Stenting vs Medical Therapy for Stable Coronary Artery Disease: A Minefield for Meta-analyses?
We read with interest the meta-analysis by Stergiopoulos and Brown1 comparing stenting vs medical therapy for stable coronary artery disease (CAD). However, we believe that the review methodology is flawed, and thus the implications are potentially wrong.
First, the study search was limited to MEDLINE/PubMed, which has very limited coverage and comprehensiveness. The incomplete scope of the search strategy greatly increases the likelihood that publication bias and selective reporting undermine the meta-analysis.2 At a minimum, The Cochrane Collaboration CENTRAL Registry of Randomized Trials, EMBASE, and key international conference proceedings (eg, America College of Cardiology and American Heart Association scientific sessions) should have been systematically queried. Thus, we recommend that the authors perform additional and more thorough searches for potentially eligible studies.
Second, Stergiopoulos and Brown1 pooled data from different studies despite striking clinical and statistical heterogeneity for 2 key end points (ie, unplanned revascularization and persistent angina). Whereas this approach can be envisioned with a hypothesis-generating scope, it should never have hypothesis-testing implications.3 Indeed, analyses for unplanned revascularization and persistent angina are clearly flawed by this analytical bias. The only meaningful approach to address this situation would be to perform meta-regression analyses, which are not reported by the authors.4 Thus, we recommend that Stergiopoulos and Brown perform such additional analyses.
Third, throughout the article, but particularly in the abstract and in the title, a clear and explicit disclosure is lacking: all included studies were extremely selective in their inclusion and exclusion criteria. For instance, the Occluded Artery Trial (OAT) excluded from randomization patients with symptomatic heart failure, shock, angiographically significant left main or 3-vessel CAD, angina at rest, and severe ischemia on stress testing. Similarly, the Clinical Outcomes Utilizing Revascularization and Aggressive Drug Evaluation (COURAGE) trial excluded from randomization subjects with persistent class IV angina, significant left main CAD, ejection fraction less than 30%, or markedly positive stress test result. Thus, the external validity of the meta-analysis by Stergiopoulos and Brown,1 even accepting at face value their quantitative estimates, is very limited.
In conclusion, while synthesizing clinical evidence on coronary stenting in stable CAD is a laudable goal, this should not come at the price of scientific validity and clinical accuracy. Accordingly, the meta-analysis by Stergiopoulos and Brown,1 while addressing an important clinical topic, cannot provide clear guidance for researchers and practitioners owing to its methodological limitations.