Impact of revascularisation on outcomes in chronic coronary syndromes: a new meta-analysis with the same old biases?

This editorial refers to ‘Cardiac mortality in patients randomised to elective coronary revascularisation plus medical therapy or medical therapy alone: a systematic review and meta-analysis', by E. P. Navarese et al., https://doi.org/10.1093/eurheartj/ehab246. Placing a finger on the scale of scientific objectivity creates an illusory benefit of revascularisation in the chronic coronary syndromes. Placing a finger on the scale of scientific objectivity creates an illusory benefit of revascularisation in the chronic coronary syndromes. “A wise man proportions his beliefs to the evidence” David Hume—Scottish Philosopher (1711-1776) The development of coronary angiography by Sones in 1958 provided, for the first time, the ability to visualize obstructive epicardial coronary artery disease (CAD) in vivo. Shortly thereafter, natural history studies demonstrated that the extent of obstructive CAD correlated with an increased incidence of myocardial infarction (MI) and mortality which, as a result of conflating association with causation, formed the basis for the enduring belief that the epicardial stenosis is the proximate cause of death and MI in patients with stable CAD,1 now referred to as chronic coronary syndromes (CCS). When coronary artery bypass graft (CABG) surgery and percutaneous coronary intervention (PCI) were developed in the 2 decades that followed, it became virtually axiomatic that the ability to bypass or dilate flow-limiting stenoses would prevent MI and reduce mortality. Shortly thereafter, investigators began comparing CABG and PCI to standard medical therapy of the time in randomized controlled trials (RCTs) to understand and quantify the potential benefit to patients. Early medical therapy consisted of nitrates and beta-blockers, with or without aspirin, and was directed at symptom relief and exercise tolerance, but had no impact on preventing atherothrombotic events, including MI and death. As newer and more robust medical therapies evolved over time and were shown individually in placebo-controlled trials to reduce cardiovascular events, a new era of disease-modifying medical therapy consisting of pharmacologic secondary prevention, risk factor control, and lifestyle modifications emerged which interrupted the progression of atherothrombotic disease and improved clinical outcomes. In the aggregate, this approach was first termed “optimal medical therapy” (OMT) in 20072 and is now the benchmark against which revascularisation or any new therapeutic advance must be compared to understand its role in current-day CCS management. Because many RCTs employed composite endpoints with small sample sizes and low event rates, meta-analysis was often employed to compare the effect of an initial strategy of revascularisation compared to medical therapy on the individual components of composite outcomes. A fundamental tenet of meta-analysis is appropriate selection of studies with minimal statistical and clinical heterogeneity, though the latter can be present in the absence of the former3. Thus, to have contemporary relevance, it is essential that meta-analyses be limited to studies that are not widely disparate temporally and therapeutically to assess the comparative effectiveness of different treatment approaches. A 1994 meta-analysis by Yusuf of 7 RCTs compared CABG, which bypasses both obstructive lesions and non-obstructive vulnerable atheromas thus mitigating the impact of plaque rupture and thrombosis, to non-disease modifying medical therapy in 2649 patients, and found CABG was superior in reducing mortality.4 One of the first meta-analyses to evaluate PCI by Katritsis and Ioannidis in 2005 included 11 studies enrolling 2,950 CCS patients5 that randomized patients to balloon angioplasty or anti-anginal therapies alone where the two treatments were compared in apposition and showed no benefit of PCI on all-cause mortality or nonfatal MI. Thereafter, the Clinical Outcomes Utilizing Revascularization and Aggressive druG Evaluation (COURAGE) trial was the first RCT to employ OMT combined with PCI using bare metal stents vs. OMT alone as the comparator and likewise showed no difference in death or nonfatal MI.2 Subsequently, to provide a more contemporary assessment of the benefit of initial PCI plus OMT vs. OMT alone, Stergiopoulos and Brown published a meta-analysis in 2012 that limited RCTs to those in which at least 50% of PCIs were performed with stents.6 In 8 studies randomizing 7229 patients, they too found no reduction in death or nonfatal MI by addition of PCI to OMT. A more recent meta-analysis by Bangalore and colleagues compared initial revascularisation with either PCI (including drug-eluting stents) or CABG plus medical therapy to medical therapy alone.7 Among 14 pooled studies comprising 14877 randomized patients followed for a weighted mean of 4.5 years, they found no benefit of revascularisation on all-cause mortality or nonfatal MI. Overall, the consistency of this body of RCT and meta-analysis evidence over time convincingly refutes the long-held belief that flow-limiting epicardial stenoses are the principal driver of adverse downstream events such as MI and death in CCS and that revascularisation reduces these events. The latest meta-analysis by Navarese et al published in the current issue of the European Heart Journal confirms once again that initial revascularisation by PCI or CABG when added to medical therapy fails to reduce all-cause mortality or nonfatal MI compared to initial medical therapy.8 Remarkably, these findings were demonstrated despite the clinical heterogeneity introduced by inclusion of outdated studies from the 1970’s that compared CABG to non-disease-modifying medical treatment in patients with obstructive CAD (including left main CAD) and studies of patients not revascularized after an acute MI with persistent ischemia or angina, two settings of historical significance but no longer relevant to current practice. Comingling old, disparate studies that include post-MI and left main CAD patients with dedicated CCS trials is an indiscriminate meta-analytic sleight of hand that, akin to combining apples and oranges, has been used previously to tip the balance in favor of revascularization (see Graphical abstract).9,10 It appears the authors chose to downplay the neutral effects of revascularisation on all-cause mortality and nonfatal MI. Instead, they “cherry-picked” the primary and secondary outcomes of cardiac death and spontaneous MI that were reduced by revascularisation. By selecting outcomes that have not been primary endpoints in other meta-analyses on this topic,4–7,10 they recycle the well-worn narrative that routine revascularisation of CCS patients improves outcomes, yet seemingly ignore the contrary findings from contemporary landmark RCTs of almost 11,000 patients in the OMT era from COURAGE2 to International Study of Comparative Health Effectiveness with Medical and Invasive Approaches (ISCHEMIA).11 The authors provide two justifications for choosing cardiac death as the primary endpoint even though this endpoint was not reported in nearly one-third of the RCTs included in their meta-analysis. First, they cite the Academic Research Consortium (ARC)-2 consensus document12 to support the use of cardiac death as their primary endpoint. However, the ARC-2 document actually considers all-cause mortality the most unbiased and patient-centric method to report deaths and explicitly states that the basis for coronary device evaluation should be overall cardiovascular outcomes from the patient perspective, including death, MI, stroke, and repeat revascularisation procedures. Second, the authors cite an editorial by Seto13 to claim that long-term all-cause mortality tends toward the null and is therefore an irrelevant endpoint. However, the Seto editorial specifically refers to a single study with 10-year follow-up in older patients (average age: 68 years) while the current meta-analysis includes studies with an average 5.7-year follow-up and an average age of 61 years. It is therefore unlikely that competing risks of non-cardiovascular death obscured any effect of revascularisation on all-cause mortality in a significantly younger patient population with a relatively short follow-up. Importantly, our objection to the use of cardiac mortality as a primary endpoint notwithstanding, if only the trials relevant to current practice published since the introduction of OMT in 2007 are meta-analyzed using data from the Navarese paper, there is no longer a reduction in cardiac death from revascularisation (Figure1). Not only is clinical heterogeneity reduced by limiting the analysis to relevant contemporary trials performed in the OMT era, but statistical heterogeneity is decreased from 21% in the Navarese study to 0%. Rate ratios and 95% confidence intervals for cardiac mortality with revascularisation plus medical therapy vs medical therapy alone in randomized trials published since the introduction of optimal medical therapy (OMT). Size of data markers is proportional to weight in the meta-analysis. PCI, percutaneous coronary intervention. CABG, coronary artery bypass graft surgery. Heterogeneity: I2 = 0%, P = 0.60. Rate ratios and 95% confidence intervals for cardiac mortality with revascularisation plus medical therapy vs medical therapy alone in randomized trials published since the introduction of optimal medical therapy (OMT). Size of data markers is proportional to weight in the meta-analysis. PCI, percutaneous coronary intervention. CABG, coronary artery bypass graft surgery. Heterogeneity: I2 = 0%, P = 0.60. The authors likewise claim that procedural MI was “sparsely and heterogeneously reported” but otherwise provide no justification for their choice of “spontaneous” MI as opposed to all nonfatal MIs, including both spontaneous and procedural MIs, as a secondary endpoint. The recent Bangalore meta-analysis documented a 2.5-fold increased risk of procedural MI following revascularization.7 The usual justification for excluding these events is that procedural (Type 4 A and Type 5) MIs are mostly benign “enzyme leaks” and not associated with “prognostically-important” downstream mortality. However, this remains a largely unsettled issue in the cardiology literature with mounting evidence from recent studies suggesting that procedural MI following PCI is both common and associated with a significantly increased risk of subsequent ischaemic cardiac events, including death.14,15 In summary, trial-level meta-analysis of agnostically selected studies with minimal clinical and statistical heterogeneity that assess relevant, patient-centric endpoints can provide accurate, objective, and clinically meaningful estimates of the effect size of one therapy vs. another. However, meta-analysis can also be influenced by decisions regarding study design that may be analogous to putting a finger on the scale of scientific objectivity. In their provocative 1991 viewpoint entitled “Can meta-analysis be trusted?” Thompson and Pocock warned that use of “statistical tricks” as we find in the current study can erode trust in meta-analysis.3 We hope readers of this paper will be both circumspect and wise in interpreting the totality of evidence that should inform our beliefs and decision-making in the guideline-endorsed management of CCS patients—namely, an “OMT first” approach to reduce symptoms and events and one which considers revascularisation only for improvement of quality of life in those whose symptoms are either unresponsive or refractory to OMT. Conflict of interest: none declared. The opinions expressed in this article are not necessarily those of the Editors of the European Heart Journal or of the European Society of Cardiology.