Prognostic relevance of elevated plasma osmolality on admission in acute decompensated heart failure with preserved ejection fraction: insights from PURSUIT-HFpEF registry

Background Complicated pathophysiology makes it difficult to identify the prognosis of heart failure with preserved ejection fraction (HFpEF). While plasma osmolality has been reported to have prognostic importance, mainly in heart failure with reduced ejection fraction (HFrEF), its prognostic meaning for HFpEF has not been elucidated. Methods We prospectively studied 960 patients in PURSUIT-HFpEF, a multicenter observational study of acute decompensated HFpEF inpatients. We divided patients into three groups according to the quantile values of plasma osmolality on admission. During a follow-up averaging 366 days, we examined the primary composite endpoint of cardiac mortality or heart failure re-admission using Kaplan–Meier curve analysis and Cox proportional hazard testing. Results 216 (22.5%) patients reached the primary endpoint. Kaplan–Meier curve analysis revealed that the highest quantile of plasma osmolality on admission (higher than 300.3 mOsm/kg) was significantly associated with adverse outcomes (Log-rank P = 0.0095). Univariable analysis in the Cox proportional hazard model also revealed significantly higher rates of adverse outcomes in the higher plasma osmolality on admission (hazard ratio [HR] 7.29; 95% confidence interval [CI] 2.25–23.92, P = 0.0009). Multivariable analysis in the Cox proportional hazard model also showed that higher plasma osmolality on admission was significantly associated with adverse outcomes (HR 5.47; 95% CI 1.46–21.56, P = 0.0113) independently from other confounding factors such as age, gender, comorbid of atrial fibrillation, hypertension history, diabetes, anemia, malnutrition, E/e′, and N-terminal pro-B-type natriuretic peptide elevation. Conclusions Higher plasma osmolality on admission was prognostically important for acute decompensated HFpEF inpatients. Supplementary Information The online version contains supplementary material available at 10.1186/s12872-021-02098-z.


Introduction
There are many common problems in heart failure (HF) that are linked to hospitalization and mortality [1]. Heart failure with preserved ejection fraction (HFpEF) accounts for approximately half of all HF cases, and this rate is increasing [2]. Because of their pathophysiological complexity [3], the precise mechanisms involved in HFpEF with a poor prognosis are not fully understood.
Plasma osmolality is easily estimated with a blood sample as [4]: Although the components of the formula, namely sodium [5], blood urea nitrogen [6], serum glucose [7], and other parameters interacting with osmolality such as serum albumin [8] and renal function [9] have been proven to affect the prognosis of HF, little has been elucidated about the prognostic meaning of osmolality itself in acute decompensated HF (ADHF).
Plasma osmolality has been reported to be influenced by well-known prognostic factors such as arginine vasopressin (AVP), the renin-angiotensin-aldosterone system (RAAS), and natriuretic peptides [10][11][12], which suggests that osmolality itself could be also associated with the prognosis of HF. On one hand, Vaduganathan et al. reported that lower osmolality was associated with poor outcomes in HF with reduced ejection fraction (HFrEF) from a post hoc analysis of the EVEREST trial [13]. Kaya et al. also reported that low osmolality on admission correlated with a poor prognosis in HFrEF patients [14]. On the other hand, independent from left ventricular ejection fraction (LVEF), Arévalo-Lorido et al. reported higher osmolality in ADHF patients could predict worse outcomes accompanied by higher comorbidities through the National Registry of Heart Failure (RICA) [15].
Based on these previous reports, the aim of this study was to investigate further the prognostic meaning of plasma osmolality, particularly in acute decompensated HFpEF patients.

The PURSUIT-HFpEF registry
This prospective, multicenter, observational cohort study was performed in 1008 consecutive hospitalized HFpEF patients. Details of the PURSUIT-HFpEF (The Prospective mUlticenteR obServational stUdy of patIenTs with Heart Failure with preserved Ejection Fraction) (1) 2 × [Serum sodium] + [blood urea nitrogen]/2.8 /18 registry have been described previously [16]. Briefly, in collaboration with 31 hospitals in Japan, this large-scale registry aimed to collect and record a comprehensive range of clinical data to define the pathophysiology and prognostic factors of HFpEF patients. Inclusion criteria were acute decompensated HFpEF diagnosed by the Framingham criteria for HF and the following: 1) LVEF ≥ 50% and 2) N-terminal pro-B-type natriuretic peptide (NT-proBNP) ≥ 400 ng/L or brain natriuretic peptide (BNP) ≥ 100 ng/L on admission. Major exclusion criteria were age < 20 years, severe valvular diseases, acute coronary syndrome on admission, life expectancy of < 6 months due to prognosis of non-cardiac diseases, and previous heart transplantation. The anonymized data were transferred to the data center of Osaka University Hospital for analysis via data capturing system connected with electronic medical records [17]. Written informed consent was received from each participating patient. This study, including the procedure for enrollment, conformed to the principles of the Declaration of Helsinki and was approved by the institutional review board of each participating facility, including the official institutional review board committee of Osaka University Hospital (approved on February 24, 2016). It was registered under the Japanese UMIN Clinical Trials Registration (UMIN000021831).

Study population
A total of 1024 inpatients with HFpEF were registered from June 2016 to February 2020. Of all the participants, 16 (1.6%) patients died in hospital. We should unfortunately exclude additional 48 patients due to missing of plasma osmolality on admission (missing of serum sodium; 1, blood urea nitrogen; 2, and glucose; 46). We finally analyzed remaining 960 (93.8%) patients discharged alive whose plasma osmolality was calculated on admission.

Plasma osmolality, nutrition status, plasma volume estimation and echocardiographic measurements
Plasma osmolality was estimated [4] with a blood sample (Eq. 1). Nutrition status was estimated with the Geriatric Nutritional Risk Index (GNRI), which was calculated using serum albumin and body mass index as described previously [18]. Systemic plasma volume was estimated with plasma volume status (PVS) using hematocrit and body weight as described previously [19]. Comprehensive echocardiographic examinations were performed by trained cardiac sonographers according to the American

Keywords: HFpEF, Plasma osmolality, Prognosis
Society of Echocardiography guidelines [20]. LVEF was calculated with the biplane Simpson's method using apical two-and four-chamber views.

Follow-up and endpoints
The primary endpoint of the present study was a composite of cardiac mortality or re-admission for HF during the follow-up period. The secondary endpoints were defined as each event of cardiac mortality and HF re-admission. The duration of the follow-up period was calculated from the day of discharge until an endpoint, or at the time of the last patient contact (including teleconferencing).

Statistical analysis
Data are presented as median and interquartile range of 25-75% for continuous variables and frequency/percentage for categorical variables. Continuous variables were compared using Kruskal-Wallis test (and Steel-Dwass test for between each groups) and categorical variables were compared using Fisher's exact test (with Bonferroni adjustment for between each groups). The distributions of plasma osmolality on admission and at discharge were compared with F-test. The correlation of plasma osmolality with sodium, urea nitrogen, glucose, and estimated glomerular filtration rate (GFR) were analyzed with linear regression models. The clinical endpoint was assessed with the Kaplan-Meier curve analysis and compared with the log-rank test. Univariable Cox proportional hazards regression models were used to calculate hazard ratios (HR) and 95% confidence intervals (CIs) for each endpoint. Multivariable Cox regression tests for plasma osmolality of our interest were performed using covariates of clinical importance as follows: age, gender, hypertension history, diabetes mellitus, hematocrit, GNRI, E/e' , and log-transformed NT-proBNP with and without estimated GFR. All statistical tests were 2-sided and P < 0.05 was regarded as statistically significant. Statistical analyses were performed using JMP ® Pro 13.2.1, (SAS Institute Inc., Chicago IL, USA) or EZR version 1.51 (Saitama Medical Center, Jichi Medical University, Saitama, Japan).

Characteristics of the study population
Distributions of plasma osmolality on admission and at discharge are shown in Fig. 1. Compared with the distribution at discharge, that on admission was significantly wide and shifted to higher levels (P < 0.0001, F-test). While the normal osmolality range is known to be 275-295 mOsm/kg [21], the median on admission was 297 mOsm/kg. Because plasma osmolality estimation consists of sodium, urea nitrogen, and glucose, osmolality had strong linear correlation with sodium (r = 0.797, Fig. 1 The distributions of plasma osmolality. Distributions of plasma osmolality on admission (A) and at discharge (B). F-test revealed that the distributions were significantly different between on admission and at discharge (P < 0.0001) P < 0.0001), and had mild correlation with urea nitrogen (r = 0.475, P < 0.0001) and with glucose (r = 0.221, P = 0.0001) (Additional file 1: Figure S1A, S1B, S1C).
Demographic and clinical characteristics of the 960 patients are summarized in the left column of Tables 1 and 2. The study population had a median age of 83 years; 55% were female. Hypertension (85%) was the most prevalent comorbidity followed by atrial fibrillation, dyslipidemia, and chronic kidney disease (46%, 41%, and 40%, respectively). The medians of NT-proBNP and estimated GFR were 3,250 ng/L and 45 mL/min/1.73 m 2 on admission, respectively. In the first-step treatment, more than half of all patients were treated with a bolus injection of diuretics (57%); non-invasive positive pressure ventilation was used in 13%. The most frequent prescription at discharge was a loop diuretic (79%), which was the most increased treatment during hospitalization.
According to the quantiles of plasma osmolality on admission (293.2 and 300.3 mOsm/kg), we divided patients into three groups. Background and general information on admission are described in the right column of Table 1. Among the components for the osmolality calculation (serum sodium, blood urea nitrogen, and glucose), serum sodium and blood urea nitrogen were significantly elevated in accordance with the elevation of plasma osmolality. In the higher osmolality groups, hypertension, diabetes mellitus, dyslipidemia and chronic kidney disease were prevalent. The higher osmolality groups showed renal dysfunction, and NT-proBNP of the highest quantile group (Q3) was significantly higher than those in other groups. Echocardiography on admission showed generally comparable between groups. In acute phase treatment, intravenous usage of carperitide was more frequent in the Q3 group. At discharge (right column of Table 2) higher osmolality groups on admission still had higher osmolality at discharge, and the medians plasma osmolality of Q2 and Q3 had decreased to just around the upper limit of normal. The higher osmolality groups also had significantly lower estimated GFR compared with lower osmolality groups.

Plasma osmolality and prognosis
Among 960 patients, 216 patients (22.5%) suffered composite endpoint with a mean ± standard deviation (SD) follow-up of 366 ± 356 days. As far as the secondary endpoint, 62 patients (6.5%) reached cardiac mortality in 444 ± 378 days, 204 (21.3%) re-admitted for HF in 366 ± 356 days. Receiver operating curve analysis provided that the ideal cut-off value of plasma osmolality on admission for predicting the primary endpoint was 299 mOsm/kg (area under the curve; 0.563, sensitivity; 0.50, specificity; 0.62, P = 0.0046), which was almost equal to the median (297 mOsm/kg) of the whole cohort.
The Kaplan-Meier curves revealed that higher plasma osmolality was significantly associated with the primary endpoint (Log-rank P = 0.0095) (Fig. 2). Univariable Cox regression tests revealed that the significance was observed particularly between the highest osmolality group (Q3) versus the lowest osmolality group (Q1) (HR 1.61; 95% CI 1.16-2.23, P = 0.0120) ( Table 3). Regarding the secondary endpoint, in the Kaplan-Meier curve analyses, HF readmission was also significantly more frequent in the higher osmolality group (Log-rank P = 0.0425), which was not in case with cardiac mortality (Log-rank P = 0.0937) (Fig. 2). Through univariable Cox regression tests for clinically important parameters on admission, higher age, lower hematocrit, lower eGFR, higher NT-proBNP, higher E/e' , and higher plasma osmolality were associated with the primary endpoint (left column of Table 4). Whereas only the age was shown to be independently associated with the primary endpoint through multivariable Cox regression analysis (center column of Table 4), plasma osmolality was also found to be also independently associated when eGFR was avoided from the confounders (right column of Table 4). Because plasma osmolality showed mild linear correlation with eGFR (r = − 0.379, P < 0.0001, Additional file 1: Figure S1D), it should be possible that plasma osmolality and eGFR had conflict in the multivariable analysis.
We further examined the event risk of a composite endpoint among the quantiles stratified by plasma osmolality at discharge. The Kaplan-Meier curve showed that the event risk was not associated with the osmolality at discharge in this cohort (Log-rank P = 0.1976, Additional file 1: Figure S2).

Discussion
In this study, we showed that higher plasma osmolality on admission was significantly associated with the adverse outcomes for HFpEF patients. Although a few reports have also indicated that plasma osmolality had prognostic meaning for HF patients, their descriptions were so scattered that we were unable to reach a consensus on how to deal with this marker. Thus, our present finding in a prospective cohort that "higher plasma osmolality on admission" was associated with the adverse outcomes in "hospitalized decompensated HFpEF" patients is notable.

Prognostic difference in plasma osmolality between HFpEF and HFrEF
Though a sub-analysis of the EVEREST trial for HFrEF patients, Vaduganathan et.al showed that normal osmolality at discharge was associated with improved outcomes [13]. Kaya et.al investigated clinical implication of plasma osmolality on admission for HFrEF patients [14]. They presented the third quartile of normo-to-hyperosmolality (mean of 293 mOsm/kg) as having the smallest adverse outcome rates, while the lowest quartile (mean of 280 mOsm/kg) showed the worst outcomes, followed by the highest quartile (mean of 301 mOsm/kg). According to these studies, plasma osmolality in the normal range seemed to be quite important for HFrEF patients. This finding should also be related to the particular prognostic importance of hyponatremia in HFrEF [22]. Contrary to these reports, Arévalo-Lorido et.al reported that the frequency of adverse outcomes increased in accordance with the increase in osmolality on admission in ADHF [15], similarly to our findings. Although their registry did not group subjects by LVEF, about 70% of the patients had LVEFs > 45%, indicating that HFrEF was underrepresented in that cohort. Taken our present findings together with those of Arévalo-Lorido et.al, we conclude that the elevation of plasma osmolality on admission raises the predictability of adverse outcomes in decompensated HFpEF patients.

Cause of higher plasma osmolality in HFpEF patients
Different from HFrEF patients, higher plasma osmolality on admission was related to adverse outcomes in HFpEF patients. It should be noted that the plasma osmolality on admission in our HFpEF cohort (median of 297 mOsm/ kg, Table 1) was generally higher than that of a previous reported HFrEF cohort (median of approximately 290 mOsm/kg) [14]. In an experimental study [12], excessive RAAS activation was proven to cause osmolality elevation in the acute phase of a rapid pacing HF model. RAAS activation could cause sodium reabsorption through modulation of the GFR, tubuloglomerular feedback, glomerulotubular balance, and distal tubular reabsorption [23], which could increase plasma osmolality. Relative hypovolemia in the higher osmolality groups compared with the lower osmolality groups was    not likely to be the cause of RAAS activation because PVS was comparable between groups (Table 1). AVP is known to be an another cause of volume retention, and increased AVP activity causes a decrease in osmolality accompanied by hyponatremia in HFrEF patients [24]. In contrast, age-related attenuation of the AVP response [25] could be more common in elderly HFpEF patients than in younger HFrEF patients. Excessive RAAS activation compared to AVP activity might contribute to the higher plasma osmolality in HFpEF compared to HFrEF patients.
Regarding echocardiographic parameters reflecting hemodynamics, tricuspid regurgitation pressure gradient was not different at all between three groups, and inferior vena cava collapsibility was significantly different but did not coincide with the prognostic result, namely in which collapsibility was significantly impaired in the Q1 group (Table 1). Although E/e′ tended to be higher in the poor prognostic Q3 group, multivariable Cox regression models showed independent prognostic importance of plasma osmolality from E/e' (Table 4). Considering these results, prognostic predictability based on the osmolality classification seemed to be unrelated to the hemodynamic observed in echocardiography. Despite these echocardiographic parameters, poor prognostic Q3 group showed frequent usage of intubation, non-invasive Fig. 2 Kaplan-Meier survival curves stratified with quantiles of plasma osmolality on admission. Composite endpoint was defined as cardiac mortality or heart failure re-admission. Kaplan-Meier survival curves for composite endpoint (A), cardiac mortality (B) and heart failure re-admission (C). HF, heart failure; Osm, plasma osmolality (mOsm/kg) on admission positive pressure ventilator, carperitide, and calcium channel blocker injection for initial treatment, frequent prior usage of loop diuretics, and higher NT-proBNP elevation on admission. The frequent usage of loop diuretics might directly affect the increase of plasma osmolality on admission [26]. These aspects ensured that Q3 group was in more decompensated hemodynamic status on admission paradoxically from the echocardiographic observations, and it seemed reasonable that those patients served poor outcomes.

Prognostic implication of higher plasma osmolality on admission
We showed that higher plasma osmolality on admission was associated with poorer prognosis in HFpEF patients. The prognostic impact of the AVP system in HF has not been fully elucidated. Because of the short half-life of AVP, it is not practical to measure plasma AVP as a prognostic marker. In this point, copeptin has attracted attention owing to its creation from prepro-vasopressin at the same time as AVP and longer half-life [27]. Some reports have shown the prognostic implications of copeptin for HF [28,29]. Although plasma osmolality on admission is not necessarily determined by the AVP system, there is a report consistent with our findings. Hage et al described the prognostic meaning of copeptin in a prospective HFpEF cohort (KaRen-study) and clarified that copeptin was elevated in HFpEF patients and had partial prognostic implications, which were blunted after adjustment for NT-proBNP [30]. The relevance of neurohormonal balance and pathophysiology in HFpEF should be further investigated.

What are the clinical implications?
The following variables have reported as prognostic markers in the acute phase in HFpEF patients: TRPG [31], lung congestion observed as B-lines ('comets') on lung ultrasound [32], soluble suppression of tumorigenesis-2  with NT-proBNP [33] and cystatin C [34]. In addition to these factors, our findings showed that higher plasma osmolality also has important prognostic implications in the acute phase of HFpEF. The higher osmolality groups presented even higher plasma osmolality than lower osmolality groups at discharge (Table 2), which showed those who had extremely elevated osmolality in the acute phase may suffer from some unfavorable factors which permanently raise the plasma osmolality. It is possible that those who have higher osmolality both on admission and at discharge are exposed to excessive RAAS activation, and the immediate and sustainable handling of this overactivation should be considered. Although various RAAS blockers have shown definite clinical benefits in HFrEF patients, including angiotensin converting enzyme inhibitors [35], angiotensin II receptor blockers [36], angiotensinneprilysin inhibitors [37], and mineral corticoid-receptor antagonists [38], the benefits in HFpEF patients are controversial [39]. We propose that further investigation to determine whether these approaches are particularly favorable to HFpEF patients with higher plasma osmolality is warranted.

Limitations
Several limitations of this study should be mentioned. First, we diagnosed included patients as HFpEF based on the presence of symptom and/or signs of HF, LVEF measurement, and elevated natriuretic peptides. There were lacking for the key structural alterations such as left atrial volume index, left ventricular mass index and E/e′ elevation and stress test assessment, which are proposed to be necessary for HFpEF diagnosis in the 2016 ESC guidelines [1] Second, although our results showed that elevated plasma osmolality on admission was associated with poor outcomes, we could not examine whether extremely decreased plasma osmolality affected prognosis because only 42 (4.4%) subjects had < 275 mOsm/kg on admission. Of note, this finding that excessively low plasma osmolality may be rare in acute decompensated HFpEF patients is important. Third, there have been several formulas which are able to calculate plasma osmolality, and Fazekas et al. reported a formula developed by Zander showed excellent concordance with measured osmolality [40]. Zander's formula included lactate and bicarbonate to calculate osmolality, however, we have not measured these parameters in our study. We selected the formula consisted of sodium, blood urea nitrogen and glucose, which was also used in the previous article investigated among HFrEF patients [14]. Fourth, plasma osmolality on admission was measured in the period between admission and approximately 48 h after admission. We were not able to assure whether the osmolality was measured prior to any initial treatments including loop diuretics administration and to avoid those initial treatment effects. Fifth, the present study was a multicenter prospective Asian cohort with quite elder patients (median age of as high as 83 years), which would limit the generalizability of the current findings for other races. Sixth, despite multivariable analysis, residual confounding from unmeasured factors may have affected the results. Finally, although we speculated that RAAS activity, and not AVP activity, was responsible for the poor outcomes, we did not measure either urine osmolality or neurohormonal factors substituting for RAAS.

Conclusion
We show here higher plasma osmolality on admission was associated with the composite endpoint of cardiac mortality or re-admission for HF in HFpEF patients. Further investigation to confirm the results of this small study and to support our understanding of the pathophysiological meaning of plasma osmolality in HFpEF patients is warranted.