Impact of Negative Fluid Balance on Mortality and Outcome of Patients with Confirmed COVID-19

Purpose Maintaining the proper fluid balance is a fundamental step in the management of hospitalized patients. The current study evaluated the impact of negative fluid balance on outcomes of patients with confirmed COVID-19. Methods We considered the negative fluid balance as a higher output fluid compared to the input fluid. The fluid balance was categorized into four groups (group 4: −850 to −500 ml/day; group 3: −499 to −200 ml/day, group 2: −199 to 0 ml/day, and group 1 : 1 to 1000 ml/day) and included ordinally in the model. The outcomes were all-cause mortality, length of hospitalization, and improvement in oxygen saturation. Results The fluid balance differed significantly among nonsurvivors and survivors (MD: −317.93, 95% CI: −410.21, −225.69, and p < 0.001). After adjusting for potential confounders, there was a significantly lower frequency of mortality in patients with negative fluid balance compared to the controls (aRR: 0.69, 95% CI: 0.57, 0.84, and p < 0.001). Similarly, the length of hospitalization was significantly shorter in the negative fluid balance group in comparison to the control group (aMD: −1.01, 95% CI: −1.74, −0.28, and p=0.006). Conclusion We determined that the negative fluid balance was associated with favorable outcomes in COVID-19 patients. The negative fluid balance was associated with the reduced mortality rate and length of hospitalization as well as improvement in oxygen saturation. Moreover, the NT-proBNP >781 pg/mL and fluid balance >−430 mL might be the predictors for positive fluid balance and mortality, respectively.


Introduction
In December 2019, severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) emerged in Wuhan, China, and spread rapidly worldwide. Death through COVID-19 could result from acute respiratory distress syndrome (ARDS) and respiratory failure, cytokine storm and subsequent multiorgan failure, and cardiovascular events. Although respiratory failure is one of the major causes of death among COVID-19 patients, the high mortality rates among the intensive care unit (ICU)admitted patients suggest the multifactorial nature of COVID-19 [1]. Moreover, comorbidities increase the risk of mortality in COVID-19 patients. Twenty-six percent of deaths have an underlying disease, and sixty-fve percent have cardiovascular disorders. Patients with cardiovascular disease are at a higher risk of severe disease development [2].
Increased production of infammatory cytokines and bradykinin contributes to the increased permeability of pulmonary vessels and fuid accumulation in lung interstitial and alveolar space. Tis process leads to pulmonary hypertension and exerts an additional overload on the right ventricle [3]. It is worth mentioning that cardiomyopathy and heart failure due to COVID-19 could be associated with fuid overload [4]. Brain natriuretic peptide (BNP) and Nterminal (NT)-pro hormone BNP (NT-proBNP) are quantitative markers representing cardiac hemodynamic stress and have an important role in diagnosis and following up the patients with heart failure [5]. Acute renal failure (ARF) is another adverse event, especially in critically ill patients [6]. Te ARF leads to fuid overload and electrolyte imbalance, which might deteriorate the pulmonary edema [7]. In the same line, ARDS is frequently associated with shock, and proper fuid management contributes to adequate tissue perfusion while avoiding positive fuid balance and tissue edema is necessary [8]. Moreover, it has been suggested that angiotensin-converting enzyme II (ACE-II) is higher in COVID-19 patients compared to the healthy population, which upregulates aldosterone and contributes to sodium and water retention [9].
Studies have showed that early resuscitation with intravenous fuids is essential for the improvement of hemodynamic stability, tissue perfusion, and mortality reduction in ICU admitted patients. Moreover, studies showed that negative fuid balance might be associated with promising outcomes in ARF, heart failure, and ARDS in non-COVID-19 patients. However, the negative fuid balance efcacy in COVID-19 patients is not well known. Additionally, it has been suggested that positive fuid balance could deteriorate the patients' general condition by afecting the renal, cardiac, and pulmonary functions. Te fuid overload is associated with pulmonary and peripheral edema, increased chance of respiratory failure, delayed weaning from mechanical ventilation, and poor prognosis [8]. We aimed to characterize the association between negative fuid balance and mortality, length of hospitalization, and oxygen saturation in COVID-19 confrmed patients. We also hypothesized that the level of NT-proBNP might be a potential predictor for furosemide administration.

Design and Setting.
Tis retrospective cohort study was conducted in two university hospitals in the Babol city, north of Iran, from July 2020 to May 2021. Tis study has been approved by the Ethics Committee of Babol University of Medical Sciences (the ethical code: IR.MUBA-BOL.REC.1399.425). Before enrolment, written informed consent was obtained from study participants.

Patient Characteristics.
All adult patients (both general ward and ICU) in the two university hospitals were screened for inclusion and exclusion criteria. We defned the cohort as hospitalized patients with positive reverse transcriptasepolymerase chain reaction (RT-PCR) for SARS-CoV-2. An infectious disease specialist verifed the positive RT-PCR tests and confrmed the infection with clinical symptoms and computed tomography (CT) scan fnding (ground glass opacifcation). Inclusion criteria included (1) age ≥18 years old and (2) normal and acceptable range of serum electrolytes. Te exclusion criteria were (1) age <18 years old; (2) pregnancy; (3) development of major arrhythmic events, including sudden cardiac death (SCD), sudden cardiac arrest (SCA), ventricular fbrillation (VF), and supraventricular tachycardia (sVT); and (4) changing the treatment protocol such as the application of plasmapheresis or adding a new antiviral agent. We defned critical illness as the presence of respiratory failure, septic shock, and/or multiorgan failure. Te treatment protocol was remdesivir, dexamethasone, and heparin. Moreover, 166 patients received invasive ventilation and six patients needed dialysis.

Data Collection at Recruitment.
We retrieved patients' characteristics such as demographics, clinical features, comorbidities, and laboratory fndings on admission from medical records. A trained team of researchers reviewed and retrieved data for each patient.

Fluid Status Assessment.
Te daily net fuid balance was calculated by subtracting daily fuid output from daily fuid input. Fluid input included oral, enteral, and intravenous fuid. We used calibrated containers for measuring the oral and enteral fuid volume. Fluid output included urine and fuid loss from drains and tubes. Urine volumes were collected and measured via Foley catheter and urine bag. Fluid input and output were calculated (milliliters per day) and charted every 24 hours. We considered the negative fuid balance as a higher output fuid compared to the input fuid. We applied oral and intravenous fuid restriction and administered furosemide (loop diuretic). We adjusted the fuid balance based on the presence of congestion, vital signs, clinical fndings, and laboratory fndings such as the creatinine level. After determining the proper fuid balance in each patient, we maintained the established amount during the hospitalization. However, we overlooked minor changes in fuid balance during hospitalization.

Outcomes and Defnition.
Te primary outcome was allcause mortality. Te length of hospitalization and improvement in O 2 saturation were selected as the key secondary endpoints.
2.6. Statistical Analysis. All statistical analyses were performed with STATA 16 (Stata Corp, College Station, TX, USA) and GraphPad Prism 9 software (GraphPad Software Inc., La Jolla, CA, USA). Te baseline characteristics were compared between the two groups using the independent sample t-test for continuous data and a chi-square test for categorical data. Prior to statistical comparisons, all data were tested for normal distribution by the Kolmogorov-Smirnov test and visually by the Q-Q plot. We used receiver operating characteristic (ROC) curves to identify the classifcation threshold of NT-proBNP and fuid balance.
We reported diagnostic indices such as area under the curve (AUC), sensitivity, specifcity, and likelihood ratio (LR). We estimated the risk ratios (RRs) and 95% confdence intervals (CIs) for mortality using a negative log-binomial regression model. Te linear mixed efects models were ft to assess changes from the baseline within the groups and diferences of those changes between groups with respect to oxygen saturation during hospitalization and the length of hospital stay. Fluid balance was categorized into four groups and included ordinally in the model (Tables 1 and 2). Te magnitude of the efect is presented as the adjusted mean diference (aMD) and its 95% confdence interval. Te independent variables included in the model were fuid balance, age, gender, prior heart failure, chronic kidney disease, and critical illness. Te signifcance was defned as p < 0.05.

Results
Te recruitment of patients was commenced in July 2020 and ended in May 2021. During eleven months, a total number of 315 patients were evaluated and 294 patients were included (according to inclusion and exclusion criteria). Te demographic and baseline characteristics of the patients are summarized in Table 3. One hundred forty-two patients (48.29%) were female, the mean age of the patients was 64.33 ± 16.38 years (median: 66, IQR: 77 − 56 years), and the proportion of patients with critical illness was 63.6%. Two hundred forty patients (81.63%) had at least one pre-existing condition. Te mean age was 66.77 years (SD � 16.56) in deceased patients and 63.01 (SD � 16.21) in the survivors (p � 0.064). Furthermore, deceased patients had almost the same frequency of pre-existing medical problems (82/101, 81.18%) compared with the control group (158/193, 81.86%) (p � 0.886). Te laboratory results showed that interleukin-6 (IL-6) (MD: 55.4, 95% CI: 8.71, 102.09; p � 0.020) and creactive protein (CRP) (MD: 26.14, 95% CI: 2.71, 49.57; p � 0.029) were signifcantly higher in deceased patients compared to the survivors. No statistically signifcant differences existed in the creatinine levels on the day of hospital admission and before hospital discharge or death (Change diference: 0.06, 95% CI: −0.1, 0.1; p � 0.943). Te mean ± (SD) of NT-proBNP in all patients at the time of admission was 8254 ± (10061.2) pg/mL. Moreover, we measured the NT-proBNP before discharging survived patients. Te NT-proBNP was signifcantly lower in survived patients before discharge compared to the time of admission (MD: 4758.5, 95% CI: 3725.2, 5791.7; p < 0.001).
One hundred ninety-six patients (66.66%) achieved negative fuid balance. In the two hundred ninety-four patients included in this study, the mean ± (SD) of fuid balance was −268. Te unadjusted result showed a signifcantly lower frequency of mortality in patients with negative fuid balance compared to the controls (RR: 0.31, 95% CI: 0.19, 0.51, and p < 0.001). Moreover, after adjusting for potential confounders, there was a signifcantly lower frequency of mortality in patients with negative fuid balance compared to the controls (aRR: 0.69, 95% CI: 0.57, 0.84, and p < 0.001).
After the enrolment day, the length of hospitalization was signifcantly shorter in the negative fuid balance group compared to the positive fuid balance group (MD: −0.73, 95% CI: −1.45, −0.01, and p � 0.046). Similarly, after adjusting for potential confounders, the length of hospitalization was signifcantly shorter in the negative fuid balance group in comparison to the controls (aMD: −1.01, 95% CI: −1.74, −0.28, and p � 0.006) ( Table 2). Figure 3 shows the mean of O 2 saturation for patients with negative and positive fuid balance at each time-point. Both groups showed a signifcant increment in oxygen saturation over time. However, during the hospitalization, the negative fuid balance group showed a signifcantly greater increment compared to the positive fuid balance group (aMD: 1.10, 95% CI: 0.41, 1.80, and p � 0.035).

Discussion
We showed that negative fuid balance was associated with a decreased mortality rate, even after adjusting for the potential confounders such as the presence of critical illness. We categorized the fuid balance into four groups and included them ordinally in the model. For every one level increase in the fuid balance category, there is a decrease of ∼30% in the mortality rate (e.g., −850 to −500 ml group had ∼30% lower mortality rate compared to −499 to −200 ml group). Te same model was applied for the length of hospitalization, and as is shown in Table 2, for every one level increase in the fuid balance group, there is a decrease of one day in the length of hospitalization. Tese fndings, alongside the improvement in oxygen saturation, provide a promising approach for managing COVID-19 in critically ill patients.
Te ventricle pressure-volume relationship is curvilinear, and the atrial pressure increases rapidly as the ventricle reaches the plateau of the Frank-Starling curve. Elevated atrial pressure contributes to pulmonary and venous hydrostatic pressure overload, associated with the intravascular fuid shift to interstitial space and disturbed capillary blood fow and oxygen saturation [10]. Te above efect is more remarkable in patients with COVID-19 due to the active infammatory process in the lungs, which disturbs the lungs' function. Furthermore, many patients with COVID-19 develop ARDS, which contributes to increased capillary permeability and pulmonary edema and puts patients in more danger against resuscitation-induced lung congestion [11]. Moreover, following the aggressive fuid resuscitation, increased atrial pressure contributes to elevated venous pressure in abdominal organs (retrogradely), including the liver, kidneys, and intestine. Te kidneys are susceptible to increased interstitial pressure and tubular compression due to renal intracapsular tamponade, which intensifes adverse events [11].
Both negative and positive fuid balance groups had low oxygen saturation at the beginning of admission. However, at the time of admission, the negative fuid balance had lower oxygen saturation compared to the positive fuid balance group. Pulmonary edema secondary to ARDS or heart failure could be the possible reason for low oxygen saturation in patients [12,13]. Moreover, the negative fuid balance group showed a signifcantly higher oxygen increment compared to the positive fuid balance group. Tis fnding suggests that negative fuid balance might apply benefcial efects by reducing the pulmonary edema. In the same line, Kevorkian et al. reported that the combination of corticosteroids and furosemide improves the patients' outcomes, especially in aged patients with comorbidities and who are at a risk of pulmonary edema (BNP >100 ng/mL). However, they evaluated the efcacy of furosemide in noncritically ill COVID-19 patients [14].
Te present study showed that males were associated with higher mortality rates compared to females. Tis fnding is consistent with other studies. Te studies suggested that COVID-19 has higher severity and mortality in males compared to females, especially in ICU-admitted patients [15,16]. Tis fnding has been reported in several countries [17]. Moreover, a meta-analysis, including 120 studies and 125,546 participants, showed that severity and mortality are higher among males [18]. Several mechanisms have been suggested for the abovementioned fndings, including sexbased diferences in the expression of ACE2 receptor and transmembrane protease serine 2 (TMPRSS2) and sex-based diferences in immunological response [19]. However, the benefcial efects of negative fuid balance remained statistically signifcant after adjusting the gender as a confounder.
Hypoxemic respiratory failure (ARDS) is the most common cause of admission to the ICU and is responsible for more than 80% of deaths [2]. It has been shown that the elevated level of cytokines such as IL-6 is associated with poor prognosis and an increased risk of ARDS [20]. In the same line, in the present study, IL-6 was signifcantly higher in deceased patients compared to survived patients and might imply the role of ARDS in disease prognosis. Hemodynamic instability frequently occurs in patients with ARDS, especially in ICU [21,22]. More than 60% of patients with ARDS develop hemodynamic instability, and catecholamine administration is necessary in 65% of cases [21,23]. Similar to other conditions with hemodynamic instability, it might be thought that aggressive fuid administration should be the mainstay of the treatment strategy. In several studies, it has been noted that aggressive fuid resuscitation is associated with poor outcomes in critically ill non-COVID-19 patients [24][25][26]. Nevertheless, increased intravascular volume next to fuid resuscitation Te fnal multivariable models were adjusted for the aforementioned risk factors in the table. * Fluid balance was categorized into four groups (group 4: −850 to −500 ml/day, group 3: −499 to −200 ml/day, group 2: −199 to 0 ml/day, and group 1 : 1 to 1000 ml/day (positive fuid balance)). Fluid balance was included ordinally in the model. * * Signifcant at p value <0.05. Te fuid balance and critical illness are signifcant (p value < 0.05).    [27]. It has been shown that AKI is a common (5.1% to 29%) complication in critically ill COVID-19 patients, associated with high mortality, especially in ICU-admitted patients [28]. Moreover, critically ill children with COVID-19 could develop AKI, and elevated IL-6 is a common fnding in these patients, emphasizing the role of cytokine storm as an underlying cause of acute renal injury [29]. High-levels of IL-6 and CRP, as we saw in this study, might be a predictor of renal injury in critically ill patients [30]. Fluid overload is an independent risk factor for ICU mortality, and it necessitates preventing fuid overload in COVID-19 patients, especially in critically ill patients with impaired renal function [31,32]. In the present study, no statistically signifcant diferences were observed in creatinine levels between the survivor and nonsurvivor groups on the day of hospital admission and before hospital discharge or death. However, the creatinine levels were higher than the healthy population, which might indicate the COVID-19-related AKI.
Several studies evaluated the impact of loop diuretic administration and negative fuid balance on ICU-admitted patients and conditions such as AKI. Moreover, there is no consistency between these studies. Some suggested benefcial efects [33,34], while others noted harmful or no efects [35][36][37]. Our results are in the same line with the study of Francisco Santos et al. [9], which evaluated the role of negative fuid balance in COVID-19 patients' outcomes. However, we have a much bigger sample size (294 vs. 20). Moreover, we adjusted the critical illness as an important confounder and determined the role of NT-proBNP level as a predictive tool for loop diuretic administration.
In addition to respiratory failure and shock, critically ill patients with COVID-19 might manifest heart failure [13]. Studies showed that an elevated level of NT-proBNP is strongly associated with a higher mortality rate in COVID-19 patients [38,39]. Te sensitivity analysis suggested that furosemide administration is helpful in the concentration of NT-proBNP >781 pg/mL. In other words, NT-proBNP >781 pg/mL is associated with positive fuid balance. Moreover, the fndings showed that the patients who had survived had a statistically signifcant reduction in the level of NT-proBNP. Tis fnding might be due to improved cardiac function through reduced volume overload.  Physicians should be vigilant during the aggressive fuid administration in COVID-19 patients who have declined cardiac output secondary to cytokine storm, heart failure, or other causes that has no benefcial efects, and overload might lead to adverse results due to pulmonary infltration that contributes to diminished respiratory function.
As our study showed, the negative fuid balance in critically ill patients was associated with better outcomes. Murphy et al. noted that the administration of adequate and early resuscitation fuid therapy besides conservative late fuid therapy, in patients with acute lung injury, is associated with better outcomes compared to the patients with only one of these approaches or none of the two [40]. Terefore, a dynamic approach and optimal timing (negative fuid balance secondary to early and adequate resuscitation) are the keys to fuid management of patients with hemodynamic instability.
We eliminated an important confounder, critical illness, which was not adjusted in similar studies. Additionally, we used a large database with high granularity. However, the current study has several limitations. First, this is a retrospective study, and we cannot eliminate the residual confounding due to unobserved factors. Furthermore, we did not include insensible perspiration in the fuid balance calculation. In addition, we did not evaluate the NT-proBNP levels in deceased patients. Moreover, we are unable to derive a defnite mechanism for the benefcial efects of diuretics and negative fuid balance. Accordingly, we highly recommend prospective studies in this context. It is essential to explore whether more precise fuid management through invasive methods such as central venous pressure measurement or pulmonary arterial catheterization or noninvasive methods such as bioimpedance spectroscopy or echocardiography is necessary or not in a particular subset of patients, including those with underlying heart or renal failure.
In summary, we determined that negative fuid balance was associated with favorable outcomes in COVID-19 patients. Te increase in oxygen saturation and decrease in mortality and length of hospitalization may be due to the improvement in the cardiovascular and pulmonary function and tissue perfusion. Although we cannot securely determine that negative fuid balance is benefcial, it is highly likely that negative fuid balance afords COVID-19 patients' material beneft. In the absence of compelling contradictory data from a randomized, blinded clinical trial, we should encourage maintaining negative fuid balance in COVID-19 patients, especially critically ill patients and higher NT-proBNP levels.