Author + information
- Received November 3, 2015
- Revision received February 17, 2016
- Accepted February 19, 2016
- Published online July 1, 2016.
- Alicia Mecklai, MDa,b,
- Haris Subačius, MAc,
- Marvin A. Konstam, MDd,
- Mihai Gheorghiade, MDc,
- Javed Butler, MD, MPH, MBAe,
- Andrew P. Ambrosy, MDf and
- Stuart D. Katz, MD, MSb,∗ ()
- aWeill Cornell Medical College, New York, New York
- bNew York University School of Medicine, New York, New York
- cNorthwestern University Feinberg School of Medicine, Chicago, Illinois
- dThe CardioVascular Center of Tufts Medical Center and Tufts Medical Center, Boston, Massachusetts
- eStony Brook School of Medicine, Stony Brook, New York
- fDuke University Medical Center, Durham, North Carolina
- ↵∗Reprint request and correspondence:
Dr. Stuart D. Katz, Leon H. Charney Division of Cardiology, New York University Langone Medical Center, 530 First Avenue, Skirball 9R, New York, New York 10016.
Objectives The aim of this study was to characterize the association between decongestion therapy and 30-day outcomes in patients hospitalized for heart failure (HF).
Background Loop diuretic agents are commonly prescribed for the treatment of symptomatic congestion in patients hospitalized for HF, but the association between loop diuretic agent dose response and post-discharge outcomes has not been well characterized.
Methods Cox proportional hazards models were used to estimate the association among average loop diuretic agent dose, congestion status at discharge, and 30-day post-discharge all-cause mortality and HF rehospitalization in 3,037 subjects hospitalized with worsening HF enrolled in the EVEREST (Efficacy of Vasopressin Antagonism in Heart Failure: Outcome Study With Tolvaptan) study.
Results In univariate analysis, subjects exposed to high-dose diuretic agents (≥160 mg/day) had greater risk for the combined outcome than subjects exposed to low-dose diuretic agents (18.9% vs. 10.0%; hazard ratio: 2.00; 95% confidence interval: 1.64 to 2.46; p < 0.0001). After adjustment for pre-specified covariates of disease severity, the association between diuretic agent dose and outcomes was not significant (hazard ratio: 1.11; 95% confidence interval: 0.89 to 1.38; p = 0.35). Of the 3,011 subjects with clinical assessments of volume status, 2,063 (69%) had little or no congestion at hospital discharge. Congestion status at hospital discharge did not modify the association between diuretic agent exposure and the combined endpoint (p for interaction = 0.84).
Conclusions Short-term diuretic agent exposure during hospital treatment for worsening HF was not an independent predictor of 30-day all-cause mortality and HF rehospitalization in multivariate analysis. Congestion status at discharge did not modify the association between diuretic agent dose and clinical outcomes.
Congestive signs and symptoms related to excess total body sodium and water are common in patients with heart failure (HF) despite the use of chronic diuretic therapy (1). Current guidelines recommend adjustment of diuretic agent doses as needed to increase renal sodium and water excretion in patients with clinical evidence of congestion with the goal of achieving a euvolemic state (2,3). Factors that may contribute to need for higher diuretic agent doses to maintain clinical euvolemia include excess dietary sodium intake, altered gastrointestinal absorption of oral diuretic agents, decreased secretion of loop diuretic agents into the nephron, comorbid renal insufficiency, and glomerular and renal tubular changes that limit the increase in fractional sodium excretion in response to loop diuretic agents (4,5).
Although loop diuretic agents are the most commonly used class of drugs for the treatment of congestive signs and symptoms (1), the effects of diuretic agents on clinical outcomes in patients with HF remain uncertain. Several prior studies have demonstrated an association between higher dose diuretic agents and greater risk for adverse clinical outcomes in both outpatient and inpatient HF populations (6–9), with putative mechanisms of harm related to greater risk for arrhythmic death due to electrolyte depletion, hypovolemia, worsening renal function, and/or increased neurohormonal activation (4,10,11). In contrast, other studies demonstrated that higher diuretic agent doses are associated with improved survival in hospitalized patients with HF with evidence of hemoconcentration in response to therapy (12–15).
Given these divergent data, we hypothesized that the relationship between diuretic agent dose and clinical outcomes might differ according to clinical response to diuretic therapy, as evidenced by the post-treatment volume status of the patient (presence or absence of clinical signs and symptoms of congestion at hospital discharge). To test this hypothesis, we performed a post hoc analysis to determine the association between short-term in-hospital diuretic agent use and post-discharge 30-day clinical outcomes in patients discharged with and without signs of persistent congestion in the EVEREST (Efficacy of Vasopressin Antagonism in Heart Failure: Outcome Study With Tolvaptan) trial.
The rationale, design, and main results of the EVEREST trial have been previously reported (16,17). In brief, the EVEREST trial was an international, multicenter, randomized, double-blind, placebo-controlled trial comparing the safety and efficacy of tolvaptan, a selective vasopressin-2 receptor antagonist, versus matching placebo, in addition to standard care, in 4,133 subjects hospitalized for worsening HF. Eligible subjects were ≥18 years of age with left ventricular ejection fractions ≤40%, 2 or more signs or symptoms of volume overload (dyspnea, pitting edema, jugular venous distension), and New York Heart Association functional class III or IV symptoms who were hospitalized for an exacerbation of chronic systolic HF. Key exclusion criteria included systolic blood pressure <90 mm Hg, hemodynamically significant primary valvular disease, acute coronary syndrome, serum creatinine ≥3.5 mg/dl, treatment with hemofiltration or dialysis, end-stage HF (i.e., candidates for ventricular assist devices or continuous positive intravenous inotropic therapy), or life expectancy <6 months. Eligible subjects were randomized within 48 h of admission to receive either 30 mg of oral tolvaptan daily or matching placebo for a minimum of 60 days. Concomitant HF medications including diuretic agents were administered according to the discretion of the treating physician at each participating site. The clinical trial was conducted in full accordance with the Declaration of Helsinki, with Institutional Review Board or ethics committee approval at all study sites. For the current post-hoc analysis, we included 3,037 subjects who survived until hospital discharge with available loop diuretic agent dose information recorded during the first 72 h of hospitalization and recorded clinical signs and symptoms of congestion at study entry.
Average daily loop diuretic agent dose (milligrams per day) was calculated during the first 72 h of hospitalization as estimates of oral furosemide equivalents on the basis of relative potency (1 mg bumetanide = 20 mg torsemide = 40 mg furosemide) and adjusted for the differential bioavailability related to the route of administration (1 mg intravenous furosemide = 2 mg oral furosemide). For example, a diuretic agent exposure of furosemide 40 mg twice daily by intravenous administration would be recorded as a total daily dose of 160 mg in oral furosemide equivalents.
Congestion status was determined by study investigators at hospital discharge or on hospital day 7 (whichever occurred earlier) on the basis of the following signs or symptoms: presence of rales on a 4-point scale (0 = no rales, 1 = rales at bases, 2 = rales at bases to 50% way up the posterior lung fields, and 3 = rales at bases to >50% way up the posterior lung fields), presence of orthopnea on the basis of a 4-point scale (0 = none, 1 = seldom, 2 = frequent, and 3 = continuous), presence of jugular venous distension ≥6 cm (yes or no), or presence of pedal or sacral edema (yes or no). Congestion status at hospital discharge was determined as a cumulative summed score adapted from a previous EVEREST publication. The revised score added rales and sacral edema to the previous congestion score on the basis of orthopnea, jugular venous pressure, and pedal edema; little or no congestion at discharge was defined as a summed score of ≤1 (18).
Values are reported as mean ± SD for continuous variables with normal distributions, median (interquartile range) for continuous variables not normally distributed, and percentages for categorical variables. Average daily loop diuretic agent dose was expressed as a categorical variable (low dose, <160 mg/day; high dose, ≥160 mg/day) on the basis of a commonly used clinical definition to differentiate “low-dose” from “high-dose” diuretic agents and the empirical observation that a significant association between diuretic agent dose and outcome events was present only in subjects treated with diuretic agent doses higher than this cutoff value. Differences in selected clinical characteristics between groups on the basis of average daily loop diuretic agent dose were assessed with parametric or nonparametric tests for independent samples for continuous variables as appropriate for the variable distribution and chi-square tests for categorical variables. The primary outcome endpoint for this post hoc analysis was the 30-day combined endpoint of all-cause mortality and rehospitalization for HF as determined by an independent event adjudication committee. To address our hypothesis that the association between short-term diuretic agent exposure and 30-day outcomes would differ by discharge congestion status (effect modification), the statistical plan was designed to first test the associations between average daily loop diuretic agent dose and discharge congestion status on the combined endpoint separately and then test for interaction between these 2 predictor variables. Univariate log-rank tests were used to evaluate time to event associations with the combined endpoint, and Cox proportional hazards regression was used to calculate the hazard ratios (HRs) and their 95% confidence intervals (CIs). Kaplan-Meier curves were constructed for patients with low and high average daily loop diuretic agent doses. Associations between average daily loop dose and the combined endpoint were also assessed using multivariate Cox proportional hazard models to estimate the independent association of average loop diuretic agent dose and the combined outcome endpoint. Multivariate regression models were adjusted for other pre-selected covariates of disease severity that were used in previous EVEREST multivariate analyses (15,18): study drug assignment, age, sex, geographic region, left ventricular ejection fraction, medical comorbidities (diabetes, hypertension, and chronic kidney disease), history of ventricular arrhythmias, serum sodium, B-type natriuretic peptide (pg/ml), serum blood urea nitrogen (mg/dl), estimated glomerular filtration rate (ml/min/1.73 m2, derived from the abbreviated MDRD [Modification of Diet in Renal Disease] equation ), serum albumin (mg/dl), systolic blood pressure (mm Hg), QRS duration (ms), New York Heart Association functional class, presence of an automatic implantable cardioverter-defibrillator, baseline medication use (beta-adrenergic receptor blocker, angiotensin-converting enzyme inhibitors or angiotensin receptor blockers, mineralocorticoid receptor antagonists, digoxin, and positive inotropic agents), and the presence of atrial flutter or fibrillation on baseline electrocardiography. Estimated effect sizes are reported as HRs and 95% CIs. Assumptions of proportional hazards were verified. All statistical analyses were performed using SAS version 9.4 (SAS Institute, Cary, North Carolina) at a conventional 2-tailed 5% significance level.
Baseline characteristics by diuretic agent exposure
Average daily diuretic agent dose during the first 72 h of hospitalization varied over a wide range in the 3,037 subjects with available loop diuretic agent dosing data (Figure 1). Table 1 describes the baseline characteristics grouped by the average daily diuretic agent dose during the first 72 h of hospitalization (low dose [<160 mg] or high dose [≥160 mg]). Mean age and sex distribution were comparable with the full EVEREST trial population. High average daily diuretic agent dose was associated with clinical markers of more severe disease, including lower left ventricular ejection fraction, more severe New York Heart Association symptoms, lower systolic blood pressure, lower estimated glomerular filtration rate, higher blood urea nitrogen levels, and higher B-type natriuretic peptide levels (p ≤ 0.001 for all vs. low average daily diuretic agent dose). Subjects receiving the high average daily diuretic agent dose had greater frequency of comorbidities, including diabetes mellitus, chronic kidney disease, and chronic lung disease (p < 0.001 for all vs. low average daily diuretic agent dose). Average diuretic agent dose differed by geographic region. A majority of subjects enrolled in North America received high average daily diuretic agent doses, whereas a majority of the subjects in the Eastern European sites received low average daily diuretic agent doses. The natural logarithm of the average daily diuretic agent dose did not differ in subjects treated with placebo compared with that in subjects treated with tolvaptan (untransformed mean values: 89.4 mg [95% CI: 85.6 to 93.3 mg] vs. 87.8 mg [95% CI: 83.9 to 91.8 mg], respectively; p = 0.55).
Association between short-term diuretic agent exposure and 30-day combined endpoint
There were 380 events (12.5%; 67 deaths, 340 rehospitalizations, and 27 patients with both) in the study population over a follow-up duration of 30 days. Figure 2 demonstrates the number of 30-day combined outcome events according to categories of short-term diuretic agent exposure. Among patients with average diuretic agent dose <160 mg/day, there was no association between diuretic agent dose and 30-day risk for the combined outcome event (HR for log-linear analysis of diuretic agent dose: 1.09; 95% CI: 0.90 to 1.33; p = 0.37). However, among those with average diuretic agent doses ≥160 mg/day, there was a significant association between diuretic agent dose and 30-day risk for the combined outcome event (HR for log-linear analysis of diuretic agent dose: 1.22; 95% CI: 1.05 to 1.42; p = 0.001). Univariate analysis demonstrated that those subjects exposed to the high-dose diuretic agents had greater risk for the 30-day combined endpoint compared with those exposed to low-dose diuretic agents (18.9% vs. 10.0%; HR: 2.00; 95% CI: 1.64 to 2.46; p < 0.0001) (Figure 3). After adjustment for pre-specified covariates of disease severity in multivariate analysis, diuretic agent dose was no longer associated with increased risk for the combined endpoint (HR: 1.11; 95% CI: 0.89 to 1.38; p = 0.35) (Table 2). Study treatment allocation (tolvaptan vs. placebo) did not modify the association between diuretic agent dose and outcomes (p for interaction = 0.26).
Interaction between short-term diuretic agent exposure and discharge congestion status
Of the 3,011 patients with information on discharge congestion status, 2,063 (69%) had little or no congestion at hospital discharge. Discharge congestion status (little or no congestion vs. persistent congestion) did not modify the effect of diuretic agent exposure on the combined endpoint (Table 3) (p for interaction = 0.84). A sensitivity analysis with a more stringent definition of discharge decongestion status (score <1 and inclusion of dyspnea score in the summed score) yielded comparable findings (p for interaction = 0.13).
The EVEREST trial provides a unique data resource with detailed information on in-hospital diuretic agent use, serial clinical assessments of congestion status at study entry and hospital discharge, and adjudicated clinical outcomes in a large contemporary cohort of patients hospitalized for worsening symptoms of HF. Two salient findings are evident from the current post hoc analysis of the EVEREST data: 1) higher average diuretic agent doses during the first 3 days of hospitalization were not associated with increased risk for the 30-day combined endpoint of all-cause mortality or rehospitalization for HF when adjusting for other known covariates of disease severity, and 2) the relationship between short-term diuretic agent exposure and 30-day combined outcomes did not vary according to clinically determined congestion status at hospital discharge.
Consensus guideline statements recommend titration of diuretic agent dose to achieve euvolemia in patients with HF and signs and symptoms of congestion in both ambulatory and acute care settings (2). This recommendation is based on expert opinion, because there are few prospective controlled trials to support the clinical benefits and safety of diuretic agent use for optimization of volume status in patients with HF (20). The impetus for the current post hoc analysis was the ongoing concern that diuretic agents used for short-term relief of congestive signs and symptoms during hospitalization for worsening HF might also be conferring increased risk for adverse outcomes after hospital discharge. Prior observational studies have reported an association between increased diuretic agent dose and increased mortality risk in patients hospitalized with worsening HF symptoms (8,9). In these reports, increased diuretic agent dose was associated with markers of increased disease severity, but diuretic agent dose remained an independent predictor of mortality in multivariate models with adjustment for potential confounding factors. In the EVEREST study population, we found that diuretic agent dose was associated with markers of increased disease severity and all-cause mortality or HF rehospitalization at 30 days in univariate analysis, but not in the multivariate model with adjustment for numerous known prognostic covariates used in previous analyses of the EVEREST database. Our multivariate analysis indicates that short-term exposure to higher diuretic agent doses during hospitalization is a marker of more severe disease and is not independently associated with greater risk for adverse outcomes. The discrepancy between the results of our multivariate analysis and those of past studies could be attributable to differences in the methods used to calculate diuretic agent exposure, differences in follow-up duration, and the number of potential confounder variables incorporated into multivariate models. Differences in international patterns of diuretic agent use and global regional heterogeneity in health care systems, standards of care, and event rates in the EVEREST population may have also contributed to the divergent findings of the present analysis versus prior studies of U.S. populations (21,22). The results of our multivariate analysis are concordant with a report from the ALARM-HF (Acute Heart Failure Global Survey of Standard Treatment) international registry, which reported no association between in-hospital diuretic agent dose and 30-day mortality in a propensity-matched analysis (23). The lack of association between diuretic agent dose and mortality or HF rehospitalizations in our multivariate analysis is also consistent with the findings of the prospective DOSE (Diuretic Optimization Strategies Evaluation) trial, which reported no difference in the risk for post–hospital discharge adverse events in hospitalized patients with HF randomly assigned to lower versus higher doses of diuretic agents (24).
To our knowledge, the present post hoc analysis of the EVEREST database is the first to specifically test the hypothesis that the relationship between short-term diuretic agent exposure during HF hospitalization and 30-day clinical outcomes might differ according to the clinically assessed congestion status at hospital discharge. Our findings indicate that the relationship between diuretic agent dose and outcomes is not modified by the presence or absence of persistent congestion on the basis of clinical assessment at hospital discharge. Two prior post hoc analyses of the placebo arm of the EVEREST cohort reported on the association between measures of discharge congestion status (on the basis of a composite clinical congestion score and the presence of hemoconcentration) and mortality, but neither of these reports provided information on diuretic agent dose (15,18). Our finding of lack of effect modification on the basis of clinical assessment of discharge congestion status is consistent with prior reports demonstrating that physical assessment underestimates the degree of intravascular volume overload in patients with HF (25–29). The empirically derived congestion score used in this analysis was adapted from a previous EVEREST post hoc analysis, but the optimal clinical application of the score or its individual components has not been determined. Persistent clinical or subclinical volume overload is frequently present at hospital discharge and may contribute to the high rate of post–hospital discharge events (25). It is also possible that our negative finding could be attributable to beta error, as the power for detection of a significant interaction is limited by the sample size and event rates. Future trials to compare the clinical utility of physical assessment versus alternate measures for assessment and management of optimal volume status (hemoconcentration, direct blood volume measurement, thoracic or total body water content, and/or data derived from implantable hemodynamic monitors) are needed to define the therapeutic strategies associated with best clinical outcomes (25,30,31).
Strengths of the present analysis include the detailed characterization of loop diuretic agent exposure during hospitalization, extensive characterization of clinical and biochemical markers of disease severity for inclusion in the multivariate analysis, and prospectively adjudicated clinical outcomes.
There are several caveats to consider in the interpretation of the presented data. Our analysis is based on short-term exposure to diuretic agents in a hospital setting and therefore is not generalizable to long-term diuretic agent use in ambulatory HF populations. Diuretic agent dose information was available in approximately 75% of the subjects enrolled in the EVEREST study. Although the demographics of our subset with diuretic agent dose information were comparable with those of the overall EVEREST study population, selection bias in the study sample and the entry criteria for EVEREST may limit the generalizability of our findings. Renal function was relatively well preserved in the EVEREST study population, so the relationship between diuretic agent dose and outcomes in patients with more severe comorbid chronic kidney disease cannot be determined from our analysis. Moreover, our analysis did not include information on change in renal function, electrolytes, or other blood biomarkers, so it does not provide insight into potential mechanisms linking diuretic agent dose and outcomes. Investigator assessment of congestion was not centrally adjudicated in the EVEREST study and therefore may be subject to interobserver variability. However, the fidelity of the reported signs of congestion in the EVEREST database is supported by the observation that subjects enrolled in EVEREST with hemoconcentration, a proposed biomarker of effective decongestion therapy, had decreased signs and symptoms of congestion compared with subjects without hemoconcentration (15). Our outcomes analysis was limited to 30 days post-discharge, as we considered this interval to be a biologically plausible time span for assessment of the health effects of short-term diuretic agent exposure and also a clinically relevant time span for hospital-based treatment programs targeting improved post-discharge outcomes. Post-discharge diuretic agent dose and other post-discharge medical therapies may have influenced the 30-day outcomes. Finally, the post hoc nature of the analysis does not permit causal inference related to the observed associations, as unmeasured confounders related to the complex medical decision-making process for determination of diuretic agent dose might have contributed to our findings. Further work is needed to characterize the association between diuretic agent exposure during hospitalization, congestion status at discharge, and post-discharge treatment on longer term clinical outcomes and health care resources utilization.
This post hoc analysis of the EVEREST study database demonstrates that higher short-term diuretic agent exposure during hospitalization for worsening HF is not independently associated with risk for the combined endpoint of all-cause mortality and HF rehospitalization over 30 days. Clinically determined congestion status at hospital discharge did not modify the association between average diuretic agent dose and 30-day outcomes. These findings support current consensus guideline recommendations. Additional work is needed to define the optimal techniques for assessment of euvolemia and determine the best strategies to improve post-discharge outcomes in patients hospitalized with worsening symptoms.
COMPETENCY IN MEDICAL KNOWLEDGE: Loop diuretic agents are commonly used to reduce congestive signs and symptoms in patients hospitalized for worsening HF, but limited data are available to guide clinicians with regard to the selection of the optimal loop diuretic agent dose, assessment of the clinical response to diuretic therapy, and safety of loop diuretic agents in clinical practice.
TRANSLATIONAL OUTLOOK: Our analysis demonstrates that use of high-dose loop diuretic agents is a marker of more advanced HF but not an independent predictor of adverse outcomes in patients hospitalized with worsening HF. These data are in agreement with the findings of a randomized study of lower dose versus higher dose loop diuretic agents in hospitalized patients with HF and support current HF treatment consensus guideline recommendations to adjust the dose of loop diuretic agents as needed to achieve euvolemia.
The EVEREST program was funded by Otsuka, and database management was provided by the trial sponsor. This work was supported in part by grant NIH/NCATS UL1 TR000038 from the National Center for Research Resources, National Institutes of Health. Dr. Katz has served as a member of the speakers bureau for Otsuka America Pharmaceuticals. Dr. Konstam has received research support from Otsuka America Pharmaceuticals. Dr. Gheorghiade has served as a consultant for Otsuka America Pharmaceuticals. All other authors have reported that they have no relationships relevant to the contents of this paper to disclose.
- Abbreviations and Acronyms
- confidence interval
- heart failure
- hazard ratio
- Received November 3, 2015.
- Revision received February 17, 2016.
- Accepted February 19, 2016.
- 2016 American College of Cardiology Foundation
- Yancy C.W.,
- Lopatin M.,
- Stevenson L.W.,
- et al.
- Yancy C.W.,
- Jessup M.,
- Bozkurt B.,
- et al.
- McMurray J.J.,
- Adamopoulos S.,
- Anker S.D.,
- et al.
- Testani J.M.,
- Chen J.,
- McCauley B.D.,
- Kimmel S.E.,
- Shannon R.P.
- Gheorghiade M.,
- Orlandi C.,
- Burnett J.C.,
- et al.
- Ambrosy A.P.,
- Pang P.S.,
- Khan S.,
- et al.
- Blair J.E.,
- Zannad F.,
- Konstam M.A.,
- et al.
- O’Connor C.M.,
- Fiuzat M.,
- Swedberg K.,
- et al.
- Butman S.M.,
- Ewy G.A.,
- Standen J.R.,
- Kern K.B.,
- Hahn E.
- Gheorghiade M.,
- Follath F.,
- Ponikowski P.,
- et al.
- Miller W.L.,
- Mullan B.P.
- Packer M.,
- Abraham W.T.,
- Mehra M.R.,
- et al.