This study is a post-hoc analysis of a multicenter, open-label, parallel randomized controlled trial (IMPROVE-HF). In that trial, patients with AHF and renal dysfunction at presentation were 1:1 randomized to a) a standard loop diuretics dosage based on routine clinical evaluation, or b) carbohydrate antigen 125 (CA125)-guided therapy. The inclusion and exclusion criteria were published previously2 and are presented in table 1 of the supplementary data. This study was registered with ClinicalTrials.gov (NCT02643147). A detailed description of the trial design is presented elsewhere.2,9

The study was conducted according to the principles of the Declaration of Helsinki and the International Conference on Harmonisation Guidelines for Good Clinical Practice and fully conformed to national regulations. The protocol, informed consent form, participant information sheet, and all applicable documents were approved by the appropriate Ethics Committee (Comite de Ética del Hospital Clínico Universitario de Valencia) and by the Agencia Española del Medicamento y Productos Sanitarios (AEMPS). All patients provided written informed consent. All analyses were performed by an independent company (MedStats Consulting, United States).

Once the diagnosis of AHF was confirmed, patients were screened and randomized within the first 24hours. During this visit, clinical characteristics and biomarkers were assessed. Scheduled follow-up visits were performed at 24hours, 72hours, and 30 days after randomization (final visit).

Eligible patients were randomized to receive intravenous diuretics with the dosage based on either conventional clinical evaluation or on CA125 values. In the conventional clinical evaluation arm, the diuretic strategy was based on the presence of signs and symptoms of systemic congestion according to current guideline recommendations.10 In the active arm, higher diuretic doses were recommended in patients with CA125> 35 U/mL. In contrast, lower doses were recommended when CA125 ≤ 35 U/mL. The diuretic strategies in both arms are summarized in table 2 of the supplementary data.

UNa+ was measured at patient randomization and at 24hours. The mean time from admission to randomization was 6±3hours, and the median and interquartile range [IQR] dose of intravenous furosemide received before randomization was 40 [20-60] mg. Based on previous studies,3,8,11 UNa+ at admission was dichotomized at 50 mmol/L (≤ 50 vs> 50 mmol/L).

The primary end point was to assess whether a single determination of UNa+ at admission predicts all-cause mortality and all-cause readmission rates. In addition, we sought to evaluate if changes in UNa+ at 24hours (ΔUNa24h) provided any additional prognostic advantage over baseline UNa+ values. For this purpose, the ΔUNa24h was tested against clinical end points at UNa+ ≤ 50 and> 50 mmol/L (UNa+50). Ambulatory follow-up was performed in the HF units of each center. End points were adjudicated by reviewing the electronic discharge records of our regional health care system. Only unplanned readmissions were included. HF-related readmissions were those with worsening or acute HF as the main diagnosis at discharge. The researchers in charge of end point adjudications were all blinded to the exposure.

Baseline continuous characteristics according to UNa+50 are presented as mean±standard deviation and median [interquartile range] as appropriate; categorical variables are presented as frequencies with percentages. Continuous variables were compared according to UNa+50 with either the t test or rank-sum test for independent samples depending on the variable distribution. Discrete variables were compared using the chi-square test.

A total of 160 patients were included in this study between March 2015 and December 2016 at 9 centers in Spain. The mean age of the study population was 78±8 years, 66.9% were male, and 46.9% had left ventricular ejection fraction <50%. The median [IQR, p25-p75] levels of N-terminal pro-brain natriuretic peptide and CA125 were 7765 pg/mL [3507-15404] and 58 U/mL [22-113], respectively. The mean BUN and creatinine levels and estimated glomerular filtration rate (eGFR) were 47.3±16.6mg/dL, 1.98±0.52mg/dL, and 34±8.5mL/min/1.73 m2, respectively. By design, all patients had renal dysfunction (eGFR <60mL/min/1.73 m2), and 50 patients (31.3%) had an eGFR <30mL/min/1.73 m2 at admission. Chronic kidney disease (eGFR <60mL/min/1.73 m2 using serum creatinine obtained at the latest available outpatient visit prior to admission and assessed during a stable phase of the disease) was present in 106 patients (66.3%). All patients received intravenous furosemide on admission, and the median [IQR] dose was 40 [20-60] mg. The accumulated doses at 24 and 72hours postrandomization were 190 [120-320] and 410 [250-640], respectively. Detailed characteristics of the study sample are presented in table 1.

The median [IQR] of UNa+ content at randomization was 90 mmol/L [65-111], and UNa+ ≤ 50 mmol/L was present in 23 patients (14.4%). Patients with UNa+ ≤ 50 mmol/L were younger, exhibited worse New York Heart Association (NYHA) class at randomization, and had higher baseline levels of CA125. Nevertheless, no significant differences were found in natriuretic peptide values or in clinical signs of congestion vs those with UNa+> 50 mmol/L (table 1). Moreover, the 2 groups were balanced regarding prior diuretic treatment (table 1). Intravenous diuretic therapy according to UNa+50 is presented in table 2. There was a statistical trend for a higher intravenous furosemide equivalent dose (FED) during the first 24hours in patients with UNa+ ≤ 50 mmol/L. No differences were found at 72hours. Accumulated diuresis during the first 24 and 72hours did not significantly differ according to UNa+50. Time to cessation of intravenous furosemide was longer in patients with UNa+ ≤ 50 mmol/L (table 2).

The mean eGFR was significantly increased at 24hours (34.0±8.5 vs 35.4±8.5mL/min/1.73 m2, P=.004). Likewise, the mean UNa+ was decreased vs baseline [median delta change of−5.5 (−25 to 12.6) mmol/L]. In total, 98 patients (61.3%) had a decreased/unchanged ΔUNa24h. Patients with a decreased/unchanged ΔUNa24h showed a trend for higher rates of death (3.9 vs 2.6 per 10 person-years, P=.094) but not for all-cause readmissions (17.9 vs 14.0 per 10 person-years, P=.279).

By modeling the effect of the ΔUNa24h on the risk of mortality and readmissions, we found a significant interaction with baseline values of UNa+50. The ΔUNa24h was inversely associated with all-cause mortality only in the group with UNa+ ≤ 50 mmol/L. Conversely, no effect was found in the group with UNa+> 50 mmol/L (figure 2). The same differential effect was found for total readmissions (figure 3). The interactions between ΔUNa24h and baseline UNa+ were no longer significant when the baseline values of UNa+ were dichotomized at 60 mmol/L (P value for interaction=.080) and 70 mmol/L (P value for interaction=.837).

Figure 2.

Differential effects of ΔUNa24h on all-cause mortality according to UNa+ at baseline. IRR depiction along the continuum of the ΔUNa24h according to the 2 levels of UNa+ at baseline. A, UNa+ at baseline ≤ 50 mmol/L. B, UNa+ at baseline> 50 mmol/L. ΔNaU24h modeled with fractional polynomial [3] and centered at the “0” value. The omnibus P value for the interaction, P=.031. The analysis was adjusted by hospital center, randomization variable, age, sex, systolic blood pressure, glomerular filtration rate, N-terminal pro-brain natriuretic peptide, urine volume at 24hours, and FED at 24hours. ΔUNa24h, change in UNa+ at 24hours; FED, furosemide equivalent dose; IRR, incidence rate ratio; UNa+, urinary sodium.

(0.13MB).

Figure 3.

Differential effects of ΔUNa24h on all-cause recurrent admissions according to UNa+ at baseline. IRR depiction along the continuum of the ΔUNa24h according to the 2 levels of UNa+ at baseline. A, UNa+ at baseline ≤ 50 mmol/L. B, UNa+ at baseline> 50 mmol/L. ΔUNa24h modeled with fractional polynomial [0] and centered at the “0” value. The omnibus P value for the interaction, P=.002. Analysis adjusted by hospital center, randomization variable, prior admission for heart failure, ischemic etiology, systolic blood pressure, glomerular filtration rate, blood urea nitrogen, hemoglobin, N-terminal pro-brain natriuretic peptide, urine volume at 24hours, and FED at 24hours. ΔUNa24h, change in UNa+ at 24hours; FED, furosemide equivalent dose; IRR, incidence rate ratio; UNa+, urinary sodium.

(0.12MB).

UNa+ has emerged as a promising biomarker in HF.3–7 From a pathophysiological point of view, lower UNa+ may result from a decrease in the glomerular filtration rate.14 However, most of the evidence indicates that lower UNa+ mainly identifies patients with renal tubular dysregulation secondary to sustained neurohormonal activation.15,16 Thus, lower UNa+ has been shown to identify patients at higher risk of adverse events and diuretic resistance.3–7 This study confirms these findings in patients with AHF and concomitant renal dysfunction. In agreement with recent work, we found that early assessment of UNa+ is strongly associated with the risk of long-term mortality, despite a thorough patient adjustment. Interestingly, we also found a significant and adjusted association of spot UNa+ with the burden of total hospitalizations in this high-risk population that enrolled a subset of patients with severe renal dysfunction at presentation.

Although we cannot compare the discriminative accuracy of spot UNa+ at admission vs UNa24h, we believe that spot determinations offer important advantages over UNa24h. First, the spot assessment might provide early clinical information with potential prognostic and therapeutic implications. Second, spot UNa+ at admission or during the early course of admission is much easier to implement in real-world clinical practice.

Serial assessment of UNa+ during hospitalization for AHF has been suggested to be useful for monitoring the diuretic response and eventually guiding the intensity of depletive treatment. Indeed, a recent position statement from the Heart Failure Association of the European Society of Cardiology (ESC-HF) recommended UNa+ measurement as part of a stepped pharmacological care algorithm1 in patients with AHF and congestion. According to that document, a spot UNa+ <50–70 mmol/L 2hours after the starting dose of loop diuretics should alert clinicians to the risk of diuretic resistance and motivate prompt interventions.1 Although this “tubular stress test” is plausible and attractive, there is little evidence supporting the clinical role of UNa+ serial measurement in patients with AHF syndromes. Biegus et al.8 recently reported that longitudinal spot UNa+ analysis during consecutive days of decongestion provides further clinical and prognostic information. They showed that patients with decreased/unchanged UNa+ excretion at 48hours had less effective decongestion, poorer diuretic efficacy, and higher risk of 1-year mortality. However, several aspects should be acknowledged when interpreting longitudinal UNa+ dynamics. First, urinary composition is characterized by a progressive decline in UNa+ along the course of diuretic therapy.7 Once extracellular volume overload has been reduced, natriuresis also declines to allow sodium excretion to once again equal intake.17 Although this phenomenon is frequently maladaptive in AHF (sodium retention in the setting of persistent congestion), it might be difficult to ascertain whether the observed decline in UNa+ indicates diuretic resistance or, in contrast, represents a physiological response to extracellular volume contraction. In addition, and regardless of the prognostic usefulness, we already envision that, in the absence of specific and well-established treatment implications, it may be difficult to widely implement serial assessment in clinical practice. However, we should also note that isolated assessment of UNa+ in the first hours or days of decompensation also provides useful prognostic information, as reported here and in other studies.8,18,19

According to the present findings, early changes in UNa+ may provide meaningful prognostic information in patients with lower UNa+ values at baseline. This subset of patients identified a subgroup with ominous prognosis whose UNa+ monitoring may provide useful clinical information on the treatment response. In contrast, in the presence of higher values at baseline (UNa+> 50 mmol/L), early UNa+ changes may not provide additional information to better define patient risk. Further research is required to confirm the present findings and, more importantly, elucidate the role of UNa+ as a therapeutic target.

Several limitations to our study need to be acknowledged. First, this was a post-hoc analysis of the IMPROVE-HF trial, which was not designed to evaluate the prognostic value of UNa+ in AHF. Therefore, all of the results should be considered exploratory and hypothesis-generating. Second, due to the limited sample size, some of the negative results could be explained by type II error (insufficient statistical power). Third, the IMPROVE-HF cohort was primarily composed of patients with AHF and concomitant renal dysfunction. Therefore, it is unclear how the results will apply to the broader AHF population. In addition, with the present data, we cannot accurately distinguish between patients with acute vs chronic renal dysfunction at presentation. Fourth, although patients were recruited in 9 centers in Spain and the follow-up was performed by a multidisciplinary team (cardiologists and internal medicine specialists), it is unknown how these results will apply to other health care systems with a different HF management organization.20 Fifth, we did not evaluate the effect of UNa+ on other surrogates of decongestion, such as weight and venous pressures. Sixth, we cannot account for differences in sodium intake before admission or during decompensation. As such, we cannot evaluate its effect as a confounder. Seventh, we used spot urine samples and not 24-hour urine collection. Therefore, we cannot compare the prognostic value of spot vs 24-hour collection. Eighth, baseline UNa+ is potentially confounded by the timing of measurement and the last administration of intravenous diuretics. This period was not recorded in this study. Lastly, the threshold for the identification of a poor natriuretic response remains arbitrary. Further studies should confirm the optimal cutoffs.