Key considerations in a meta-analysis of physiology-guided percutaneous vs surgical revascularization in multivessel coronary artery disease

Abstract

We read the meta-analysis by Kataveni et al with great interest; however, we have identified two major methodological errors. First, the population from the Fractional Flow Reserve Versus Angiography for Multivessel Evaluation 3 (FAME 3) trial appears to have been double-counted through inclusion of both Fearon et al (primary FAME 3 trial) and Dillen et al (a FAME 3 bifurcation substudy), likely violating the independence assumptions required for meta-analytic pooling. Second, the study by Di Gioia et al (2020) has been misclassified as a randomized controlled trial, whereas it is a registry-based, non-randomized study. Because the authors’ conclusions rely on the inclusion of “three randomized controlled trials” with minimal heterogeneity, these errors may overstate the certainty of the evidence and introduce bias into the pooled estimates. We request that the authors address these concerns, perform a re-analysis as required, and revise the manuscript accordingly. Ultimately, these issues reflect the general paucity of high-quality randomized data in this field.

Key Words: Fractional flow reserve; Percutaneous coronary intervention; Coronary artery bypass grafting; Randomized controlled trial; Meta-analysis

Core Tip: The recent meta-analysis by Kataveni et al was timely in a field with scarce randomized data. However, the authors appear to have included two manuscripts derived from the same Fractional Flow Reserve Versus Angiography for Multivessel Evaluation 3 population and misclassified the registry-based study by Di Gioia et al as a randomized trial. These issues warrant further clarification and underscore the need for future editorial and peer-review oversight in this area.

TO THE EDITOR

We read an article by Kataveni et al[1] with great interest, given its focus on a very important clinical issue. On initial review, the analytical framework appears to broadly adhere to PRISMA guidelines, and the methodology seems rigorous. However, upon closer inspection, the results and conclusions appear substantially compromised by two major concerns: Inclusion of a duplicate study population and misclassification of study design.

First, there is a very high likelihood of double-counting of the Fractional Flow Reserve Versus Angiography for Multivessel Evaluation 3 (FAME 3) population. Table 1 lists Fearon et al[2] (FAME 3) as a randomized trial with 757 patients vs 743 patients, and also includes Dillen et al[3] 2025, a FAME 3 substudy of bifurcation lesions, with the same sample sizes (757 vs 743). Despite this, Kataveni et al[1] state that the baseline characteristics of patients in the study by Dillen et al[3] were “NA,” although these data are available in the original publication[3]. Had these baseline data been shown, the population overlap would have been evident. Treating a substudy nested within a parent randomized controlled trial (RCT) as an independent dataset violates the independence assumptions of meta-analyses and likely results in double-weighing of FAME 3, thereby biasing pooled estimates and artificially narrowing confidence intervals.

Second, we believe that the study by Di Gioia et al[4] (2020) has been misclassified. The manuscript repeatedly states that all three included studies are RCTs, and in Table 1, Di Gioia et al[4] is labelled as an RCT. However, the original publication in Circulation: Cardiovascular Interventions describes a registry-based, nonrandomized comparison rather than an RCT. Indeed, the authors explicitly acknowledge in their limitations that the patients were not randomized and that the analysis shared the intrinsic limitations of observational data. This misclassification contradicts the stated inclusion criteria of RCT-only studies and undermines both the risk-of-bias assessment and the Grading of Recommendations Assessment, Development, and Evaluation (GRADE) certainty judgments.

These two issues are critical to the validity of the meta-analysis. As the synthesis is predicated on “three independent RCTs” with minimal heterogeneity, any duplicate inclusion would violate the independence assumption fundamental to meta-analytic methods. Such duplication would result in double-counting of the same patient population, artificially inflating the precision of pooled effect estimates and inappropriately narrowing confidence intervals. In parallel, misclassification of study design will compromise the validity of risk-of-bias assessments, spuriously inflate the certainty of evidence ratings under GRADE criteria, and potentially introduce confounding bias into the pooled estimates.

The distinction between randomized and observational evidence is not merely semantic but represents qualitatively different levels of evidence for causal inference. Accordingly, the manuscript’s current claims of moderate-to-high certainty for several outcomes may be based on weaker evidence than reported and should therefore be carefully revisited.

I respectfully request that the authors and editors: (1) Clarify the relationship between the study by Dillen et al[3] and the FAME 3 trial and, if overlap is confirmed, retain only the main trial to avoid double-counting; (2) Correct the study design classification of the study by Di Gioia et al[4] through direct verification against the original publication rather than relying on database categorizations or other indirect methods. If reclassification is required, please consider either excluding this study from the RCT-only meta-analysis or reframing the synthesis to appropriately incorporate mixed study designs, with appropriate sensitivity analyses and cautious interpretation; and (3) If any corrections are warranted, rerun all analyses to re-evaluate heterogeneity, risk of bias, and GRADE assessments, which are meaningful only in correctly classified and non-overlapping meta-analyses. After reanalysis, the authors may need to temper their conclusions appropriately.

It is possible that after appropriate de-duplication and correct study classification, only two studies would remain, with FAME 3 as the sole RCT, substantially weakening the rationale for a conventional trial-level meta-analysis. In such a scenario, the manuscript could acknowledge this limitation and convey to the clinical community that currently available randomized evidence is insufficient to inform guideline recommendations or routine clinical decision-making regarding the choice between coronary artery bypass grafting and fractional flow reserve-guided percutaneous coronary intervention beyond the FAME 3 trial. This could also serve as a call for well-designed future prospective RCTs to address this important clinical question.

Given the magnitude of these concerns, we urge the Journal to consider issuing an editorial note of concern, to reassess the peer-review process that allowed these potential flaws to pass undetected despite the standard publication review checklist, and to acknowledge the critical role of post-publication scrutiny. If reanalysis confirms that the principal conclusions do not hold, formal corrections should be considered to protect the integrity of the scientific record and the Journal’s reputation.