Hierarchical Endpoints in Prior Heart Failure Trials and the HEART-FID Trial

Abstract

Background:

Clinical trials in heart failure (HF) traditionally use time-to-event analyses focusing on death and hospitalization for HF. These time-to-first event analyses may have more limited abilities to assess the probability of benefiting from a therapy, especially if that benefit manifests as improved functional status rather than reduced risk of death or HF hospitalization. Hierarchical endpoints including clinical outcomes and patient status measures allow for ranked evaluation of outcomes in one metric assessing whether patients randomized to intervention or control are more likely to derive an overall benefit while also allowing more patients to contribute to the primary outcome.

Methods:

We review the rationale for using hierarchical endpoints in HF trials, provide examples of HF trials that used this type of endpoint, and discuss its use in HEART-FID, the largest HF trial to date implementing a hierarchical endpoint analysis for the primary outcome.

Results:

Using a hierarchical endpoint as the primary outcome allows for inclusion of different types of outcomes in one ranked endpoint, making it possible to more holistically assess the potential utility of a new therapy on patient well-being and outcomes.

Conclusions:

Hierarchical endpoints assess the potential utility of a new therapy on patient well-being and outcome more holistically than time-to-first-event analysis. Trials that would not have been feasible due to decreasing rates of death and hospitalization in the HF population, can use hierarchical endpoints to successfully power studies to identify promising HF therapies. HEART-FID used hierarchical endpoints to better determine the role of intravenous FCM in patients with HF.

Trial Registration:

ClinicalTrials.gov (NCT03037931).

INTRODUCTION

Clinical trials in heart failure (HF) traditionally use time-to-first event analyses focusing on death and hospitalization for HF.^{1, 2} However, these analyses have clinical and statistical limitations. Clinically, time-to-first event analyses may be less sensitive to changes in clinical status, emphasize timing over severity of event (e.g., an early hospitalization for HF would be worse than a later death), and do not capture recurrent events.³ Statistically, time-to-first event endpoints in HF trials require large numbers of participants for analyses to be adequately powered since improvements in therapy have contributed to decreasing event rates.

Hierarchical endpoints and summary measures, such as win ratios, rank events in order of importance, such that death outranks HF hospitalization, and HF hospitalization outranks change in clinical status.⁴ This enables both clinical events and important changes in patient status to be considered in the analysis. It also allows comparatively more patients to contribute to the endpoint, as many patients may have changes in their clinical status that do not result in death or hospitalization. The increased number of patients contributing to the endpoint simultaneously facilitates patient contribution to trial outcomes and may offer advantages for statistical power, allowing for smaller and/or shorter duration trials and for the most severe events to drive the analysis. This latter point is highlighted when the endpoint includes death.

Retrospective analyses comparing a post hoc evaluation of hierarchical endpoints with traditional hazard ratios generated in time-to-first event analyses found that results were consistent: one analysis evaluated 16 large cardiovascular outcomes trials with composite endpoints and found post hoc calculations of a win ratio based on the ranked individual components of each composite endpoint generated similar results to hazard ratios reported in each primary analysis.⁵ These findings increased confidence that hierarchical endpoints are a valid alternative or complement to more traditional time-to-first event analyses, even in trials not designed explicitly for hierarchical endpoint analysis.

Hierarchical analyses are increasingly used in HF trials.^6–8 These endpoints may be particularly appropriate for interventions with a large anticipated benefit in functional status, such as intravenous ferric carboxymaltose (FCM) use in patients with HF and iron deficiency. Acceptance by regulatory agencies of quality of life metrics into the approval for HF therapies may also increase the use hierarchical endpoints. We review the rationale for using hierarchical endpoints in a HF trial, describe examples of HF trials that used hierarchical endpoints, and discuss the use of a hierarchical endpoint in the Randomized Placebo-controlled Trial of FCM as Treatment for Heart Failure with Iron Deficiency (HEART-FID) trial and its impact on interpretation of trial results.⁹

METHODS

Data from analyses are obtainable from the sponsor company on reasonable request. The primary endpoint for HEART-FID was a hierarchical endpoint including death, total number of HF hospitalizations (both at 12 months), and change in 6-minute walk test (6MWT) distance at 6 months, though the trial was powered to assess superiority for its more traditional secondary endpoint, a composite of time to cardiovascular death or first hospitalization for HF.⁹ The HEART-FID trial, designed to provide definitive insight into the efficacy and safety of intravenous FCM in patients with iron deficiency and HF with reduced ejection fraction (HFrEF), leveraged a hierarchical endpoint to evaluate the impact of intravenous FCM on clinical events and changes in functional capacity. The study was approved by the institutional review boards at participating sites and all patients provided written informed consent.

DISCUSSION

Clinical Outcomes in Heart Failure: What Matters?

Patients, clinicians, investigators, and regulators are motivated to identify and use therapies that positively impact a patient’s clinical status (functional capacity, overall well-being, or symptom burden) and/or reduce their risk for major clinical events like death or hospitalization, including recurrent hospitalizations. The U.S. Food and Drug Administration has issued endpoint guidance for HF drug development and prior expert panels have summarized considerations regarding the use of functional or symptomatic endpoints.^{10, 11} The occurrence of major clinical events and changes in clinical status are distinct but important issues for patients, providers, and the healthcare community. Incorporation of these endpoints requires the application of standardized definitions to ensure consistent evaluation across trials.

Major clinical events occuring during study follow-up are typically identified through study-specific visits and medical record review and then undergo central, blinded adjudication. Though regional variations in hospitalization exist, death and hospitalization for HF are objective measures of outcome. As an event, all-cause death is unambiguous, and with appropriate pre-specified definitions for cardiovascular death and HF hospitalization, these events can be consistently adjudicated.¹² In contrast, patient clinical status may represent a more ephemeral concept: what is the best way to assess a patient’s overall well-being in a standardized way that transcends social, demographic, and other potential regional variations?

Fortunately, standardized and validated tools to assess clinical status for patients with HF are available. These tools generally rely on either quality of life or symptom burden or use metrics like exercise tolerance to evaluate functional status. In clinical practice and clinical research, New York Heart Association (NYHA) functional class is often used. This clinician-assigned metric differentiates patients by the degree to which they have symptoms with physical exertion, ranging from no symptoms or restrictions (NYHA class I) to symptoms even at rest (NYHA class IV).¹³ However, the NYHA classification system is subjective, relying on clinician assessment, and insensitive to small changes in clinical status. Another tool frequently used in HF is the Kansas City Cardiomyopathy Questionnaire (KCCQ). The KCCQ is a self-administered 23-item questionnaire that explores patient-reported health status and the impact of that status on physical abilities, symptom burden, self-care, social factors, and quality of life.¹⁴ Use of the KCCQ score is particularly appealing in HF trials, given its granularity across multiple facets of clinical status, and the fact that it was explicitly designed to evaluate status of patients with HF. KCCQ (and other patient-reported measures) can also be incorporated into hierarchical endpoints with positioning at different levels in the hierarchy depending on the specific trial objectives/hypotheses.

Another option is to use exercise capacity as a marker for functional status. Exercise capacity can be assessed by 6MWT distance or other metric, generating an easily quantifiable granular result that can be tracked over time. Though the 6MWT has been shown to correlate with the KCCQ, it may also represent a more objective metric of patient status that is more easily standardized across different regions and patient populations, since it measures an objective data point (meters walked) rather than relying on patient-reported symptom burden.¹⁵

Clinical Reasons for Hierarchical Endpoints in Clinical Trials

Ideally a HF treatment (pharmacotherapy, device, or behavioral intervention) would do one or more of the following: reduce mortality, maintain or improve clinical stability (prevent or minimize HF events), and/or improve symptoms/functional status. When designing a clinical trial, incorporating clinical events and clinical status measures into endpoints may allow for the most comprehensive assessment of an intervention’s impact. This strategy also acknowledges that death and hospitalization may be more important outcomes for some stakeholders, but that symptom burden is high in HF and improvement in that burden is meaningful to patients.¹⁶

By incorporating a ranked evaluation of mortality, HF hospitalization, and clinical status within a single metric, a hierarchical composite endpoint enables a more comprehensive evaluation of important clinical changes for patients in the trial. In a time-to-first event analysis looking at a composite of death and HF hospitalization, a patient with a HF hospitalization occurring early in the trial would be considered to have faired “worse” than a patient with a death occurring later in the trial, because their event occurred earlier. Similarly, a patient with a single early HF hospitalization that lasted for 2 days would be ranked as having faired worse than a patient who had multiple recurrent HF admissions later in the trial, even if the total number of admissions (and days admitted) was far larger for the second patient. By using a hierarchical endpoint, it is possible to include death and HF hospitalizations within the same outcome, but to also acknowledge that death at any time during the trial would be ranked as a worse outcome than HF hospitalization, even if it occurs later. Hierarchical endpoints allow for the incorporation of more granular data regarding hospitalization, including total number of hospitalizations or total number of days hospitalized, to be incorporated into the outcome rather than relying on the timing of the first event.

Depending on the specific study population, many patients within a clinical trial may not have a major clinical event during the study follow-up period. For instance, in recent trials of empagliflozin and sacubitril-valsartan in HFrEF, 22–24% of patients experienced death from a cardiovascular cause or a HF hospitalization over median follow-up times that ranged from 16 to 27 months.^{1, 2} However, the observation that over three-quarters of all patients in these trials did not experience one of these events does not mean that three-quarters of patients were non-responders to empagliflozin or sacubitril-valsartan. In PARADIGM-HF, patients with HFrEF who were never hospitalized for HF still had significantly larger improvements in their KCCQ scores at 8 months if they were randomized to sacubitril-valsartan vs enalapril; though these patients did not contribute to the primary endpoint in this trial, these patients demonstrated a beneficial response to the drug.¹⁷ Similarly, in EMPEROR-Reduced, patients with HFrEF randomized to empagliflozin were both more likely to show a meaningful improvement in KCCQ score and less likely to have a deterioration in their KCCQ score than patients randomized to placebo.¹⁸ These findings, though important to providers and patients, are not captured in a traditional time-to-first event analysis. Though markers of functional status are sometimes captured as secondary endpoints, these endpoints are often considered exploratory. If these secondary endpoints are considered within a hierarchical statistical analysis, they are still likely to be disregarded if a primary endpoint driven by clinical events does not meet a threshold value for statistical significance. In contrast, a hierarchical endpoint can capture differences in each ranked outcome and allow for changes in both clinical events and functional status to contribute not only to a point estimate for effect size, but also to statistical significance. For instance, a subsequent analysis of EMPULSE showed that a calculated hazard ratio for time to cardiovascular death or HF event numerically favored empagliflozin but did not meet statistical significance (HR 0.69, p=0.10). In contrast, a win ratio that incorporated change in KCCQ score found a similar magnitude of benefit and captured enough events to show statistical significance (win ratio 1.36, p=0.005).¹⁹

Incorporation of clinical status into a primary endpoint potentially makes it possible to better identify the likelihood that patients might derive a benefit from an intervention, even if no clinical events occur, and allows more patients with true changes in clinical status to contribute to the observed treatment effect of a study drug. Conversely, prioritizing major clinical events in an endpoint, and not solely relying on changes in clinical status may allow for early identification of safety concerns with a more complete estimate of potential benefit. A trial of levosimendan in patients with advanced HF found that although patients had greater improvements in clinical status on the drug, they also had a numerically higher risk of death.²⁰ If only change in clinical status had been evaluated, the impact on mortality might have been minimized or only assessed as a safety signal. By incorporating both death and clinical status into one endpoint, both can be appropriately incorporated into evaluation of the potential treatment effect.

Statistical Reasons for Hierarchical Endpoints

There are also statistical reasons to support the use of a hierarchical endpoint in clinical trials. Hierarchical endpoints may be more sensitive in discerning a treatment effect (i.e., identifying those who experienced a change in symptom burden or clinical status in addition to changes in major clinical events), the inclusion of more components in the outcome may also increase power by capturing changes in functional status as events rather than only death or hospitalization.^{21, 22} As cardiovascular morbidity and mortality have decreased over time with the addition of efficacious therapies, the number of patients and length of clinical trials has increased in order to allow for an adequate number of events, increasing cost and delaying time to trial results. By including changes in clinical status, more patients may show a difference in treatment between the two groups, provided that such a difference exists, possibly allowing for a similar trial to be completed in less time and/or with fewer patients. In addition, the hierarchy can be pre-specified to capture the “worst” outcome for a patient not merely the first event.

Potential Limitations to Hierarchical Endpoints

There have been concerns that hierarchical endpoints could be impacted by length of follow-up, especially if some outcomes were more likely to show a difference at earlier timepoints of follow-up.²³ For example, a trial with a relatively short follow-up period might have a hierarchical endpoint driven by short-term events rather than long-term events, which are more likely to be clinical endpoints such as death or hospitalization. This could potentially be problematic if the intervention had a discrepant effect on the short- and long-term outcomes (e.g., if it led to benefit on a short-term endpoint but harm on a longer-term endpoint). For example, a trial of an inotrope that increased exercise capacity but also increased risk for life-threatening arrhythmias using a hierarchical endpoint of death, hospitalization for HF, and change in 6MWT might find an overall benefit if the trial was run for only 2 weeks, based on improvement in 6MWT, but an overall harm if the trial was run for a year, based on higher rates of death with inotrope therapy. It is therefore conceivable that over time the hierarchical endpoint could have different findings depending on whether length of follow-up allowed for adequate accrual of long-term events. Individual reporting of each component of the hierarchical endpoint in order to understand relative benefit and the frequency of each event can help to identify any discordance between different outcomes. For hierarchical endpoints that are assessed using win ratios, trials with a high number of ties may find the use of win odds beneficial. Win odds may be somewhat more difficult to interpret, however, but this method incorporates wins, losses, and ties as compared with the win ratio which only examines wins and losses.¹⁹

Hierarchical Endpoints in Heart Failure Trials

Hierarchical endpoints have been used in several recent clinical HF trials. These trials each leveraged a hierarchical endpoint as their primary outcome in order to incorporate information not only about death and first HF hospitalization into the primary endpoint, but also information about changes in HF biomarkers, recurrent hospitalizations, and change in symptom burden (Table 1).^6–8

Table 1.

Examples of HF trials utilizing hierarchical endpoints

Trial	Trial Characteristics	Trial Length and Size	Hierarchical Endpoint
FIGHT6	GLP1 agonist after hospitalization for HFrEF	300 patients 6 month follow-up	Ranked score of time to death, time to HF hospitalization, and time-averaged change in NT-proBNP
ATTR-ACT7	Tafamidis use in patients with transthyretin amyloid cardiomyopathy	441 patients 30 month follow-up	Hierarchy of all-cause mortality and frequency of cardiovascular hospitalizations
EMPULSE8	Empagliflozin initiation during admission for acute HF	530 patients 90 day follow-up	Hierarchy of death, HF events and time to first HF event, or a ≥5 point change in KCCQ total symptom score

ATTR-ACT indicates Transthyretin Amyloidosis Cardiomyopathy Clinical Trial; EMPULSE, EMPagliflozin in patients hospitalized with acUte heart faiLure who have been StabilizEd trial; FIGHT, Functional Impact of GLP-1 for Heart Failure Treatment trial; GLP1, glucagon-like peptide 1; HF, heart failure; HFrEF, heart failure with reduced ejection fraction; KCCQ, Kansas City Cardiomyopathy Questionnaire; NT-proBNP, N-terminal pro-B-type natriuretic peptide.

FIGHT

The placebo-controlled Functional Impact of GLP-1 for Heart Failure Treatment (FIGHT) trial evaluated the use of liraglutide in patients with recent HF hospitalization. The primary endpoint included a hierarchy of death, time to HF hospitalization, and time-averaged change in N-terminal pro-B-type natriuretic peptide (NT-proBNP).⁶ In this hierarchical analysis, patients who died were ranked based on time of death; patients who did not die were ranked based on time to first HF hospitalization. Patients with neither a death nor a HF hospitalization were ranked based on change in NT-proBNP indexed to length of follow-up. The study found no significant difference in this hierarchical endpoint between those randomized to liraglutide vs placebo. Nearly half of all patients (43%) experienced either death or HF hospitalization during the 6-month follow-up period, and no significant difference in NT-proBNP was observed between the two groups. Importantly, this was a neutral trial, and the point estimates favored placebo over liraglutide. Though use of a hierarchical endpoint may increase sensitivity to discerning a difference in clinical outcomes, it is clear that this endpoint also allows for identification of ineffective therapies.

ATTR-ACT

The Transthyretin Amyloidosis Cardiomyopathy Clinical Trial (ATTR-ACT) assessed the use of tafamidis in patients with transthyretin amyloid with cardiac involvement. The primary endpoint was a hierarchy of all-cause mortality and frequency of cardiovascular hospitalizations.⁷ The trial found that patients randomized to tafamidis vs placebo had significantly better outcomes at 30 months. Overall, 34.9% of patients died and 55.5% had a cardiovascular hospitalization. By including total number of cardiovascular hospitalizations in the primary endpoint, the investigators acknowledged that recurrent hospitalizations may be more clinically relevant than a single hospitalization occurring early in trial follow-up.

EMPULSE

The Multicentre, Randomised, Double-blind 90-day Superiority Trial to Evaluate the Effect on Clinical Benefit, Safety, and Tolerablilty of Once Daily Oral EMPagliflozin 10mg Compared to Placebo, Initiated in Patients Hospitalised for Acute Heart Failure (de Novo or Decompensated Chronic HF) Who have been Stabilised (EMPULSE) evaluated the use of empagliflozin initiation for patients hospitalized for HF, regardless of ejection fraction. The primary outcome was clinical benefit, evaluated based on occurrence of death, total number of HF events, time to first HF event, and change in KCCQ score at 90 days.⁸ These outcomes were assessed hierarchically in a stratified win ratio. Patients within each strata (de novo or acute or chronic HF) randomized to empagliflozin were compared with those randomized to placebo: if within a pair of patients, one patient died, then the patient who survived “won.” If neither patient in the pair died during the 90-day follow-up, then the patient with the fewest number of HF hospitalizations won, and so on. After all patient pairs had been assessed, the total number of empagliflozin “wins” was divided by the number of empagliflozin “losses.” Patients randomized to empagliflozin vs placebo were 36% more likely to show a clinical benefit (stratified win ratio 1.36, 95% CI 1.09–1.68, p=0.005) (Figure 1), demonstrating the likelihood of study drug empagliflozin “winning” overall or for each individual outcome within the hierarchy of death, HF event frequency, time to first HF event, and change in KCCQ score.

Figure 1.

Results of the stratified win ratio from EMPULSE.⁸

CI indicates confidence interval; HFE, heart failure event; KCCQ-TSS, Kansas City Cardiomyopathy Questionnaire Total Symptom Score.

Likely in part because trial follow-up was only 90 days, only 6.2% of patients died and only 12.6% experienced a worsening HF event, with only 15.7% experiencing either death or a HF event. However, most patients (79.6%) had changes in their KCCQ score enabling an analysis for clinical benefit using a relatively small trial population and short follow-up time. Importantly, patients randomized to empagliflozin were shown to have a clinical benefit comparable to what was previously reported in other trials of empagliflozin in chronic HF, despite the smaller sample size and shorter trial duration, and despite the use of a hierarchical endpoint rather than a time-to-first event analysis.^{1, 24}

Use of Hierarchical Endpoint in HEART-FID

The HEART-FID trial leveraged a hierarchical primary endpoint in the assessment of intravenous FCM in patients with HFrEF and iron deficiency. The primary outcome was a hierarchical composite of death, total number of HF hospitalizations, and 6MWT distance, with clinical events assessed over 12 months and change in 6MWT assessed at 6 months (Figure 2).⁹ A longer follow-up period for death and hospitalizations for HF (12 months vs. 6 months) was chosen to help ensure that differences in this outcome were captured and that findings were not driven solely by changes in 6MWT distance. This design allowed for a larger number of clinical events (by inclusion of recurrent hospitalization events) than would typically be included in a conventional time-to-first event analysis over the same follow-up period, and also ensured that the maximum number of patients have the opportunity to contribute to the primary endpoint. The use of 6MWT as the metric to assess functional status is consistent with prior studies of FCM in HF, enabling comparisons across trials.^{25, 26} The trial enrolled approximately 3000 patients, with a minimum follow-up of one year. It was powered to identify changes in the hierarchical endpoint and to assess changes in a secondary endpoint that is a time-to-first event analysis for a composite of cardiovascular death and hospitalization for HF.⁹ The trial was therefore well-positioned to assess not only the impact of intravenous FCM on morbidity and mortality in patients with HFrEF, but also the net impact on their health by incorporating both clinical events and change in functional status in response to FCM therapy.

Figure 2:

Win ratio: FCM vs. placebo.

Each patient randomized to intervention (intravenous ferric carboxymaltose) is compared to each patient randomized to placebo (intravenous saline), from the first imputed dataset. If one patient dies, the one who survived “wins.” If both patients die, then the one who survived longest “wins.” If neither patient died, then whichever patients had the fewest number of hospitalizations for heart failure “wins.” If neither patient died or was hospitalized for heart failure, then the patient with the greatest improvement (or smallest worsening) in 6MWT distances wins.

CI indicates confidence interval; FCM, ferric carboxymaltose; HF, heart failure; mos, months; 6MWD: 6-minute walk distance.

In addition to the primary hierarchical endpoint analysis to measure the likelihood of a better outcome with intravenous FCM vs placebo, results of HEART-FID were presented using an unmatched win ratio to aid in clinical interpretation (Figure 2). The primary outcome, as assessed by the Wilcoxon-Mann-Whitney test, evaluated whether patients randomized to intravenous FCM vs placebo were likely to have statistically different outcomes overall and generated a p-value indicating statistical significance. A complimentary win ratio analysis was used to describe how much more likely a patient is to “win” or “lose” in terms of ranked outcomes if randomized to intravenous FCM vs placebo, therefore providing not only a p-value but also an estimated treatment effect for the hierarchical endpoint. Importantly, HEART-FID reported both a hierarchical endpoint and results on each individual outcome within the composite endpoint in order to enable better understanding of which specific outcomes might contribute to a possible difference in the primary endpoint between intervention and placebo arms. These results were further strengthened by a traditional Cox hazard model examining time to cardiovascular death or first HF hospitalization. As a sensitivity analysis, a hierarchical endpoint will also be analyzed using data collected over the duration of the trial.

Ultimately, HEART-FID narrowly missed a prespecified p value of 0.01 with an observed p value of 0.019, indicating that there was not a statistically significant difference in the hierarchical endpoint.²⁷ Similar results were found when the results were assessed as a win ratio, with patients overall having 10% more “wins” with intravenous FCM as compared with placebo, but a 99% confidence interval that crossed unity (win ratio 1.10, 99% CI 0.99 to 1.23). While the observed differences in the primary endpoint were driven by wins in death (20% more wins with FCM; mortality representing 18% of win ratio decisions), the other components contributed to a larger proportion of the decisions in the analysis (i.e., 82% of win ratio decisions were made for either HF hospitalization or 6MWD). Specifically, there were also 11% more wins with FCM for 6MWD (64% of decisions) and a similar percentage of wins with FCM (compared with placebo) for HF hospitalization (18% of decisions). When change in 6MWT was not considered and the impact of study drug was considered using a more traditional time-to-event analysis in a secondary outcome of time to cardiovascular death or first HF hospitalization, there was a nominal reduction in events with FCM that was not statistically significant (HR 0.93, 96% CI 0.81 to 1.06). Taken together, these results suggest that there is likely a modest clinical benefit for intravenous FCM in patients with HFrEF.

Despite the noted advantages of the hierarchical endpoint in HEART-FID, there were several potential limitations that warrant consideration. First, the endpoint focused on all-cause rather than cardiovascular mortality which may be less precise for some therapeutic agents. However, by assessing all deaths, it avoids the challenge when a therapy has discordant cardiovascular and non-cardiovascular effects, as well as concerns for competing risk. Second, the prespecified hospitalization endpoint focused on total number of HF hospitalizations rather than duration of hospitalization. However, length of stay was incorporated into sensitivity analyses in HEART-FID. Third, different stakeholders (clinicians, patients, payers) may have less familiarity with this type of endpoint which may challenge broad interpretation. The HEART-FID investigators have prioritized dissemination efforts around enhancing familiarization with this endpoint. Fourth, the prespecified timepoints of assessment at 12 months for clinical outcomes and 6 months for 6MWT distance were selected based on prior trial data with FCM, prior HFrEF trials, and the aim to be clearly translatable to clinical practice; however, these timepoints do not provide longer-term data on these hierarchical components. Nonetheless, the key secondary endpoint of time to cardiovascular death or first HF hospitalization over the duration of follow-up helped to providethese insights. Fifth, the ranked hierarchy of importance of outcomes (death being worse than a hospitalization) was determined by the investigators, and may not be consistent with the values of patients and other stakeholders.

Conclusion

Hierarchical endpoints provide an important opportunity in cardiovascular research to better understand the impact of an intervention on both HF events and changes in clinical status. By incorporating both types of outcomes into one ranked endpoint, it becomes possible to assess more holistically the potential utility of a new therapy on patient well-being and outcomes. The incorporation of a marker of clinical status allows for a more holistic assessment of whether a patient is likely to benefit from an intervention as compared with placebo, and provides the opportunity for more patients to contribute to the primary outcome (compared with a time-to-first-event analysis). Statistically, this endpoint also allows for robustly powered trials, which may enable the implementation of trials that otherwise would not have been feasible given decreasing rates of death and hospitalization in the HF population. The potential role for hierarchical endpoints has been demonstrated in several prior HF trials and represents a robust endpoint that can successfully identify promising therapies in HF. The HEART-FID trial built on the past experiences of these successful HF trials, and utilized a hierarchical primary endpoint to better determine the role of intravenous FCM on not only clinical events in HF, but also on patient functional status.

What is New?

• Traditional time-to-event endpoints are often used in heart failure (HF) trials to assess the impact of a therapy on death and hospitalization for HF.

• These endpoints often require large trials run over longer periods of time in order to accrue enough events, and also fail to assess the impact of a therapy on a patient’s functional status

• The use of hierarchical endpoints is therefore of great interest, because it allows for “ranked” evaluation of likelihood for a better outcome across multiple outcomes, including death and hospitalization for HF, while also incorporating an assessment of changes in functional status

What are the Clinical Implications?

• Hierarchical endpoints may allow for a more comprehensive evaluation of the impact of a therapy on patient outcomes and well-being

• The Randomized Placebo-controlled Trial of FCM as Treatment for Heart Failure with Iron Deficiency (HEART-FID) trial represents the largest HF trial to date that implemented a hierarchical endpoint as its primary outcome in order to assess the impact of IV ferric carboxymaltose on death, hospitalization for HF, and change in 6-minute walk test.

Nonstandard Abbreviations and Acronyms:

6MWT

6-minute walk test

FCM

ferric carboxymaltose

GLP-1

glucagon-like peptide 1

HF

heart failure

HFrEF

heart failure with reduced ejection fraction

KCCQ

Kansas City Cardiomyopathy Questionnaire

NT-proBNP

N-terminal pro-B-type natriuretic peptide

NYHA

New York Heart Association