Author + information
- Received January 25, 2016
- Revision received March 16, 2016
- Accepted March 16, 2016
- Published online September 1, 2016.
- Aditi Mallick, MDa,
- Parul U. Gandhi, MDb,
- Hanna K. Gaggin, MD, MPHa,c,
- Nasrien Ibrahim, MDa and
- James L. Januzzi, MDa,c,∗ ()
- aCardiology Division, Massachusetts General Hospital, Boston, Massachusetts
- bCardiology Division, VA Connecticut Healthcare System, West Haven, and Yale School of Medicine, New Haven, Connecticut
- cHarvard Clinical Research Institute, Boston, Massachusetts
- ↵∗Reprint requests and correspondence:
Dr. James L. Januzzi, Cardiology Division, Massachusetts General Hospital, 32 Fruit Street, Yawkey 5984, Boston, Massachusetts 02114.
Objectives The goal of this study was to define and assess the significance of worsening heart failure (WHF) in patients with chronic ambulatory heart failure with reduced ejection fraction (HFrEF).
Background WHF has been identified as a potentially relevant clinical event in patients with acute heart failure (HF) and is increasingly used as an endpoint in clinical trials. No standardized definition of WHF exists. It remains uncertain how WHF relates to risk for other HF events or how treatment may affect WHF.
Methods A total of 151 symptomatic patients with chronic HFrEF were randomized to standard of care HF management or a goal to lower N-terminal pro–B-type natriuretic peptide (NT-proBNP) concentrations ≤1,000 pg/ml in addition to standard of care. WHF was prospectively defined as: 1) new or progressive symptoms and/or signs of decompensated HF; and 2) unplanned intensification of diuretic therapy.
Results Over a mean follow-up of 10 months, 45 subjects developed WHF. At baseline, patients developing incident WHF had higher ejection fraction (31% vs. 25%; p = 0.03), were more likely to have jugular venous distension and edema (p < 0.02), were less likely to receive angiotensin-converting enzyme inhibitors or received these agents at lower doses (p < 0.04), and also received higher loop diuretic doses (p < 0.001). Occurrence of WHF was strongly associated with subsequent HF hospitalization/cardiovascular death (hazard ratio, landmark analysis: 18.8; 95% confidence interval: 5.7 to 62.5; p < 0.001). NT-proBNP–guided care reduced the incidence of WHF in adjusted analyses (hazard ratio: 0.52; p = 0.06) and improved event-free survival (log-rank test p = 0.04).
Conclusions In chronic HFrEF, WHF was associated with substantial risk for morbidity and mortality. NT-proBNP–guided care reduced risk for WHF.
Heart failure (HF) is a heterogeneous clinical diagnosis, encompassing a variety of underlying pathophysiologic processes. Diagnosis and treatment of HF have improved (1), although patients affected by the diagnosis nonetheless experience considerable morbidity and mortality. Risk for death after HF hospitalization rises considerably, as high as 35% by 1 year (2), and this risk essentially doubles with each subsequent hospitalization (3). Accordingly, a better understanding of the risk factors for adverse outcomes is needed.
Part of the challenge in the care of patients with HF is the fact that the diagnosis is a heterogeneous clinical entity, whose manifestations and outcomes remain difficult to predict. This challenge is particularly germane in the present era of HF care, in which increased focus on optimization of chronic HF care and simultaneous prevention of HF hospitalization represents a major effort. Emerging tools for identifying impending risk for HF events include implantable hemodynamic monitoring, as well as serial measurement of prognostic biomarkers (4,5); however, clinical history and physical examination remain highly valuable for such prognostication. In this manner, recent attention has focused on the phenomenon of worsening heart failure (WHF), which may signal HF deterioration and unfavorable prognosis (6). WHF is increasingly being used as an accepted inclusion criterion (7) and endpoint in clinical trials (8) as an indirect predictor of HF outcomes such as hospitalization or death.
Despite growing use of WHF to include subjects in trials or as an endpoint, most studies have generally relied on clinician judgment to define the presence or absence of WHF, leading to substantial subjectivity; to our knowledge, no standardized definition of WHF has been accepted. Furthermore, the importance of WHF in the ambulatory population with heart failure and reduced ejection fraction (HFrEF) remains poorly defined, as much of the data regarding WHF have focused on those with hospitalized HF. Finally, treatment strategies that might favorably influence incidence of WHF are not defined.
In the PROTECT (Pro-BNP Outpatient Tailored Chronic HF Therapy) study, WHF was strictly and prospectively defined, and it was used as an endpoint for the trial (9). This approach provides an opportunity to examine protocol-defined WHF in a contemporary cohort of patients with HFrEF. The hypothesis of the present study was that WHF would be associated with significant subsequent adverse outcomes and that HF management guided by N-terminal pro–B-type natriuretic peptide (NT-proBNP) would reduce WHF.
All study procedures were approved by the local institutional review board. Informed consent was obtained from participants.
PROTECT study design
The design and results of the PROTECT study have been published previously (5,9). PROTECT was a prospective, randomized, single-center trial of 151 patients with New York Heart Association functional class II to IV systolic HF (left ventricular ejection fraction [LVEF] ≤40%). In brief, patients were eligible if they were >21 years of age and had experienced a decompensated HF event within 6 months before enrollment. Patients were excluded if they had severe renal disease, inoperable aortic valvular heart disease, life expectancy <1 year due to causes other than HF, cardiac transplantation or revascularization indicated or expected within 6 months, severe pulmonary disease, or coronary revascularization (percutaneous coronary intervention or coronary artery bypass graft) within the previous 3 months.
After enrollment, patients were randomized to receive either standard HF management (with a goal of minimizing HF symptoms and achieving maximal dosages of therapies with proven mortality benefit in HF) or standard HF management plus treatment adjustments to reduce NT-proBNP concentrations ≤1,000 pg/ml. To achieve this goal, patients received up-titration of guideline-directed medical therapy according to clinical judgement (both arms) with or without supplemental testing for NT-proBNP; in the context of a therapy change, repeated office visits were made within 4 weeks. At entry to the study, patients underwent a 2-dimensional echocardiogram (10), which was repeated in study completers at a mean of 10 months from enrollment; both the technician performing the echocardiogram as well as the staff member interpreting the study were blinded to study arm or NT-proBNP values.
The primary endpoint of the PROTECT study was total cardiovascular (CV) events (including WHF) over a 1-year period.
Protocol definition of WHF
Online Table 1 details the PROTECT protocol definition of WHF, which was defined as: 1) new or progressive symptoms/signs of decompensated HF (including significant weight gain, worsening dyspnea or fatigue, newly elevated jugular venous pressure, new cardiac S3 gallop rhythm, the development of pulmonary rales, hepatic congestion, cool extremities, or lower extremity edema); and 2) unplanned intensification of oral or intravenous decongestive therapy with loop diuretic agents or the addition of a thiazide diuretic agent to loop diuresis.
Baseline patient characteristics were assessed and analyzed as a function of the presence or absence of subsequent incident WHF after enrollment. These factors included demographic characteristics, background medical history, physical examination at baseline, and drug therapy at baseline. As before, we expressed dosages of angiotensin-converting enzyme (ACE) inhibitors in “lisinopril total daily equivalents,” dosages of beta-blockers in “metoprolol succinate total daily equivalents,” dosages of mineralocorticoid receptor antagonists in “spironolactone total daily equivalents,” and dosages of loop diuretic agents in “furosemide total daily equivalents” (5,11). Results of diagnostic testing at enrollment, including LVEF, renal function (estimated glomerular filtration rate determined by using the Modification of Diet in Renal Disease equation), as well as concentrations of several prognostic biomarkers (including NT-proBNP, soluble ST2 [sST2], highly sensitive troponin T [hsTnT] , and galectin-3 ), were also considered. Dichotomous variables were compared by using chi-square tests, and continuous variables were compared by using the Student t test or Mann-Whitney U test as appropriate.
Univariate followed by multivariable logistic regression analyses were used to identify baseline predictors of the occurrence of subsequent WHF; all baseline characteristics, medical history, medications, and echocardiographic data were considered for the univariate model, with a retention p value of 0.05 used to identify candidates for inclusion in the multivariable model. Biomarkers were considered for this model, both as log-transformed continuous variables as well as dichotomous variables, using previously described thresholds (4,5,12) of 1,000 pg/ml and 2,000 pg/ml for NT-proBNP, 14 ng/ml for hsTnT, 35 ng/ml for sST2, and 17.8 ng/ml for galectin-3.
To better understand the prognostic importance of WHF, we then considered the composite outcome of HF hospitalization or CV death relative to the presence or absence of incident WHF after enrollment. To do so, univariate regression was used to identify candidate baseline predictors of HF hospitalization/CV death subsequent to enrollment. Dichotomous clinical variables were entered into the model as such and included demographic characteristics and medical history. Use of guideline-directed medical therapy was entered both dichotomously as well as relative to the log-transformed doses used. As described earlier, to better explore the predictive relationship between biomarker concentrations, we log-transformed each biomarker concentration, as well as biomarkers expressed in dichotomous fashion. Multivariable Cox proportional hazards analysis was then performed, including univariate predictors with a retention p value of <0.05 along with WHF, with hazard ratios (HRs) and 95% confidence intervals (CIs) for HF hospitalization/CV death landmarked after the WHF event.
Treatment strategies were next considered as a function of subsequent WHF. The effect of HF drug therapy on the likelihood of developing WHF was first assessed and the effects of NT-proBNP–guided HF care on WHF then considered. First, total burden of WHF was considered as a function of treatment allocation, with logistic regression expressing reduction in total WHF events in odds ratios (ORs) and 95% CIs. After this assessment, Cox proportional hazards analysis was used to evaluate the effects of NT-proBNP–guided therapy on the occurrence of WHF. Lastly, Kaplan-Meier curves were constructed to evaluate the effects of NT-proBNP versus standard of care (SOC) management relative to the time-to-first WHF event and compared by using the log-rank test.
All statistics were performed by using Stata version 13.0 software (Stata Corp., College Station, Texas); all p values are 2-sided, with values <0.05 considered significant.
Patient characteristics in those with WHF versus without WHF
Of the 151 study subjects, the baseline factors of demographic characteristics, medical history, medication use, medication doses at baseline, treatment allocation, laboratory results for various biomarkers, and, when available, echocardiographic data were compared across those with and without WHF (Table 1). These data show that patients who went on to subsequently develop WHF were largely similar to those who did not in terms of age, race, and baseline systolic blood pressure, among other characteristics. Those developing WHF were more likely to have higher LVEF (31% vs. 25%; p = 0.03), jugular venous distension (51% vs. 29%; p = 0.01), and peripheral edema (49% vs. 24%; p = 0.002) on physical examination. Therapeutically, those developing incident WHF later in the study were less likely to be taking ACE inhibitors (53% vs. 72%; p = 0.03) at baseline and, if they were taking these drugs, they were on lower dosages (2.5 mg vs. 5 mg; p = 0.03). In addition, patients who eventually developed WHF were also more likely to be taking a loop diuretic agent at baseline (100% vs. 88%; p = 0.01) and at higher dosages (60 mg vs. 40 mg; p < 0.001) (Table 2). Overall, those developing WHF were less likely to receive guideline-directed medical therapies or received them at lower doses, and were more likely to receive loop diuretic agents at higher doses or to have thiazide diuretic agents added. Data were not available regarding intravenous medication use.
Those who developed WHF also had worse renal function at baseline (estimated glomerular filtration rate 50.2 ml/min/m2 vs. 63.7 ml/min/m2; p < 0.001) and higher concentrations of known prognostic markers, including NT-proBNP, hsTnT, sST2, and galectin-3 (all p < 0.05) (Table 1).
WHF in the PROTECT study
Among the 151 patients in the study, 45 (29.8%) had WHF, and the average number of WHF events per patient was 0.56, with a range of 0 to 6. Of those with WHF, the majority (80%) had 1 to 2 events. The median time from enrollment to worsening HF event was 128 days (interquartile range: 57 to 194 days) with a range of 6 to 521 days. Variables of interest, including baseline demographic characteristics, medications, medication doses, treatment allocation, echocardiographic data, and time in response for each biomarker assessed, were analyzed to identify predictors of subsequent WHF (Table 3). In univariate logistic regression analyses, multiple potential predictors were identified; however, after adjustment in multivariable logistic regression analyses, only log-transformed sST2 (HR: 4.6; 95% CI: 2.4 to 8.8; p < 0.001) and higher baseline LVEF (HR: 3.1; 95% CI: 1.1 to 8.7; p = 0.03) retained significance.
WHF and outcomes
Patients with WHF had a significantly higher rate of subsequent HF hospitalization than those without (53% vs. 1%; p < 0.001) and a significantly higher rate of composite outcome of HF hospitalization or cardiovascular death (56% vs. 6%; p < 0.001) (Table 4). Thus, of 45 patients with WHF, 25 subsequently had a HF hospitalization or died after WHF as defined according to the PROTECT protocol, whereas of the 106 patients without WHF, only 6 had an HF hospitalization or death.
Multivariate Cox proportional hazards models for the composite of CV death and worsening HF as the outcome were constructed by using stepwise selection from baseline variables (Table 5), highlighting the predictive power of baseline NT-proBNP (HR: 1.9; 95% CI: 1.3 to 2.8; p < 0.001) for this outcome, as well as spironolactone usage, ACE inhibitor usage, and baseline New York Heart Association functional class. The model was further refined with subsequent addition of the incident WHF; in this model, landmarked incident WHF powerfully predicted risk for subsequent HF hospitalization/CV death (HR, landmark analysis: 18.8; 95% CI: 5.7 to 62.5; p < 0.001).
NT-proBNP–guided therapy and WHF
Among patients treated with biomarker-guided therapy, WHF occurred in 23% of patients compared with 37% of patients in the SOC arm; considering frequency of total WHF events, NT-proBNP–guided care was associated with near-significant reduction of total WHF events, compared with SOC (HR: 0.50; 95% CI: 0.25 to 1.0; p = 0.06). Using first incident WHF as the dependent variable, NT-proBNP–guided therapy significantly reduced the occurrence of WHF (HR: 0.52; 95% CI: 0.28 to 0.96; p = 0.04) in Cox proportional hazards analyses. Lastly, Kaplan-Meier analysis found that NT-proBNP–guided care significantly improved the probability of survival without WHF (p = 0.03) (Figure 1).
The principal findings of the present study illustrate that WHF (identified by using the protocol definition as given in the PROTECT study) is both common and predictive of poor outcomes (particularly subsequent HF hospitalization) among patients with chronic HFrEF. In addition, elevated sST2 serum concentrations and higher baseline LVEF were independent predictors of WHF. Although data are not available regarding use of intravenous rescue diuretic therapy, patients who developed WHF were less likely to be taking ACE inhibitors at high doses and more likely to be taking higher doses of oral loop and thiazide diuretic agents at follow-up, despite similar blood pressure readings. Finally, outpatient management guided by reduction of NT-proBNP seemed to reduce the risk of WHF.
Our results are novel in 3 ways. First, the PROTECT study protocol definition of WHF (incorporating factors from history, physical examination, laboratory test results, and therapeutic intervention) provides useful possibilities relative to clinical trial design of treatments for HF care. Although utilized as an inclusion criterion (7) and endpoint in clinical trials of HF (8), standardized research criteria for the definition of WHF are lacking; recent research has called for standardized endpoints in clinical trials of HF (1,13), and our data define and validate the endpoint of WHF. In this regard, given strong links to consequent HF hospitalization or cardiovascular death, the PROTECT study definition of WHF identifies a high-risk patient phenotype, which may be leveraged for future clinical trial design for HF therapeutics. Although the benchmark outcome measure of HF hospitalization/CV death will likely continue to represent a gold standard, addition of WHF as a relevant high-risk CV event may allow for minimizing of sample size. Second, on a clinical level, given the importance of HF hospitalization as a sentinel event in the care of patients with chronic HF, our results may help clinicians to immediately identify patients likely to benefit from closer follow-up intervals, more frequent screening, and additional ancillary services, such as home telemonitoring. Because studies of home monitoring and treatment approaches have been generally disappointing (14), identifying patients most likely to benefit would be expected to be useful. WHF seems amenable to reduction with different treatment strategies, and such patients might therefore be ideal for more intensive surveillance. In this regard, third, no previous data have evaluated the effect of NT-proBNP–guided care on WHF; we showed that NT-proBNP–guided care reduced WHF events by 50% and prolonged WHF-free survival. Given the importance of HF hospitalization to patients and health care systems alike, the 50% reduction in burden and frequency of WHF associated with NT-proBNP care is of substantial significance, particularly because HF care is increasingly being shifted from hospitals to ambulatory centers; with the greater focus on natriuretic peptide–guided care, such data are important.
Interestingly, elevated concentrations of sST2 (a biomarker believed to play a pivotal role in myocardial fibrosis and left ventricular remodeling ) were associated with increased risk of developing WHF. This outcome is in accordance with previously reported data that sST2 may be predictive of deleterious left ventricular remodeling (4) and confirms that some biomarkers may be more useful than others to more objectively risk-stratify patients with HF. Because sST2 has been shown to predict a broad range of adverse HF outcomes independently of NT-proBNP or other biomarkers, its role for more broad use in HF monitoring and treatment programs deserves further attention.
Limitations of this study include the fact it is small, with the potential for type I or II error, and was performed at a single tertiary care urban medical center. Nonetheless, results of the study are significant and consistent throughout. Our results should be validated in other, prospectively designed studies using the PROTECT study definition of WHF. Of note, most events in this study were HF readmissions (i.e., that an episode of WHF was mostly predictive of the need for a subsequent hospital admission). The relationship between WHF and cardiovascular mortality could not be adequately assessed in this study due to the small number of events. In terms of the statistical analysis, hierarchical modeling was not performed because only 22 patients experienced >1 WHF event. Moreover, the number of events was a categorical/descriptive variable (none, exactly 1, or >1) with the 3 categories all being mutually exclusive on the basis of approximately equal follow-up; thus, the data were analyzed independently. Lastly, we lack data regarding use of intravenous loop diuretic agents used as rescue therapy for those with WHF; outpatient intravenous diuretic therapy was not in substantial use in our institution at the time of the trial, however.
The strengths of this study include its well-defined cohort and its strictly defined definition for WHF. Our findings support the central finding of the PROTECT trial that biomarker-guided therapy is superior to SOC for management of chronic outpatient HFrEF. Our results also lend further understanding to the observation that biomarker-guided therapy may improve quality of life (16), possibly by leading to a more stable HF phenotype, with better functional capacity, and may foster more significant reverse left ventricular remodeling (10,17).
WHF is both common and morbid in patients with ambulatory chronic HFrEF. WHF can serve as a useful intermediate outcome, predictive of existing, harder outcomes of HF hospitalization, and CV death. Although recent data have focused on the importance of WHF in acute HF syndromes, our study is the first, to our knowledge, to specifically examine this question in patients with chronic ambulatory HF. Treatment strategies affecting WHF such as NT-proBNP–guided care seem to reduce risk for other events in parallel. Further studies are needed to corroborate and extend these findings.
COMPETENCY IN MEDICAL KNOWLEDGE: For clinicians, this research defines the sentinel event of WHF as a combination of symptoms, signs, and therapeutic response. The results indicate that the phenomenon of WHF may be useful for identifying patients at higher risk for HF hospitalization or death. In this vein, better characterization of WHF may be advantageous for delivery systems and stewards of population health management by helping to identify patients likely to benefit from closer follow-up intervals, more frequent screenings, and additional ancillary services.
TRANSLATIONAL OUTLOOK: For investigators, defining WHF allows for standardization of language and a benchmark against which novel diagnostics, pharmacotherapies, and device therapies can be compared. In addition, the independent predictors of WHF included the biomarker sST2, which further supports its important role as a risk predictor in HF.
For a supplemental table, please see the online version of this article.
Dr. Gaggin is supported in part by the Clark Fund for Cardiac Research Innovation; has received grant support from Roche and Portola; consulting income from Roche Diagnostics, American Regent, Amgen, Boston Heart Diagnostics, and Critical Diagnostics; and research payments for Clinical Endpoint Committees for EchoSense. Dr. Ibrahim is supported by the Dennis and Marilyn Barry Fellowship in Cardiovascular Research. Dr. Januzzi has received grants and consulting fees from Roche; grants from Singulex and Prevencio; consulting fees from Critical Diagnostics, Sphingotec, and Philips; was a member of the Data and Safety Monitoring Board of Amgen; received consulting and Clinical Events Committee fees from Novartis, Boeringer Ingelheim, and Janssen; and is supported in part by the Hutter Family Professorship in Medicine in the Field of Cardiology. Drs. Mallick and Gandhi have reported that they have no relationships relevant to the contents of this paper to disclose. John R. Teerlink, MD, served as Guest Editor for this paper.
- Abbreviations and Acronyms
- angiotensin-converting enzyme
- confidence interval
- heart failure with reduced ejection fraction
- hazard ratio
- highly sensitive troponin T
- left ventricular ejection fraction
- N-terminal pro–B-type natriuretic peptide
- soluble ST2
- worsening heart failure
- Received January 25, 2016.
- Revision received March 16, 2016.
- Accepted March 16, 2016.
- American College of Cardiology Foundation
- Dharmarajan K.,
- Hsieh A.F.,
- Kulkarni V.T.,
- et al.
- Gaggin H.K.,
- Szymonifka J.,
- Bhardwaj A.,
- et al.
- Januzzi J.L.,
- Rehman S.U.,
- Mohammed A.A.,
- et al.
- Davison B.A.,
- Metra M.,
- Cotter G.,
- et al.
- Gheorghiade M.,
- Greene S.J.,
- Butler J.,
- et al.,
- SOCRATES-REDUCED Investigators and Coordinators
- Weiner R.B.,
- Baggish A.L.,
- Chen-Tournoux A.,
- et al.
- Gaggin H.K.,
- Motiwala S.,
- Bhardwaj A.,
- Parks K.A.,
- Januzzi J.L.
- Gaggin H.K.,
- Truong Q.A.,
- Rehman S.U.,
- et al.