they ranged from 1.27 for patients with NYS score ≥3% to 1.42 for patients with NYS score <3%, and if these ratios had persisted with larger sample sizes, SAVR would have had significantly lower mortality. Given the observational nature of our study, we await the results of clinical trials (e.g., PARTNER II and SURTAVI [Medtronic CoreValve Surgical Replacement and Transcatheter Aortic Valve Implantation]) that are currently underway to compare TAVR and SAVR outcomes for medium-risk patients (42).
First, since this is an observational study, its results are subject to potential selection biases in comparison to the results from randomized controlled trials. We attempted to minimize this bias by using the best available methods to propensity match patients on the basis of numerous patient characteristics. The statistic used for measuring the quality of the match (percent standardized difference in prevalence) indicates that none of the numerous matching characteristics exceeded the recommended limit of 10%. Also, the propensity analyses were further adjusted by using Cox proportional hazards models to adjust for differences that remained after propensity matching.
Thus, we did everything possible to minimize selection bias with the clinical registry database we had. However, it should be noted that we did not have access to some important variables for risk adjustment, including frailty, severe hepatic disease without failure, severe pulmonary hypertension, and cognitive impairment in the propensity matching process. We attempted to capture frailty by matching registry data to New York’s administrative dataset and then using the presence of a variety of conditions as proxies for frailty. However, these conditions may be undercoded, particularly among patients with several severe comorbidities.
It should be noted that the procedures were all performed in hospitals that have both SAVR and TAVR programs, meaning that the surgeon/cardiologist teams decided which procedure to use. It is likely that lower-risk patients who underwent TAVR appeared to be at higher risk in some regard and higher-risk patients who underwent SAVR appeared to be at lower risk in the opinion of the clinicians. The bias introduced by heart team choice cannot be removed. Another weakness of the study is that it compares a mature versus a new therapy, so it may not be indicative of future relative outcomes.
Also, we did not have access to socioeconomic status, which may have been related to selection bias. It is also important to note that the process of propensity matching eliminated many high-risk TAVR patients and low-risk SAVR patients because they could not be matched with similar patients undergoing the other procedure. Prior to propensity matching, the median NYS score was 4.91 for TAVR patients and 2.13 for SAVR patients, and after matching the respective median scores were 3.74 and 3.93.
In addition, we were unable to compare transfemoral and transapical TAVRs with SAVR because the vast majority of TAVR patients underwent transfemoral TAVRs (only 62 patients in the matched cohort underwent transapical TAVRs).
WHAT IS KNOWN? Numerous randomized controlled trials and observational studies have shown that TAVR outcomes are frequently as good as or better than SAVR outcomes. However, for the most part, these studies were limited to very high-risk patients or examined short-term outcomes. Also, little population-based data from an entire region are available.
WHAT IS NEW? In New York, the percentage of aortic valve replacements done using the transcatheter approach increased from 13.0% in 2011 to 26.4% in 2012, primarily among low- to medium-risk patients. This increase was largely responsible for an increase of 26.5% in the total number of aortic valve replacements performed. There was no significant difference in 1-year mortality between TAVR and SAVR patients (15.6% vs. 13.1%; HR: 1.30 [95% CI: 0.89 to 1.92]), and there were no differences in mortality after classifying procedures as low and high risk.
WHAT IS NEXT? With the increasing use of TAVR among lower-risk patients, it will be important to determine which types of patients experience superior outcomes for each type of aortic valve replacement as utilization patterns and procedural improvements occur.
The authors would like to thank Kimberly S. Cozzens, Rosemary Lombardo, and the cardiac surgery departments of the participating hospitals for their tireless efforts to ensure the timeliness, completeness, and accuracy of the registry data.
Dr. Sundt is a scientific advisor for Thrasos Innovation; and is the principal investigator for the PARTNER trial (Edwards Lifesciences) and the PERIGON trial (Medtronic) at Massachusetts General Hospital. Dr. Ruiz has received a grant and speaker honoraria from St. Jude Medical Center. Dr. Smith is the surgical principal investigator of the PARTNER trial, for which he receives reimbursement from the sponsor (Edwards Lifesciences) for travel and customary expenses related to trial management. All other authors have reported that they have no relationships relevant to the contents of this paper to disclose.
- Abbreviations and Acronyms
- New York State
- surgical aortic valve replacement
- Society of Thoracic Surgeons
- transcatheter aortic valve replacement
- Received September 1, 2015.
- Revision received November 12, 2015.
- Accepted December 17, 2015.
- American College of Cardiology Foundation
- Eltchaninoff H.,
- Prat A.,
- Gilard M.,
- et al.
- D’Errigo P.,
- Barbanti M.,
- Ranucci M.,
- et al.,
- for the OBSERVANT Research Group
- Walther T.,
- Schuler G.,
- Borger M.A.,
- et al.
- Clavel M.A.,
- Webb J.G.,
- Rodes-Cabau J.,
- et al.
- Wenaweser P.,
- Pilgrim T.,
- Kadner A.,
- et al.
- Moat N.E.,
- Ludman P.,
- de Belder M.A.,
- et al.
- Tamburino C.,
- Capodanno D.,
- Ramondo A.,
- et al.
- Thomas M.,
- Schymik G.,
- Walther T.,
- et al.
- Chieffo A.,
- Buchanan G.L.,
- Van Mieghem N.M.,
- et al.
- Mylotte D.,
- Osnabrugge R.L.,
- Windecker S.,
- et al.
- Rodés-Cabau J.,
- Webb J.G.,
- Cheung A.,
- et al.
- Thyregod H.G.H.,
- Steinbruchel D.A.,
- Ihlemann N.,
- et al.
- Piazza N.,
- Kalesan B.,
- van Mieghem N.,
- et al.
- Reardon M.J.,
- Adams D.H.,
- Kleiman N.S.
- ↵New York State Department of Health. Adult Cardiac Surgery in New York State: 2008–2010. August 2012. Available at: http://www.health.ny.gov/statistics/diseases/cardiovascular/heart_disease/docs/2008-2010_adult_cardiac_surgery.pdf. Accessed July 16, 2015.
- ↵Centers for Medicare and Medicaid Services. Methodology and updates reports: mortality measures. Available at: http://www.qualitynet.org/dcs/ContentServer?c=Page&pagename=QnetPublic%2FPage%2FQnetTier4&cid=1163010421830, Accessed July 17, 2015.
- Rosenbaum P.R.,
- Rubin D.B.
- Rosenbaum P.R.
- Klein J.P.,
- Moeschberger M.L.