Network meta-analysis on the comparative effectiveness and safety of transcatheter aortic valve implantation with CoreValve or Sapien devices versus surgical replacement.

INTRODUCTION: Surgical replacement for aortic stenosis is fraught with complications in high-risk patients. Transcatheter techniques may offer a minimally invasive solution, but their comparative effectiveness and safety is uncertain. We performed a network meta-analysis on this topic.
METHODS: Randomized trials on transcatheter aortic valve replacement vs surgery were searched. The primary outcome was all cause death. Risk estimates were obtained with Bayesian network meta-analytic methods.
RESULTS: Four trials with 1,805 patients were included. After a median of 8 months, risk of death and myocardial infarction was not different when comparing surgery versus transcatheter procedures, irrespective of device or access. Conversely, surgery was associated with higher rates of major bleeding (odds ratio vs CoreValve=3.03 [95% credible interval: 2.23-4.17]; odds ratio vs transfemoral Sapien =1.82 [1.21-2.70]; odds ratio vs transapical Sapien =2.08 [1.20-3.70]), and acute kidney injury (odds ratio vs CoreValve =2.08 [1.33-3.32]; odds ratio vs transapical Sapien =2.78 [2.21-99.80]), but lower rates of pacemaker implantation (odds ratio vs CoreValve =0.41 [0.28-0.59]), and moderate or severe aortic regurgitation (odds ratio vs CoreValve =0.06 [0.02-0.27]; odds ratio vs Sapien=0.17 [0.02-0.76]). Strokes were less frequent with CoreValve than with transfemoral Sapien (odds ratio =0.32 [0.13-0.73]) or transapical Sapien (odds ratio =0.33 [0.10-0.93]), whereas pacemaker implantation was more common with CoreValve (odds ratio vs surgery =2.46 [1.69-3.61]; odds ratio vs transfemoral Sapien =2.22 [1.27-3.85]).
CONCLUSIONS: Survival after transcatheter or surgical aortic valve replacement is similar, but there might be differences in the individual safety and effectiveness profile between the treatment strategies and the individual devices used in transcatheter aortic valve implantation.

Introduction

The burden of degenerative aortic stenosis continues to increase, and many patients with severe aortic stenosis are too sick or old to undergo surgery safely [1, 2]. Accordingly, minimally invasive means to treat patients with aortic stenosis have been recently developed, including self-expandable or balloon-expandable aortic prostheses enabling transcatheter aortic valve implantation (TAVI) [3,4,5].

The evidence base on these novel technologies has accrued recently, thanks to the completion of several important randomized clinical trials (RCT) with careful adjudication of clinically relevant endpoints [4, 5]. There is favorable data on the comparative risk-benefit balance of TAVI in comparison to medical therapy in patients with aortic stenosis but with prohibitive operative risk [4]. However, comparative data on TAVI vs surgical aortic valve replacement (SAVR) in those with heightened but not prohibitive risk are less clear cut given differing impacts on meaningful outcomes (e.g. stroke risk being higher with TAVI vs bleeding risk being higher with SAVR).

As systematic reviews exploiting network meta-analytic tools can provide more precise and robust as well as less optimistic effect estimates [6, 7], we aimed to conduct a comprehensive mixed treatment comparison of TAVI vs SAVR for severe aortic stenosis in patients at increased surgical risk.

Methods

The present review was performed in keeping with the Cochrane Collaboration, Quality of Reporting of Meta-analyses (QUOROM) and Preferred Items for Reporting of Systematic Reviews and Meta-analyses (PRISMA) [8,9,10]. All reviewing activities were independently performed by two reviewers, with divergences resolved after consensus.

MEDLINE/PubMed was searched for RCTs on TAVI in patients with severe aortic stenosis at high but not prohibitive surgical risk according to Biondi-Zoccai and colleagues on April 15, 2014 [11]. Additional queries were conducted in the Cochrane Library, Google Scholar, and Scopus. No language restriction was enforced.

Initially retrieved citations were screened first at the title and abstract level, and then appraised as full text. Studies were included if reporting was on a randomized trial of TAVI (using different TAVI devices or techniques) or TAVI vs SAVR in patients at high but not prohibitive surgical risk. Duplicate reports or studies focusing on patients at risk so high as to contraindicate SAVR were not included [12].

Several design, baseline, procedural, and outcome data were abstracted from shortlisted studies. Outcomes of interest at the longest available follow-up up to 12 months were: all causes of death; stroke; acute myocardial infarction; acute kidney injury; major bleeding; permanent pacemaker requirement; moderate or severe aortic regurgitation. The valve Academic Research Consortium definitions were used throughout [12]. In case of incomplete outcome data, effect estimates were computed imputing figures from Kaplan-Meier curves [13]. The internal validity of included studies was appraised according to the Cochrane Collaboration approach, separately evaluating the risk of selection, performance, attrition and adjudication bias [8]. Pairwise meta-analysis was first performed with a descriptive scope and to appraise statistical inconsistency and small study effects with RevMan (Cochrane Collaboration, Copenhagen, Denmark). Specifically, I-squared >50% was considered evidence of moderate or severe inconsistency. Network meta-analysis was conducted, after appraisal of evidence geometry according to van Valkenhoef et al. [14], within a Bayesian framework and with Markov chain Monte Carlo (MCMC) resampling using WinBUGS (University of Cambridge, Cambridge, UK). A fixed-effect method was used throughout given the prevalent star-shaped network, computing point estimates (95% credibility intervals) for odds ratios (OR) as well as the probability of each treatment being the best one (Pbest). Indeed, star-shaped evidence networks are inefficiently analyzed with random effects as the random effect term of the model may lead to lack of convergence of chains or very large credible intervals [6]. As the evidence network was prevalently star-shaped, we thus chose a priori to use a fixed effect model. In addition, the validity of this choice was confirmed also post hoc on deviance information criterion (DIC) estimates, always favoring the fixed effect model. Inferential estimates were based on 150,000 iterations after a burn-in phase of 50,000 iterations and graphical evidence of convergence of 3 independent MCMC chains. Consistency between direct and indirect estimates was appraised comparing consistency and inconsistency models [15].

Model fit was appraised by comparing values of DIC for fixed-effects and random-effects models. Results were also summarized graphically with boxplots (displaying 95% and 75% credible intervals, as well as median values). Specifically, boxplots were used as they are the default graphical representation provided by the WinBUGS package which also provides the quantitative statistical estimates, and thus any other graph (which should be obtained from a separate graphical package) would lack the real graphical precision provided by WinBUGS. Second, boxplots provide more precise and comprehensive information on the real posterior probability distribution which is used for statistical inference (as they provide median, 1st quartile, 3rd quartile, 2.5% quintile, and 97.5% quintile values).

Results

A total of 3,678 citations were initially retrieved (76 from the Cochrane Library, 3,130 from Google Scholar, 204 from MEDLINE/PubMed, and 268 from Scopus). Four RCTs were finally included, with a total of 1,805 patients (Table 1, Figure 1).

Design and patient features in the included studies.

Evidence network, showing the four alternative treatments and individual studies with different colors.

SAVR = surgical aortic valve replacement; TA = transapical; TF = transfemoral; STACCATO = Prospective, Randomized Trial of Transapical Transcatheter Aortic Valve Implantation vs Surgical Aortic Valve Replacement in Operable Elderly Patients with Aortic Stenosis; CHOICE = Randomized Comparison of Transcatheter Heart Valves in High Risk Patients With Severe Aortic Stenosis: Medtronic CoreValve Versus Edwards SAPIEN X.

Specifically, the PARTNER Cohort A trial randomized 699 subjects after defining their eligibility to transapical vs transfemoral TAVI with the Sapien (Edwards Lifesciences, Irvine, CA, USA) device to Sapien vs SAVR [4]. The Prospective, Randomized Trial of Transapical Transcatheter Aortic Valve Implantation vs Surgical Aortic Valve Replacement in Operable Elderly Patients with Aortic Stenosis (STACCATO) trial randomized patients to Sapien-TA vs SAVR, but was stopped prematurely after enrolment of 70 patients given an excess of adverse events in the Sapien-TA group [5]. The US CoreValve trial randomized 795 patients to CoreValve (Medtronic, Minneapolis, MN, US) vs SAVR [14]. In this trial the CoreValve could be implanted via the TF, transubclavian or direct aortic route, but most cases (323 out of 390 [82.8%]) were performed with a TF access. In this trial stroke was reported only with Kaplan-Meier estimates, and thus these were used to input effect estimates for the systematic review. Finally, the Randomized Comparison of Transcatheter Heart Valves in High Risk Patients With Severe Aortic Stenosis: Medtronic CoreValve Versus Edwards SAPIEN XT (CHOICE) trial compared Sapien vs CoreValve for TF access only [16]. Pairwise meta-analyses are reported only as hypothesis-generating data in the . Network meta-analysis comparing the effectiveness and safety of SAVR, CoreValve, Sapien-TF and Sapien-TA after a median of 8 months showed that the risk of death was similar irrespective of treatment (Table 2, Figure 2).

Network effect estimates for clinical outcomes at up to 1 year of follow-up.

Risk of death, reported as boxplots for odds ratios (with values <1 favoring the first treatment, and values >1 favoring the second treatment).

SAVR = surgical aortic valve replacement; TA = transapical; TF = transfemoral.

The risk of stroke was lower with CoreValve than with Sapien (Figure 3).

Comparative risk of stroke, reported as boxplots for odds ratios (with values <1 favoring the first treatment, and values >1 favoring the second treatment).

SAVR = surgical aortic valve replacement; TA = transapical; TF = transfemoral.

Acute myocardial infarction was similar irrespective of treatment (Figure 4).

Comparative risk of acute myocardial infarction, reported as boxplots for odds ratios (with values <1 favoring the first treatment, and values >1 favoring the second treatment).

SAVR = surgical aortic valve replacement; TA = transapical; TF = transfemoral.

Acute kidney injury was less likely with CoreValve than with SAVR (Figure 5).

Comparative risk of acute kidney injury, reported as boxplots for odds ratios (with values <1 favoring the first treatment, and values >1 favoring the second treatment). SAVR = surgical aortic valve replacement; TA = transapical; TF = transfemoral.

Major bleeding was less common with any TAVI device or access in comparison to SAVR, with CoreValve superior to Sapien-TF, and possibly also superior to Sapien-TA (Figure 6).

Comparative risk of major bleeding, reported as boxplots for odds ratios (with values <1 favoring the first treatment, and values >1 favoring the second treatment).

SAVR = surgical aortic valve replacement; TA = transapical; TF = transfemoral.

Permanent pacemaker implantation was required in more instances after CoreValve implantation than after SAVR or Sapien implantation (Figure 7).

Comparative risk of permanent pacemaker implantation, reported as boxplots for odds ratios (with values <1 favoring the first treatment, and values >1 favoring the second treatment).

SAVR = surgical aortic valve replacement; TA = transapical; TF = transfemoral.

The analysis for moderate or severe aortic regurgitation, which was limited by the fact that the PARTNER trial has not reported detailed data on both Sapien-TF and Sapien-TA, but only cumulative data, suggested that SAVR was better than both CoreValve and Sapien, but Sapien was nonetheless better than CoreValve (Figure 8).

Comparative risk of moderate or severe aortic regurgitation, reported as boxplots for odds ratios (with values <1 favoring the first treatment, and values >1 favoring the second treatment).

SAVR = surgical aortic valve replacement.

As an additional sensitivity analysis, we computed effect estimates for stroke excluding data imputed graphically (i.e. stroke data from the US CoreValve trial) [14], obtaining results similar in magnitude and direction to the overall analysis.

Discussion

To the best of our knowledge this is the first meta-analysis comparing different treatment for patients with severe aortic stenosis in terms of survival and peri-procedural complications. When compared with SAVR, TAVI appeared to be associated with a lower risk of periprocedural complications but an increased risk of moderate-to-severe aortic regurgitation, and a neutral effect on long-term survival.

The burden of degenerative aortic stenosis continues to increase due to increased life expectancy, yet many patients who would be candidates for SAVR are considered to be at too high a risk to safely undergo SAVR and therefore die of untreated severe aortic stenosis [17].

From a pathophysiologic point of view the hemodynamic overload generated by severe aortic stenosis imposes mechanical and neurohormonal challenges on cardiac walls, initially triggering compensatory left ventricular hypertrophy, but eventually activating complex biological responses evolving into maladaptive remodelling with relevant differences between genders [18]. As a result, this multifaceted mechanism culminates in tissue remodelling, leading to a progressive loss of regional and global cardiac function that after a long latency phase evolves rapidly to progress finally into a high rate of death. For decades surgical replacement of the aortic valve has been the sole suitable therapeutic option in patients with severe aortic stenosis even in the presence of significant co-morbidity. To date, large improvements in minimally invasive technology such as TAVI have been recently developed and have increased dramatically in the last decade. After the first-in-man implantations, TAVI has become an established procedure in patients with aortic stenosis and high surgical risk. Before adding to the complexity of data analysis by including second and third generation TAVI devices, we performed a comprehensive network meta-analysis to best characterize the pros and cons of the different strategies and possibly fill any remaining gaps concerning the effectiveness and safety of this procedure.

All the procedures, from CoreValve to Sapien-TF or TA and SAVR, showed the same protective effect on prognosis, with a significant improvement in survival compared with non-intervention. In-hospital operative complications (e.g. bleeding) were reduced by TAVI when compared to SAVR, apart from pacemaker implantation, while TAVI increased the risk of aortic regurgitation after the procedure. Rates of stroke did not differ among surgical and transcatheter interventions, while CoreValve proved superior to both TA and TF. Differently from other procedures like percutaneous coronary interventions vs coronary artery bypass graft, clinically relevant strokes did not differ between the more invasive surgical approach and the percutaneous one as previously reported in another pairwise systematic review, when including both observational and randomized controlled studies [19, 20].

This may be related to the different profile of patients, who are older with a larger burden of comorbidities and to the different interventions with a higher risk of embolization especially during the manipulation of aortic arch. Our findings of reduced rates of stroke for CoreValve may relate to the size of the delivery system, the need for rapid ventricular pacing with Sapien, or its self-expanding mechanism.

As expected, rates of complications like bleedings and acute kidney injury were reduced by TAVI. According to various definitions, they interest at least 25% of patients undergoing TAVI sharing similar risk factors such as reduced renal function and diabetes mellitus with an ominous impact on prognosis, both in-hospital and at follow up [21,22,23,24,25,26].

Similarly, as largely described, the very superficial location of the left bundle branch in the uppermost part of the leftward ventricular septum and its close proximity to the aortic valve complex probably represents the most important reason to explain atrio-ventricular blocks and consequently pacemaker implantation after TAVI [27]. In this setting, Corevalve, when compared to Sapien, has been related to a higher incidence of atrio-ventricular block because of the deeper implantation of the lower one third of the prosthesis stent frames, which are characterized by high radial forces for secure anchoring of the stent [28]. Although it increased costs, this complication was not shown to have a negative impact on prognosis. Further analyses will be required to compare intensive care unit stay, total hospital stay, need for prolonged intubation or hemodialysis.

Finally, aortic regurgitation was shown more frequent in patients with TAVI than with aortic surgery, without differences between TAVI approaches. Our analysis confirmed the similar results of TAVI (both TF and TA) for aortic regurgitation, although all performed inferiorly to surgery. Despite being challenging to correctly evaluate, degree of aortic regurgitation has been shown to negatively affect prognosis, and therefore the reduction in in-hospital complications with TAVI vs SAVR may be offset by late events related to the presence of moderate-to-severe aortic regurgitation.

Moreover, while this finding may be questionable for high risk patients without alternatives as it may show a different impact on patients with a low or intermediate risk consequently with a longer life expectancy due to the lower burden of comorbidities [29]. Future research should be focused on this complication, both to correctly address patients without risk factors for aortic regurgitation and to the development of new valves with a lower risk of regurgitation.

Limitations of this work include those typical of all systematic reviews, pairwise meta-analyses, and mixed treatment comparisons [6]. In addition, the prevalently star-shaped evidence base implies that most inference is based on indirect comparisons, which require of course future confirmation or disproval in head-to-head randomized trials.

Finally, the assumption underlying a sizable portion of our analyses that patients undergoing TF and TA TAVI are reasonably similar in risk, notwithstanding peripheral artery disease burden, may not be shared by all clinicians and researchers.

Conclusion

The long-term outlook after TAVI is promising in comparison to SAVR, but there might be differences in the individual safety and effectiveness profile between the competing treatment strategies and the individual devices used in transcatheter aortic valve implantation.