Author + information
- Received October 24, 2018
- Revision received January 11, 2019
- Accepted January 14, 2019
- Published online April 29, 2019.
- Daniel Z. Hodson, BAa,∗,
- Matthew Griffin, MDa,∗,
- Devin Mahoney, BSa,
- Parinita Raghavendra, MSa,
- Tariq Ahmad, MD, MPHa,
- Jeffrey Turner, MDb,
- F. Perry Wilson, MD, MSb,
- W.H. Wilson Tang, MDc,
- Veena S. Rao, PhDa,
- Sean P. Collins, MDd,
- Wilfried Mullens, MD, PhDe and
- Jeffrey M. Testani, MD, MTRa,∗ ()
- aSection of Cardiovascular Medicine, Yale University School of Medicine, New Haven, Connecticut
- bSection of Nephrology, Yale University School of Medicine, New Haven, Connecticut
- cDepartment of Cardiovascular Medicine, Cleveland Clinic Lerner College of Medicine, Cleveland, Ohio
- dDepartment of Emergency Medicine, Vanderbilt University Medical Center, Nashville, Tennessee
- eDepartment of Cardiology, Ziekenhuis Oost Limburg, Genk–Biomedical Research Institute, Faculty of Medicine and Life Sciences, Hasselt University Diepenbeek, Belgium
- ↵∗Address for correspondence:
Dr. Jeffrey M. Testani, Section of Cardiovascular Medicine, Yale School of Medicine, 135 College Street, Suite 230, New Haven, Connecticut 06510.
Objectives This study sought to describe sodium excretion in acute decompensated heart failure (ADHF) clearly and to evaluate the prognostic ability of urinary sodium and fluid-based metrics.
Background Sodium retention drives volume overload, with fluid retention largely a passive, secondary phenomenon. However, parameters (urine output, body weight) used to monitor therapy in ADHF measure fluid rather than sodium balance. Thus, the accuracy of fluid-based metrics hinges on the contested assumption that urinary sodium content is consistent.
Methods Patients enrolled in the ROSE-AHF (Renal Optimization Strategies Evaluation-Acute Heart Failure) trial with 24-h sodium excretion available were studied (n = 316). Patients received protocol-driven high-dose loop diuretic therapy.
Results Sodium excretion through the first 24 h was highly variable (range 0.12 to 19.8 g; median 3.63 g, interquartile range: 1.85 to 6.02 g) and was not correlated with diuretic agent dose (r = 0.06; p = 0.27). Greater sodium excretion was associated with reduced mortality in a univariate model (hazard ratio: 0.80 per doubling of sodium excretion; 95% confidence interval: 0.66 to 0.95; p = 0.01), whereas gross urine output (p = 0.43), net fluid balance (p = 0.87), and weight change (p = 0.11) were not. Sodium excretion of less than the prescribed dietary sodium intake (2 g), even in the setting of a negative net fluid balance, portended a worse prognosis (hazard ratio: 2.02; 95% confidence interval: 1.17 to 3.46; p = 0.01).
Conclusions In patients hospitalized with ADHF who were receiving high-dose loop diuretic agents, sodium concentration and excretion were highly variable. Sodium excretion was strongly associated with 6-month mortality, whereas traditional fluid-based metrics were not. Poor sodium excretion, even in the context of fluid loss, portends a worse prognosis.
Dysregulation of sodium homeostasis is central to the pathophysiology of heart failure and leads to a positive sodium balance (1–3). Sodium is the primary osmolyte in the extracellular fluid compartment, and the active renal regulation of total body sodium content is used by the kidney to regulate total body fluid content passively. Importantly, the downstream effects of sodium retention include fluid retention, which in turn results in the traditional congestive symptoms associated with acute decompensated heart failure (ADHF) (4–7). Loop diuretic agents, the cornerstone of ADHF therapy, exert their effect by antagonizing renal sodium transporters to increase sodium excretion (8) along with the passive loss of fluid.
Current practice guidelines related to therapeutic monitoring focus on fluid balance and changes in weight (5,6), which serve only as proxies for sodium excretion. These metrics correlate poorly with each other (7,9), are often difficult to assess accurately, and have not consistently been linked to meaningful outcomes despite their frequent use as endpoints in major clinical trials (10–12). In addition to the fact that fluid and weight loss only indirectly query the physiology of interest (sodium excretion), there is inherent difficulty in obtaining accurate fluid intake and output, even under idealized study conditions (13,14). Furthermore, there is a common assumption that urinary sodium concentration is predictably hypotonic in patients taking a loop diuretic agent, sometimes compared with the sodium content of “half-normal saline” (15,16). There is an accumulating body of published reports challenging the supposition that urinary sodium concentration remains constant throughout the course of a hospitalized patient receiving diuretic agents (17–19).
Given the central role of sodium in both the pathophysiology and the therapy of congestion in ADHF, as well as the known limitations of urine output and weight, we sought to explore urinary sodium further as a potential metric to monitor ADHF treatment. We hypothesized that sodium excretion would be highly variable among patients receiving loop diuretic therapy, thus making any assumption of a consistent urine sodium concentration, such as half-normal saline, unreliable. Next, we hypothesized that higher degrees of sodium excretion, but not weight loss or negative fluid balance, would be associated with improved survival.
The ROSE-AHF (Renal Optimization Strategies Evaluation in Acute Heart Failure) trial dataset has been previously described (20–22). The data, analytic methods, and study materials are available through the Biologic Specimen and Data Repository Information Coordinating Center (BioLINCC). The original trial consisted of 360 patients hospitalized with ADHF with baseline renal dysfunction (estimated glomerular filtration rates [eGFRs] 15 to 60 ml/min/1.73 m2). All patients received open-label, intravenous (IV) loop diuretic treatment. The recommended total daily dose was equal to 2.5 times the total daily oral outpatient furosemide (or equivalent) dose at 7 days before admission. The study found no difference in the primary endpoints of cumulative urine output or change in cystatin C through 72 h between the intervention groups (low-dose dopamine or low dose nesiritide) and the pooled placebo group (21).
All patients received high-dose loop diuretic agents as defined earlier and were placed on standard 2 g/day sodium and 2 l/day fluid restriction. Sodium excretion was measured during the 3 intervention days by using 24-h urine collections, and patients were followed for clinical outcomes through 6 months after randomization. Our main analysis focuses on the 316 patients with complete data for the first 24 h (day 1) of sodium excretion. Day 1 sodium excretion was compared with day 2, day 3 (when available), and cumulative 72-h sodium excretion to assess within-patient variability. We focused on data from the first 24 h (day 1) of the 72-h intervention period because the first 24 h after meeting ROSE-AHF inclusion and exclusion criteria was a time when aggressive fluid and sodium removal was a universal therapeutic goal. All patients were by definition volume overloaded at this point because they just met criteria for congestion. We would expect fluid-sodium balance to track with outcomes only if physicians were attempting to achieve a negative sodium or fluid balance. Notably, in the DOSE-AHF (Diuretic Optimization Strategy Evaluation in Acute Heart Failure) trial (13), 31% of patients in the high-dose arm (same dosing as ROSE-AHF) were switched to oral therapy after 48 h. In addition, day 1 offered the lowest number of missing data.
Definitions and calculations
Patients were classified as having a positive sodium balance if their measured sodium excretion on day 1 was less than the amount in the prescribed diet (2 g = 87 mmol). If they excreted between 2 and 4 g (87 to 174 mmol), they were considered intermediate responders, and excretion of >4 g (174 mmol) was deemed to be an excellent response.
The definitions for fluid and weight response were similar to the definitions for sodium. Net fluid balance was calculated from recorded fluid intake and urine output. Patients had a positive fluid balance on day 1 if their fluid intake exceeded their urine output. Intermediate responders for fluid had a fluid balance between 0 and 2 l negative, and intermediate responders for weight lost between 0 and 2 kg. A net negative fluid balance of 2 l or greater was considered excellent, as was weight loss of >2 kg.
To test the assumption that urinary sodium concentration is consistent, such as half-normal saline, we defined estimated sodium excretion as a product of total urine output multiplied by the sodium concentration of half-normal saline (77 mmol/l = 1.77 g/l).
The primary outcomes of interest were: 1) the variability in 24-h sodium excretion; and 2) 6-month all-cause mortality. Categorical variables were compared using the chi-square test, and continuous variables were compared using Student’s t-test, analysis of variance, Wilcoxon signed rank test, Mann-Whitney U test, or Kruskal-Wallis test on the basis of examination of the distribution. Correlations are reported as Spearman’s rho. The Bland and Altman method was used to analyze the differences between the measured sodium excretion and estimated sodium excretion using the half-normal saline assumption. As the current reference standard, measured sodium excretion was used for the x-axis. However, the bias, or mean of the differences, allowed for an artificially deflated measure of the actual bias at the individual level, so we also calculated the absolute value of the difference between estimated and measured sodium excretion.
Associations between the metrics of diuretic response and 6-month all-cause mortality were examined in univariate and multiple regression Cox proportional hazards modeling and through Kaplan-Meier analysis. Multiple regression analyses adjusted for the following baseline characteristics: age, sex, race, heart rate, blood pressure, left ventricular ejection fraction, log baseline N-terminal pro–B-type natriuretic peptide, eGFR, angiotensin-converting enzyme inhibitor or angiotensin receptor blocker use, beta-blocker use, and aldosterone antagonist use. Candidate covariates entered in the model were baseline characteristics with univariate all-cause mortality associations with p ≤ 0.10. Covariates that had a p > 0.10 but a theoretical basis for potential confounding were forced into the model. Models were built using backward elimination (likelihood ratio test) where all covariates with a p < 0.10 were retained.
Laboratory values from the Heart Failure Network (HFN) core laboratory were preferentially used, and when they were missing, local site laboratory values were substituted. GFR rate was estimated by the Modification of Diet in Renal Disease equation (23), as specified by the ROSE-AHF protocol. Statistical analysis was performed with IBM SPSS Statistics version 24 (IBM Corp., Armonk, New York), and statistical significance was defined as 2-tailed p < 0.05 for all analyses.
Baseline characteristics for the 316 patients included in this analysis are presented in Table 1 and mirrored the overall ROSE-AHF study group: patients tended to be white men with ischemic cardiomyopathy, multiple comorbid conditions, and a high prevalence of physical examination findings consistent with volume overload. Given that there were no significant differences in sodium excretion across the treatment groups in the ROSE-AHF trial (p = 0.75 for low-dose dopamine vs. placebo; p = 0.52 for low-dose nesiritide vs. placebo), we elected to analyze the groups together. The median eGFR for these patients was 43.7 ml/min/1.73 m2 (interquartile range: 32.5 to 55.4 ml/min/1.73 m2). When stratified by natriuretic response, those patients with a positive sodium balance (who excreted <2 g) tended to have higher baseline N-terminal pro–B-type natriuretic peptide levels and a lower eGFR (Table 1). A considerable percentage of patients (28.5%) exhibited positive sodium balance through 24 h (Central Illustration), and this percentage was similar on day 2 (29.7%) and day 3 (32.4%). However, net positive fluid balance (n = 26; 8.2%) and weight gain (n = 50; 15.8%) occurred less commonly, and only 7 (2.2%) participants exhibited both net fluid and weight gain. There was only a modest correlation between sodium excretion and weight loss (r = 0.53) and between sodium excretion and net fluid loss (r = 0.67) (Online Table 1).
Variability in sodium excretion with diuresis
Sodium excretion through 24 h was highly variable across patients (Figures 1A and 1B) despite aggressive diuretic dosing (median dose on day 1 = 200 mg IV furosemide equivalents; interquartile range: 100 to 280 mg). There was no overall correlation between sodium excretion and diuretic dose (r = 0.06; p = 0.27), and there was no difference in diuretic dose among those patients with positive sodium balance, intermediate response, and excellent response (p = 0.68). This lack of correlation between diuretic dose and sodium output persisted after controlling for eGFR (p =0.33). Within-patient analysis confirmed considerable variability in sodium excretion between hospital days with a modest correlation between day 1 and day 2 sodium excretion (r = 0.44; p < 0.001), but a weaker correlation between day 1 and day 3 (r = 0.18; p = 0.002) (Online Table 2).
Measured sodium excretion and the assumption of a consistent urine sodium concentration
When the total urine output was multiplied by the concentration of half-normal saline, sodium excretion was overestimated in 69.9% of patients, and there were large underestimations and overestimations at the individual level. The assumption overestimated by >1 g in 47.1% of patients and by >2 g in 24.3%, and it underestimated by >1 g in 17.7% of patients and by >2 g in 8.2% (Figures 1C and 1D). Figure 1C presents a Bland and Altman plot of estimated and measured absolute sodium excretion. Notably, the 95% limits of agreement were wide (−155.1 to 226.5 mmol or −3.6 to 5.2 g). The bias, or mean of the differences, indicated that the assumption overestimated by 35.7 mmol (800 mg), whereas the mean of the absolute value of the differences (Figure 1D) was considerably larger 78.2 mmol (1.8 g). Similarly, when comparing the estimated sodium concentration with the 77 mmol/l half-normal saline assumption, there was poor agreement (Figure 1B).
Prognostic value of the various metrics of diuretic response
Sodium excretion was significantly associated with 6-month all-cause mortality. For each doubling of sodium excretion, the odds of death were reduced by 20% in both univariate (hazard ratio [HR]: 0.80 per doubling of sodium excretion; 95% confidence interval [CI]: 0.66 to 0.95; p = 0.01) and multiple regression (HR: 0.80; 95% CI: 0.65 to 0.97; p = 0.02) modeling. As sensitivity analyses (Online Table 3), the results using cumulative sodium excretion over the first 48 and 72 h, respectively, were similar in both magnitude and direction (HR: 0.75; 95% CI: 0.57 to 0.98; p = 0.04 for 48-h excretion and HR: 0.74; 95% CI: 0.51 to 0.94; p = 0.02 for 72-h excretion). However, total urine output (adjusted HR/l: 0.93; 95% CI: 0.77 to 1.11; p = 0.43), net fluid balance (adjusted HR/l: 0.98; 95% CI: 0.83 to 1.17; p = 0.87), and weight loss (adjusted HR/kg: 0.94; 95% CI: 0.85 to 1.04; p = 0.25) were not significantly associated with survival. Positive sodium balance was associated with significantly decreased survival in Kaplan-Meier analysis (Central Illustration, Figure 2A), whereas net positive fluid balance (p = 0.57) and weight gain (p = 0.44) were not significantly associated with survival.
To elucidate the relative importance of fluid versus natriuretic responses further, we examined patients who exhibited discordant responses. The 71 patients who exhibited both positive sodium balance and net fluid loss still experienced significantly decreased 6-month survival (Central Illustration, Figure 2B) (p = 0.01). This association remained significant after controlling for baseline covariates and loop diuretic dose (HR: 2.02; 95% CI: 1.17 to 3.46; p = 0.01).
The primary findings of this study are as follows: 1) urinary sodium excretion during aggressive diuresis in patients with ADHF and renal dysfunction is highly variable both across patients and within the same patient across different days; 2) as a consequence of such variability, the assumption of consistent salt content, such as half-normal saline, of loop diuretic–induced urine does not accurately predict 24-h sodium excretion; 3) sodium excretion, but not metrics of fluid or weight response, carries significant prognostic value; and 4) even in the context of high diuretic dosing and documented net fluid loss, a positive sodium balance is relatively common and is associated with significantly worse survival in patients with ADHF. In conjunction with the well-established central pathophysiological role of sodium in volume retention and its treatment, the current observations reinforce that sodium excretion is a critical variable in the treatment of patients with ADHF and requires additional study.
Despite recognition of their shortcomings, fluid and weight assessments remain the standard quantitative metrics guiding treatment for heart failure in clinical practice, as well as endpoints in clinical trials (10,13,24). Notably, although essentially measuring the same underlying parameter, net fluid balance and weight loss correlate poorly, even in the setting of rigorous National Institutes of Health–funded prospective trials on diuretic therapy (25). This is likely in part the result of the logistical challenges of accounting for every milliliter of a patient’s fluid intake. Even when providers officially place patients on fluid restriction, nearly every hospital room has a faucet. Conversely, there is not such a ubiquitous and easily accessible source of sodium in hospitals.
Furthermore, it is not clear that change in body weight performs any better than measuring fluid intake and output. Patients may be weighed using different types of scales (bed, standing, sling), scales with different calibrations, and at different times of the day relative to meals, voids, or bowel movements. The variability of bowel movements is of even greater significance in patients undergoing fluid restriction, which has been linked to constipation (26). Finally, important nuances such as whether the patient’s telemetry box or shoes were included in the measurement are often overlooked. Sodium, conversely, is not affected by these factors.
The results of this analysis are consistent with the primacy of sodium in the pathophysiology of heart failure. Given the predominance of semipermeable membranes that allow water, but not ions, to move freely, fluid loss is a secondary effect of sodium excretion. As highlighted by our analysis, urinary sodium concentration is highly variable, thus uncoupling fluid excretion from sodium excretion. By measuring only fluid excretion without urinary sodium or urinary sodium concentration, patients with lower urinary sodium concentrations are not always assessed as poor responders, as long as their urine output is deemed adequate. We would argue that there is an important difference between a patient who produces 2 l of urine with a low urinary sodium concentration and another patient who produces the identical amount of urine with a higher sodium content. This difference is captured when measuring total urinary sodium excretion, but not with strictly fluid metrics. However, it has yet to be demonstrated that routine performance of 24-h urine collection to measure sodium excretion is feasible outside a clinical trial setting.
To address this issue, our group has developed a sodium prediction equation to forecast sodium excretion following a bolus dose of loop diuretic agent (27). Much in the way that 24-h creatinine collections have largely been supplanted by eGFR derived from an equation and a spot serum creatinine value, we have developed a formula that can estimate the amount of sodium excretion from a dose of loop diuretic agent by using a spot urine sodium and creatinine level. We are currently in the process of validating this equation in a larger cohort (Diagnosing and Targeting Mechanisms of Diuretic Resistance in Heart Failure; NCT02546583). Our current analysis provides justification for a shift in emphasis away from fluid-based metrics toward sodium-based metrics, and the forthcoming tool may help to inform therapy for these complex patients. One strategy would be to titrate loop diuretic dose to measured sodium excretion, instead of solely using fluid and weight metrics to assess response.
Importantly, this was a post hoc analysis of the ROSE-AHF trial, which was not designed to compare different metrics of monitoring diuresis. As such, these results should be considered hypothesis-generating. The ROSE cohort was primarily composed of white men, and by design, all patients had an eGFR <60 ml/min/1.73 m2. As a result, it is unclear how the results will apply to the broader ADHF group. In addition, although the 24-h urine collection is the gold standard to evaluate sodium balance, such a collection by definition comprises all urine produced in a day. This limits inferences made specifically about the composition of urine produced solely as the result of the loop diuretic agent. Although patients were placed on a 2-g sodium diet, individuals may have actually consumed more or less than 2 g, thus limiting the accuracy of assumptions made regarding sodium balance.
In line with the central role of sodium in the pathophysiology of ADHF and its treatment, sodium excretion was more strongly associated with 6-month survival than either net fluid balance or weight change. However, the sodium content of urine in patients with ADHF was highly variable, thereby indicating that an assumption of a consistent urine sodium concentration is inadequate, and urine sodium content needs to be determined with each diuretic dose for each individual patient. Additional research is needed to better understand the role of urine sodium content in personalizing ADHF therapy.
COMPETENCY IN MEDICAL KNOWLEDGE: In patients with ADHF who were receiving high-dose loop diuretic agents, sodium excretion was highly variable. Higher degrees of sodium excretion were associated with improved survival, whereas the traditional fluid-based metrics, such as weight change and net fluid balance, were not. Patients with a positive sodium balance, even when accompanied by a net negative fluid balance, had worse survival compared with all patients who had a net negative sodium balance.
TRANSLATIONAL OUTLOOK: Direct measurement of urinary sodium excretion may provide a viable alternative to urine output and fluid balance to guide diuretic therapy in ADHF. Additional research is required to better understand how measurement of urinary sodium content can guide heart failure management.
↵∗ Mr. Hodson and Dr. Griffin contributed equally to this work.
This work was supported by National Institutes of Health (NIH) grants K23HL114868, L30HL115790, R01HL139629, R21HL143092, R01HL128973 (Dr. Testani), K23DK097201 (Dr. Wilson), and T32 training grant 5T32HL007950 (Dr. Griffin). Dr. Tang has received consultant fees from Sequana Medical and MyoKardia. Dr. Collins has been a consultant for Novartis, Vixiar, Ortho Clinical, and Medtronic; and has received research support from Ortho Clinical. All other authors have reported that they have no relationships relevant to the contents of this paper to disclose.
- Abbreviations and Acronyms
- acute decompensated heart failure
- confidence interval
- estimated glomerular filtration rate
- hazard ratio
- Received October 24, 2018.
- Revision received January 11, 2019.
- Accepted January 14, 2019.
- 2019 American College of Cardiology Foundation
- Mullens W.,
- Verbrugge F.H.,
- Nijst P.,
- Tang W.H.W.
- Nijst P.,
- Verbrugge F.H.,
- Grieten L.,
- et al.
- Gheorghiade M.,
- Pang P.S.
- Ponikowski P.,
- Voors A.A.,
- Anker S.D.,
- et al.
- Writing Committee Members,
- Yancy C.W.,
- Jessup M.,
- et al.
- Kociol R.D.,
- McNulty S.E.,
- Hernandez A.F.,
- et al.
- Verbrugge F.H.,
- Nijst P.,
- Dupont M.,
- Penders J.,
- Tang W.H.,
- Mullens W.
- Zazzeron L.,
- Ottolina D.,
- Scotti E.,
- et al.
- Chen H.H.,
- AbouEzzeddine O.F.,
- Anstrom K.J.,
- et al.
- Jain A.K.,
- Chen H.H.
- Stevens L.A.,
- Coresh J.,
- Feldman H.I.,
- et al.
- Costanzo M.R.,
- Guglin M.E.,
- Saltzberg M.T.,
- et al.
- Testani J.M.,
- Hanberg J.S.,
- Cheng S.,
- et al.