Author + information
- Peter E. Carson, MD∗ ()
- ↵∗Address for correspondence:
Dr. Peter E. Carson, Department of Cardiology, Washington VA Medical Center, 151D, 50 Irving Street Northwest, Washington, DC 20422-0001.
In the society that we live in, many endeavors consist of contests that culminate in a “winner.” These include a myriad of sporting events, be them baseball or basketball, or even a grand collection of events such as the Olympics. Countries engage in such activities as elections with a candidate being declared a “winner” who then attempts to govern. There are a series of rules that contestants, players, or candidates follow to be declared the “winner.” In medicine, perhaps we can consider our version to be clinical trials in which an intervention may be compared with another intervention, or with no intervention at all (placebo), and a set of rules exist for this contest and there are criteria to determine what is an effective intervention, ergo a “winner.” In all of these endeavors, there is notoriety for a winner—in medicine such a winner may become a standard of medical therapy with benefits for patients and eventually for corporate sponsors. Therapies that are not favorable are often forgotten. One may ask, “What if the criteria for winning were changed?” For example, some in the United States may ask whether, in the recent presidential election, the winner might have been required to win the popular vote or maybe for that matter both the popular and the electoral votes.
In this issue of JACC: Heart Failure, Shen et al. (1) have contributed a well-written and well-analyzed addition to the medical literature that examines what the authors describe as the “evolution of trial endpoints in heart failure.” In this case, the effort involves a look at expanding the comparison of outcomes within the BEST (Beta-blocker Evaluation of Survival Trial) from the primary outcome of all-cause mortality to the more commonly used current composite outcome of cardiovascular death or heart failure hospitalization, but then further expanding the outcome to include emergency room visits for heart failure and finally recurrent heart failure hospitalizations. The authors demonstrate that the assessment of these different outcomes first alters the BEST trial from a neutral trial to one that is favorable for the effects of the study drug bucindolol, and also that the expanding composite outcomes decreases sample size and time to follow-up for the statistical outcomes. Hence bucindolol became a “winner.” The resulting clinical trial would have been smaller—no mean issue, as the size and cost of current studies is seen by some as unsustainable.
It is true that the BEST trial was a creature of its time in its primary outcome. When the BEST trial was undertaken, the government agencies that provided much of the sponsorship were wedded to all-cause mortality as a primary outcome (although they were not alone in this—the COPERNICUS [Carvedilol Prospective Randomized Cumulative Survival] trial had this primary outcome also and it was a co-primary outcome in the MERIT-HF [Metoprolol CR/XL Randomized Intervention Trial in Congestive Heart Failure]) (Online Appendix). Current heart failure clinical trials commonly use the more contemporary endpoint of cardiovascular death or heart failure hospitalization. However, there is complexity with this endpoint as well, which has been characterized in numerous forums (2). For example, cardiovascular deaths would seem a safe categorization for a cardiovascular drug; however, a common approach to fatal events without information has been to categorize these as “unknown” mode of death, which are then statistically analyzed as cardiovascular based on assumptions from previous clinical trials that the majority of deaths were in this classification. However, the routine acceptance of unknown deaths as cardiovascular may be in doubt. A draft document (3) that seems to be favored by the U.S. Food and Drug Administration and industry states “the appropriate classification and analysis of undetermined causes of death depends on the population, the intervention under investigation, and the disease process”—a reasonable thought given that current clinical trials enroll older populations than previously, and studies involving subjects with heart failure and preserved ejection fraction report a lower proportion of cardiovascular events. This document also suggests rigorous definition of heart failure hospitalizations requiring specific documentation of symptoms, signs, or treatment in an approach similar to the Framingham Heart Study definition of heart failure (Online Appendix). The adoption of this approach promises to be an interesting challenge in clinical trials involving areas of the world where documentation may be less than optimal. In the case of the BEST trial, which was designed as a mortality trial, documentation of additional clinical events such as hospitalizations and emergency room visits would only have been available as adverse event reports, would have made this new level of granularity problematic. It should be noted that the current manuscript used investigator classification of events, and a previous analysis using an adjudication committee reported fewer heart failure hospitalizations (4). This may in part explain why, although the BEST trial enrolled heart failure patients with substantial risk, heart failure hospitalization rates of 49.5 (placebo) and 40.8 (bucindolol) per 100 patient-years are higher than other reports in heart failure with reduced ejection fraction (heart failure with reduced ejection fraction studies) (5) (Online Appendix). Interestingly, the effect size with an adjudication committee was similar to that for investigator-reported events despite fewer events, suggesting that, at least in this database, the use of an adjudication committee may also provide a means to conduct a smaller, shorter study.
The expanded endpoints in the Shen et al. (1) paper include emergency room visits for heart failure and recurrent heart failure hospitalizations. There is a growing literature concerning the consideration of such endpoints with post hoc analyses evaluating relative risk of patients with such events (compared with those patients with no events) and their inclusion into primary outcomes. For example, a recent publication from the PARADIGM-HF (Prospective Comparison of ARNI with ARB Global Outcomes in Heart Failure With Preserved Ejection Fraction) study (6) reported that, although isolated (without a subsequent heart failure hospitalization) emergency room heart failure visits were uncommon, they were associated with a poor prognosis and imparted some addition to the first-event analysis, thereby influencing trial size and duration. The Shen et al. (1) paper echoes these findings, albeit from a different area and 1 geographic area—largely the United States. This data, taken in concert with the previous MADIT-CRT (Multicenter Automatic Defibrillator Implantation Trial With Cardiac Resynchronization Therapy) trial (Online Appendix) and PARADIGM-HF study publications, lend impetus to prospective trials including well-defined emergency room visits—with criteria and documentation similar to heart failure hospitalizations—as part of either primary or secondary analyses.
The analysis of recurrent heart failure events has been increasingly used in post hoc assessments in clinical trials and is very well done in the Shen et al. paper—discussing and utilizing 4 different methods. Ironically, the concept of “win ratio” is part of recurrent event analyses—though it is different from “winning” as discussed previously. There is, however, one difficulty, in that the BEST trial database contains a curve: there is not a constant hazard and that removes the ability to use the statistical methods (e.g., negative binomial, joint frailty model) used in previous publications—the CORONA (Controlled Rosuvastatin Multinational Trial in Heart Failure), EMPHASIS (Eplerenone in Mild Patients Hospitalization and Survival Study in Heart Failure), CHARM-Preserved (Candesartan in Heart failure Assessment of Reduction in Mortality and morbidity) trials (Online Appendix). The authors persevered, eventually using the Lin, Wei, Ying, and Yang model as the principal method of analysis. The statistical analyses demonstrate that there is not a constant event rate and a constant treatment effect over time. This very interesting finding may be linked a rapid up-titration—faster than those of other beta-blocker trials in heart failure. However, the effect may also be related to the sympatholytic effect of bucindolol, which has been reported as prominent in African Americans with a defined pharmacogenetic profile (Online Appendix).
The larger questions associated with an analysis such as this are the following:
1. Does the analysis of expanded composite utilizing recurrent heart failure hospitalizations identify important treatment effects that would not be apparent by a first-event analysis? The current manuscript discusses previous studies—in 2 studies, the MADIT-CRT and EMPHASIS studies, the use of a composite including recurrent hospitalization analysis confirmed the initial first-event analysis. In 2 neutral trials, the CORONA and CHARM-Preserved trials, which were neutral for their primary endpoints but were modestly favorable with the addition of recurrent event analyses, only the primary outcome was influenced only in the CHARM-Preserved trial. In the Shen et al. (1) paper, had the trial been done using the contemporary first-event composite, or using recurrent events, or adding emergency room heart failure visits, the results would have been favorable.
2. Does the analysis of expanded composites lower the barrier for demonstrated efficacy? Leaving aside the notion that there have been a number of efforts to reinterpret the BEST trial—due to the subgroup findings that black and class IV patients particularly had poor outcomes with bucindolol—how significant would these redone results have been? For better or for worse, the results would still not really be comparable to the 3 beta-blocker trials in which the tested agents reduced all-cause mortality and a myriad of cause-specific outcomes, For example, even the less selective composite outcome of all-cause mortality or heart failure hospitalization was reduced by 31% in the COPERNICUS and the MERIT-HF trials but by 20% according the current data in the BEST trial. Bucindolol might reasonably have been considered by regulators, and health care professionals in a similar fashion as nebivolol based on the SENIORS (Study of Effects of Nebivolol Intervention on Outcomes and Rehospitalization in Seniors With Heart Failure) trial data (Online Appendix) as good, but not good enough.
3. Do expanded primary outcomes and recurrent event analyses improve the efficiency of a clinical trial? There is no doubt that there is concern over whether it is sustainable to undertake clinical trials in a mature field such as heart failure where new therapies (particularly in heart failure with reduced ejection fraction) are added on to successful therapies due to the costs, patient numbers, and duration that such clinical trials must have. There is also concern about analyses in heart failure with preserved ejection fraction where lower event rates and a heterogeneous population represent additional challenges. The current manuscript does clearly demonstrate advantages with an expanded primary outcome—whether by adding emergency room visits or recurrent heart failure hospitalizations—on the potential size and duration of the trial, assuming that the intervention tested has a consistent treatment effect on components of an expanded composite.
Additionally, the assertion that the expanded endpoints better reflect the burden of heart failure is a sound one. The best test of these approaches of course would be a prospective clinical trial. To this end, the PARAGON-HF (Efficacy and Safety of LCZ696 Compared to Valsartan, on Morbidity and Mortality in Heart Failure Patients With Preserved Ejection Fraction; NCT01920711) study, utilizing recurrent heart failure hospitalizations as part of its composite primary outcome along with cardiovascular deaths, will be helpful. Further, a new trial, the EMPEROR (EMPagliflozin outcomE tRial in patients with chrOnic heaRt failure), which is just beginning, will include a measure of heart failure emergency room visits as part of its primary outcome. These ongoing studies will help us to better understand the benefits of expanded endpoints. Now to return to better defining a winning election candidate…
For an expanded references section, please see the online version of this paper.
↵∗ Editorials published in JACC: Heart Failure reflect the views of the authors and do not necessarily represent the views of JACC: Heart Failure or the American College of Cardiology.
Dr. Carson has reported that he has no relationships relevant to the contents of this paper to disclose.
- Shen L.,
- Jhund P.S.,
- Mogensen U.M.,
- et al.
- ↵Hicks KA, Hung HM, Mahaffey KW, et al., on behalf of the Standardized Data Collection for Cardiovascular Trials Initiative Draft Definitions for CDISC. Standardized Definitions for Cardiovascular and Stroke Endpoint Events in Clinical Trials. Available at: https://www.cdisc.org/system/files/all/standard/Draft%20Definitions%20for%20CDISC%20August%2020%2C%202014.pdf. Accessed August 20, 2014.
- Okumura N.,
- Jhund P.S.,
- Gong J.,
- et al.