OUP user menu

Weight changes after hospitalization for worsening heart failure and subsequent re-hospitalization and mortality in the EVEREST trial

John E.A. Blair, Sadiya Khan, Marvin A. Konstam, Karl Swedberg, Faiez Zannad, John C. Burnett Jr, Liliana Grinfeld, Aldo P. Maggioni, James E. Udelson, Christopher A. Zimmer, John Ouyang, Chien-Feng Chen, Mihai Gheorghiade
DOI: http://dx.doi.org/10.1093/eurheartj/ehp144 1666-1673 First published online: 2 May 2009

Abstract

Aims Increases in body weight (BW) are important determinants for hospitalization in ambulatory patients with heart failure (HF), but have not yet been explored in patients hospitalized for worsening HF. We explore the relationship between change in BW after hospitalization for worsening HF and risk for repeat hospitalization and mortality in the EVEREST trial.

Methods and results The EVEREST trial randomized 4133 patients hospitalized for worsening HF and low ejection fraction (≤40%) to tolvaptan, a vasopressin antagonist, or placebo. Following discharge, BW was assessed at 1, 4, and 8 weeks, and every 8 weeks thereafter. A time-dependent Cox proportional Hazard model explored the relationship between change in BW at 60, 120, and 180 days from discharge and the risks of HF hospitalization, cardiovascular (CV) hospitalization, and all-cause mortality. For subjects re-hospitalized for heart failure at 60, 120, and 180 days after discharge, mean BW increase prior to the event was 1.96, 2.07, and 1.97 kg, respectively, compared with 0.74, 0.90, and 1.04 kg in patients without re-hospitalization (P < 0.001 all groups). A similar pattern was observed with CV hospitalization. However, increases in BW were not predictive of all-cause mortality.

Conclusion Increases in BW after hospitalization for worsening HF was predictive of repeat hospitalization events, but not mortality in the post-discharge period.

  • Heart failure hospitalizations
  • Body weight
  • Outcomes

Introduction

Acute heart failure syndromes (AHFS), defined as rapid development of signs and symptoms of heart failure (HF) are a major public health problem, accounting for over 1.1 million hospitalizations, or 390 per 100 000 people annually in the USA alone.1,2 Although existing therapies for AHFS, such as vasodilators and diuretics, effectively relieve signs and symptoms during the hospitalization, high mortality (10%) and re-hospitalization (30%) rates remain in the first 60–90 days post-discharge.3 In these patients, congestion or fluid overload due to elevated left ventricular (LV) filling pressures manifests as dyspnoea, jugular venous distension (JVD), and/or oedema. Congestion appears to be the main reason for admission and re-admission for HF,4 and is often reflected by an increase in body weight (BW).57

The correlation between changes in BW and outcomes has not been well-studied in a prospective manner with systematic measurement of BW in the post-discharge period after hospitalization for worsening HF. The purpose of this analysis is to explore the role of increases in BW that occurs after hospitalization for worsening HF to predict repeat hospitalization and mortality.

Methods

Patients

The design of the EVEREST programme has been reported previously.810 This was an international, multicentre, randomized, double-blind, placebo-controlled trial that examined the efficacy and safety of tolvaptan when added to optimal medical therapy in patients hospitalized for worsening HF. Two identical trials examining the short-term efficacy of clinical signs and symptoms during the inpatient period (trials A and B) were embedded into a long-term outcome study.

Patients ≥18 years of age, with left ventricular ejection fraction (LVEF)≤40%, hospitalized for worsened HF, with two or more signs of fluid overload were considered and randomized within 48 h of hospitalization to oral tolvaptan (30 mg/day) or matching placebo, in addition to standard HF medical therapies. Exclusion criteria were cardiac surgery within 60 days of enrolment, cardiac mechanical support, biventricular pacemaker placement within 60 days, expected survival of less than 6 months, acute myocardial infarction at the time of hospitalization, haemodynamically significant uncorrected cardiac valvular disease, end-stage HF, dialysis, systolic blood pressure (SBP) <90 mmHg, serum creatinine >3.5 mg/dL, serum potassium >5.5 mEq/L, haemoglobin <9 g/dL, and comorbid conditions with <6 months life-expectancy. In addition, all patients continued to receive standard therapy for HF, at the discretion of the treating physician.

Patient follow-up

Between 7 October 2003 and 3 February 2006, a total of 4133 patients were randomized, with follow-up until 6 July 2006 for a median follow-up period of 9.9 months per patient. All patients discharged had follow-up clinic visits at weeks 1, 4, 8, and every 8 weeks thereafter. The first outpatient visit occurred 7 days after discharge for those subjects discharged on or before Day 10 or on Day 17 after randomization for those still in the hospital on Day 10. BW was measured per protocol at baseline, on all inpatient days, and on all follow-up clinic visits using a standardized scale.

Endpoint assessment

The primary long-term efficacy endpoints for the EVEREST trial were time to all-cause mortality and the combined endpoint of cardiovascular (CV) mortality or HF hospitalization. Data on all CV hospitalizations were also gathered as part of secondary endpoints. For the current analysis, endpoints measured were all-cause mortality, HF hospitalization, CV hospitalization, and the combined endpoint of all-cause mortality and CV hospitalization. These events were measured cumulatively at specific time points (60, 120, and 180 days). An independent events committee adjudicated all endpoints.

Statistical analysis

Post hoc analysis was performed on patients with and without events up to 180 days. Baseline demographics, physical and laboratory findings, medical history, and medical revascularization, and device therapies were compared between patients with and without HF hospitalization using the Kruskal–Wallis test for continuous variables and the Pearson χ2 test for categorical variables. The primary analysis was a comparison of BW (kg) between patients without events to those with events at the above time points. Four events (HF hospitalization, all-cause mortality, CV hospitalization, and all-cause mortality/CV hospitalization) at three time points (60, 120, and 180 days) were studied. Since BW did not remain constant in the post-discharge period, the change from discharge or inpatient Day 7 to the last available BW prior to an event was used for patients with events, while the mean BW change after discharge or inpatient Day 7 was used for patients without events. The BW changes within each group were summarized, and the means and standard deviations were reported. Change in BW from Day 7 or discharge to the BW immediately prior to the clinic visit leading to the event (remote clinic visit), if available, was also compared with the mean change in BW of the group without events. Test characteristics for the use of body weight to predict HF hospitalizations were calculated using standard sensitivity and specificity analysis. A Cox proportional Hazards model was used to study the change in weight from discharge (as a time-dependent variable) and the risks of HF and CV hospitalization and all-cause mortality, accounting for the following baseline covariates: age, SBP, LVEF, NYHA class, blood urea nitrogen (BUN), Cr, sodium, brain natriuretic peptide (BNP), N-terminal pro-BNP, chronic kidney disease, β-blocker, angiotensin-converting enzyme (ACE)-inhibitor, angiotensin-receptor blockers (ARB), and aldosterone blocker use. These variables were obtained by selecting the variables that had a P < 0.05 for predicting overall mortality and those that were clinically relevant.

The sponsor performed database management according to pre-specified plan, and conducted all final analyses using SAS software, version 8.2 (SAS Institute Inc., Cary, NC, USA). All authors had full access to the data, take responsibility for its integrity, and have read and agree to the manuscript as written.

Results

Of the 4133 patients enrolled in EVEREST, 367 were excluded due to missing BW data at baseline or at discharge and two were excluded because of death at or prior to discharge, leaving 3764 patients for analysis. The first clinic visit occurred at a mean of 6.6 ± 4.7 days after discharge, however, 183 (5.0%) patients were still hospitalized at the scheduled week 1 evaluation and included in the analysis.

Baseline characteristics

Patient characteristics at randomization were analysed based on presence or absence of HF hospitalization within 180 days of discharge and are summarized in Table 1. Compared with patients without HF hospitalizations, patients with HF hospitalizations were more likely to be from North America, have lower blood pressure, higher NYHA class, more jugular venous distention and less pedal oedema, lower ejection fraction, and no difference in BW at randomization. Patients who were subsequently hospitalized also had greater impairment in renal function (as measured by BUN and creatinine), lower serum sodium and higher serum BNP, and a longer QRS (≥120 ms). Comorbidities, such as diabetes and chronic kidney disease, as well as prior device and revascularization procedures were more prevalent in patients hospitalized compared with those without hospitalization at 180 days, however the presence of coronary artery disease was similar in both groups. Medication use at randomization was similar between the two groups, with the exception of lower use of ACE-inhibitors and ARBs, higher use of lipid-lowering agents, and vasoactive medications in patients with HF hospitalizations. A similar pattern was noted at discharge, however among patients who were not re-hospitalized, spironolactone use was higher (Table 2).

View this table:
Table 1

Baseline characteristics

No HF hospitalizationHF hospitalizationP-value*
Demographics
Number of patients (%)2895 (76.9)869 (23.1)
Region, no. (%)
 North America740 (25.6)348 (40.1)<0.001
 South America499 (17.2)152 (17.5)
 Western Europe340 (11.7)130 (15.0)
 Eastern Europe1316 (45.5)239 (27.5)
Age, mean (SD), years65.50 (11.6)65.87 (12.1)0.42
Men, no. (%)2151 (74.3)664 (76.4)0.21
Ethnicity, no. (%)
 White2513 (86.8)704 (81.0)<0.001
 Black173 (6.0)100 (11.5)
 Other209 (7.2)65 (7.5)
Weight, mean (SD), kg83.36 (18.6)83.16 (18.9)0.78
Physical and laboratory findings
Dyspnoea, no. (%)2630 (91.0)799 (92.1)0.35
Systolic BP, mean (SD), mmHg122 (20)116 (18)<0.001
Diastolic BP, mean (SD), mmHg74 (13)70 (12)<0.001
Heart rate, mean (SD), b.p.m.80 (16)79 (14)0.05
NYHA class, no. (%)
 III1794 (62.4)462 (53.2)<0.001
 IV1081 (37.6)407 (46.8)
JVD ≥10 cm, no. (%)738 (25.6)257 (30.1)0.01
Pedal oedema, no. (%)1733 (75.4)454 (69.0)<0.001
Ejection fraction, mean (SD), %28.3 (7.9)25.58 (8.1)<0.001
Serum BUN, mean (SD), mg/dL27.9 (14.2)34.38 (17.52)<0.001
Serum creatinine, mean (SD), mg/dL1.31 (0.47)1.48 (0.49)<0.001
Serum sodium, mean (SD), mEq/L140.0 (4.4)138.7 (4.7)<0.001
Serum BNP, median (IQR), pg/mL567.0 (246.0–1294.0)1018.4 (480.9–1895.8)<0.001
QRS ≥120 ms, no. (%)1611 (55.7)601 (69.2)<0.001
Medical history, no. (%)
 Previous hospitalization for HF2203 (76.4)756 (87.1)<0.001
 CAD2023 (69.9)619 (71.2)0.45
 Previous MI1423 (49.2)466 (53.6)0.02
 Hypertension2090 (72.2)591 (68.0)0.02
 Hypercholesterolaemia1348 (46.8)469 (54.3)<0.001
 Mitral valve disease842 (29.2)327 (37.7)<0.001
 Atrial arrhythmia1443 (50.1)485 (56.0)0.002
 Diabetes1045 (36.1)394 (45.3)<0.001
 Chronic kidney disease620 (21.4)344 (39.6)<0.001
 Severe COPD260 (9.0)102 (11.7)0.02
Revascularization and device use, no. (%)
 Previous PCI449 (15.5)219 (25.2)<0.001
 Previous CABG547 (18.9)225 (25.9)<0.001
 No ICD/Pacemaker2494 (86.2)647 (74.5)<0.001
 Pacemaker401 (13.9)222 (25.6)<0.001
 ICD325 (11.2)201 (23.1)<0.001
  • SD, standard deviation; IQR, interquartile range; BP, blood pressure; NYHA, New York Heart Association; JVD, jugular venous distention; BUN, blood urea nitrogen; HF, heart failure; MI, myocardial infarction; COPD, chronic obstructive pulmonary disease; PCI, percutaneous coronary intervention; ICD, implantable cardioverter-defibrillator; CAD, coronary artery disease.

  • *P-value assessed by Kruskal–Wallis test for continuous variables and the Pearson's χ2 test for categorical variables.

View this table:
Table 2

Medications

No HF hospitalizationHF hospitalizationP-value
At randomization (%)n = 2895n = 869
 ACE-I/ARBs86.179.9<0.001
 Beta-blocking agents68.466.70.36
 Spironolactone57.455.60.35
 Digoxin44.847.00.26
 Diuretics97.097.70.30
 Lipid-lowering agents34.441.7<0.001
 Nitroglycerin15.119.30.003
 Amiodarone14.820.3<0.001
 Inotropic agents4.06.9<0.001
 IV nitroglycerin5.46.20.35
 Nesiritide5.710.0<0.001
Discharge (%)n = 2895n = 869
 ACE-I/ARBs87.080.4<0.001
 Beta-blocking agents75.974.20.32
 Spironolactone60.156.40.05
 Digoxin45.648.80.10
 Diuretics93.294.70.12
 Lipid-lowering agents36.843.00.001
 Nitroglycerin8.912.9<0.001
 Amiodarone15.921.7<0.001
 Inotropic agents1.02.20.008
 IV nitroglycerin0.10.11.000
 Nesiritide1.73.50.003
  • HF, heart failure; ACE-I, angiotensin converting enzyme-inhibitor; ARB, angiotensin receptor-blocker; IV, intravenous.

Univariate analysis

At all three time points, mean BW increased for all patients. However, in patients re-admitted with HF, the change in BW post-discharge measured at the visit prior to admission was significantly more than the mean change in BW post-discharge in patients without HF admissions at all three time points (1.96 vs. 0.74 kg, 2.07 vs. 0.90 kg, and 1.97 vs. 1.04 kg at 60, 120, and 180 days, respectively; P < 0.001 in all groups). However, at a remote clinic visit prior to the event, BW was not significantly increased in patients with HF hospitalization compared to those without at all three time points (Figure 1). A similar pattern was found for CV hospitalization and the composite endpoint of all-cause mortality and CV hospitalization. In patients who died post-discharge, there was no significant difference in the change in BW leading to the event or remote from the event compared to patients without events. Test characteristics of changes in BW for prediction of HF hospitalization were calculated (Table 3). In general, the larger the change in BW, the greater the specificity for predicting re-hospitalization. For example, specificity is 71% for predicting 60-day HF hospitalization when weight gain is ≥2 kg, and increases to 89% when the weight gain is ≥4 kg.

Figure 1

Body weight (BW) changes and outcomes. Post-discharge changes in BW are plotted as the mean BW in patients without events, the mean BW at the clinic visit before an event, and at a remote clinic visit prior to the event in patients with events at 60, 120, and 180 days. (A) Represents hospitalization for heart failure (HF). (B) All-cause mortality. (C) All-cause mortality and cardiovascular hospitalization, and (D) cardiovascular hospitalization. Mean values are plotted with standard error from the mean. P-values are a comparison of changes in BW between patients with no event and the visit prior to the event.

View this table:
Table 3

Analysis of test characteristics of body weight (BW) changes

BW change (kg)Number of hospitalizationsSensitivity (%)Specificity (%)PPV (%)NPV (%)
60 day474
 ≥0.557.448.713.988.8
 ≥1.054.056.415.289.5
 ≥1.546.064.915.989.3
 ≥2.040.370.716.589.1
 ≥2.534.477.318.089.1
 ≥3.030.681.619.489.1
 ≥3.523.686.219.888.7
 ≥4.020.989.121.688.7
120 day708
 ≥0.560.744.220.182.9
 ≥1.055.549.820.482.8
 ≥1.548.757.421.082.9
 ≥2.042.562.620.882.4
 ≥2.537.669.422.282.8
 ≥3.034.274.323.583.0
 ≥3.527.579.623.882.6
 ≥4.024.782.424.582.5
180 day869
 ≥0.561.743.424.779.0
 ≥1.056.547.324.478.4
 ≥1.550.353.824.678.3
 ≥2.044.058.023.977.5
 ≥2.538.164.024.177.5
 ≥3.033.968.524.477.5
 ≥3.528.173.424.077.3
 ≥4.025.477.125.077.5
  • PPV, positive predictive value; NPV, negative predictive value.

The median (inter-quartile range) number of days from the prior clinic visit to an event was 9 (3–16), 12 (4–23), and 14 (5–30) days for hospitalizations at 60, 120, and 180 days, respectively. The median number of days from the remote visit to the event was much longer (Figure 2).

Figure 2

Median times from clinic visits to heart failure (HF) hospitalization events. Boxes represent the median (inter-quartile range), and bars represent the maximum and minimum days from the visit to HF hospitalization for (A) the visit immediately prior to hospitalization, and (B) the visit remote to the hospitalization at 60, 120, and 180 days. For ranges with outliers, the bars represented 25th percentile − 1.5× IQR and 75th percentile + 1.5 × IQR.

Multivariate analysis

After adjusting for important covariates, increases in BW remained predictive of re-admission for worsening HF at 60, 120, and 180 days, with a hazard ratio (HR), (95% confidence interval) of 1.072 (1.041–1.103), 1.057 (1.033–1.081), and 1.036 (1.014–1.057) per 1 kg increase, respectively. Although BW increases also remained predictive of CV hospitalization, and combined all-cause mortality/CV hospitalization, increases in BW did not predict all-cause mortality (Figure 3).

Figure 3

Multivariate analysis. Hazard ratios (HR) and 95% confidence intervals are plotted at 60, 120, and 180 days for heart failure (HF) hospitalization, CV hospitalization, combined all-cause mortality/cardiovascular hospitalization, and all-cause mortality, for every 1 kg gain in body weight (BW). The analysis was adjusted for age, systolic blood pressure, left ventricular ejection fraction, NYHA class, blood urea nitrogen, Cr, sodium, brain natriuretic peptide (BNP), N-terminal proBNP, chronic kidney disease, β-blocker, angiotensin-converting enzyme-inhibitors, angiotensin receptor-blockers, and aldosterone blocker use at baseline.

Discussion

The current study demonstrates that acute increases in BW can predict re-hospitalization (but not mortality) days before the event in patients recently admitted for worsening HF and are at high risk for re-hospitalization and death.

One case–control study in ambulatory HF patients demonstrated that gradual elevations in BW as early as 2 weeks prior to hospitalization are strongly associated with the admission for HF.6 A survey of patients hospitalized with worsening HF demonstrated that this weight gain was also associated with increased dyspnoea, orthopnoea, and oedema, which all preceded hospitalization by 1–2 weeks.7 A prospective study in outpatients with chronic HF determined that a weight gain of ≥2 kg over 48–72 h was 97% specific but only 9% sensitive for independently assessed clinical deterioration.11

Hospitalizations12 for HF are costly and impart a significant financial and social burden on health-care systems and are primarily related to congestion rather than a low cardiac output state.5,1316 LV dysfunction resulting in both high LV filling pressure and renal abnormalities contribute to fluid retention, resulting in signs and symptoms of HF such as rales, a third heart sound (S3), JVD, and peripheral oedema.4 Retrospective analysis of the Studies of Left Ventricular Dysfunction (SOLVD) trial demonstrated that the presence of JVD or S3 predicted hospitalization for HF in stable patients with congestive HF.17 Continuous monitoring of right ventricular (RV) haemodynamics in chronic HF patients demonstrated that HF hospitalization was preceded by a >20% increase in RV pressures in nine of 12 hospitalized chronic HF patients, occurring approximately 4 days prior to the onset of typical HF symptoms.18 Often, these clinical and haemodynamic changes correlate with the increase in BW observed in these patients prior to hospitalization.6,7 However, unlike these methods (JVD, S3, invasive haemodynamics), the measurement of BW is precise and accurate and can be obtained by the patient without special equipment or training.

It is important to note that although increases in BW in the immediate pre-hospitalization period predicts re-hospitalizations, a reduction in BW as a result of a specific intervention (e.g. diuretics, vasopressin antagonists) may not necessarily prevent re-hospitalizations.19,20 A pilot study in which 280 patients admitted with worsening HF and LVEF ≤ 35% showed no difference in 180-day hospital re-admission, in patients randomized to a home BW, and symptom monitoring system after discharge.21 Furthermore, in EVEREST, a reduction in BW post-discharge that likely reflected a reduction in intravascular volume was not associated with reductions in hospitalization or mortality.10

There are different modalities for fluid removal to reduce BW in HF, including loop diuretics,22,23 ultrafiltration,24 vasopressin antagonists,10 and adenosine-blocking agents.25 Although the optimal agent for fluid removal in AHFS is yet to be established, it is clear that the net effect of the therapeutic intervention on outcomes will depend on the mechanism of action (how it removes fluid) and the potential undesirable effects (e.g. activation of neurohormones, worsening renal function). The composition (e.g. sodium or water) of the fluid removed as well as the origin of the fluid (e.g. intravascular or interstitial) should also be considered. Diuretics are known to effectively remove fluid, reduce intravascular volume, and decrease BW, but in the process may worsen renal function and further activate neurohormones.22,23 It is possible that vasopressin antagonists, which induce free water excretion, may decrease BW by primarily reducing extravascular volume, and minimally reduce intravascular volume (PCWP).26

Increases in BW should be viewed as a marker for implementation and/or optimization of therapies that are known to improve LV function and outcomes in HF (e.g. β-blockers, cardiac resynchronization therapy), since the underlying mechanism behind BW increases in HF is LV dysfunction resulting in elevated filling pressures. In fact, titration of specific HF therapies guided by BNP levels, a more laborious and costly measurement of congestion, resulted in improved outcomes in selected groups.27,28

Limitations

This study was a post hoc analysis. However, BW data were systematically collected in 90% of patients enrolled in the trial, and there was adequate follow-up with independent adjudication of events. In addition, potential confounders were accounted for in the multivariate analysis. Another limitation was that BW was not measured on a daily basis post-discharge. This, however, represents a more real-world situation of patient management after discharge for AHFS and provides relevant clinical information for potential intervention in this time period. Finally, the EVEREST trial was a highly selected population of patients hospitalized for worsened chronic HF and EF ≤40%, and had relatively few comorbidities. This limits the generalizability to the entire population of patients with acute HF syndromes.

Conclusion

Increases in BW in the post-discharge period are a major predictor for repeat hospitalization for HF but not for mortality, and should be followed closely after hospitalization for worsening HF. However, a decrease in BW cannot be used as an indiscriminate target for reducing hospitalization events. Improvement in outcomes as a result of BW reduction will depend on the mechanism utilized to remove fluid, and will need to balance potential adverse effects of the therapeutic intervention.

Funding

The EVEREST programme was funded by Otsuka.

Conflict of interest: J.B. is a consultant for Debiopharm. K.S. has received research grants from AstraZeneca, Servier, and Amgen; is a consultant for Cytokinetics, Servier, and Novartis; and has received honoraria from AstraZeneca, Otsuka, Amgen, and Servier. M.K. has received research grants and contracts from, is a consultant for, and has received honoraria from Otsuka. F.Z. has received research grants from Bayer; is a consultant for Servier and Johnson & Johnson; and has received honoraria from AstraZeneca, Pfizer, Boehringer Ingelheim, Novartis, Abbott, sanofi-aventis, and Otsuka. J.B. has received research grants from the National Institutes of Health, Microbia, and Theravance; is a consultant for Abbott, Bayer, Otsuka, Wyeth, and Astellas; and has received honoraria from Scios, Otsuka, and Orqis. L.G. has received research grants from GlaxoSmithKline, Otsuka, Amgen, and Bristol; is a consultant for Cordis; and has received honoraria from GlaxoSmithKline, Otsuka, Cordis, Amgen, and Bristol. A.M. has received research grants from the National Institutes of Health, Italian Ministry of Health, AstraZeneca, Novartis, Pfizer, Takeda, Società Prodotti Antibiotici, Sigma Tau, sanofi-aventis, and GiennePharma; is a consultant for Novartis and Daiichi Sankyo; and has received honoraria from AstraZeneca, Novartis, Takeda, Società Prodotti Antibiotici, Sigma Tau, sanofi-aventis, Servier, and Otsuka. J.U. has been a consultant for and received research grants and honoraria from Otsuka. C. Zimmer, J.O., and C.C. are employees of Otsuka. M.G. has received research grants from the National Institutes of Health, Otsuka, Sigma Tau, Merck, and Scios Inc; is a consultant for Debbio Pharm, Errekappa Terapeutici, GlaxoSmithKline, Protein Design Laboratories, and Medtronic; and has received honoraria from Abbott, AstraZeneca, GlaxoSmithKline, Medtronic, Otsuka, Protein Design Laboratories, Scios Inc, and Sigma Tau.

Footnotes

  • Presented at the 2008 European Society of Cardiology Congress, and summarized by Dr Helmut Drexler in the highlight session ‘Heart failure and cardiomyopathies.’

References

View Abstract