Author + information
- †Division of Cardiology, San Francisco General Hospital, San Francisco, California
- ‡Department of Medicine, and the Department of Epidemiology and Biostatistics, University of California, San Francisco, San Francisco, California
- ↵∗Reprint requests and correspondence:
Dr. Dhruv S. Kazi, San Francisco General Hospital, Division of Cardiology, 1001 Potrero Avenue, Room 5G1, San Francisco, California 94110.
Nothing can stop an idea whose time has come.
—Victor Hugo (1)
Percutaneous replacement or repair of heart valves have been some of the most exciting innovations in the field of interventional cardiology in the past decade. Since the early animal experiments by Henning-Rud Andersen in 1992 (2) and the first-in-human deployment by Alain Cribier in 2002 (3), transcatheter aortic valve replacement (TAVR) has moved from being an investigational procedure reserved for patients without surgical options to an effective and, when performed through transfemoral access, a cost-effective alternative in patients at high risk for surgical complications (4). Newer generations of the devices have lower profile delivery systems, incorporate adaptable seals to prevent paravalvular regurgitation, and may be repositionable to facilitate accurate deployment (5). Accumulating evidence of efficacy, durability, and safety has led to calls for expanding the indications for TAVR, and several ongoing trials are examining outcomes in intermediate-risk patients (6). With increasing demand from an aging population as well as a rapidly improving supply of centers offering the procedure, the number of TAVRs performed is projected to rise exponentially over the coming decade.
The dissemination of a device or surgical procedure that was previously only offered only in research settings involves 3 phases. The primary concern in the first phase is ensuring that practitioners in the community can reproduce the results obtained by pioneering operators at high-volume academic centers. This can be challenging because of the learning curve associated with complex procedures and a strong correlation between higher procedure volumes and better clinical outcomes. The second phase of dissemination relates to the increasing real-world experience with the device, which yields additional insights into its effectiveness and durability and may lead to the recognition of rare or delayed complications that were not observed in the initial trials (which typically have fewer patients and shorter follow-up periods). This eventually results in improved patient selection, refinements in the technology or procedure, or changes in the protocol for post-intervention management. The third and final stage of technology dissemination is maturation and widespread adoption, when the focus shifts to decreasing practice variability, and controlling costs.
Initial post-approval data from the United States suggest that TAVR has sailed smoothly past the first phase. Data from nearly 250 centers participating in the controlled rollout of TAVR have shown that the community experience mirrors the findings of randomized clinical trials (7). This initial success is likely the result of intense site training, close supervision of early adopters by the manufacturer, and a team-based approach to care delivery. The work by Généreux et al. (8) in this issue of the Journal is a striking reminder that TAVR is now well into in the second phase of technology dissemination, in which careful analysis of the accumulating evidence will yield valuable insights that should be harnessed to optimize long-term outcomes.
The investigators examine the association between major late bleeding complications (MLBCs) after TAVR, defined as major bleeding occurring between 30 days and 1 year after the procedure, and all-cause mortality. The investigators collated a dataset of 2,401 patients who were treated with TAVR and survived to 30 days, including patients from PARTNER (Placement of Aortic Transcatheter Valves) randomized trial cohorts A (high risk but operable, n = 236) and B (inoperable, n = 164), and 2 continued-access registries (n = 1,911). During a median follow-up period of 1 year, 142 patients (5.9%) experienced MLBCs, on average approximately 4 months after the initial procedure. The most frequent bleeding sources were gastrointestinal (41%) and neurological (16%), including falls resulting in traumatic intracranial bleeds (6%). After adjusting for demographic, clinical, and procedural characteristics, 4 factors predicted an increased risk for MLBCs during follow-up: lower baseline hemoglobin, increased left ventricular mass, the presence of atrial fibrillation (AF), and a moderate to severe paravalvular leak. Compared with those who did not experience MLBCs during follow-up, patients who did were at increased risk for all-cause death, death from cardiac causes, major stroke, rehospitalization, and the development of renal failure requiring dialysis. After multivariate adjustment for baseline, clinical, and procedural characteristics, MLBCs were associated with nearly 4-fold risk for all-cause mortality (adjusted hazard ratio: 3.83; 95% confidence interval: 2.62 to 5.61; p < 0.001). The findings were consistent within subgroup analyses by sex, age, and surgical risk.
When faced with novel experimental results, a scientific mind asks 3 questions: 1) Are these results real? 2) Is this association causal? 3) What are the implications of these findings?
Are the Findings Real?
The interpretation of results from nonrandomized data necessitates caution, primarily because of concerns about residual confounding. Contemporary statistical techniques perform well at adjusting observed differences between the 2 groups, but unobserved differences may bias the results. For instance, if doctors systematically offer a surgical procedure to their least frail patients, and in as much as frailty is not perfectly captured by the variables in the dataset, the procedure may appear to be more effective than it actually is (9). Despite the meticulous nature of this analysis, some concerns about unobserved confounders remain. For instance, in what appears to be a major oversight in the design of the original studies, data about anticoagulant use (a predictor of bleeding complications) were not prospectively collected. The authors use AF as a surrogate for anticoagulation due to the high anticoagulant use in older patients with hypertension and valvular disease. However, AF is an imperfect surrogate for anticoagulation, and the analysis fails to account for patients taking anticoagulant agents for other indications, such as thromboembolic disease. Nevertheless, the strength of the association between MLBCs and mortality (with a near quadrupling of the hazard of death even after multivariate adjustment) would suggest that the association is real.
Is the Association Causal?
Another concern with observational analyses is that correlation does not imply causation. Does bleeding increase the risk for mortality, or are high-risk patients more likely to bleed? Evidence from other settings suggests that the association between bleeding and long-term mortality may be causal, mediated by the physiologic effect of the bleed itself and subsequent interventions, such as transfusion of blood products (10) or cessation of antiplatelet agents or anticoagulant agents. In the case of observational studies, the plausibility of the findings may also make a case for a causal association. Of the 4 factors that predicted post-TAVR MLBCs, lower baseline hemoglobin and the presence of AF or atrial flutter (as a marker of anticoagulant use) are intuitive. Also, increased left ventricular mass may be a marker of end-organ damage from long-standing hypertension. The association between moderate to severe paravalvular leak and MLBCs is novel and interesting: the investigators hypothesize that this may represent an acquired thrombophilia from turbulence-induced cleavage of proaggregation proteins. If future investigations confirm the association between paravalvular regurgitation and MLBCs, the underlying mechanism and strategies to address it deserve further investigation. Definitive evidence of causation, however, will only be available if future randomized trials to reduce MLBCs demonstrate improved survival.
What are the Implications?
Finally, if these findings are reproduced in other cohorts, the observed association between MLBCs and long-term mortality after TAVR has important clinical implications. Since the introduction of TAVR, an intense focus on technological and procedural improvements has substantially reduced peripro cedural bleeding. These findings argue that bleed reduction strategies must also be extended to the first year post-procedure. In the future, nested clinical trials or well-designed observational analyses must examine the impact of improved patient selection and the effectiveness of pre- or post-procedure bleed reduction interventions on long-term outcomes after TAVR. Following on the impressive reductions in procedural complications, the focus must shift to improving long-term outcomes. Small refinements in the TAVR protocol could yield valuable clinical and economic dividends by reducing delayed complications.
There is palpable excitement among patients and their cardiologists for safe and effective transcatheter valve repair and replacement. This technology represents a disruptive innovation that has changed the way we deliver care for advanced valvular disease, and its explosive growth argues that this is indeed an idea whose time has come. But as Généreux et al. (8) remind us, even as we embrace the technology, careful attention to accumulating evidence offers a unique opportunity to continue to optimize long-term outcomes after TAVR.
↵∗ Editorials published in the Journal of the American College of Cardiology reflect the views of the authors and do not necessarily represent the views of JACC or the American College of Cardiology.
Dr. Kazi has reported that he has no relationships relevant to the contents of this paper to disclose.
- 2014 American College of Cardiology Foundation
- Hugo V.
- Cribier A.,
- Eltchaninoff H.,
- Bash A.,
- et al.
- Reynolds M.R.,
- Magnuson E.A.,
- Lei Y.,
- et al.
- Sorajja P.,
- Pedersen W.
- Généreux P.,
- Cohen D.J.,
- Mack M.,
- et al.
- Farzaneh-Far R.,
- Bolger A.F.
- Nuis R.J.,
- Rodes-Cabau J.,
- Sinning J.M.,
- et al.