The Correlation Between Early Net Fluid Balance and The Clinical Outcomes of Patients Receiving Extracorporeal Cardiopulmonary Resuscitation

Background: Large amounts of uid administration are often needed during the extracorporeal cardiopulmonary resuscitation (ECPR) course. Finally, uid overload is often present. Methods: Adult patients on ECPR admitted to our center from May 2015 to December 2020 were included. Net uid balance for consecutive 4 days after ECPR was recorded. The primary outcome was to intensive care unit(ICU) survival . We used multivariable logistic regression to assess the association between uid status and clinical outcome. Results: A total of 72 patients were enrolled and divided into two groups: the survivor group and the non-survivor group. The overall rate of ICU survival was 44.4%. Daily uid balance(DFB) in the survivor group was lower than that in the non-survivor group at day 4 (-11.47[-19.74,8.7] vs -5.08[-12.94,13.9]ml/kg, P=0.046), as was CFB over the rst 4 days (-36.03[-51.45,19.03] vs -7.22[-32.79,21.02] ml/kg, P=0.009). Both continuous renal replacement therapy(CRRT) and CFB from days 1-4 were signicantly correlated with ICU survival(OR=14.617[95%CI:1.344,48.847], P=0.028; OR=1.261[95%CI1.091,1.375], P=0.003, respectively). CFB from day 1-4 was determined to have a roughly linear association with the log odds of ICU survival. Conclusions: Early negative uid balance maybe associated with ICU survival in patients receiving ECPR.


Background
The standard therapy for cardiac arrest (CA) is cardiopulmonary resuscitation (CPR).Recent studies have reported 20% survival to hospital discharge in patients after in-hospital CA (IHCA) and 10% after outhospital CA (OHCA) [1][2][3].To improve the neurological outcome and survival rate in patients with CA, extracorporeal cardiopulmonary resuscitation (ECPR) has been proposed in some selected cases of refractory CA.Once ECPR has been initiated, frequent blood and uid transfusion may be required to maintain the targeted mean arterial pressure (MAP) and blood ow of extracorporeal membrane oxygenation (ECMO).Excessive positive uid administration may, in turn, result in cardiogenic/noncardiogenic pulmonary edema, acute kidney injury (AKI), systemic capillary leakage and so on.In light of this, the restrictive uid balance strategy has been advocated in patients with septic shock, ARDS and AKI [3][4][5].In terms of uid management in patients receiving ECMO treatment, there are few studies reporting the association between uid balance and clinical outcomes [6][7].
To date, no consensus on uid administration strategies has been achieved in patients on ECPR.Given the high incidence of positive uid administration and potential adverse effects caused by inappropriate uid management, it is important to understand the association of uid balance status with clinical outcomes.
Our objective was to determine the prevalence of positive uid balance in patients receiving ECPR and to evaluate the association between uid balance status and survival to intensive care unit (ICU) discharge.

Methods
Study design, setting, and subjects Our retrospective study was conducted in a 17-bed ICU of an inner city, university-a liated teaching hospital.Our ICU ECMO database was accessed to identify all patients from May 2015 to December 2020 who received ECPR.The need for written consent was waived because of the retrospective design of the study.

Study participants and inclusion criteria
A patient's eligibility for ECPR was assessed by the physician on duty who was leading the ECMO team, and this eligibility was based on the following factors: age<70 years, witnessed CA, shockable rhythm, <5 min of estimated no-ow time, <30 min of low-ow time, and no malignant tumor.
The inclusion criteria in our study were as follows: adult patients on ECPR, duration of ECMO≥96 hours.
The exclusion criteria were patients under 18 years of age, pregnancy, duration of ECMO<96 hours, ECMO modality switching and intracranial hemorrhage.

Data collection and outcomes measures
The patients' demographic and laboratory data were retrieved from electronic medical records.The data included age, sex, weight, comorbidities, location of CA, ECMO parameter settings, continuous renal replacement therapy (CRRT) and intra-aortic balloon pump (IABP) initiation, duration of mechanical ventilation (MV), ICU length of stay (LOS), and vasoactive inotropic score (VIS).VIS was calculated as ([(epinephrine+noepinephrine) ug/kg/min]×100+[(dobutamine+dopamine) ug/kg/min]×100+[milrinone µg/kg/min]×15+[vasopressin IU/kg/min]×10000 [8].The amount of input uid and output uid were continuously recorded within 4 days of ECMO initiation.The daily uid balance (DFB) was calculated by the difference between uid inputs and outputs and was divided by the patient's actual body weight.Weight was measured using the automated-weighing bed system upon admission and daily at 8 am.The cumulative uid balance (CFB) was de ned as the cumulative total input uid minus the cumulative total output uid and obtained by the addition of each daily uid balance from ECMO commencing until the day of evaluation.Survival was observed at ICU discharge, and patients were classi ed into the survivor group or non-survivor group according to their clinical outcome.

Statistical analysis
Continuous variables not normally distributed were expressed as median and IQR and compared with Mann-Whitney U test.Normally distributed continuous variables were expressed as mean±SD and compared with Student's t-test.Categorical variables were expressed as percentages and compared with the chi-square test.Kendall's tau-b correlation was used to quantify the relationship between continuous variables and categorical variables.Univariable and multivariable analyses were performed to identify factors associated with mortality at ICU discharge.Associations between clinical outcomes and net uid balance were evaluated with the Cochran-Armitage trend test for binary variables.Statistical signi cance was set at a two-sided P value of less than 0.05.All data were analyzed using GraphPad Prism v8.0 and SPSS software v24.0.

Characteristics of Study Subjects
A total of 84 patients admitted to the ICU for ECPR between May 2015 and December 2020 were enrolled.Eight patients were excluded due to a duration of ECMO support less than 96 hours, three patients were excluded because of ages under 18 years, and one patient was excluded due to ECMO modality switching.Finally, 72 patients were eligible for further analysis (Fig. 1).
The baseline characteristics of the patients are presented in Table 1.The mean age was 42.6±16.3years, and 66.67% (48/72) were males.The average time on V-AECMO was 164.8(98.1,212.3)hours in all patients, while the duration of MV was 173.5(93,233) hours.To investigate the factors associated with survival to ICU discharge, we categorized the enrolled patients into survivor and non-survivor groups according to clinical outcome at ICU discharge.Compared with the non-survivor group, those in the survivor group were more likely to have a lower APACHE II score and to be predisposed to receive CRRT  versus 39[95%CI: 24-64], P=0.033; 37.5% versus 90%; [95%CI:1.3%-39.4%],P=0.001, respectively).In addition, the duration of MV was signi cantly shorter in the survivor group (100.5
The unadjusted association of CFB from day 1-4 with ICU survival was assessed with univariable logistic regression and natural cubic splines.CFB from day 1-4 was determined to have a roughly linear association with the log odds of ICU survival (Fig. 2).

Discussion
There is robust evidence in papers that discourages excessive uid accumulation in critically ill patients with sepsis/septic shock [9][10][11][12].Similar results were also reported in patients on V-VECMO or V-AECMO [6][7][13][14].However, it is unclear whether patients on ECPR should also receive restrictive uid administration.Patients receiving ECPR often require large-volume uid infusion during and after ECPR initiation to guarantee appropriate ECMO blood ow and perfusion pressure [15][16].Therefore, positive uid balance or excessive uid accumulation is sometimes inevitable.To date, there are scarce data that investigate the relationship between early net uid balance status and clinical outcomes in patients on ECPR.
In clinical practice, uid resuscitation is often the cornerstone of early management to stabilize hemodynamics in critically ill patients.Vincent has proposed four distinct phases of uid resuscitation: rescue, optimization, stabilization and de-escalation in patients with sepsis. 17Strategies of uid management emphasizes not only uid resuscitation but also "reversed uid resuscitation", which equals de-escalation.Undoubtedly, intensivists around the world have attached great importance to the rst two phases when critically ill patients manifest macrocirculation and/or microcirculation dysfunction.
However, the last phase is not always given priority to by physicians after patients achieve hemodynamic improvement.Therefore, a positive uid balance commonly occurs in patients, especially during the early phase.Brotfain reported that patients with sepsis/septic shock who had less positive cumulative uid balance had lower ICU and in-hospital mortality (P 0.001 for both ICU and in-hospital mortality, OR:1.04[95%CI:1.02to 10.6]; OR:1.06[95%CI:1.03 to 1.08]) [11].Chao WC also found that a positive CFB from days 1-4 was independently associated with a higher 30-day mortality in critically ill patients with in uenza (HR:1.088[95%CI:1.007,1.074])[18].
Similar conclusions have been drawn by several authors regarding V-AECMO patients with refractory cardiac shock.In a retrospective analysis, Besnier et al. demonstrated that patients with more positive uid balance at day 1 had higher mortality (OR:14.34[95%CI:1.58 to 129.79]).A threshold of 38.8 ml/kg uid balance predicted mortality with a sensitivity of 60% and a speci city of 83% [19].Another retrospective multicenter study enrolling 723 patients on V-AECMO revealed a signi cantly increased risk of 90-day mortality in patients with higher CFB during the rst 3 days after ECMO initiation (HR:1.76[95%CI:1.37 to 2.27], P<0.001).Further analysis found that the relative HR of mortality started to increase signi cantly when CFB exceeded 82.3 ml/kg [20].Therefore, higher uid balance was consistently linked to poor outcomes.
Similarly, CA is frequently considered as a sepsis-like syndrome [21].Patients with CA typically encounter circulatory collapse and subsequently undergo a systemic in ammatory response, including pathologic vasodilation, increased capillary leakage, and low albumin levels.Large-volume uid and other drugs are often administered to maintain intravascular volume and improve cardiac output, especially during the early stage during and after ECPR [22][23].As mentioned above, liberal uid resuscitation is actually not recommended in critically ill patients with sepsis.Thus, it is still unknown which uid administration strategy is preferred in patients receiving ECPR.In this study, excessive CFB from day 1-4 was found to be independently associated with ICU survival.To date, few studies have evaluated the impact of uid balance in patients receiving ECPR.A study conducted by Staudacher admitted 195 patients on V-AECMO due to refractory cardiogenic shock and found no evidence to support a liberal uid strategy [7].Of note, 149 of recruited patients developed IHCA or OHCA, which implied that the study might have comprised several patients receiving ECPR.However, further subgroup analysis was not available.
In our center, as long as macrocirculation and microcirculation got improvement, "reversed uid resuscitation" was always put on the agenda during the ECMO course.Therefore, except for the daily uid balance of non-survivors at day 1, a negative net uid balance was acquired at all other investigated points within 4 days of ECMO commencement.Moreover, the difference in CFB between the two groups became more statistically signi cant over time.In addition, the application of CRRT was more common in the non-survivor group, which we thought accounted for more uid administration in the non-survivor group.Strict uid control has several advantages, including effective preload reduction, decreased cardiac-wall stress, avoidance of uid congestion and high hydrostatic pressure, tissue edema alleviation, and so on.Actually, the conception of "less is more" should be recommended by intensivists during the management process of critically ill patients [24][25][26].The core value of the conception is that physiological indexes and medical interventions matching the patient's current pathophysiological state are strongly suggested.In other words, overtreatment should be avoided.Taking uid resuscitation for instance, as long as both macrocirculation and microcirculation perfusion are su cient, restrictive uid management should be considered.Furthermore, if permitted, "reversed uid resuscitation" should also be taken into consideration [27], which implies that it is strongly recommended to initiate the phases of stabilization and de-escalation as early as possible [19].

Limitations
Our study had a number of limitations.First, the de nition of uid inputs only included intravenous uids, and enteral-nutritional-solution intake was not taken into account.On the other hand, uid outputs included urine, adding the net uid balance via CRRT.Drainage of serosal effusions was ignored.Second, all patients in our study had colloid uid administration in conjunction with crystalloid uid.Given the different effects of volume expansion between colloid and crystalloid uids [28][29], the same amount of uid inputs with different crystal-colloid ratios may nally lead to different volume changes.Third, this was a single-center retrospective observational trial.Therefore, our ndings cannot indicate causality.Further studies are warranted to con rm these ndings.

Conclusion
We found a signi cant association between CFB during the rst 4 days after ECMO treatment and survival to ICU discharge in this observational cohort study.Restrictive uid administration may be an appropriate alternative for guiding the management of patients on ECPR.

Table 1 .
Baseline data analysis: univariate comparisons of demographic factors and outcomes

Table 2 .
Details of daily and cumulative uid balance DFB: Daily uid balance, CFB: Cumulative uid balance, CI: Con dence interval

Table 3 .
Univariate and multivariate logistic regression analysis for ICU survival DFB: Daily uid balance, CFB: Cumulative uid balance, CI: Con dence interval OR: odds ratio