The evolution of hypovolemic and euvolemic hyponatremia coincides with an inammatory status in patients with COVID-19

BACKGROUND Hyponatremia is a frequent electrolytic disorder in hospitalized patients with Coronavirus Disease 2019 (COVID-19). Inammatory status could be associated with hyponatremia in patients with COVID-19. We evaluate the hospital evolution of hyponatremia and inammation markers in COVID-19 patients. METHODS A case-series of adult patients with COVID-19 pneumonia and hyponatremia at admission were included and retrospectively followed during hospitalization. Data of the volemic status, general characteristics, inammation, and clinical outcomes, were collected and analyzed. Comparative and multivariate analyzes were developed, and Hazzard ratio (HR) with 95%-condence intervals (95% CI) were calculated.


Introduction
Hyponatremia is observed in 19.7% of hospitalized patients (1) and is associated with an increase in hospital mortality (1,2).
It is known that pneumonias are related with the development of hyponatremia via non-osmotic stimulation of the arginine-vasopressin release. Hyponatremia in patients with pneumonia is often euvolemic, and caused by the syndrome of inappropriate secretion of antidiuretic hormone (SIADH), which has been observed in approximately a third of hospitalized pneumonia patients (3). The study complied with accepted standards of good clinical practice of the World Medical Association Declaration of Helsinki, in accordance with the guidelines of the local Research Ethics Committee of the Hospital Clínico San Carlos Pneumonia caused by Coronavirus disease 2019 (COVID-19) is often accompanied by hyponatremia. The prevalence of hyponatremia in COVID-19 patients has been reported in a range from 9.9-30% (4-7). The mechanisms behind its etiopathogenesis remain unclear. SIADH has been suggested as a frequent cause(8). Berni et al. showed an inverse relationship between levels of serum sodium (SNa) and interleukin-6 (IL-6) in COVID-19 patients (9), proposing an in ammatory via implicated in the development of hyponatremia. However, in ammatory mechanisms on the development of hyponatremia would be expected when the latter is caused by SIADH (10). Whether an in ammatory status is also implicated in other volemic types of hyponatremia has not been elucidated.
We reviewed COVID-19 patients with hyponatremia and analyzed the relationship between their in ammatory markers with the evolution of the SNa during the hospitalization, in addition to describe the main clinical characteristics that accompanied hyponatremia.

Methods
It is a retrospective study of a series of consecutive cases of adult patients with COVID-19 pneumonia who had hyponatremia at hospital admission and were followed during the hospitalization. COVID-19 pneumonia was con rmed by a positive reverse-transcriptase polymerase chain reaction test plus clinical and/or radiological picture of pneumonia. All patients were admittance during April 1st to April 30th 2020 in three wards of different tertiary hospitals of Madrid, Spain, where the "COVID-19 teams" -conformed by physicians of different medical specialties-attended COVID-19 patients during the rst wave of the pandemic. Each of the authors of the current study was part of a COVID-19 team of her/him respective hospital ward during this period.
Hyponatremia was de ned as a SNa < 135 mmol/L after correction for glycemia. Hypovolemia was considered when three or more of the following symptoms/signs were observed: thirst, orthostatic symptoms, blood pressure ≤ 90/60 mmHg, heart rate ≥ 90 bpm, urinary sodium ≤ 30 mmol/L [5,6] or a rise in serum creatinine (SC) and/or serum urea (SU) accompanying the descent in SNa [7,8]. The absence of the previous criterion was assumed as euvolemia. Hypervolemia was considered present when explicit signs of hydric overload or edematous heart failure were registered in the clinical records.
Variables were collected from admission day, 2nd -3th day of hospitalization, 7th -10th day of hospitalization, and from the day of discharge of the patients, when available. Data was collected according to how it appeared registered in clinical records. Collected variables were age, sex, SNa, serum potassium (SK), SC, SU, volemia, serum in ammatory markers (IM): D-dimer, brinogen, C-reactive protein, ferritin, lactate dehydrogenase, IL-6; comorbidities: obesity, hypertension, diabetes mellitus, chronic obstructive pulmonary disease/smoke habit, obstructive sleep apnea/hypopnea syndrome, asthma, pulmonary brosis, any cancer, obstructive uropathy and advanced/moderate chronic kidney disease (glomerular ltration rate < 60 ml/min); previous drug use, and the clinical outcomes: hospital death, admittance to the intensive care unit (ICU), and length of hospital stay (LOS). Treatment variables were not collected since an important heterogeneity and a lack of details of this variable were observed during the review of the clinical records, therefore, this data would have had no value.

Statistical analysis
Univariate analysis for quantitative variables was developed with Mann-Whitney U or Kruskal-Wallis tests (when non-parametric) and T-student or ANOVA tests (when parametric). Univariate analysis for categorical variables was performed with Chi-squared or Fisher tests. Correlation studies with Pearson and Spearman tests were executed for quantitative variables.
Multivariate analyses with logistic regressions and Cox regression including variables with a p value < 0.1 in the univariate analysis and others of clinical relevance, were tested to evaluate the association of the variables with clinical outcomes, calculating odd ratio (OR) and Hazard ratio (HR) respectively.
Multivariate analysis was developed with the forward steps Wald´s method. A two-tailed p value < 0.05 was considered statistically signi cant. Ninety-ve percent con dence intervals (95% CI) were calculated. Statistical analysis was performed using SPSS version 25 (IBM Corp., N.Y., USA).

Ethical issues
The study complied with accepted standards of good clinical practice of the World Medical Association Declaration of Helsinki, in accordance with the guidelines of the local Research Ethics Committee.

Results
A total of 172 patients with COVID-19 pneumonia were attended in the 3 wards by the respective COVID-19 teams. Of them, hyponatremia was observed in 49 (28.5%) patients. No data suggesting "hypervolemia" or registered as such was found in the clinical records of these patients. Whereas the data of hypovolemia was found in 27/49 (55.1%) and of euvolemia in 22/49 (44.9%) patients. The rate of cases with SNa < 130 mmol/L at admission were similar between hypovolemic and euvolemic (22.2 vs. 18.2%, p = 1). No patient in the entire group had SNa < 125 mmol/L at admission. Thirty-four of the 49 patients (69.4) were treated with systemic glucocorticoids during the rst week of the hospitalization. The patients´ general characteristics at admission grouped according to volemic status are described in Table 1.  During hospital evolution, hyponatremia rate progressively decreased from the 2nd -3th hospitalization day to the discharge of patients ( Fig. 1 In accordance to the univariate analysis, patients with obesity had a higher risk to remain hyponatremic at the 2nd -3th hospitalization day (OR 5.1, 95%CI: 1.02 to 25.54), whereas those with chronic obstructive pulmonary disease/smoke habit were in a lower risk (OR 0.29, 95%CI: 0.18 to 0.46). With the exception of volemic status at admission -as shown in previous paragraph-, nor the comorbidities nor previous treatment at admission, nor administration of glucocorticoids during hospitalization were associate with hyponatremia at 7th -10th day. Likewise, no difference of means or medians of IM was found between those who persisted with hyponatremia at 7th -10th day and those who did not.
On the end of the data collection, from the 49 initial followed patients, 37 (75.5%) were discharged, with 2 (4.1%) previously admitted to ICU, 5 (10.2%) remained hospitalized, and 7 (14.3%) dead. Death occurred before the 7th day in all these patients. Patients who dead were > 70 years old, and 3 of them were hyponatremic at their 2nd -3th hospitalization day.
No comorbidity or previous treatment at admission, neither sex, age, initial volemic status, IM at admission, hyponatremia at 2nd -3th day nor administration of glucocorticoids during hospitalization were statistically associated with death in the univariate analysis. In a multivariate analysis including age, sex, and the admission variables: SNa, D-Dimer, lactate dehydrogenase, and C-reactive protein, a higher SNa level was found to be independently associated with a decrease in the risk of mortality were found between hypovolemic and euvolemic ( Table 2). We did not nd signi cant correlations between IM levels and SNa at admission when all patients were analyzed together. However, SNa from hypovolemic patients was correlated with their brinogen levels at admission (r: -529, p = 0.035). We observed that C-reactive protein, lactate dehydrogenase and brinogen levels statistically decreased from admission to the 7th -10th hospitalization day in the entire cohort (Table 3). Parallelly, an increase in SNa levels as well as a reduction in the hyponatremia rate was also observed at the same time. However, we did also not encounter correlations between IM levels and SNa at the 7th -10th hospitalization day, nor when patients were analyzed in the global cohort nor when were analyzed according to their initial volemic status. Although IM levels had a trend to be higher in hypovolemic than euvolemic at the 7th -10th hospitalization day, no statistical differences were found (Table 2). Furthermore, in the 11 patients who remained hyponatremic at the 7th -10th hospitalization day, no correlation was found between IM levels and SNa.

Discussion
The results of the current study suggest that the evolution of both hypovolemic or euvolemic hyponatremia during the hospitalization of the patients with COVID-19 pneumonia is linked to their in ammatory status. Indeed, the decrement of IM levels during hospitalization was accompanied by an increase in SNa and a reduction in the rate of hyponatremia at the same time. This permits us, therefore, hypothesize that the presence of hyponatremia in COVID-19 patients, either hypovolemic or euvolemic, could depend of the in ammation generated in these patients.
As in pneumonia caused by other infections (11,12), hyponatremia is also a risk factor for a poor hospital prognosis in patients with COVID-19 pneumonia (4,5,9,13). In fact, a lower SNa at admission have been related with a higher risk for mortality (5). The current study found similar results. Although most of the patients in our cohort had mild hyponatremia, with no patient with SNa below 125 mmol/L, we observed that per each increment in mmol/L of SNa at admission, the risk of mortality reduced by a 57%. However, the mechanisms behind the impact of the decreased SNa levels or hyponatremia on hospital mortality are totally elucidated.
A common discussion is about whether hyponatremia, especially when mild, is a direct cause for increased mortality risk or whether it is only another "indicator" of a worse clinical situation. Based on our results, we hypothesize that the latter would be the most probable setting, and some ndings of our study support that idea. First, we observed a parallel decrement of the rate of hyponatremia as IM levels were decreasing. This suggests that an in ammatory status could predispose to the development of hyponatremia and that the degree of hyponatremia could be a re ection of the degree of the in ammation. Thus, when in ammation decreases, a recovering in the water/sodium balance would be expected, as did in our cohort. In ammatory pathways via different cytokines, such as tumoral-necrosis factor-α, IL-1 or IL6, are involved in a non-osmotic release of arginine-vasopressin (14). In fact, it has been reported that the risk of hyponatremia is higher in the setting of an acute in ammatory phase, especially when patients are receiving hypotonic uids (15,16), indicating an inadequate antidiuresis. Beukhof et al. in a case-control study of patients with hyponatremia, observed that those who developed hyponatremia had a coincident increase in C-reactive protein levels(16). Likewise, IL-6, a usual IM, has been strongly corelated with arginine-vasopressin concentrations in blood, and directly related with the development of SIADH, both in children (17) and adults (18,19). Recently, Berni et al. (9) found an inverse moderate correlation between IL-6 levels and SNa in 29 COVID-19 patients. Furthermore, they observed a signi cative increment of SNa when hyponatremic patients with elevated IL-6 levels were treated with tocilizumab (humanized monoclonal antibody against the IL-6 receptor). Second, we found that patients who remained hyponatremic during hospitalization were not at high risk of death. This is in contrast with the expected if mild hyponatremia could induce a direct damage and therefore increases the mortality risk. And third, the early death of the patients, with only 3/7 of them with hyponatremic at the 2nd -3th hospitalization day, suggests that the clinical severity was principally caused by COVID-19 pneumonia itself rather than hyponatremia. Therefore, in agreement with Berni et al., we suspect that hyponatremia is another sign of a major severity in COVID-19 patients, which is depending of the in ammatory involvement affecting respiratory function. In fact, higher respiratory rates(4,5)and lower oxygen levels (4) have been reports as more frequent in hyponatremic than eunatremic COVID-19 patients.
However, we cannot deny that even mild hyponatremia could impair the metabolic cellular response to in ammation in different tissues, and therefore, perpetuate the damage already stablished by the in ammation itself. Thus, the role of hyponatremia seems to be more complex. In fact, in the HOPE-COVID-19 study(5) -with a predominantly mild hyponatremia cohort-hyponatremia was also associated with a higher risk for sepsis. Furthermore, in that study the combination of sepsis with hyponatremia increased the mortality risk of these patients (5). Sepsis, as an in ammatory status, can cause Na + and Ca + + channel dysfunction (20). Hyponatremia itself can also alter the normal function of Na + and Ca + + channels (21,22).A combination of both factors could thus harm the mechanisms of voltage-dependent cellular conduction and therefore impair several cellular functions, which can lead to organ/tissue dysfunction (e.g., in the kidney (23)) and consequently induce multi-organ failure. Therefore, we encourage physicians to properly correct hyponatremia in all clinical situations.
One would expected that SIADH, an euvolemic hyponatremia is the main cause of hyponatremia in COVID-19 patients, as some authors have proposed (8). Furthermore, it would be in accordance with the commented above. However, we nd a similar rate of hypovolemic and euvolemic hyponatremia. Unfortunately, our study lacked the power to determine speci c etiologies of hyponatremia, since nor adrenal axis nor the withdrawal of SIADH-related drugs, nor urine ions studies, nor data of clinical losses of volemia were fully evaluated and registered in the clinical records of our cohort. Nevertheless, the fact that the IM levels were similar between hypovolemic and euvolemic, both at admission and at 7th -10th hospitalization day, indicates that in both types of hyponatremia, the in ammatory status played a similar role. Therefore, although other factors may be involved in the development of hyponatremia depending on whether it is hypovolemic or euvolemic, it is a fact, in our opinion, that the in ammatory status of patients with COVID-19 pneumonia would favor its presence.
A weakness of the study is the size of the series. Other is the lack of a more detailed information about hospital treatment of the patients. However, at that moment of the pandemic, different protocols were used in different hospital, so these data would have been heterogeneous and biased. Nevertheless, the current data was enough to permit the main objective to be evaluated. Unfortunately, IL-6 was not measured in all patients, which decreased the statistical potency to assay this variable. Likewise, no hypervolemic hyponatremia was encountered in our registry, therefore whether hypervolemic hyponatremia has the same in ammatory characteristics than hypovolemic or euvolemic hyponatremia remains unknown. The main strength is that this study highlights the role of hyponatremia as a clinical marker of severity and in ammation in COVID-19 patients, which could help physicians stratify patient risk in order to take clinical decisions.
In conclusion, in our COVID-19 series, the evolution of admission hyponatremia during the hospitalization coincided with the improvement of IM levels. This evolution was similar in both hypovolemic and euvolemic hyponatremia. Therefore, the implication of the in ammatory status on the development of hyponatremia in COVID-19 patients is proposed.