Observational studies are needed to complete the body of evidence for left main coronary artery disease revascularization strategies

This commentary refers to ‘PCI or CABG for left main coronary artery disease: the SWEDEHEART registry’, by J. Persson et al., https://doi.org/10.1093/eurheartj/ehad369 and the discussion piece ‘Revascularization for left main coronary artery disease’, by F. Alfonso et al., https://doi.org/10.1093/eurheartj/ehad644.

We appreciate the critique provided by Alfonso et al.¹ regarding our recent publication.

There is no doubt that the results from well-conducted randomized clinical trial (RCT) merit large weight when evaluating the body of evidence for left main coronary artery disease (LM CAD) revascularization strategies. However, a caveat with RCTs is that most patients with LM CAD are not included in RCTs. For example, five Swedish sites participated in the NOBLE trial and 77 patients with LM CAD were included between 2008 and 2015. During the same period, there were 7791 patients with LM CAD in Sweden who were revascularized with coronary artery bypass grafting (CABG) or PCI.² Consequently, 0.1% of patients with LM CAD in Sweden during that period were included, and 99.9% were not. Furthermore, ∼4600 patients have been included in four RCTs investigating PCI vs. CABG for LM CAD, none of which have been designed to investigate differences in mortality.² The RCT cohorts evaluating revascularization strategies for LM CAD are selected subsets, and the external validity is consequently restricted.³ It is therefore important to investigate revascularization strategies for LM CAD in all-comer populations not included in RCTs.

Firstly, we disagree with the notion that our interpretation of the results based on the instrumental variable (IV) analysis is flawed. IV analysis, which has been recognized with a Nobel Prize in 2021,⁴ serves as a foundational method for drawing causal inferences from observational datasets.⁵ IVs function in a manner akin to ‘natural experiment’, serving as proxies that, while not directly affecting outcomes, influence treatment decisions, thereby mimicking the randomization seen in controlled trials.⁶ It is noteworthy that our mortality risk estimate for PCI vs. CABG closely aligns with EXCEL’s hazard ratio (HR) 1.34, with a significant overlap in the 95% confidence intervals (CIs).

Secondly, the IV mortality estimate for PCI vs. CABG in our study, HR 1.5 (95% CI 1.1–2.0), would be nullified by an unmeasured confounder with the following characteristics: an estimated prevalence of 40% in the PCI population, 20% in the CABG population, and an estimated relationship of HR 6.0 with mortality. The sensitivity analysis results underscore our findings’ robustness, as such a confounder is unlikely in this population. Thirdly, data from SWEDEHEART, which compares PCI and CABG for LM CAD treatment in patients between 2015 and 2022, are planned for an upcoming publication. Preliminary insights indicate that the primary findings remain the same even with the escalated adoption of new-generation DES.

Finally, while we respect the findings of the cited meta-analysis, it is crucial to emphasize the power of a study. A meta-analysis, if underpowered, may fail to discern a genuine mortality difference between the two revascularization techniques. To put things in perspective, whereas the aforementioned meta-analysis reported 550 deaths, our SWEDEHEART study documented 1794.

As physicians and researchers, we are obliged to conduct large nationwide observational studies on revascularization strategies for LM CAD due to high validity for large groups of patients, not to compete with RCTs but to complete the body of evidence.

Declarations

Disclosure of Interest

J.P. has received unrestricted grants from Abbott Inc., unrelated to the present work. T.I. and E.O. have no conflicts of interest.

References