Pursuit of Myocardial Ischemia for Therapeutic Decision-Making in Patients With Diabetes and Stable Ischemic Heart Disease: Reconciling Randomized Controlled Trials and Observational Studies

The choice between medical therapy and revascularization for management of stable ischemic heart disease (SIHD) remains among the most polarizing debates in cardiovascular (CV) medicine. This debate is fueled by two opposing pathophysiological models of SIHD. One is the epicardial stenosis–centric model that drives noninvasive stress tests to detect ischemia, which, in turn, justifies coronary angiography followed by revascularization, if feasible. The other is a conservative or biologic model that centers on biochemical, inflammatory, and atherothrombotic factors (1). In each model, measurement of risk factors and both lifestyle interventions and aggressive pharmacologic secondary prevention are targeted to reduce incident CV events and residual risk. In the epicardial stenosis model revascularization is touted to additionally reduce risk. Observational studies have suggested that early revascularization of patients with moderate to severe ischemia on single-photon emission computed tomography–myocardial perfusion imaging (SPECT-MPI) with a threshold of >10% ischemia decreases short-term cardiac death and long-term all-cause mortality (2–7) (Table 1). However, a series of randomized trials over a period of more than 2 decades (post-2000) have failed to verify the additional therapeutic benefit of revascularization for SIHD over and beyond guideline-directed medical therapy (GDMT) (8–15) (Table 2).

Whether an invasive strategy with revascularization reduced outcomes among patients with stable coronary artery disease (CAD) with inducible ischemia using SPECT-MPI was first explored in the Clinical Outcomes Utilizing Revascularization and Aggressive Drug Evaluation (COURAGE) trial, and the results showed no clinical benefit compared with GDMT among 2,287 patients with significant obstructive CAD. A nuclear substudy in 314 patients showed a significantly greater ischemia reduction following revascularization in those with moderate to severe ischemia, but this did not translate into significant reduction in death or myocardial infarction (16).

The International Study of Comparative Health Effectiveness With Medical and Invasive Approaches (ISCHEMIA) trial was designed to overcome a key limitation of previous trials by ensuring enrollment of high-risk patients. A central guiding principle was to include only patients with >10% ischemia burden defined by SPECT-MPI (11). The rationale was based on the results from a single-center nonrandomized study conducted in the 1990s showing that revascularization was associated with lower risk of cardiac death only in those with >10% ischemia (2). Although the point estimate of the survival curves crossed at the 10% threshold, the 95% CI overlapped substantially, and a significant survival benefit was not observed until ischemia exceeded 20%. As in the COURAGE trial (17), the extent of myocardial ischemia did not predict outcome benefit with revascularization in the ISCHEMIA trial (11).

In this issue of Diabetes Care, Kuronuma et al. (7) retrospectively explore the long-term association of survival benefit from early revascularization with the magnitude of ischemia in patients with or without diabetes in 41,982 patients who underwent SPECT-MPI from 1998 to 2017 at a single center. About a fifth of the cohort had diabetes, and <8% underwent revascularization (74% percutaneous coronary intervention [PCI] and 26% coronary artery bypass graft surgery [CABG]). Less than 9% had moderate to severe ischemia, whereas 12% had it in an earlier study (2). Propensity score matching followed by a multivariable Cox proportional hazards model was utilized to adjust for known confounding factors. The key findings are the following: 1) patients with diabetes had greater amounts of moderate to severe ischemia, higher rates of early revascularization, and higher annual mortality than those without diabetes; 2) association between early revascularization and mortality showed benefit among patients with moderate (with diabetes) or severe (with and without diabetes) ischemia; 3) the threshold of ischemic burden associated with survival benefit from early revascularization was lower in patients with diabetes (summed difference score [SDS] >8.6%) than in patients without diabetes (SDS >12.1%). Exclusion of subjects with left ventricle (LV) dysfunction (ejection fraction [EF] <35%) shifted this threshold to 16.2% (without diabetes) and 10.2% (with diabetes). Another notable finding is that in an analysis that was not propensity matched, no significant difference in mortality was observed between CABG or PCI. This finding is in contrast to the mortality advantage seen with CABG over PCI in patients with diabetes reported in the Comparison of Two Treatments for Multivessel Coronary Artery Disease in Individuals With Diabetes (FREEDOM) trial (18) and which underlies the 2021 guideline recommendation for coronary revascularization favoring CABG over PCI in patients with diabetes (19).

Despite strong pathophysiological plausibility, why have the randomized controlled trials failed to reproduce the results of observational studies that show patients with moderate-severe ischemia on SPECT-MPI or PET-MPI have an improved outcome with early coronary revascularization? How might one reconcile the positive findings in the current study with the null results in recent trials, including the ISCHEMIA trial? The authors have provided a detailed accounting of potential reasons for differences. Two among them merit further discussion.

First, the authors suggest that the high likelihood of “referral bias” might have precluded inclusion of high-risk patients in the ISCHEMIA trial, leading to a dilution of a survival advantage with revascularization. However, clinical outcomes in the ISCHEMIA-Chronic Kidney Disease (ISCHEMIA-CKD) trial, a parallel trial, were three- to fourfold higher than those for the ISCHEMIA trial, yet no treatment advantage with revascularization was found (12). No treatment heterogeneity was present in those with or without diabetes (a higher-risk cohort) in both ISCHEMIA-CKD and ISCHEMIA trials. In the recently reported Study of Efficacy and Safety of Percutaneous Coronary Intervention to Improve Survival in Heart Failure (REVIVED-BCIS2) trial (15), which included a much-higher-risk cohort, including patients with multivessel CAD and LVEF ≤35% and 14% having left main stenosis (who were excluded from the ISCHEMIA trials), an incremental treatment advantage for PCI was not evident over GDMT (Table 2). No treatment by diabetes interaction was found in this trial (15).

Second, this study used propensity matching to avoid confounding, but this may not have fully accounted for all differences in clinical factors and treatment comparisons. The authors assert, without supporting evidence, that these unmeasured confounding factors are likely to be distributed evenly between patients with and without diabetes. Therefore, their finding of different ischemic thresholds for survival benefit from revascularization between patients with and without diabetes would not be explained by these unmeasured confounders. However, the issue is not one of different thresholds among the two cohorts but whether the mortality advantage associated with early revascularization in either cohort is a valid finding or is subject to residual confounding.

A fundamental concern with observational data is that unmeasured confounding might be mistaken for a treatment effect. For this reason, a sensitivity analysis might be conducted to assess how strong a relationship would have to be between an unmeasured confounder and the treatment assignment, as well as between the unmeasured confounder and the outcome, to explain away an observed treatment effect. One recently proposed approach is the E value (20). For example, one can calculate the magnitude of confounding that would be necessary to fully explain the estimated hazard ratio of 0.69 (95% CI 0.49, 0.98) in patients with diabetes and moderate ischemia (SDS >10–14%). An E value of 1.91 tells us that a confounder, or set of confounders, would have to be associated with a 91% increase in the risk of mortality to explain the observed hazard ratio. One could also calculate the E value for the lower CI, which calculates to 1.13. This value is not high enough to claim immunity against unmeasured confounding. Table 1 provides calculated E values from 6 major observational studies in the last 2 decades, including the current study. The E values for the lower confidence limit (the rate-limiting estimate) range from 1.13 to 1.94, which are not implausible enough to dismiss the threat of residual confounding.

Some variables that are known in clinical practice to have a major impact on the choice for revascularization (e.g., extensive coronary disease, the presence of chronic total coronary occlusions, patient frailty, and comorbid conditions such as liver or chronic lung disease, quality of life, and socioeconomic status) were not available for this analysis. It is quite possible patients with these adverse prognostic factors have been preferentially recommended against revascularization, leading to a contrast in survival in favor of revascularization. Although the sensitivity analysis suggests that either a single powerful variable or several confounding variables acting in concert could account for the between-group difference in the rate of survival, such confounders could also increase the difference. Thus, the magnitude or the direction of the bias cannot be estimated with certainty.

What is the take-home message for practicing clinicians, and how should we integrate the findings of this study and randomized controlled trials into our clinical decision-making?

The 2021 chest pain guidelines endorse stress testing, including SPECT-MPI for diagnosis of myocardial ischemia, risk stratification, and guiding therapeutic decision-making (21). Based on the totality of the evidence, patients with established obstructive CAD who have stable chest pain should be started on an initial strategy of therapeutic lifestyle intervention aimed at risk factor control plus GDMT regardless of the presence or absence of diabetes. Revascularization can be reserved for a later time in the minority of patients whose symptoms or quality of life is either refractory or unacceptable on medical therapy with little risk that an unfavorable event will intervene. Early revascularization should be avoided in patients with minimal and mild ischemia. For patients with moderate to severe ischemia, the observational studies favor early revascularization with a lower ischemic threshold for benefit in patients with diabetes. However, the randomized controlled trials show that for patients presenting with daily, weekly, or monthly angina, early revascularization provides a prompt and durable improvement in symptoms compared with conservative management, but it does not reduce CV risk even in patients with diabetes. Importantly, the prevalence of moderate to severe ischemia has been declining appreciably over the last 3 decades to <5%, likely driven by stabilization of vulnerable coronary plaque with aggressive risk factor modification and GDMT. Routine screening for CAD, including stress testing, is not recommended in asymptomatic patients with diabetes as it does not improve outcomes as long as risk factors are controlled (22).

Our pursuit of a prognostically important ischemia prompted by promising results in observational studies has failed to yield dividends in randomized controlled trials. This disconnect is likely due to potential bias introduced by unmeasured confounding in the former rather than referral bias in the latter. A key lesson, therefore, is that only high-quality, unbiased evidence ideally generated from large, prospective, and rigorously conducted randomized controlled trials should inform and guide clinical practice.