Author + information
- ↵⁎Reprint requests and correspondence:
Dr. A. Laurie W. Shroyer, Stony Brook University, Surgery, School of Medicine T19-020A, 100 Nicholls Road, Stony Brook, New York 11794-8191
- coronary artery bypass grafting
- surgical ventricular reconstruction
- randomized clinical trial
- mortality prediction
- ischemic cardiomyopathy
Clinical trial comparisons of treatments can provide important evidence to improve health policy, patient care management, and clinician and patient informed decisions. Assuming a clinical trial has been well designed and successfully implemented (i.e., internal validity has been established), its findings may provide important treatment-related information to guide future care for the study-relevant populations.
A key challenge in reviewing any clinical trial findings is to evaluate to what degree the published results have applicability for general clinical practice. In light of the STICH (Surgical Treatment for Ischemic Heart Failure) trial publications, therefore, in this editorial we examine the generalizability of the STICH trial's surgical ventricular reconstruction (SVR) hypothesis conclusions more broadly (1).
Two key factors related to the STICH trial that warrant careful consideration include an assessment of patient selection and treatment appropriateness (2). To evaluate the STICH trial's patient selection, it would have been optimal to have access to an international registry of all patients with ischemic cardiomyopathy to comprehensively compare the STICH versus non-STICH patient risk factors, processes of care, and outcomes. A fundamental concern is that STICH patients volunteering as trial participants may not be representative of the broader patient population. Specifically, the STICH trial did not include SVR candidates who did not require coronary artery bypass grafting (CABG) procedures, as well as the patients who definitely needed both SVR and CABG procedures. Pending population registry access, comparisons between participating and nonparticipating centers might be performed, as well as comparisons between enrolled patients and screened (but not enrolled patients) at participating sites. As a template to evaluate study generalizability, these types of direct comparisons were performed in the Veterans Administration–based PSOCS (Processes, Structures, and Outcomes of Care in Cardiac Surgery) multicenter study (3).
Heart Failure Impact
Heart failure is a major cause of death and disability worldwide, which imposes a substantial economic burden (4). The prevalence rate for heart failure has been estimated at 2% in the U.S. (5) and Europe (6). Prevalence rates up to 13% have been reported for the elderly (7).
Heart failure represents a complex clinical syndrome with progression modulated by the presence of coexisting conditions such as coronary disease (4). Over the past 2 decades, improved treatments for acute myocardial infarction have improved survival. Survivors often have reduced left ventricular (LV) function, with immediate ramifications for functional capacity, as well as future progressive remodeling changes occurring in LV chamber size and/or LV geometry. Because preliminary data from a case-control study indicated promising improvements in LV function after SVR (8), the STICH trial was designed to evaluate whether CABG either alone or in combination with SVR might improve outcomes in patients with heart failure.
The STICH Trial
The STICH trial had a complex study design whereby distinct patient subpopulations were enrolled to support the 2 different hypotheses proposed. In the current publications, patients with ischemic heart failure were enrolled with coronary artery disease amenable to CABG, with ejection fractions <35%, as well as dominant anterior LV dysfunction amenable to SVR. Exclusion criteria included a recent myocardial infarction, need for aortic valve replacement, planned percutaneous coronary intervention, or coexisting noncardiac disease with a projected life expectancy <3 years.
During enrollment, the STICH trial relied heavily on the clinical judgment of the study physicians to ensure that the potential benefits and risks to patients were equivalent. Because varying physician thresholds for equipoise in treatment assignments were permitted, there was wide diversity in the characteristics of the 2,136 STICH patients enrolled across the 127 participating sites in 26 different countries (9).
Despite the pragmatic approach taken to ensure clinician equipoise, the study's enrollment process was reportedly “challenging.” Given the extreme variations reported for enrollment rates across centers and countries, a healthy skepticism arises related to evaluating the generalizability of this study's more complicated patient selection process using the tiered stratum randomization design. To address this concern, in this issue of the Journal, Jones et al. (9) have designed an innovative approach to classify each individual patient's risk at randomization (RAR) using a multivariate model. On the basis of RAR categorical classifications (low, moderate, and high RAR groupings), the RAR match rates were compared between high-volume versus low-volume enrolling centers and countries. As anticipated, the high-volume enrolling sites generally enrolled lower-risk patients. Similarly, the low-volume enrolling sites generally enrolled higher-risk patients. Additionally, a country's specific population characteristics dominated the RAR level for enrolled patients. However, substantial overlap of RAR categories was demonstrated. For high-volume versus low-volume enrolling sites, 85% of enrolled patients had similar predicted RAR classifications. Similarly, 76% of patients across high- versus low-volume enrolling countries shared similar RAR risk profiles.
In this issue of the Journal, Zembala et al. (10) extracted STICH-eligible patient records (the closest match of patients with ischemic heart failure meeting STICH inclusion and exclusion criteria) from the Society of Thoracic Surgeons (STS) National Cardiac Database, identifying 104,135 STICH-eligible STS patients, of whom 1,092 (1%) underwent CABG with SVR. Applying an RAR multivariate regression algorithm, the randomized STICH trial patients were compared with the STS CABG plus SVR patients who were STICH eligible (for the subgroup not enrolled in the STICH trial). Again, there was a substantial degree of overlap between these risk classifications.
Albeit indirect, this novel RAR approach developed by the STICH team has provided a partial assurance that for the majority of patients enrolled, there appears to be general concordance of their providers' assessments of SVR treatment eligibility, both within the STICH trial and comparing STICH patients with STS SVR plus CABG patients. Differences for unmeasured risk factors may remain, however, as the percentage of akinesia or dyskinesia of the anterior wall was reported for only a subset of patients. If your surgical practice's profile for patients with ischemic heart failure (i.e., applying the STICH inclusion and exclusion criteria) is not substantially different, and your surgical revascularization treatments are similarly performed (i.e., per STICH protocol requirements), the STICH trial's findings now warrant your careful consideration.
Given the diversity of “real-world” clinical practice, larger multicenter clinical trials with pragmatic patient selection criteria and/or flexible treatment protocols are more likely to be representative. In contrast, complicated study designs may be less likely to have external validity and more likely to have important challenges with clinical interpretability. Trial findings are inherently affected by high-volume enrolling physicians and/or sites, which may have a differential propensity to select patients or may have different outcomes from other lower-volume enrolling providers. If large variations in center enrollment rates exist, trial findings reported should verify whether a “treatment-by-center” effect exists and, where appropriate, adjust for these effects accordingly.
Ideally, the variation in STICH patient selection would have been examined by comparing the enrolled patients with the screened (but not enrolled) patients directly (11). Future clinical trials should be proactively designed and adequately funded to capture an expanded trial registry to compare screened versus enrolled patient risk factors, procedures, and outcomes. Additionally, a randomization scheme should be used to ensure treatment arm balance for participating surgeons and within participating centers. Comparisons to national and/or international study-eligible population registries should be proactively planned as part of the original study protocol to evaluate the representativeness of the trial patients.
In summary, generalizing these recent STICH trial findings to support changing your own clinical practice should be considered only after careful review of both the patients selected and the treatments administered.
↵⁎ Editorials published in the Journal of the American College of Cardiologyreflect the views of the authors and do not necessarily represent the views of JACCor the American College of Cardiology.
- American College of Cardiology Foundation
- Flather M.,
- Delahunty N.,
- Collinson J.
- Lloyd-Jones D.,
- Adams R.J.,
- Brown T.M.,
- et al.
- Cowie M.R.,
- Mosterd A.,
- Wood D.A.,
- et al.
- McMurray J.J.,
- Stewart S.