Predicted probability of freedom from death and death or heart failure (HF) rehospitalization across levels of sodium after adjusting for important covariates. Plots are for the “average” patient using the mean values of all covariates. Lighter solid and dotted line pairs represent 95% confidence intervals. To convert serum sodium to millimoles per liter, multiply by 1.0.
Relationship between clinical events and patients with persistent hyponatremia, corrected hyponatremia, or normonatremia. Error brackets indicate exact binomial 95% confidence intervals; HF, heart failure.
Customize your JAMA Network experience by selecting one or more topics from the list below.
Gheorghiade M, Rossi JS, Cotts W, Shin DD, Hellkamp AS, Piña IL, Fonarow GC, DeMarco T, Pauly DF, Rogers J, DiSalvo TG, Butler J, Hare JM, Francis GS, Stough WG, O’Connor CM. Characterization and Prognostic Value of Persistent Hyponatremia in Patients With Severe Heart Failure in the ESCAPE Trial. Arch Intern Med. 2007;167(18):1998–2005. doi:10.1001/archinte.167.18.1998
Copyright 2007 American Medical Association. All Rights Reserved. Applicable FARS/DFARS Restrictions Apply to Government Use.2007
Mild hyponatremia is relatively common in patients hospitalized with heart failure (HF). To our knowledge, the association of hyponatremia with outcomes has not been evaluated in the context of in-hospital clinical course including central hemodynamics and changes in serum sodium level.
The ESCAPE trial (Evaluation Study of Congestive Heart Failure and Pulmonary Artery Catheterization Effectiveness) was a randomized, controlled study designed to evaluate the utility of a pulmonary artery catheter plus clinical assessment vs clinical assessment alone in guiding therapy in patients hospitalized with New York Heart Association class IV HF due to systolic dysfunction (left ventricular ejection fraction <30%). A Cox proportional hazards model with baseline serum sodium level as a continuous variable was used to examine the association of serum sodium level with 6-month postdischarge mortality, HF rehospitalization, and death or rehospitalization. A categorical analysis was also performed comparing persistent and corrected hyponatremia.
A total of 433 hospitalized patients with HF were enrolled in ESCAPE. Hyponatremia (serum sodium level ≤ 134 mEq/L) was present in 103 patients (23.8%). (To convert serum sodium to millimoles per liter, multiply by 1.0.) Of these, 71 had persistent hyponatremia (68.9%). Hyponatremia was associated with higher 6-month mortality after covariate adjustment (hazard ratio [HR] for each 3-mEq/L decrease in sodium level, 1.23; 95% confidence interval [CI], 1.05-1.43) (P = .01). After controlling for baseline variables and clinical response, we found that patients with persistent hyponatremia had an increased risk of all-cause mortality (31% vs 16%; HR, 1.82) (P = .04), HF rehospitalization (62% vs 43%; HR, 1.52) (P = .03), and death or rehospitalization (73% vs 50%; HR, 1.54) (P = .01) compared with normonatremic patients.
Persistent hyponatremia was an independent predictor of mortality, HF hospitalization, and death or rehospitalization despite clinical and hemodynamic improvements that were similar to those in patients without hyponatremia.
Patients with worsening heart failure resulting in hospitalization have a poor prognosis. In-hospital mortality occurs in 3% to 4% of these patients, and 60- to 90-day postdischarge mortality and rehospitalization rates have been reported to be 35% to 50%.1- 5 Hyponatremia, defined as a serum sodium concentration of 134 mEq/L or lower, is relatively common in patients hospitalized for worsening heart failure. (To convert serum sodium to millimoles per liter, multiply by 1.0.) In the Outcomes of a Prospective Trial of Intravenous Milrinone for Exacerbations of Chronic Heart Failure (OPTIME-CHF) trial,2 27% of the patients had hyponatremia at baseline. Similarly, in the Acute and Chronic Therapeutic Impact of a Vasopressin Antagonist in Congestive Heart Failure (ACTIV in CHF) study,6 21% had hyponatremia at baseline. Among 47 647 patients enrolled in the Organized Program to Initiate Life Saving Treatment in Patients Hospitalized for Heart Failure (OPTIMIZE-HF) registry,4 25.3% had hyponatremia at admission.
Although previous analyses have suggested that baseline hyponatremia is associated with a poor prognosis, they did not account for postbaseline changes in serum sodium concentration or the patient's response to therapy during the hospitalization.7 The Evaluation Study of Congestive Heart Failure and Pulmonary Artery Catheterization Effectiveness (ESCAPE) trial8,9 database provides a unique opportunity to evaluate the prognostic value of serum sodium concentration in the context of the patient's clinical course and changes in serum sodium concentration during hospitalization.
The study design and the primary results of the ESCAPE trial have been published.8,9 The ESCAPE protocol was reviewed and approved by an institutional review board at each participating center, and written informed consent was obtained from study participants. The ESCAPE trial was designed to evaluate the utility of the pulmonary artery catheter in patients hospitalized with New York Heart Association class IV heart failure.9 Study participants were required to have systolic dysfunction (left ventricular ejection fraction <30%), at least 1 symptom of congestion (eg, orthopnea, severe fatigue with minimal exertion, or discomfort from edema or anasarca), and 1 sign of fluid overload (eg, >10 cm of jugular venous distention, rales, or peripheral edema). Patients were randomized to receive therapy guided either by clinical assessment and a pulmonary artery catheter (n = 215) or by clinical assessment alone (n = 218). The outcomes of all-cause mortality, rehospitalization for heart failure, and a composite of death or heart failure rehospitalization at 6 months were evaluated in this analysis.
Serum sodium concentration was recorded at randomization, several times during the hospitalization, and at discharge. Cox proportional hazards models with baseline serum sodium concentration as a continuous variable were used to examine the association of serum sodium concentration with mortality, rehospitalization, and the composite of death or rehospitalization at 6 months. Each model included the covariates known to be significant predictors of outcome in the overall ESCAPE cohort. The model for mortality included age, baseline 6-minute walk distance, discharge serum urea nitrogen (SUN) level, and whether the patient received cardiopulmonary resuscitation or mechanical ventilation in the hospital. For heart failure rehospitalization and the composite end point, the model included baseline heart rate and hemoglobin level, discharge SUN level, discharge 6-minute walk distance, discharge β-blocker use, and whether the patient received cardiopulmonary resuscitation or mechanical ventilation in the hospital. In addition, each model included 3 covariates believed to be related to both serum sodium level and adverse outcomes: discharge systolic blood pressure, discharge diuretic dose, and weight change during the hospitalization. All models included the randomized therapy. These variables accounted for baseline factors as well as factors related to the patient's clinical response during the hospital course. The proportional hazards assumption was tested in each model using a sodium level × time interaction term; the assumption was met in all models. For results from the Cox models, P < .05 was considered significant.
To achieve a more thorough understanding of the association of serum sodium concentration with risk of adverse events, we also evaluated the relationship between changes in serum sodium concentration during the hospitalization and clinical outcomes. Patients were categorized with persistent hyponatremia if their serum sodium concentration values were 134 mEq/L or lower at baseline and throughout the hospital course. Patients with initial baseline serum sodium concentrations of 134 mEq/L or lower that subsequently increased to higher than 134 mEq/L during the hospitalization were classified as having corrected hyponatremia. Patients with no hyponatremia were those whose serum sodium concentrations were higher than 134 mEq/L on admission. Fifty-two patients with normal serum sodium levels developed hyponatremia during inpatient therapy in the ESCAPE trial. These patients were analyzed in the normonatremic group because baseline sodium level was thought to be a more reliable reflection of the overall clinical status, and inpatient development of hyponatremia may have resulted from subsequent therapies.
Baseline characteristics were compared between the persistent hyponatremia, corrected hyponatremia, and normonatremia groups using likelihood ratio χ2 tests for categorical variables and Wilcoxon rank sum tests for continuous variables. Because of the large number of tests performed in comparing baseline characteristics, a more stringent criterion for significance, P < .01, was used in these tests.
To assess the relationship of these serum sodium concentration categories with 6-month mortality, heart failure hospitalization, or the composite, Cox proportional hazards models were used in which the persistent hyponatremia group was compared with each of the other 2 groups. The models were otherwise unchanged.
The ESCAPE study randomized 433 patients hospitalized for decompensated heart failure. This analysis included 424 patients who had adequate serum sodium concentration data during the hospitalization and who survived to discharge. Six patients died during the index hospitalization but had adequate data to characterize their sodium level patterns. These patients were not included in the Cox models, but they were included in other summaries to the extent their data allowed. The remaining 3 patients were missing baseline serum sodium concentration data and were not included in this analysis.
Hyponatremia was present at baseline in 103 patients (23.8%), and it persisted during the hospitalization in 71 of 103 patients (68.9%). Compared with normonatremic patients, patients with persistent hyponatremia had lower baseline systolic blood pressure, higher baseline and last-measured SUN values, were more likely to be treated with spironolactone at baseline, and received larger doses of diuretics during hospitalization (Table 1). Patients whose serum sodium concentration corrected during the hospitalization had a greater decrease in SUN and a greater increase in heart rate than patients with persistent hyponatremia.
The response to in-hospital therapies was similar among the normonatremia, persistent hyponatremia, and corrected hyponatremia groups (Table 1). Changes in cardiac index, pulmonary capillary wedge pressure, and right atrial pressure measured in patients randomized to the pulmonary artery catheter group were not different among the 3 groups. Weight loss was similar, and the presence of heart failure signs at discharge including jugular venous pressure, rales, or edema was not different among the 3 groups. Improvements in heart failure symptoms and quality of life scores were also similar among the 3 groups.
After covariate adjustment, baseline serum sodium concentration was an independent predictor of 6-month mortality. The hazard ratio (HR) for mortality associated with each 3-mEq/L decrease in baseline serum sodium concentration was 1.23 (95% confidence interval [CI], 1.05-1.43) (P = .01) (Table 2). A trend was observed for the association of baseline serum sodium concentration with 6-month death or rehospitalization (HR, 1.09; 95% CI, 0.99-1.19), but it did not reach statistical significance (P = .07). No significant association was detected between baseline serum sodium concentration and 6-month rehospitalization alone (HR, 1.06; 95% CI, 0.96-1.17) (P = .29). The predicted probability of event-free survival based on the Cox model is shown in Figure 1. These data illustrate the linear relationship between baseline serum sodium and event-free survival.
Patients with persistent hyponatremia appeared to have higher rates of 6-month mortality, rehospitalization, and the composite of death or rehospitalization than patients with corrected hyponatremia or normonatremia (Table 2, Figure 2). Persistent hyponatremia was associated with a significantly higher risk of all 3 end points after covariate adjustment compared with normonatremia: 6-month mortality (HR, 1.82; 95% CI, 1.03-3.22) (P = .04); heart failure rehospitalization (HR, 1.52; 95% CI, 1.05-2.22) (P = .03); and the composite of death or rehospitalization (HR, 1.54; 95% CI, 1.09-2.17) (P = .01) (Table 2). In contrast, baseline hyponatremia was only associated with 6-month mortality but not rehospitalization alone or the composite end point.
The HRs for end point comparisons between patients with persistent hyponatremia and corrected hyponatremia suggested that those with persistent hyponatremia may be at higher risk of mortality, heart failure rehospitalization, and the combined end point. However, the total number of events in the corrected hyponatremia group was small, yielding wide CIs, and the finding was not statistically significant (Table 3).
Hyponatremia is a common disorder in patients hospitalized with worsening heart failure, and it is not corrected during the hospitalization in most patients. In the OPTIME-CHF trial,7 baseline hyponatremia was an independent predictor of mortality. However, the analysis did not adjust for factors reflective of the patient's clinical course, including changes in serum sodium level, and only baseline covariates were included in the model. Heart failure is a dynamic syndrome, and limiting covariates to baseline variables without considering clinical response to in-hospital therapies, serum sodium concentration changes during the hospitalization, or the influence of discharge medications likely leads to an incomplete evaluation of subsequent risk. In addition, the OPTIME-CHF trial follow-up was limited to 60 days, and so the ability of serum sodium concentration to predict risk beyond this time point could not be assessed.
In contrast, ESCAPE follow-up was continued for 180 days.8,9 This analysis of the ESCAPE data confirms previous analyses suggesting that baseline serum sodium level was an independent predictor of postdischarge mortality. A novel finding is that persistent hyponatremia yielded even more information in that it was not only associated with postdischarge mortality but also with heart failure rehospitalization and the composite of death or rehospitalization. Importantly, this analysis demonstrates for the first time that patients with persistent hyponatremia have a higher risk of events than normonatremic patients despite similar clinical improvements. Clinical course was assessed by including weight change, discharge blood pressure, discharge diuretic dose, discharge 6-minute walk distance, discharge SUN, and discharge β-blocker in the multivariate model. During the hospitalization, all patients appeared to improve based on weight loss, hemodynamic changes, and symptoms, regardless of baseline serum sodium concentration. Despite these improvements, persistent hyponatremia remained an independent predictor of death, heart failure rehospitalization, and the combined end point.
The cause of hyponatremia in the ESCAPE patients was likely related to hypervolemia, since all patients were volume overloaded at the time of admission.8,9 Increases in sympathetic stimulation and activation of the renin-angiotensin aldosterone system can cause renal vasoconstriction and sodium and water retention in patients with heart failure.11 Arginine vasopressin release promotes free water retention at the level of the renal collecting duct.12,13 During the hospitalization, some patients may have developed euvolemic or hypovolemic hyponatremia as a result of diuretic therapy.
The use of loop diuretics and spironolactone at discharge was higher in the patients with persistent hyponatremia, suggesting that prolonged diuretic use could have contributed to the persistent hyponatremic state. These observations may be related to the enhanced sodium excretion that occurs with the use of diuretics and spironolactone, or the higher rate of use of these drugs may reflect a greater severity of illness. It is also possible that increased diuretic doses may have resulted in increased free water intake in these patients, which may have contributed to hyponatremia. Although thiazide diuretics are most often implicated, nonthiazide agents such as furosemide and spironolactone also have been associated with hyponatremia.14
It has been demonstrated that patients with heart failure and hyponatremia have higher circulating levels of neurohormones (catecholamines, renin, angiotensin II, aldosterone, and vasopressin) than normonatremic patients.10,15- 19 In addition, patients with hyponatremia have impaired neurohormonal response to orthostasis, lower hepatic and renal plasma flows, elevated liver enzymes, and are more likely to have prerenal azotemia compared with normonatremic patients.16 Furthermore, compared with outpatients with stable disease, hospitalized patients with worsening heart failure have higher neurohormonal activation and more frequently have hyponatremia (>20% vs 7%).2,4,6,20 Low serum sodium concentration in the setting of congestion may therefore be a marker of increased neurohormonal activation and greater severity of disease.
It is also conceivable that low serum sodium concentration may play a pathogenic role in heart failure. For example, while hyponatremia is well known to exert its clinical effects on the brain, its direct pathophysiologic and clinical effects on cardiac myocytes are not known. Nor is it known if hyponatremia itself adversely affects the neurohormonal profile. Whether hyponatremia is a pathogenic factor (constituting a modifiable risk factor) or simply a marker of disease severity has an important therapeutic implication and warrants further study in a randomized clinical trial.
The finding that persistent hyponatremia is an independent predictor of risk for postdischarge adverse events suggests that it may contribute to the pathophysiology of this syndrome and be an appropriate therapeutic target. Investigational agents such as vasopressin receptor antagonists have been shown to normalize serum sodium levels in patients with hyponatremia.21,22 Data are now available from the Study of Ascending Levels of Tolvaptan in Hyponatremia (SALT-1 and SALT-2)22 that indicate sustained normalization of serum sodium concentration is possible with the vasopressin antagonist tolvaptan in patients with euvolemic or hypervolemic hyponatremia of different causes. However, it is unknown whether therapies directed at increasing serum sodium concentration will effect outcomes.
In the ACTIV in CHF study,6 no significant difference in 60-day outcomes was detected for the vasopressin receptor antagonist tolvaptan compared with placebo. In a post hoc analysis, 60-day mortality appeared to be lower in patients with corrected hyponatremia randomized to tolvaptan compared with placebo.23 However, in the recently published Efficacy of Vasopressin Antagonism in Heart Failure Outcome Study with Tolvaptan (EVEREST) trials,24- 26 it appears that correction of hyponatremia with tolvaptan (a vasopressin-2 antagonist) was not associated with improvement in clinical outcomes. Given that only 8% of the EVEREST study population had hyponatremia, the effects of tolvaptan in patients with heart failure and hyponatremia have yet to be determined.
Several limitations of our study should be considered. This is a post hoc analysis of a clinical trial database. Confounding factors that could have influenced the relationship between serum sodium concentration and outcomes may have been present that were not accounted for in this analysis. Second, the number of patients with persistent vs corrected hyponatremia was small, limiting our ability to draw conclusions from these data. However, the findings from the persistent and corrected hyponatremia comparison can be viewed as hypothesis generating. Finally, we can only detect associations, not causality from these data. Thus, we cannot determine whether hyponatremia is simply a marker of adverse outcome or if it has a specific effect on the pathophysiology of heart failure.
In conclusion, hyponatremia is a relatively common electrolyte disorder observed in patients hospitalized with heart failure, and it is often not corrected during hospitalization. In the ESCAPE trial,8,9 baseline hyponatremia was an independent predictor of 6-month mortality, while persistent hyponatremia predicted 6-month mortality as well as heart failure rehospitalization and the composite of death or rehospitalization. Persistent hyponatremia predicted these events even though these patients experienced clinical and hemodynamic improvements during the hospitalization that were similar to the improvements observed in normonatremic patients. Additional research is needed to determine if therapies that are known to normalize serum sodium levels (eg, vasopressin antagonists) are associated with improved clinical outcomes.
Correspondence: Mihai Gheorghiade, MD, Division of Cardiology, Northwestern University, Feinberg School of Medicine, Galter 10-240, 201 E Huron St, Chicago, IL 60611 (firstname.lastname@example.org).
Accepted for Publication: May 17, 2007.
Author Contributions:Study concept and design: Gheorghiade and O’Connor. Acquisition of data: Piña, DeMarco, Pauly, DiSalvo, Butler, Hare, and O’Connor. Analysis and interpretation of data: Gheorghiade, Rossi, Cotts, Shin, Hellkamp, Fonarow, Rogers, Butler, Francis, Stough, and O’Connor. Drafting of the manuscript: Gheorghiade, Rossi, Hellkamp, Rogers, and Stough. Critical revision of the manuscript for important intellectual content: Gheorghiade, Rossi, Cotts, Shin, Piña, Fonarow, DeMarco, Pauly, Rogers, DiSalvo, Butler, Hare, and Francis. Statistical analysis: Hellkamp. Administrative, technical, and material support: Rogers, Hare, and Stough. Study supervision: Gheorghiade, Piña, Fonarow, Pauly, DiSalvo, Butler, Francis, and O’Connor.
Financial Disclosure: None reported.
Funding/Support: The ESCAPE study was supported by contract N01-HV-98177 from the National Heart, Lung, and Blood Institute to Duke University Medical Center.
Role of the Sponsor: The National Heart, Lung, and Blood Institute oversaw the formulation and activities of the data and safety monitoring board in the ESCAPE study.