Technology Assessment (TA)
TA Criterion 1: The Technology Must Have Final Approval From the Appropriate Government Regulatory Bodies
On November 2, 2011, the SAPIEN Transcatheter Heart Valve model 9000TFX received FDA PMA approval for transfemoral delivery in patients with severe, symptomatic native aortic valve stenosis who have been determined to be inoperable by a cardiac surgeon.
The FDA approval requires that Edwards Lifesciences conduct two post approval studies.
A National Transcatheter Aortic Valve Replacement registry must be developed within four months of the start of the post approval studies to house study data (pre-procedure, peri-procedure, post-procedure, discharge, 30-day, and 1-year follow-up). The registry will be housed jointly by the American College of Cardiology and the Society for Thoracic Surgeons. Finally, the data will be linked to CMS data for long term follow up including annual follow up through five years post implant.
TA Criterion 1 is met.
TA Criterion 2: The Scientific Evidence Must Permit Conclusions Concerning the Effectiveness of the Technology Regarding Health Outcomes
The Medline database, Embase, Cochrane clinical trials database, Cochrane reviews database and the Database of Abstracts of Reviews of Effects (DARE) were searched using the key words "transcatheter aortic valve replacement" OR "percutaneous aortic valve replacement" OR "CoreValve" OR "Sapien." The search was performed for the period from 1945 through December 2011. The bibliographies of systematic reviews and key articles were manually searched for additional references. References were also solicited from the manufacturers and local experts. The abstracts of citations were reviewed for relevance and all potentially relevant articles were reviewed in full. We included case series describing at least 100 patients treated with TAVR, comparative studies with medical therapy, and randomized trials comparing TAVR to medical therapy.
The search identified 1047 potentially relevant studies (Figure 1). After elimination of duplicate and non-relevant references including reviews and animal studies the search identified 19 articles describing thirteen case series,[34–48] two small comparative trials[49,50] and one randomized trial.[51,52] An additional 42 early case series were excluded because they were too small to provide reliable estimates of the outcomes of interest.[33,53–93] Seven comparative studies were excluded because they compared TAVR to surgery and not to medical therapy.[94–100]
Level of Evidence: 1, 3, and 5.
TA Criterion 2 is met.
TA Criterion 3: The Technology Must Improve Net Health Outcomes
Symptomatic aortic stenosis has a high short-term mortality so total mortality should be the primary outcome of interest. Treatment of aortic stenosis is associated with an increased risk for strokes and some patients perceive that their quality of life living with the disabilities of a stroke could be worse than death, so stroke risk is a major concern. Clearly quality of life and functional ability are also major concerns. The New York Heart Association developed the system that is most commonly used to quantify the degree of functional limitation imposed by heart disease. This system assigns patients to one of four functional classes, depending on the degree of effort needed to elicit symptoms: patients may have symptoms of HF at rest (class IV), on less-than-ordinary exertion (class III), on ordinary exertion (class II), or only at levels of exertion that would limit normal individuals (class I). Two more detailed quality of life questionnaires, the Kansas City Cardiomyopathy Questionnaire (KCCQ) and the Medical Outcomes Study Short Form 12 (SF-12) were used in some of the studies. The KCCQ is a validated 23-item questionnaire focusing on five domains: symptoms, physical limitation, social limitation, self-efficacy, and quality of life.[102,103] It is scored on a 100 point scale with higher scores representing higher functioning and better quality of life. A change of more than five points is considered clinically significant.[104,105] The SF-12 includes both physical and mental health summary scores with higher scores representing better health-related quality of life. A change of greater than two points is considered clinically significant. Other important peri-procedural and longer term outcomes include myocardial infarction, major bleeding, kidney injury, major vascular complications, and valve dysfunction including aortic regurgitation. A consensus panel published clear definitions for these outcomes in 2011. Prior to that time, the definitions for these outcomes varied somewhat between studies.
These larger case series are useful for defining the expected range of benefits and harms from the use of TAVR[34–48] Table 2 describes when and where the studies were performed, which valve type was used, and a brief summary of the characteristics of patients in the studies. Several studies separated the data into groups based on whether the valve was implanted using a transfemoral approach (TF) or a transapical approach (TA). Data for the SAPIEN valve deployed using the transfemoral approach is most relevant for this assessment because the current FDA indication is for this delivery approach. Patients requiring a transapical approach usually have significant peripheral vascular disease that precludes a transfemoral approach and concomitantly more atherosclerosis in the cerebrovascular and coronary circulation, putting them at higher risk for procedural complications and death.
The valves used in these studies were implanted from 2005 through 2010. Both the CoreValve and the SAPIEN valve have gone through several generations of development over that time period based on insights gained during the early case series. About half of the participants in these studies were women and their average age was slightly greater than 80 years. The predicted risk for operative mortality varied from 19% to 34% using the EuroSCORE and from 9% to 19% using the STS score. When both scores were reported, the EuroSCORE was consistently higher than the STS score.
Table 3 summarizes the procedural outcomes and events through thirty days. The procedural success rate was generally high, ranging from 95% to 99% in the most recent series. The most important outcome, 30-day mortality, varied from 6% to 13%, likely due to differences in the underlying risk of the different patient groups studied. The risk of major stroke varied from 1% to 5%. TAVR is known to cause a number of arrhythmias and 2% and 39% of patients in these case series required insertion of a permanent pacemaker. Case series that evaluated both the CoreValve and the SAPIEN valve suggested that the requirement for a pacemaker is more common with the CoreValve.[36,41,47] Major vascular complications varied widely (2% to 28%), in part because of differences in what was classified as a vascular complication. Aortic regurgitation, both around and through the implanted valve, is a common complication, but most is trivial to mild and unlikely to cause symptoms. However the prevalence of moderate to severe aortic regurgitation has been reported to be between 2% and 42% in these case series. The rates of MIs, valve embolization, and cardiac tamponade were generally low (<1%).
Table 4 summarizes the outcomes at one year including quality of life outcomes. Many of the studies did not have sufficient follow-up to report one-year outcomes. One-year mortality ranged from 15% to 28% reflecting the age and comorbidity burden of patients with symptomatic AS. The rate of major stroke ranged from 4% to 10%. The majority of surviving patients reported NYHA Class I or II symptoms. In the one study reporting quality of life using the KCCQ, there were large improvements in the summary score and in a second quality of life measure, the EuroQoL. The need for repeat interventions on the implanted valve was between 1% and 3%, but these data were not consistently reported.[45,47]
One study, the UK TAVR registry, reported two-year outcomes in addition to one year outcomes. Their one-year survival was 79% and their two year survival was 74%, suggesting that patients who survive the first year continue to do well. The authors note that 61% of patients had paravalvular aortic regurgitation (AR) that would not have been considered acceptable following surgical aortic valve replacement (SAVR) and that moderate to severe AR was a significant predictor of mortality at one year (hazard ratio 1.7, 95% confidence interval 1.1 to 2.5). The authors suggest that design improvements aimed at reducing AR may lead to better long-term outcomes following TAVR.
Case series are useful for defining important clinical outcomes to evaluate in comparative studies, but are not useful for comparing the effectiveness of TAVR to surgical AVR or to medical therapy. The incidence of the major outcomes varied widely between studies. This likely represents a range of causes including variations in the underlying patient populations studied, skills of the team deploying the device, improvements in device design over time, and variation in the definitions used for the outcomes. The wide variation in the incidence of important outcomes was highlighted in a systematic review that included many of the smaller, early studies. The 30-day mortality ranged from 0% to 25%, MI from 0% to 15%, conversion to surgery 0% to 8%, moderate to major paravalvular leaks from 3% to 35%, and 30-day major cardiovascular and cerebral events from 3% to 35%, and six-month mortality from 18% to 48%. These large differences in event rates between studies precludes any meaningful conclusions about the role of TAVR based on case-series data. Either large, high quality comparative trials or randomized trials are needed to clearly define the relative risks and benefits of TAVR in well characterized patient populations.
Two studies comparing outcomes in high risk patients with symptomatic AS referred for TAVR who either were treated with TAVR or were managed medically are also summarized in Table 2 , Table 3 , and Table 4 .[49,50] Both studies reported much higher mortality in the patients who received medical therapy (44% versus 22% and 28% versus 13%). However, both studies were small and neither attempted to adjust for differences between the medical and TAVR groups. In one of the studies, the authors explicitly state that 28% of the patients were in the no intervention group because of "mortality before definitive treatment because of limitations in the number of patients who could be enrolled for" TAVR. Thus, all patients who died quickly were by design included in the medical treatment group – clearly biasing the findings of the study. Large comparative studies using sophisticated techniques like propensity score matching to adjust for confounding by indication have the potential to say something meaningful about the role of TAVR, but these two small studies made no attempt to adjust for potential confounders. They suggest that there may be a role for TAVR, but add little to the evaluation of comparative effectiveness.
Randomized Clinical Trials
The Placement of Aortic Transcatheter Valves (PARTNER) Trial The PARTNER trial included two parallel, randomized clinical trials, one in operable, high risk patients with transfemoral access (PARTNER A) and one in patients ineligible for surgery (PARTNER B). This assessment will only consider the PARTNER B trial as TAVR does not have FDA approval for use in patients who are eligible for surgical aortic valve replacement.
Inoperable patients were expected to have a predicted probability of death or permanent disability of at least 50% and the agreement of two cardiac surgeons that they were not suitable candidates for surgery. Patients were required to have severe aortic stenosis defined by either an aortic valve area of less than 0.8 cm2, an aortic valve gradient of at least 40 mm Hg, or a peak aortic jet velocity of at least 4.0 m per second. They also were required to have at NYHA class II, III, or IV symptoms. Patients were excluded if they did not have transfemoral access or had a bicuspid or non-calcified aortic valve, acute myocardial infarction, coronary artery disease requiring revascularization, left ventricular ejection fraction less than 20%, severe mitral or aortic regurgitation, transient ischemic attack or stroke in the prior six months, or severe renal insufficiency (creatinine > 3.0 mg/dl or preoperative renal replacement).
Eligible patients were randomized to receive the second-generation Edwards SAPIEN heart-valve system or usual care. Patients randomized to TAVR group received heparin therapy during the procedure and dual antiplatelet therapy with aspirin and clopidogrel for six months following the procedure. The primary endpoint of the trial was death from any cause.
The study characteristics and findings are summarized in Table 2 , Table 3 , and Table 4 . In brief, 179 patients were randomized to each group. Follow-up was 100% complete through one year and the primary analysis was done according to strict intention to treat principles. No blinding was reported and no sham procedure was performed. It is unclear if endpoint ascertainment and adjudication were blinded.
The average age of the participants was 83 years and a little more than half were female. Their predicted operative mortality by the logistic EuroSCORE was 28% and by the STS score was 12%. Notably, the logistic EuroSCORE risk was significantly lower in the TAVR group (26.4% versus 30.4%, p=0.004) and the STS score was also lower, but not significantly lower (11.2% versus 12.1%, p=0.14). Thus, despite randomization, the TAVR group was at lower risk for operative mortality than the standard care group. In addition, the TAVR group had less COPD (41.3% versus 52.5%, p=0.04), less atrial fibrillation (32.9% versus 48.8%, p=0.04), and a trend towards better left ventricular function (ejection fraction 53.9% versus 51.1%, p=0.06).
TAVR was successful for 97% of patients in the TAVR group, but 16% experienced major vascular complications (aortic dissection; left ventricular perforation; embolization resulting in permanent damage; vascular injury requiring surgical intervention or causing death, permanent disability, or blood transfusion of at least 3 units), 17% had major bleeding (causing death or prolonged hospitalization or requiring surgical intervention or at blood transfusion of at least 3 units), and 11.8% had moderate or severe paravalvular aortic regurgitation. In the first 30 days there were no MIs in either group, but there was a non-significant trend toward more deaths from any cause (5.0% versus 2.8%, p=0.41) and major strokes (5.0% versus 1.1%, p=0.06) in the TAVR group. Interestingly, there was a trend towards placement of fewer permanent pacemakers (3.4% versus 5.0%, p=0.60) and less need for dialysis (1.1% versus 1.7%, p=1.00) in the TAVR group.
At one year outcomes clearly favored the TAVR group. There were significantly fewer deaths from any cause (30.7% versus 49.7%, p<0.001), fewer hospitalizations (22.3% versus 41.9%, p<0.001), and fewer deaths or hospitalizations (42.5% versus 70.4%, p<0.001) in the TAVR group. There was still a trend towards more major strokes in the TAVR group (7.8% versus 3.9%, p=0.18) but it was not statistically significant. In the TAVR group, there were few re-interventions of the valve. Three patients had a repeat TAVR within 24 hours of the initial procedure, two had surgical aortic valve replacement and one had a balloon aortic valvuloplasty. In the standard therapy group, four patients received TAVR off protocol at other institutions, 17 patients had surgical aortic valve replacement, and 150 (84%) had at least one balloon aortic valvuloplasty with 30 having a second valvuloplasty during the first year of follow-up.
The quality of life outcomes at one year also favored the TAVR group. The proportion of patients reporting NYHA class I or II symptoms was higher in the TAVR group (74.8% versus 42.0%, p<0.001). In the TAVR group, the KCCQ summary score increased from 36.2 at baseline to 69.4 at one year (p<0.001), while the standard therapy group summary score only increased from 34.4 to 47.0 (p=0.20). The between group differences were clinically and statistically significantly different (p<0.001) at one, six and twelve months follow-up for the summary score and for all five domains of the KCCQ. At one year, patients in the TAVR group also reported significant increases in both the physical and mental health domains of the SF-12 (p<0.001), while no significant increases were found in the standard therapy group.
It is also noteworthy that the 30-day mortality in the PARTNER B trial was lower than that reported in any of the 13 large case series summarized in Table 2 even though most of the case series had lower average risk scores by both the EuroSCORE and STS models. For instance, the European PARTNER cohort had a 30-day mortality of 8.5%, much higher than the 2.8% mortality in the PARTNER B randomized trial. The opposite was true at one year: the mortality in the TAVR group in the PARTNER B randomized trial was higher than that reported by any of the case series except for the subgroup of patients in the European PARTNER case series treated using the transapical approach. This apparent paradox may reflect the fact that all of the TAVR procedures in the randomized trial were done via the lower risk transfemoral approach while many of the case series were a mix of patients treated using the transfemoral and transapical approaches. It also could be due to the technical skill and experience of the centers involved in the PARTNER B trial. The higher one-year mortality may be due to a higher comorbidity burden among the patients who were eligible for inclusion in this randomized trial of "inoperable" patients. However, these findings, combined with the baseline differences in risk scores in the two groups, suggest that real world outcomes are unlikely to be as impressive as those reported in the PARTNER B randomized trial.
The 50% one-year mortality in the standard therapy group was also remarkably high. There may be some element of selection bias despite randomization given the significant higher EuroSCORE and STS score in the standard therapy group. However, the absolute differences in the risk scores were not large enough to explain the 19% absolute difference in the one-year risk of death. In addition, the analysis was done by strict intention to treat and follow-up was 100% at one year, so there should be no additional selection bias due to loss to follow-up.
Finally, follow-up in the trial is relatively short. Problems with valve failure may not show up in short term clinical trials. A good example is one of the early heart valves: the Bjork-Shiley valve approved by the FDA in 1979, but withdrawn from the market in 1986 after it became clear that this version of the valve was prone to strut failures that caused death in two-thirds of the cases of failure. There are case series data beyond one-year of follow-up, but the number of patients followed is relatively low. Additional long-term follow-up will be essential to prove the durability of the one-year results of the PARTNER B trial.
In summary, inoperable patients randomized to TAVR had a significant lower risk of dying at one year compared to usual care and were less likely to be readmitted to the hospital. However, patients in the TAVR group experienced more major strokes and major vascular events. In spite of these harms, patients in the TAVR group had markedly better quality of life as assessed by NYHA class, the KCCQ, and the SF-12. An important concern that has been raised with these results is that 84% of patients in the control group received aortic valvuloplasty during follow-up. As Dr. Redberg noted in her letter, the ACC and AHA consider aortic valvuloplasty as "not useful and may be harmful" when used as primary therapy for aortic stenosis. The procedure does not appear to change the natural history of AS and is associated with a high incidence adverse events.[111–113] However, the ACC and AHA also concluded that balloon valvuloplasty may be considered for palliation in patients for whom surgery is not an option. And the patients in the PARTNER B trial were considered inoperable. Furthermore, the magnitude of harms following aortic valvuloplasty are not large enough to explain the 19% absolute difference in total mortality between the TAVR group and the standard therapy group at one year. In the authors' reply to Dr. Redberg, the investigators reported that there was only one death and two strokes within 7 days of balloon valvuloplasty among the 150 patients in the standard therapy group who were treated with that procedure.
TA Criterion 3 is met.
TA Criterion 4: The Technology Must be as Beneficial as Any Established Alternatives
For truly inoperable patients with severe symptomatic aortic stenosis, there is no established alternative to TAVR. The ACC/AHA guidelines do not recommend balloon aortic valvuloplasty in adults who are candidates for surgical aortic valve replacement, though it may be appropriate for palliation in some inoperable patients. The procedure is associated with a moderate reduction in the pressure gradient across the aortic valve, but the valve area rarely increases to greater than 1.0 cm2.[115,116] Serious complications (stroke, aortic regurgitation, MI) occur in more than 10% of patients[112,113,116,117] and restenosis usually occurs within six to 12 months.[115,116,118] Most importantly, the natural history of patients with aortic stenosis who are treated with balloon aortic valvuloplasty is the same as that of untreated aortic stenosis.[111,113,118]
The PARTNER B trial compared TAVR to "standard therapy," which in the context of this trial was valvuloplasty as 60% of patients in the standard therapy group received valvuloplasty within one month of randomization and an additional 24% received valvuloplasty during follow-up. As discussed under TA criterion 3, very few patients in the standard therapy group experienced serious adverse events within one week of valvuloplasty (one death, two strokes). Thus the marked decrease in total mortality and large improvements in quality of life in the TAVR arm of the PARTNER B trial at one year compared to standard therapy demonstrated that TAVR is more beneficial than established alternatives as long as patients are aware of the short term increased risk of death, stroke, aortic regurgitation, and major vascular events and long-term uncertainties about valve durability.
The key assumption underlying this reasoning is that patients receiving TAVR are truly inoperable. This was clearly not the case for all patients in the PARTNER B trial. Twelve patients in the standard therapy group had surgical aortic valve replacement during follow-up. In spite of their prohibitively high operative risk, their one-year mortality was only 33%. This was much lower than the remainder of the standard therapy group and about equivalent to the one-year mortality of the TAVR group. The PARTNER B trial demonstrated that TAVR is at least as beneficial as medical therapy or valvuloplasty, but only in truly inoperable patients. Care must be taken to ensure that appropriate patients are selected to receive TAVR.
TA Criterion 4 is met.
TA Criterion 5: The Improvement Must be Attainable Outside of the Investigational Setting
TAVR is a technically difficult procedure with a steep learning curve. In the Vancouver single center case series, outcomes improved over time. The investigators compared results in the first 135 patients to the results in the subsequent 135 patients. The overall success rate increased from 93% to 98% and the 30-day mortality decreased from 13% to 6%. There were fewer cases of device embolization, coronary occlusion, stroke, and major vascular injury in the second half of their case series. Some of the improvement may reflect changes in the design of the valvular implants and their delivery systems. In addition, patients in the second half of the case series were at significantly lower risk for poor outcomes (STS score 8.5 versus 10.5, p<0.01). In fact, in the much larger, multicenter UK TAVR registry, there was no significant difference between proctored and non-proctored cases or between the first twenty cases at a center and the subsequent cases. However, a large Italian study did show significantly higher 30-day mortality in the first third of the cases in their series. Clearly care must be taken to ensure adequate training and proctoring of physicians performing TAVR.
There are consensus recommendations that have been made jointly by the American College of Cardiology Foundation and the Society of Thoracic Surgeons. In brief, they recommend that programs utilizing TAVR should be performed by a limited number of specialized heart centers with multidisciplinary teams that include at least one primary cardiologist, one interventional cardiologist, and one cardiac surgeon. All personnel should receive appropriate training and credentialing and follow standard protocols set up by expert consensus groups for evidence-based patient selection, procedural details, and complication management. Finally, a registry should be established to track appropriate use and patient outcomes. If followed, these recommendations should ensure that appropriate patients are selected and that highly skilled teams perform the procedure.
TA Criterion 5 is met.
CTAF © 2012 California Technology Assessment Forum