Evidence comparing transcatheter aortic valve implantation (TAVI) with surgical aortic valve replacement (SAVR) for treating aortic stenosis (AS) is based primarily on seven randomized clinical trials (RCTs). Since the first RCT was conducted in 2007, patient inclusion criteria have gradually expanded to include those with lower surgical risk profiles. These studies have demonstrated not only the non-inferiority but, in some cases, the superiority of TAVI compared to conventional treatment, shaping the guidelines for managing severe symptomatic AS in the 2021 ESC/EACTS Guidelines on Valvular Heart Disease.

Randomization helps control various biases, such as confounding factors in assigning treatment options. However, certain limitations persist, including selection criteria that do not accurately represent the target population, deviations from assigned treatment (DAT) post-randomization (crossover, intention-to-treat vs. as-treated analyses, etc.), protocol deviations in relation to additional procedures besides AS treatment, and patient loss during follow-up. These unresolved issues introduce biases that may affect the validity of RCTs. Quantifying these biases in TAVI vs. SAVR RCTs had not been adequately addressed until now.

The aim of this study was to determine whether randomization effectively protects TAVI vs. SAVR RCTs from other biases that may arise at different stages of study design. To this end, a systematic literature review was conducted from January 1, 2007, to June 6, 2022, across MEDLINE, Embase, and the Cochrane Central Register of Controlled Trials. Two independent researchers extracted data following PRISMA guidelines. A random-effects meta-analysis was conducted, including eight eligible RCTs with a total of 8,849 participants, who were randomized to TAVI (n = 4,458) or SAVR (n = 4,391) with a maximum follow-up of five years. An analysis of the impact of various biases on the aggregate outcomes of the selected studies was conducted:

The pooled DAT rate was 4.2%, favoring TAVI (pooled relative risk [RR] vs. SAVR = 0.16; p < 0.001), indicating a higher rate of continued treatment for patients initially assigned to TAVI than vice versa.

The pooled loss-to-follow-up rate was 4.8%. Meta-regression showed a significant association between the rate of follow-up loss and the length of follow-up, although there was an imbalance in favor of TAVI (RR = 0.39; p < 0.001), with a lower rate of patient attrition compared to the SAVR group.

The pooled average rate of patients receiving additional procedures besides AS treatment was 10.4%: 4.6% in the TAVI group and 16.5% in the SAVR group (RR = 0.27; p < 0.001). The most frequent associated procedure was myocardial revascularization, which was also less common in the TAVI group (RR = 0.40; p < 0.001).

The authors conclude that a significant proportion of DAT, follow-up loss, and additional concomitant procedures skew results in favor of TAVI, posing a systematic, selective imbalance that may affect the validity of comparative TAVI vs. SAVR RCTs.

COMMENTARY:

This review is based on eight RCTs: PARTNER 1A, 2A, and 3; the US CoreValve Pivotal High Risk Trial; SURTAVI; the Evolut Low Risk Trial; NOTION; and UK TAVI. It originates from suspicions that early RCTs comparing TAVI with SAVR outcomes in high-risk surgical patients exhibited potential methodological errors. In subsequent trials conducted with intermediate- and low-risk surgical patients, these biases appeared to persist. Barili, the lead author of this meta-analysis and a strong proponent of TAVI, demonstrated with this meta-analysis something that had only previously been suggested: a systematic, selective imbalance, with a lower rate of DAT, patient loss to follow-up, and additional procedures performed in the TAVI group.

The first issue involved DAT. After randomization, most patients assigned to TAVI continued with the assigned treatment, while a higher proportion of patients in the SAVR group deviated from their assigned treatment. The cause of this phenomenon is unknown, but one hypothesis could be how this type of clinical trial is “presented” to potential participants. Patient expectations generated before randomization toward the experimental treatment (in this case, TAVI) may have been frustrated upon learning they were assigned to conventional (surgical) treatment, potentially leading to voluntary study withdrawal in greater numbers due to the apparent disadvantage of conventional surgery versus a less invasive option.

The second relevant finding concerns the significant imbalance in patients undergoing additional associated procedures between the TAVI and SAVR groups (4.6% vs. 16.5%, respectively). Although multiple associated procedures have been enumerated, the most common was myocardial revascularization, affecting 4.5% of patients in the TAVI group compared to 10.8% in the SAVR group. The 2021 ESC/EACTS Guidelines on Valvular Heart Disease state that this higher incidence of myocardial revascularization in the surgical group should ultimately be seen as a protective factor. This assumption is not entirely accurate, as an additional procedure increases perioperative risk (as demonstrated by various surgical risk scales) and, while it may reduce the risk of future clinical events, this has not yet been proven. The most plausible explanation for the higher proportion of concomitant procedures in the SAVR group could be the significant flexibility that surgeons have to modify the surgical technique when encountering other concurrent conditions, as well as the learned tendency to address all surgical issues once the patient is on the operating table.

Finally, and no less important, is the finding of significantly lower patient loss to follow-up in the TAVI arm. Follow-up losses reached as high as 20% at five years in the SAVR group and consistently increased over time (1.4% at one year vs. 8.9% at five years). These dropout biases, particularly when follow-up loss is high, can compromise study validity, especially if clinical events contributing to the primary endpoint (such as death and stroke) are infrequent. The cause of this phenomenon remains unclear. Perhaps again, the fact that patients are receiving conventional surgery rather than the experimental treatment could be a reason, especially if their recovery is uneventful, leading them to skip follow-up visits. The same ESC/EACTS 2021 Guidelines on Valvular Heart Disease adopted an age-based intervention approach, recommending TAVI for patients over 75 years old (class IA). Prosthesis durability is a highly relevant topic, and many RCTs are extending follow-up to assess long-term outcomes. However, due to limited survival or lack of follow-up continuity, most trials will lack the statistical power to detect ten-year differences, which would prevent clarifying this issue. Therefore, regardless of their results, these studies should not be used to further lower the age threshold for TAVI.

We all agree that RCTs provide the best evidence available, but they are not infallible sources of all answers. Even the highest-quality research studies have limitations, and the key lies precisely in those “details.” This meta-analysis serves as a reminder for clinicians that, even with RCT-based information, caution is needed in interpretation, and sound clinical judgment and common sense should prevail.

REFERENCE:

Barili F, Brophy JM, Ronco D, Myers PO, Uva MS, Almeida RMS, et al.; Risk of Bias in Randomized Clinical Trials Comparing Transcatheter and Surgical Aortic Valve Replacement: A Systematic Review and Meta-analysis. JAMA Netw Open. 2023 Jan 3;6(1). doi: 10.1001/jamanetworkopen.2022.49321.