Author + information
- Received August 3, 2020
- Accepted August 4, 2020
- Published online October 26, 2020.
- Justin A. Ezekowitz, MBBCH, MSca,
- Christopher M. O’Connor, MDb,c,
- Richard W. Troughton, MDd,
- Wendimagegn G. Alemayehu, PhDa,
- Cynthia M. Westerhout, PhDa,
- Adriaan A. Voors, MD, PhDe,
- Javed Butler, MD, MPH, MBAf,
- Carolyn S.P. Lam, MBBS, PhDe,g,
- Piotr Ponikowski, MDh,
- Michele Emdin, MD, PhDi,
- Mahesh J. Patel, MDj,
- Burkert Pieske, MDk,
- Lothar Roessig, MDl,
- Adrian F. Hernandez, MD, MHSb and
- Paul W. Armstrong, MDa,∗ ()
- aDivision of Cardiology, Department of Medicine, Canadian VIGOUR Centre, University of Alberta, Edmonton, Canada
- bDivision of Cardiology, Department of Medicine, Duke Clinical Research Institute, Duke University, Durham, North Carolina
- cInova Heart and Vascular Institute, Falls Church, Virginia
- dDepartment of Medicine, University of Otago, Christchurch, New Zealand
- eDepartment of Cardiology and Thorax Surgery, University of Groningen, University Medical Center Groningen, Groningen, the Netherlands
- fDepartment of Medicine, University of Mississippi Medical Center, Jackson, Mississippi
- gDepartment of Cardiology, National Heart Centre Singapore and Duke-National University of Singapore
- hWroclaw Medical University, Department of Heart Disease, Wroclaw, Poland
- iCardiothoracic Department, Fondazione Toscana Gabriele Monasterio, Pisa, Italy; and Institute of Life Sciences, Scuola Superiore Sant'Anna, Pisa, Italy
- jMerck & Co. Inc., Kenilworth, New Jersey
- kDepartment of Internal Medicine–Cardiology, Charité University Medicine, German Heart Center, Berlin, Germany
- lBayer AG, Wuppertal, Germany
- ↵∗Address for correspondence:
Dr. Paul W. Armstrong, 4-120 Katz Group Centre for Pharmacy and Health Research, University of Alberta, Edmonton, Alberta T6G 2E1, Canada.
Objectives The purpose of this study was to examine the treatment effect of vericiguat in relation to N-terminal pro-B-type natriuretic peptide (NT-proBNP) levels at randomization.
Background Vericiguat compared with placebo reduced the primary outcome of cardiovascular death (CVD) or heart failure hospitalization (HFH) in patients with HF with reduced ejection fraction (HFrEF) in the VICTORIA (A Study of Vericiguat in Participants With Heart Failure With Reduced Ejection Fraction) trial. Because an interaction existed between treatment and the primary outcome according to pre-specified quartiles of NT-proBNP at randomization, we examined this further.
Methods This study evaluated the NT-proBNP relationship with the primary outcome in 4,805 of 5,050 patients as a risk-adjusted, log-transformed continuous variable. Hazard ratios (HRs) and 95% confidence intervals (CIs) are presented.
Results Median NT-proBNP was 2,816 pg/ml (25th to 75th percentile: 1,556 to 5,314 pg/ml). The study treatment effect varied across the spectrum of NT-proBNP at randomization (with log2 transformation, p for interaction = 0.002). A significant association between treatment effects existed in patients with levels <4,000 pg/ml and remained evident up to 8,000 pg/ml. A 23% relative risk reduction occurred in the primary endpoint with NT-proBNP ≤4,000 pg/ml (HR: 0.77; 95% CI: 0.68 to 0.88). For NT-proBNP values ≤4,000 pg/ml (n = 3,100), the HR was 0.78 (95% CI: 0.67 to 0.90) for HFH and 0.75 (95% CI: 0.60 to 0.94) for CVD. For NT-proBNP ≤8,000 pg/ml (n = 4,133), the HR was 0.85 (95% CI: 0.76 to 0.95) for the primary outcome, 0.84 (95% CI: 0.75 to 0.95) for HFH, and 0.84 (95% CI: 0.71 to 0.99) for CVD. For NT-proBNP >8,000 pg/ml (n = 672), the HR was 1.16 (95% CI: 0.94 to 1.41) for the primary outcome.
Conclusions A reduction in the primary composite endpoint and its CVD and HFH components was observed in patients on vericiguat compared with subjects on placebo with NT-proBNP levels up to 8,000 pg/ml. This provided new insight into the benefit observed in high-risk patients with worsening HFrEF. (A Study of Vericiguat in Participants With Heart Failure With Reduced Ejection Fraction [HFrEF] [MK-1242-001] [VICTORIA]; NCT02861534)
The measurement of N-terminal pro-B-type natriuretic peptide (NT-proBNP) is a cornerstone for diagnosis, risk stratification, and follow-up of patients with heart failure (HF) (1). Accordingly, NT-proBNP levels have been widely used as an inclusion criterion for clinical trials and as surrogate metrics within studies to gauge the relationship between the treatment effect on this biomarker and overall outcomes (2). Because patient, trial, mechanism of treatment, or other factors may modulate the relationship between natriuretic peptides and outcomes, further understanding of these relationships is crucial for the interpretation of interventions and their impact on outcomes.
The VICTORIA (A Study of Vericiguat in Participants With Heart Failure With Reduced Ejection Fraction) trial evaluated vericiguat compared with placebo in high-risk patients with HF and reduced ejection fractions (HFrEFs) who had a recent HF hospitalization (HFH) or intravenous diuretic therapy. The primary findings were a reduction in the primary composite outcome of cardiovascular death (CVD) or HFH with vericiguat compared with placebo (3). In a pre-specified subgroup analysis, a significant interaction was noted between NT-proBNP quartiles at randomization and the primary endpoint. Specifically, those patients in the highest quartile of NT-proBNP (>5,314 pg/ml) appeared to have less benefit from vericiguat compared with placebo; conversely, those in the lowest 3 quartiles derived greater benefit from vericiguat compared with placebo. Caution must be exercised in interpreting treatment effects from such subgroups because of the uncertainties about their biological plausibility versus the play of chance (4).
To further understand the NT-proBNP subgroup results, we explored the relationship of NT-proBNP across the spectrum of levels at randomization with the treatment effect of vericiguat compared with placebo in the VICTORIA trial. We also described the relationship of other baseline factors that potentially modified the interpretation of this pre-specified NT-proBNP subgroup interaction.
The design, baseline characteristics, and results of the VICTORIA trial were previously published (3,5,6). In brief, the trial included 5,050 patients with worsening chronic HF (New York Heart Association functional classes II to IV), a left ventricular EF <45%, elevated natriuretic peptide levels, and recent HF decompensation. Patients were randomly assigned in a 1:1 ratio to receive vericiguat or placebo. Patients in sinus rhythm had to have BNP ≥300 pg/ml or NT-proBNP ≥1,000 pg/ml; patients with atrial fibrillation had to have BNP ≥500 pg/ml or NT-proBNP ≥1,600 pg/ml. Guideline-based HF therapies were encouraged before inclusion, including sacubitril/valsartan. The trial protocol was approved by regulatory agencies in participating countries, as well as the ethics committees and institutional review boards at participating sites. All patients provided written informed consent.
Natriuretic peptide assessment
Of the 5,050 participants in VICTORIA, 245 did not have an evaluable NT-proBNP at randomization, leaving a remaining 4,805 patients for analysis (Supplemental Table 1). Per protocol, NT-proBNP was measured by a central laboratory on the Roche Elecsys assay (Roche Diagnostics, Mannheim, Germany). Analytical range of the assay was 10 to 175,000 pg/ml.
The primary outcome of VICTORIA was the composite endpoint of time to CVD or first HFH. We also explored the individual components of the composite outcomes (CVD and HFH). All clinical outcomes observed up to the primary analysis cutoff date (June 18, 2019) were included.
The distribution of NT-proBNP was right skewed, with values ranging from 10 to 175,000 pg/ml. The relationship of NT-proBNP (as a continuous variable) to the primary composite outcome was assessed using the restricted cubic spline with 4 knots (7) and was observed to be nonlinear (p < 0.001) (Supplemental Figure 1). We transformed NT-proBNP on a logarithmic scale (base 2), revealing a linear relationship with the log-hazard of the primary outcome. The relative risk change in this transformed variable was interpreted as per doubling of NT-proBNP. To estimate these relative associations with the clinical outcomes, Cox proportional hazard models were used, and hazard ratios (HRs) and 95% confidence intervals (CIs) were reported, along with absolute risk reduction. We anticipated differences in the patient characteristics across the spectrum of NT-proBNP at randomization. Therefore, associations with clinical outcomes were stratified for region and/or race and reported as unadjusted and adjusted for the modified Meta-Analysis Global Group in Chronic Heart Failure (MAGGIC) risk score (6,8). We also explored broader adjustment using an internally validated set of variables from VICTORIA for the primary composite endpoint and CVD, which contained 16 and 19 covariates, respectively (Supplemental Table 2A). A description of the patient characteristics that are in the VICTORIA prediction models but do not appear in Table 1 are provided in Supplemental Table 2B. The interaction of the randomized study treatment assignment (vericiguat or placebo) with NT-proBNP at randomization was included in the models, and p values for the interactions were reported).
Cut points were approximated from the treatment effect across the spectrum of (log-transformed) NT-proBNP, in which the upper confidence limit of the treatment effect did not include 1.00 (i.e., 4,000 pg/ml) and below which the point estimate of the treatment effect was <1.00 (i.e., 8,000 pg/ml). These cut points were then used to summarize the relative treatment estimate (HR; 95% CI) above and below these values. To provide context to these findings, patient characteristics were summarized according to ≤4,000, >4,000 to 8,000, and >8,000 pg/ml. These characteristics are also provided according to quartiles of NT-proBNP in the Supplemental Appendix. Median (25th to 75th percentiles) for continuous variables and frequency (percentages) for categorical variables were used to summarize data.
Patients with missing values of NT-proBNP (n = 245) at randomization were excluded from the present study, and their data were not imputed (Supplemental Table 1). Data for patients with a missing MAGGIC score (71 of 4,805 patients) were not imputed, and these patients were excluded from the risk-adjusted analyses. All analyses were performed using SAS version 9.4 (SAS Institute, Inc., Cary, North Carolina). A 2-sided test result with p < 0.05 was considered statistically significant.
Treatment effect of vericiguat compared with placebo by baseline NT-proBNP
The relationship between NT-proBNP as a continuous variable and the treatment effect of vericiguat on the primary composite outcome is shown in Central Illustration A. The histograms within this Figure depict the distribution of patients by NT-proBNP intervals of 1,200 pg/ml. Across the spectrum of NT-proBNP levels, there was an association of the treatment effect up to 8,000 pg/ml (Supplemental Figures 2A and 2B). This association of treatment efficacy with vericiguat, compared with placebo, on the primary outcome was further estimated at 4,000 pg/ml (HR: 0.90; 95% CI: 0.82 to 0.99) (Supplemental Figure 2A). The treatment effect on the primary outcome for patients with NT-proBNP ≤4,000 pg/ml and those with >4,000 pg/ml produced HRs of 0.77 (95% CI: 0.68 to 0.88) and 1.05 (95% CI: 0.92 to 1.20), respectively (Table 2). These and the following findings were confirmed after adjustment for prediction models developed in the VICTORIA trial (Supplemental Table 3). The individual components of the primary outcome were similarly affected (Central Illustration B and C, Supplemental Figures 3 and 4) in patients with NT-proBNP levels ≤4,000 pg/ml at randomization for CVD (HR: 0.75; 95% CI: 0.60 to 0.94) and HFH (HR: 0.78; 95% CI: 0.67 to 0.90) (Table 2). These HRs corresponded to absolute risk reductions of 6.8, 1.3, and 5.5 events per 100 patient-years for the primary composite endpoint, CVD, and HFH, respectively.
Further exploration of this relationship showed an association of the treatment effect with vericiguat on the primary outcome in NT-proBNP levels extended to 8,000 pg/ml. The components of the primary outcome were associated with a significant effect for CVD (HR: 0.84; 95% CI: 0.71 to 0.99) and HFH (HR: 0.84; 95% CI: 0.75 to 0.95) (Table 2). These HRs corresponded to absolute risk reductions of 5.4, 0.8, and 4.6 events per 100 patient-years for the primary composite endpoint, CVD, and HFH, respectively. When NT-proBNP values >8,000 pg/ml were analyzed, this association of treatment benefit was no longer evident for the primary outcome (HR: 1.16; 95% CI: 0.94 to 1.41), CVD (HR: 1.32; 95% CI: 1.01 to 1.71), or HFH (HR: 1.17; 95% CI: 0.92 to 1.48). When these outcomes for NT-proBNP values >8,000 pg/ml were adjusted using the VICTORIA prognostic model (Supplemental Table 3), the respective HRs were 1.10 (95% CI: 0.90 to 1.35) for the primary outcome, 1.27 (95% CI: 0.97 to 1.66) for CVD, and 1.11 (95% CI: 0.87 to 1.41) for HFH.
To provide clinical context, the patient characteristics according to NT-proBNP categories (≤4,000, >4,000 to 8,000, >8,000 pg/ml) at randomization are shown in Table 1. Patients with higher NT-proBNP levels tended to be older, were in a higher New York Heart Association functional class, and had a lower body mass index, ejection fraction, and estimated glomerular filtration rates (Table 1). In addition, they were more likely to be enrolled earlier after index HFH. Although atrial fibrillation, baseline device, and beta-blocker use were similar across the NT-proBNP spectrum, lower rates of angiotensin-converting enzyme inhibitors, angiotensin receptor blockers, and mineralocorticoid receptor antagonists use were evident for patients within the upper strata of NT-proBNP values. The median MAGGIC risk score was lower among the cohort with NT-proBNP values ≤4,000 pg/ml than that in the other 2 groups.
These findings were in concert with those based on the pre-specified subgroup analysis of quartiles of NT-proBNP, which indicated higher risk patients in the highest quartile (e.g., median MAGGIC risk score of 27). Patient characteristics and the treatment effect according to the components of the primary composite endpoint are available in Supplemental Tables 4A and 4B.
Association of NT-proBNP with clinical outcomes
When the overall association between NT-proBNP values at randomization and the risk of the primary composite outcome was assessed, we observed an excess hazard per doubling of NT-proBNP with and without adjustment for the MAGGIC risk score (Table 3). This relationship was more pronounced in patients assigned to vericiguat compared with those assigned to placebo. Similar results were observed for CVD and HFH. These findings were confirmed after adjustment for prediction models developed in the VICTORIA trial (Supplemental Table 5).
The principal novel finding of the present study was that the treatment effect of vericiguat compared with placebo on the primary composite endpoint was greatest in patients with NT-proBNP levels <8,000 pg/ml at randomization, which was further amplified if these levels were <4,000 pg/ml. The benefit of vericiguat in this high-risk HF population appeared to be clinically meaningful in both relative and absolute terms and extended to both the CVD and HFH components of the primary outcome. Although patients with higher NT-proBNP levels within the range up to 8,000 pg/ml had several clinical features indicative of higher risk, the treatment effect associated with vericiguat compared with placebo persisted after risk adjustment for many of these factors. Our findings—using all the natriuretic peptide data as a continuous variable—provided further insight into that provided from the initial pre-specified NT-proBNP quartile subgroups and suggested the potential for a broader range of patients who could benefit from treatment. Moreover, there appeared to be amplification of that benefit for patients whose entry NT-proBNP levels were <4,000 pg/ml and who represented most of the study population. These findings might be useful in informing clinicians about how best to apply this new treatment in the growing population of patients with HFrEF with recent worsening of their status, despite receiving good guideline-based therapies.
Patients enrolled in the VICTORIA trial were at higher baseline risk for hospitalization or death compared with patients in contemporary clinical trials in HF, as evident by the median baseline NT-proBNP concentration of 2,816 pg/ml, and they were generalizable to the broader HFrEF population (9). In this trial population at such high risk for clinical events, NT-proBNP was predictive of the primary outcomes of CVD or first HFH, concordant with previous trials and other patient cohorts. This analysis also identified that this association between NT-proBNP and clinical outcomes was not linear by revealing a much steeper increment in the relationship for values >4,000 pg/ml. This finding deserves consideration in future clinical trials or cohort studies that will estimate prognosis and evaluate new therapies.
A secondary analysis of the PARADIGM-HF (Prospective Comparison of ARNI with ACEI to Determine Impact on Global Mortality and Morbidity in Heart Failure) trial identified that NT-proBNP levels before and after run-in, as well as 1 month after randomization, were independently associated with clinical outcomes (10). However, there was no interaction between baseline NT-proBNP and treatment efficacy in that study. A subsequent Swedish registry in patients with HFrEF analyzed at hospital discharge or as outpatients showed median NT-proBNP levels of 2,640 pg/ml (75% of patients had a NT-proBNP of <5,914 pg/ml) (11). This study also demonstrated higher cardiovascular event rates, defined as cardiovascular hospitalization or CVD, with increasing NT-proBNP values.
The VICTORIA trial did not have a run-in phase, and the values at randomization were similar to the baseline values in other studies (e.g., GUIDE-IT [Guiding Evidence Based Therapy Using Biomarker Intensified Treatment in Heart Failure]),which had a median baseline NT-proBNP of 2,653 pg/ml (12). This comparison was particularly relevant because the primary event rate of 35% in GUIDE-IT (using the same primary endpoint) identified a similarly high-risk population. Other contemporary trials, including DAPA-HF (Dapagliflozin and Prevention of Adverse Outcomes in Heart Failure) and PARADIGM-HF, had median NT-proBNP values of 1,437 pg/ml and 1,615 pg/ml, respectively, and primary endpoint rates less than one-half of that observed in the VICTORIA trial (11).
Study strengths and limitations
We studied a large event-rich population with carefully adjudicated clinical outcomes and NT-proBNP values centrally measured in a core laboratory. Although we chose not to impute values for the 245 patients missing NT-proBNP levels at randomization, there were no substantive differences between those patients and the overall cohort, and their omission was unlikely to have affected our results (Supplemental Table 1). The present study was an extension of the pre-specified subgroup analysis of the primary outcomes by quartiles of NT-proBNP and allowed for the thorough examination of this relationship across the full spectrum of NT-proBNP, which was broad (10 to 175,000 pg/ml). The relationship among higher NT-proBNP levels, reduced vericiguat efficacy, and poorer outcomes might be related to diverse factors, including patient illness severity, adequacy of guideline-based therapy, activation of neuroendocrine pathways involved in HF pathophysiology, or blunting of guanylate cyclase responsiveness in advanced HF. This characterization of continuous NT-proBNP informed the additional subgroups of NT-proBNP and should be viewed as hypothesis-generating in the context of the overall trial results. Finally, we adjusted for known determinants of NT-proBNP, but as with all secondary analyses, there might be unmeasured confounders.
In patients enrolled in the VICTORIA trial, NT-proBNP at randomization was predictive of the primary and secondary outcomes. Patients with an NT-proBNP <8,000 pg/ml had both a reduction in the primary outcome and its 2 components, CVD and HFH, which identified a patient cohort who had vericiguat might be beneficial.
COMPETENCY IN MEDICAL KNOWLEDGE: For high-risk patients with HFrEF on guideline-based therapy who had a recent HF hospitalization or intravenous diuretic therapy, vericiguat appears to benefit those patients with NT-proBNP <8,000 pg/ml. This population, especially in those with NT-proBNP <4,000 pg/mml, had a reduction in both the primary outcome and its 2 components, CVD and HFH, thereby identifying a patient cohort who had vericiguat might be beneficial.
TRANSLATIONAL OUTLOOK: Although vericiguat is not yet approved for clinical use, this treatment may provide a useful additional therapeutic option in a high-risk population with HF.
Author Relationship With Industry
This study was supported by Merck Sharp & Dohme Corp., a subsidiary of Merck & Co., Inc., and Bayer AG. Mr. Ezekowitz has received research grants from Bayer, Merck, Servier, Amgen Sanofi, Novartis, Cytokinetics, American Regent, and Applied Therapeutics; has received consulting fees from Bayer, Merck, Servier, Amgen, Sanofi, Novartis, Cytokinetics, American Regent, and Applied Therapeutics. Dr. O’Connor has received research funding from Merck; and has received consulting fees from Bayer, Dey LP, and Bristol-Myers Squibb Foundation. Dr. Troughton has received research grants from and personal fees from Merck and Roche Diagnostics. Dr. Voors has received research grants from Boehringer Ingelheim and Roche Diagnostics; and has received consulting fees from Merck, Bayer, Amgen, AstraZeneca, Boehringer Ingelheim, Cytokinetics, Myokardia, Novartis, Servier, and Roche Diagnostics. Dr. Butler has received consulting fees from Bayer, Merck, Amgen, AstraZeneca, Boehringer Ingelheim, Bristol-Myers Squibb, CVRx, G3 Pharmaceutical, Janssen, Luitpold, Medtronic, Novartis, Vifor, and Novo Nordisk. Dr. Lam has received research grants from Bayer, National Medical Research Council of Singapore, Boston Scientific, Roche Diagnostic, Medtronic, Vifor Pharma, and AstraZeneca; has received consulting fees from Merck, Bayer, Boston Scientific, Roche Diagnostic, Vifor Pharma, AstraZeneca, Novartis, Amgen, Janssen Research & Development LLC, Menarini, Boehringer Ingelheim, Abbott Diagnostics, Corvia, Stealth BioTherapeutics, Novo Nordisk, JanaCare, Biofourmis, Darma, Applied Therapeutics, MyoKardia, Cytokinetics, WebMD Global LLC, Radcliffe Group Ltd, and Corpus; has a patents pending for PCT/SG2016/050217 and 16/216,929; and is the Cofounder and non-executive director of eKo.ai. Dr. Ponikowski has received research grants from Vifor Pharma Ltd, and Servier; has received consulting fees from MSD, Novartis, Vifor Pharma Ltd, Servier, Bristol-Myers Squibb, Boehringer Ingelheim, Respicardia, AstraZeneca, Cibiem, RenalGuardSolution, and Berlin Chemie. Dr. Pieske has received research grants from MSD, Bayer, and Servier; has received consulting fees from MSD, Bayer, Servier, Bristol-Myers Squibb, MedScape, Daiichi-Sankyo, and Novartis; and has received non-financial support from MSD, Bayer, and Novartis. Dr. Hernandez has received research grants from Merck, AstraZeneca, Novartis, and Verily; and has received consulting fees from Merck, Bayer, Amgen, AstraZeneca, and Novartis. Dr. Armstrong has received research grants from Merck, Bayer, Sanofi-Aventis Recherche & Développement, Boehringer Ingelheim, and CSL Limited; and has received consulting fees from Merck, Bayer, AstraZeneca, and Novartis. All other authors have reported that they have no relationships relevant to the contents of this paper to disclose.
The authors are pleased to acknowledge Yinggan Zheng, MA, MEd, of the Canadian VIGOUR Centre for his collaboration on the statistical analysis, and Elizabeth Cook, BA, of the Duke Clinical Research Institute for her editorial assistance and preparation of the manuscript submission.
Scott Solomon, MD, served as Guest Editor for this paper.
The authors attest they are in compliance with human studies committees and animal welfare regulations of the authors’ institutions and Food and Drug Administration guidelines, including patient consent where appropriate. For more information, visit the JACC: Heart Failure author instructions page.
- Abbreviations and Acronyms
- confidence interval
- cardiovascular death
- heart failure
- heart failure hospitalization
- heart failure with reduced ejection fraction
- hazard ratio
- Meta-Analysis Global Group in Chronic Heart Failure
- N-terminal pro-B-type natriuretic peptide
- Received August 3, 2020.
- Accepted August 4, 2020.
- 2020 The Authors
- Yancy C.W.,
- Jessup M.,
- Bozkurt B.,
- et al.
- Vaduganathan M.,
- Claggett B.,
- Packer M.,
- et al.
- Armstrong P.W.,
- Pieske B.,
- Anstrom K.J.,
- et al.
- Armstrong P.W.,
- Roessig L.,
- Patel M.J.,
- et al.
- Pieske B.,
- Patel M.J.,
- Westerhout C.M.,
- et al.
- Harrell F.
- Butler J.,
- Anstrom K.J.,
- Armstrong P.W.
- Zile M.R.,
- Claggett B.L.,
- Prescott M.F.,
- et al.
- Savarese G.,
- Orsini N.,
- Hage C.,
- et al.