Positive fluid balance and mortality in adult patients treated with extracorporeal membrane oxygenation: A retrospective study

Abstract

Background

Intravenous fluid is important for resuscitation and maintenance of circuit flow in patients with extracorporeal membrane oxygenation, but fluid overload is widely recognized as detrimental in critically ill patients. This study aimed to evaluate the association between positive fluid balance and outcomes in adult patients treated with extracorporeal membrane oxygenation.

Methods

This was a retrospective observational study of a tertiary hospital from October 2010 to January 2018. Patients aged ≥18 years who received extracorporeal membrane oxygenation for ≥48 h were included. The fluid balance was determined as the difference between fluid intake and fluid output, and the cumulative fluid balance was calculated as the sum of these values on the preceding days. The primary outcome was hospital mortality.

Results

Of the 123 included extracorporeal membrane oxygenation episodes, 79 were venovenous extracorporeal membrane oxygenation. The hospital mortality rate was 31.7%. Seventy-eight patients underwent continuous renal replacement therapy during their extracorporeal membrane oxygenation course. Non-survivors had a greater cumulative fluid balance (p≤0.001) and a lower cumulative fluid output (p = 0.006) than survivors on day 7. Fluid intake was not significantly different between survivors and non-survivors (p = 0.583). In the multivariate analysis, the cumulative fluid balance (per litre) on day 7, but not on day 3, was associated with increased hospital mortality (adjusted OR: 1.17, 95% CI: 1.06–1.29, p = 0.001).

Conclusions

In adult patients treated with extracorporeal membrane oxygenation, a higher positive cumulative fluid balance on day 7 was associated with increased hospital mortality. The association between positive fluid balance and mortality was mainly influenced by lower fluid output rather than an increase in fluid intake.

Keywords

Extracorporeal life support outcome adult

Introduction

Extracorporeal membrane oxygenation (ECMO) serves as a rescue therapy for patients with acute respiratory failure or refractory cardiogenic shock. The use of ECMO has been increasing since the H1N1 influenza outbreak in 2009.¹ Despite years of experience, fluid management remains challenging. Intravenous fluids are often administered in the early resuscitation phase and for maintenance of ECMO flow. However, the adverse effects of excessive fluid are well recognized, with increased mortality in critically ill patients,^2–4 and are consistent across patients with sepsis^5,6 and acute kidney injury (AKI).^7–9 Whether this positive fluid balance is mainly influenced by excessive intravenous fluid administration or inadequate fluid removal is not well established.¹⁰ Furthermore, the data on the fluid management of patients with ECMO remain limited.

The aim of this study was to evaluate the association between positive fluid balance and outcomes in adult patients treated with ECMO.

Methods

Setting

The study was conducted in a 21-bed medical-surgical intensive care unit (ICU) of a tertiary hospital in Hong Kong. This hospital provided cardiothoracic surgical services and received ECMO referrals from neighbouring hospitals. The study was approved by the Research Ethics Committee (Kowloon Central/Kowloon East) of Hospital Authority (Reference KC/KE-17-0245/ER-1). The study was reported according to the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) statement.¹¹

Study design and data collection

We performed a retrospective analysis of consecutive patients aged ≥18 who received ECMO for more than 48 h from 1 October 2010 to 31 January 2018. The exclusion criteria were ECMO for <48 h, management of ECMO in other centres, age <18, or repeat ECMO episodes within one hospital admission.

The following data were retrieved from electronic patient records and the Clinical Information System (Intellispace Critical Care and Anesthesia; Philips): age, gender, body weight, Acute Physiology and Chronic Health Evaluation (APACHE) IV score,¹² Sequential Organ Failure Assessment (SOFA) score¹³ and type of ECMO (venovenous (VV) versus venoarterial (VA)). Data on daily fluid intake (intravenous fluid, blood products, enteral and parenteral nutrition and continuous renal replacement therapy (CRRT) replacement fluids), urine output and fluid output (urine output, dialysis effluent-dialysate from CRRT, gastrointestinal loss and drain output) were extracted from nursing charts. The daily fluid balance was calculated by subtracting the fluid output from the fluid intake. The cumulative fluid balance was expressed as the sum of the fluid balance on the preceding days. Plasma lactate on day 1 of ECMO, the use of CRRT and diuretics, and bleeding events were also recorded. The severity of AKI was defined according to the risk, injury, failure, loss of kidney function and end-stage kidney disease (RIFLE) consensus.¹⁴ Baseline creatinine was extracted using a rule described elsewhere.¹⁵ Bleeding events were defined as clinically overt bleeding recorded in the medical and/or nursing charts, which was associated with a sudden decrease in haemoglobin ≥2 g/dL over 24 h, required intervention, intracerebral haemorrhage or death. Data collection was complete, with no missing data identified.

ECMO support and management

The decision to start ECMO was made by the attending intensivist for refractory cardiogenic shock or acute respiratory failure. Absolute contraindications for ECMO support included advanced malignancy, progressive and non-recoverable cardiac or respiratory disease, chronic severe pulmonary hypertension with right ventricular failure, intracranial haemorrhage and immunosuppression. Multiple trauma and contraindications to anticoagulation were the relative contraindications for ECMO support. The standard ECMO circuit included either a Rotaflow device or Cardiohelp device (Getinge), and an oxygenator (Quadrox PLS or HLS, Getinge). The connection tubings were Bioline-coated (Getinge) which was a heparin-albumin complex coating designed to reduce blood/circuit interaction. VV ECMO and VA ECMO cannulation was performed percutaneously by intensivists, while central VA ECMO cannulation was performed by both cardiothoracic surgeons and intensivists.

Fluid administration was not protocolised but was based on the attending physician's discretion, primarily related to access insufficiency as evidenced by frequent chattering of the drainage cannula or a decrease in venous pressure. Crystalloids were used as the maintenance solutions. The choice of fluid for resuscitation included normal saline, balanced solutions and gelatin. The decision to use CRRT was made by the attending physicians. Typical indications for CRRT included a potassium level >6 mmol/L, metabolic acidosis with pH <7.2, creatinine level > 300 µmol/L, significant positive fluid balance or organ oedema. The use and dosage of loop diuretics in patients with oliguria and positive fluid balance were at the discretion of attending physicians. CRRT was performed via the ECMO circuit. In Rotaflow PLS circuit, blood entered the CRRT machine (Prismaflex™; Gambro, Lund, Sweden) distal to the pump-head and was returned proximal to the oxygenator. In Cardiohelp HLS circuit, the CRRT access line was connected to the oxygenator inlet port and the return line was connected to the post-oxygenator connector. The protocolised dosage of CRRT was 30 mL/kg/h. Unfractionated heparin was the standard anticoagulant with the dose titrated against activated clotting time.

The primary outcome was hospital mortality. Secondary outcomes included 90-day mortality, dialysis-independence at 90 days and length of the ICU and hospital stay.

Statistical analyses

Data were examined for normality by visual inspection of histograms and the Shapiro–Wilk test. Continuous variables are presented as the medians with interquartile ranges, and categorical variables are presented as numbers with percentages. Differences between groups were assessed using the Mann–Whitney U test, the chi-square test, Fisher's exact test and the Kruskal–Wallis test as appropriate.

Logistic regression was used to identify the independent risk factors for hospital mortality in ECMO patients, with results presented as odds ratios (ORs) with 95% confidence intervals (CIs). Logistic regression instead of survival analysis was chosen because a prolonged hospitalisation for patients who eventually died in the hospitals would unlikely be truly beneficial to the patients.¹⁶ Independent variables were selected with reference to the univariate analyses and biological plausibility. Collinearity was checked before modelling using a variance inflation factor with a cutoff of 2.5.¹⁷ Sensitivity analysis was performed using different ICU scoring systems to reflect disease severity, which included SOFA scores on the first day of ECMO and non-renal SOFA scores with AKI defined by RIFLE.

Statistical analyses were performed using R (version 3.4.2) with Rstudio (version 1.1.423) using the tidyverse car package.^18–20 All statistical tests were two-sided, and a p value <0.05 was considered significant.

Results

Participants

Of the 170 episodes of ECMO during the study period, 123 episodes (55 (41–62) years old, 64.2% male) were included in the analyses (Figure 1). The main characteristics of the study population are presented in Table 1. Seventy-nine patients (64.2%) received VV ECMO. Table 2 describes the indications for ECMO support. The median ICU length of stay was 14.9 (9.1–25.7) days, and the median hospital length of stay was 23.7 (14.1–39.8) days (Table 1). The overall hospital mortality was 31.7%. More VV ECMO patients survived to hospital discharge. Non-survivors had higher APACHE IV scores, higher SOFA scores on day 1, and higher plasma lactate levels. The use of diuretics was more common in survivors, while the use of CRRT was more common in non-survivors. Bleeding events were associated with increased hospital mortality. The ICU mortality was 27.6%, while the 90-day mortality was 30.9%. The leading causes of hospital mortality were infection (33.3%) and intracranial bleeding (25.6%) (Table 3). Nineteen deaths (48.7%) were due to elective ECMO termination.

Figure 1.

Patient flowchart. ECMO: extracorporeal membrane oxygenation; VA: venoarterial; VV: venovenous.

Table 1.

Patient characteristics and outcomes by hospital survival status.

	All (n = 123)	Survivors (n = 84)	Non-survivors (n = 39)	p value
Age (years)	55.0 (41.0–62.0)	53.5 (40.8–60.0)	60.0 (45.0–64.5)	0.071
Male	79 (64.2%)	53 (63.0%)	26 (66.7%)	0.853
Body weight (kg)	60.0 (55.0–70.0)	61.3 (55.0–70.0)	60.0 (55.0–67.5)	0.376
VV ECMO	79 (64.2%)	61 (72.6%)	18 (46.1%)	0.008
APACHE IV	91 (69–123)	80 (63.8–97.8)	122 (98.5–149.5)	<0.001
Already on CRRT before ECMO	20 (16.2%)	12 (14.3%)	8 (20.5%)	0.542
SOFA scores on day 1	14.0 (12.0–16.0)	14.0 (12.0–15.0)	15.0 (13.5–17.0)	0.006
Cardiovascular subscore	4.0 (4.0–4.0)	4.0 (4.0–4.0)	4.0 (4.0–4.0)	0.370
Renal subscore	1.0 (0–2.0)	1.0 (0–2.0)	2.0 (1.0–2.0)	0.005
Coagulation subscore	1.0 (0–2.0)	1.0 (0–2.0)	2.0 (1.0–2.0)	0.060
Lung subscore	4.0 (3.0–4.0)	4.0 (3.0–4.0)	4.0 (3.0–4.0)	0.874
Hepatic subscore	0 (0–1.0)	0 (0–1.0)	0 (0–1.5)	0.017
Neuro subscore	4.0 (4.0–4.0)	4.0 (4.0–4.0)	4.0 (4.0–4.0)	0.589
Non-renal SOFA scores on day 1^a	13.0 (11.0–14.0)	13.0 (11.0–14.0)	14.0 (12.0–15.0)	0.079
Plasma lactate (mmol/L) on day 1	4.1 (2.3–9.0)	3.1 (2.0–5.9)	9.1 (3.6–11.8)	<0.001
AKI defined by RIFLE
No AKI	29 (23.6%)	25 (29.8%)	4 (10.3%)	0.032
Risk	13 (10.6%)	12 (14.3%)	1 (2.6%)	0.060
Injury	18 (14.6%)	13 (15.5%)	5 (12.8%)	0.910
Failure	63 (51.2%)	34 (40.4%)	29 (74.4%)	<0.001
Use of loop diuretics	91 (74.0%)	71 (84.5%)	20 (51.3%)	<0.001
Use of CRRT	78 (63.4%)	42 (50.0%)	36 (92.3%)	<0.001
Use of loop diuretics and CRRT	15 (12.2%)	9 (10.7%)	6 (15.4%)	0.555
Bleeding event^b	54 (44.0%)	29 (34.5%)	25 (64.1%)	0.004
ECMO duration (days)	6.1 (4.5–8.7)	6.0 (4.4–7.9)	6.9 (12.6–18.0)	0.092
ICU length of stay (days)	14.9 (9.1–25.7)	14.0 (9.5–25.3)	17.8 (7.9–26.5)	0.931
Hospital length of stay (days)	23.7 (14.1–39.8)	25.4 (16.2–42.4)	20.6 (8.8–35.1)	0.020

Data are presented as the medians (interquartile range) or n (%).

AKI: acute kidney injury; APACHE: Acute Physiology and Chronic Health Evaluation; CRRT: continuous renal replacement therapy; ECMO: extracorporeal membrane oxygenation; ICU: intensive care unit; RIFLE: risk, injury, failure, loss of kidney function and end-stage renal failure consensus; SOFA: sequential organ failure assessment; VV: venovenous.

Non-renal SOFA scores were the sum of SOFA subscores, excluding the renal subscores.

Bleeding events were defined as clinically overt bleeding recorded in the medical and/or nursing charts, which was associated with a drop in haemoglobin ≥2 g/L over 24 h requiring intervention, intracerebral haemorrhage, or death.

Table 2.

Indications for ECMO support.

VV ECMO (n = 79)	n (%)	VA ECMO (n = 44)	n (%)
Acute respiratory distress syndrome	69 (87.3)	Cardiogenic shock due to coronary artery disease	18 (40.9)
Status asthmaticus	6 (7.6)	Post-cardiotomy	11 (25.0)
Pulmonary haemorrhage	3 (3.8)	Cardiac arrest	9 (20.5)
Others	1 (1.3)	Acute myocarditis	3 (6.8)
		Others	3 (6.8)

ECMO: extracorporeal membrane oxygenation; VA: venoarterial; VV: venovenous.

Table 3.

Causes of death according to type of ECMO.

VV ECMO (n = 18)	n (%)	VA ECMO (n = 21)	n (%)
Infection	9 (50.0)	Neurologic deaths – Intracranial bleeding – Ischaemic stroke – Anoxic brain damage	9 (42.9) 6 1 2
Intracranial bleeding	4 (22.2)	Infection Bowel ischaemia	4 (19.0) 3 (14.3)
Other uncontrolled bleeding	4 (22.2)	Multiorgan failure	2 (9.5)
Futility of anticipating pulmonary recovery	1 (5.6)	Complications from coronary artery disease – Recurrent left ventricular wall rupture – Recurrence of malignant arrhythmia after weaning off ECMO	3 (14.3) 1 2

ECMO: extracorporeal membrane oxygenation; VA: venoarterial; VV: venovenous.

Fluid balance

Figure 2 illustrates the daily fluid status in the first week of ECMO. Urine output was lower in non-survivors (262 mL vs. 1115 mL on day 1, p<0.001), accompanied by less fluid output and a more positive fluid balance. The daily fluid output plateaued at 2270 mL from day 3 in non-survivors, while the fluid output continued to rise in survivors (2764 mL vs. 2038 mL in non-survivors on day 4, p = 0.005). The daily fluid intake was greater in non-survivors than in survivors (5515 mL vs. 4442 mL on day 1, p = 0.060), followed by a downward trend in both groups (3370 mL in survivors vs. 3091 mL in non-survivors on day 4, p = 0.488). This resulted in an initially more positive fluid balance in non-survivors than survivors (4204 mL vs. 2640 mL on day 1, p = 0.003), followed by a gradual decrease in daily fluid balance in both groups (438 mL in survivors vs. 1028 mL in non-survivors on day 7, p = 0.013).

Figure 2.

Median fluid intake, urine output, fluid output and fluid balance in the first seven days of ECMO according to hospital survival status. Daily fluid intake included intravenous fluid, blood products, enteral and parenteral nutrition and CRRT replacement fluids. Fluid output included urine output, dialysis effluent-dialysate from CRRT, gastrointestinal loss and drain output. Daily fluid balance was calculated by subtracting the fluid output from fluid intake.

Table 4 describes the cumulative fluid status from ECMO initiation according to hospital survival status. Again, non-survivors had a more positive cumulative fluid balance than survivors. Lower urine output was consistently observed in non-survivors. The cumulative fluid level decreased significantly more in non-survivors later than in survivors on day 7. Cumulative fluid intake was similar in both groups.

Table 4.

Cumulative fluid status during ICU stay by hospital survival status.

	Overall	Survivors	Non-survivors	p value
Day 3	n = 123	n = 84	n = 39
Fluid intake	12,701 (10,668–15,655)	12,627 (10,555–14,362)	13,299 (11,070–17,555)	0.273
Urine output	3485 (627–5823)	4477 (1403–6159)	437 (78–3225)	<0.001
Fluid output	5906 (4318–7699)	6024 (4525–7726)	5575 (2949–6908)	0.089
Fluid balance	6886 (4983–9812)	6696 (4896–8569)	8714 (5164–12,114)	0.027
Day 7	n = 111	n = 80	n = 31
Fluid intake	26,846 (22,503–29,905)	26,875 (22,214–29,284)	26,326 (23,230–31,856)	0.583
Urine output	10,474 (1583–17,428)	13,704 (3884–17,901)	1247 (165–10,082)	<0.001
Fluid output	17,452 (13,687–20,915)	17,786 (15,020–21,215)	14,268 (11,057–18,655)	0.006
Fluid balance	9852 (5579–11,882)	9025 (4966–10,904)	11,729 (9054–18,705)	<0.001

Note: Fluid intake included intravenous fluid, blood products, enteral and parenteral nutrition and CRRT replacement fluids. Fluid output included urine output, dialysis effluent-dialysate from CRRT, gastrointestinal loss and drain output. Fluid balance was calculated by subtracting the fluid output from fluid intake. Fluid intake, urine output, fluid output and fluid balance were calculated with the incorporation of all the values from the preceding days. Twelve patients died before day 7. Data are presented as the medians (mL) (interquartile range).

Figure 3 illustrates the evolution of fluid management on days 1, 3 and 7 for patients on ECMO over the study period. When the study period was divided into specific time periods (2010–2013, 2014–2016 and 2017–2018), no significant difference in terms of fluid balance was observed on day 1 (p = 0.188), day 3 (p = 0.499) or day 7 (p = 0.801) over the study period.

Figure 3.

The evolution of net fluid balance on day 1, 3 and 7 for patients on ECMO over the study period. The net fluid balance was determined as the difference between the fluid intake and the fluid output on the corresponding day after ECMO initiation. VA: venoarterial; VV: venovenous.

Acute kidney injury

No patient in this study received renal replacement therapy before ICU admission. Twenty patients (16.2%) were already on CRRT at ECMO initiation. Among the patients who were not on CRRT at ECMO initiation, AKI occurred in 74 (71.8%) patients during the course of ECMO. The number of patients with RIFLE failure was higher in non-survivors than in survivors (67.7% vs. 37.5%, p = 0.009). There was no significant difference in the amount of cumulative fluid intake in patients with or without RIFLE failure on day 3 (p = 0.594) and day 7 (p = 0.081). All patients who were put on CRRT during their course of ECMO were dialysis-independent at 90 days in this study.

Multivariate analyses

The cumulative fluid balance on day 3 and day 7 was introduced alternately into separate logistic regressions because of collinearity. The cumulative fluid balance on day 3 was not associated with hospital mortality (adjusted OR: 1.09, 95% CI: 0.97–1.23, p = 0.155). However, the cumulative fluid balance on day 7 was independently associated with increased hospital mortality (adjusted OR: 1.17, 95% CI: 1.06–1.29, p = 0.001) (Table 5). The other variable associated with hospital mortality was the SOFA-hepatic subscore. There was no significant interaction between cumulative fluid balance and the use of CRRT (p = 0.617).

Table 5.

Factors associated with hospital mortality according to the logistic regression model.

	Crude OR (95% CI)	p value	Adjusted OR (95% CI)	p value
Cumulative fluid balance on day 7 (per litre)^a	1.14 (1.06–1.23)	<0.001	1.17 (1.06–1.29)	0.001
SOFA subscores on day 1 of ECMO
Cardiovascular subscore	1.23 (0.79–1.90)	0.355	0.96 (0.48–1.91)	0.911
Renal subscore	1.58 (1.12–2.23)	0.009	1.28 (0.73–2.23)	0.346
Coagulation subscore	1.36 (0.97–1.90)	0.074	0.90 (0.56–1.46)	0.668
Lung subscore	0.89 (0.65–1.22)	0.472	0.96 (0.55–1.69)	0.896
Hepatic subscore	1.93 (1.18–3.17)	0.009	3.72 (1.63–8.49)	0.002
Neuro subscore	0.94 (0.47–1.88)	0.851	0.68 (0.26–1.76)	0.423
VV ECMO	0.32 (0.15–0.71)	0.005	0.98 (0.20–4.82)	0.984
Lactate on day 1 of ECMO (mmol/L)	1.22 (1.11–1.35)	<0.001	1.19 (0.99–1.43)	0.055
CRRT within the first seven days of ECMO	6.80 (2.42–19.08)	<0.001	3.31 (0.76–14.39)	0.111
Bleeding within the first seven days of ECMO^b	2.92 (1.33–6.41)	0.008	1.92 (0.62–5.92)	0.257

CRRT: continuous renal replacement therapy; ECMO: extracorporeal membrane oxygenation; SOFA: sequential organ failure assessment; VV: venovenous.

Cumulative fluid balances were calculated by the incorporation of the values from the first seven days of ECMO.

Subgroup analysis

The results of subgroup analysis stratified according to the modes of ECMO are shown in Figure 4. Cumulative fluid balance on day 7 was associated in hospital mortality in both the VA and VV ECMO subgroups (VA ECMO, OR: 1.32, 95% CI: 1.02–1.70, p = 0.021; VV ECMO, OR: 1.19; 95% CI: 1.02–1.38, p = 0.022).

Figure 4.

Adjusted odds ratio of cumulative fluid balance for VA and VV ECMO. The cumulative fluid balance (per litre) on day 3 and day 7 was introduced alternately into separate logistic regressions with the following covariates: lactate, SOFA subscores on the day of ECMO initiation, need for renal replacement therapy and presence of bleeding events.

Sensitivity analyses

The association between the cumulative fluid balance on day 7 and hospital mortality remained significant after substituting SOFA subscores with APACHE IV scores, SOFA scores on day 1, or non-renal SOFA scores together with AKI defined by RIFLE (Table 6). After restricting the analysis to patients with RIFLE failure, the cumulative fluid balance on day 7 was still an independent risk factor for hospital mortality (Table 7). The results remained similar when the analysis was categorized by specific time periods (2010–2013, 2014–2016 and 2017–2018) (Table 8).

Table 6.

Multivariate logistic regression of factors associated with hospital mortality in sensitivity analyses.

	Model 1^a		Model 2^b		Model 3^c
	Adjusted OR (95% CI)	p value	Adjusted OR (95% CI)	p value	Adjusted OR (95% CI)	p value
Cumulative fluid balance on day 7 (per litre)	1.10 (1.01–1.20)	0.033	1.11 (1.03–1.21)	0.011	1.11 (1.02–1.21)	0.014
APACHE IV score	1.02 (1.00–1.04)	0.028	NA	NA	NA	NA
SOFA scores on day 1 of ECMO	NA	NA	1.14 (0.92–1.39)	0.228	NA	NA
Non-renal SOFA scores on day 1 of ECMO^d	NA	NA	NA	NA	1.13 (0.88–1.45)	0.354
AKI-RIFLE
No AKI	NA	NA	NA	NA	Ref.	Ref.
Risk	NA	NA	NA	NA	0.40 (0.03–5.70)	0.500
Injury	NA	NA	NA	NA	0.71 (0.10–5.06)	0.729
Failure	NA	NA	NA	NA	0.55 (0.08–3.64)	0.539
VV ECMO	1.48 (0.31–7.02)	0.621	0.86 (0.19–3.90)	0.848	0.88 (0.19–4.19)	0.876
Lactate on day 1 of ECMO (mmol/L)	1.10 (0.93–1.31)	0.270	1.13 (0.96–1.34)	0.144	1.14 (0.96–1.35)	0.136
CRRT within the first seven days of ECMO	3.18 (0.74–13.67)	0.120	4.88 (1.21–19.63)	0.026	7.66 (1.23–48.25)	0.030
Bleeding within the first seven days of ECMO	1.65 (0.58–4.69)	0.348	1.50 (0.53–4.25)	0.446	1.47 (0.52–4.18)	0.472

AKI: acute kidney injury; CRRT: continuous renal replacement therapy; ECMO: extracorporeal membrane oxygenation; NA: not applicable; Ref: reference; SOFA: sequential organ failure assessment; VV: venovenous.

Model 1 included: Cumulative fluid balance on day 7, APACHE IV score, VV ECMO, lactate on day 1, CRRT within the first seven days of ECMO and bleeding within the first seven days of ECMO.

Model 2 included: Cumulative fluid balance on day 7, total SOFA score on day 1, VV ECMO, lactate on day 1, CRRT within the first seven days of ECMO and bleeding within the first seven days of ECMO.

Model 3 included: Cumulative fluid balance on day 7, non-renal SOFA scores on day 1, AKI defined by RIFLE, VV ECMO, lactate on day 1, CRRT within the first seven days of ECMO and bleeding within the first seven days of ECMO.

Non-renal SOFA scores were calculated by subtracting the renal subscores from the total SOFA scores.

Table 7.

Multivariate logistic regression of factors associated with hospital mortality in patients with RIFLE failure (n = 63).

	Adjusted OR (95% CI)	p value
Cumulative fluid balance on day 7 (per litre)^a	1.18 (1.03–1.35)	0.015
SOFA scores on day 1 of ECMO, excluding renal subscore
Cardiovascular subscore	1.05 (0.44–2.47)	0.915
Coagulation subscore	0.93 (0.50–1.71)	0.805
Lung subscore	1.09 (0.50–2.39)	0.829
Hepatic subscore	6.94 (1.69–28.53)	0.007
Neuro subscore	0.65 (1.87–2.25)	0.496
VV ECMO	1.40 (0.20–9.92)	0.737
Lactate on day 1 of ECMO (mmol/L)	1.31 (1.03–1.67)	0.028
Bleeding within the first seven days of ECMO^b	2.16 (0.46–10.24)	0.332

ECMO: extracorporeal membrane oxygenation; SOFA: sequential organ failure assessment; VV: venovenous.

Cumulative fluid balances were calculated with the incorporation of the value from the first seven days of ECMO.

Bleeding events were defined as clinically overt bleeding recorded in the medical and/or nursing charts, which were associated with a drop in haemoglobin ≥2 g/L over 24 h requiring intervention, intracerebral haemorrhage or death.

Table 8.

Multivariate logistic regression of factors associated with hospital mortality adjusted for year.

	Adjusted OR (95% CI)	p value
Cumulative fluid balance on day 7 (per litre)^a	1.16 (1.06–1.28)	0.002
SOFA scores on day 1 of ECMO
Cardiovascular subscore	1.07 (0.53–2.16)	0.861
Renal subscore	1.22 (0.70–2.14)	0.480
Coagulation subscore	0.91 (0.55–1.48)	0.696
Lung subscore	1.03 (0.58–1.83)	0.925
Hepatic subscore	4.03 (1.68–9.69)	0.002
Neuro subscore	0.82 (0.29–2.35)	0.717
VV ECMO	0.98 (0.19–5.06)	0.982
Lactate on day 1 of ECMO (mmol/L)	1.14 (0.94–1.39)	0.170
CRRT within the first seven days of ECMO	3.57 (0.82–15.51)	0.090
Bleeding within the first seven days of ECMO^b	2.11 (0.66–6.71)	0.205
Year
2010–2013	Reference	Reference
2014–2016	1.56 (0.31–7.96)	0.592
2017–2018	3.79 (0.56–25.45)	0.170

CRRT: continuous renal replacement therapy; ECMO: extracorporeal membrane oxygenation; SOFA: sequential organ failure assessment; VV: venovenous.

Cumulative fluid balances were calculated by the incorporation of the values from the first seven days of ECMO.

Bleeding events were defined as clinically overt bleeding recorded in the medical and/or nursing charts that were associated with a decrease in haemoglobin ≥2 g/L over 24 h, required intervention or resulted in intracerebral haemorrhage or death.

Discussion

In this retrospective, observational, single-centre study of more than seven years, we examined the influence of positive fluid balance on outcomes in adult patients treated with ECMO. The key findings of this study were as follows: A more positive fluid balance was found in non-survivors; non-survivors had lower urine and fluid outputs; and fluid intake was not associated with hospital mortality. After adjusting for potential confounders, the cumulative fluid balance on day 7, but not on day 3, was independently associated with hospital mortality.

There is ample evidence showing the negative impact of fluid overload in critically ill patients. The results from the multicentre SOAP study showed that fluid balance was an independent risk factor for mortality in patients with sepsis.²¹ Another multicentre retrospective cohort study including over 18,000 patients by Balakumar et al. showed that either a positive or negative fluid balance, compared with an even fluid balance, was associated with increased one-year mortality.² To the best of our knowledge, few studies have addressed fluid management in patients on ECMO. Schmidt et al. reported that a positive fluid balance on ECMO day 3 was an independent predictor of 90-day mortality.⁴ However, there are still several knowledge gaps that have not been fully addressed in ECMO patients.

First, the effect of fluid balance is time-dependent, and its relationship with the phases of resuscitation has been observed. The framework of different fluid resuscitation phases was proposed in 2014: Rescue, Optimization, Stabilization and De-escalation.²² Shum et al. showed that fluid balance on the second plus third days and the total fluid balance during the ICU stay were positively associated with hospital death, but a positive fluid balance on the first ICU day was negatively associated with hospital mortality.²³ In the post hoc analyses of the Vasopressin in Septic Shock Trial (VASST), a more positive fluid balance, early resuscitation phase and cumulative fluid balance over four days were associated with increased mortality in patients with septic shock.⁶ In a worldwide prospective audit, Sakr et al. reported that a higher cumulative fluid balance by day 3, but not within the first 24 h after ICU admission, was associated with mortality.³ In our study, we demonstrated that the cumulative fluid balance by day 7, but not by day 3, was associated with hospital mortality. A possible explanation for this delayed effect is that patients in our cohort who required ECMO generally had greater disease severity than other patient groups and therefore required a more prolonged rescue phase for fluid resuscitation.

Second, the optimal fluid balance in ECMO patients is not well established. Both a positive and negative fluid balance was shown to be associated with mortality in the critically ill.^2,4 A positive fluid balance can be a result of excessive fluid administration, inadequate fluid excretion or both. Teixeira et al. showed that a higher mean fluid balance and a lower urine volume were associated with mortality, while the use of diuretics was associated with better survival in patients with AKI.⁸ In Balakumar's study, the mortality risk associated with a positive fluid balance was reduced by the use of RRT.² Our findings support that fluid therapy can be administered safely in ECMO patients in the resuscitation phrase, even if they have a positive fluid balance. This concurred with the findings from Kim who suggested that the risk of death did not increase until a certain ‘threshold’ of fluid overload was reached in patients on ECMO.²⁴ On the other hand, fluid removal in ECMO patients is often challenging, particularly in the sicker patients with multiorgan failure, as evident from the higher APACHE score and lactate level in patients who had the greater positive fluid balance. Intravenous fluid is important in resuscitation, maintaining intravascular volume and to avoid access insufficiency. The large volume of fluid required is secondary to the systemic inflammatory response, caused by the ECMO circuit itself and the underlying disease process, leading to peripheral vasodilation and fluid leak to the interstitial compartment.²⁵ As a result, a negative fluid balance is often possible only after the initial resuscitation phase. Further prospective study on the impact of negative fluid balance in the later stage of resuscitation would be valuable.

Third, the relationship between fluid overload and kidney injury is complex. Fluid balance has been shown to be greater in patients with rather than without AKI. Physiologically, fluid overload may play a role in the development of AKI because of organ oedema in encapsulated kidneys, which increases interstitial pressure and compromises renal blood flow.²⁶ Fluid administration to patients with endothelial dysfunction may result in both fluid overload and AKI.¹⁰ It remains unclear whether fluid overload itself causes AKI or is just a marker of disease severity. However, based on our analysis, even after adjusting for AKI or restricting the analysis to patients with RIFLE failure, the cumulative fluid balance on day 7 was still associated with hospital mortality. Therefore, the adverse effects from fluid overload are likely not solely mediated by AKI. Furthermore, whether the fluid output by native kidneys or artificial means via renal replacement therapy would improve renal outcome and mortality is uncertain.

Lastly, the mechanisms by which fluid overload causing adverse outcome remain poorly understood. It has been postulated that fluid overload may contribute to increased risk of sepsis in critically ill.²⁶ Renal interstitial oedema causing AKI may itself dampen innate immunity and increased infection risk.²⁷ Ileus secondary to gut oedema may promote bacterial translocation.^28,29 Hepatic congestion may lead to impaired synthetic function and liver plays an important role in removal of circulating endotoxin.³⁰ Further human studies are needed to elucidate the mechanisms behind.

The strength of this study was the inclusion of detailed analyses of the various components of fluid balance, namely, fluid intake, urine output and fluid output. The breakdown of fluid balance has not been reported in previous studies. This analysis enabled the exploration of the predominant factor of positive fluid balance. The results were robust, as shown in the sensitivity analyses using different disease severity scores, restriction to RIFLE failure patients, and taking into account the effect of evolution of treatment over the study period. A separate analysis of the daily fluid balance over the study period also indicated no significant change in terms of fluid management strategy in the unit.

There were several limitations of this study. The major limitation was its retrospective observational nature. We could only conclude the presence of an association, rather than causality, of various predictors with mortality because of possible undocumented confounders. However, we adjusted for relevant confounders, and the results were robust using different ICU scoring systems. Another limitation was the small sample size because ECMO is a rescue therapy for the most critically-ill. Third, 28% of the screened patients were excluded from the analysis, which may affect the representativeness of the study population. However, all excluded patients died within 48 h after ECMO initiation, while the remaining study cohort had a length of stay that was sufficient reflect the adverse effects of fluid overload. Furthermore, undetectable water loss was not taken into account, although the water loss was unlikely to be significant.³¹ Finally, the type of intravenous fluid administration was not standardized at the time of the study. Further prospective large-scale studies on dose and type of intravenous fluid are suggested.

Conclusions

In summary, this study demonstrated a significant association between positive fluid balance and hospital mortality in adult patients treated with ECMO. The association with mortality was mainly influenced by a decrease in fluid output rather than an increase in fluid intake. Further studies on the role of aggressive fluid removal are warranted.

Footnotes

Declaration of conflicting interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This study was supported by a Kowloon Central Cluster research grant from the Hospital Authority (KCC/RC/G/1819-A01). The Hospital Authority was not involved in the development of the study concept or the analysis or interpretation of the data.

References

Patroniti

Zangrillo

Pappalardo

, et al. The Italian ECMO network experience during the 2009 influenza A (H1N1) pandemic: preparation for severe respiratory emergency outbreaks. Intensive Care Med 2011; 37: 1447–1457.

Balakumar

Murugan

Sileanu

, et al. Both positive and negative fluid balance may be associated with reduced long-term survival in the critically ill. Crit Care Med 2017; 45: e749–e757.

Sakr

Rubatto Birri

PNR

Kotfis

, et al. Higher fluid balance increases the risk of death from sepsis: results from a large international audit. Crit Care Med 2017; 45: 386–394.

Schmidt

Bailey

Kelly

, et al. Impact of fluid balance on outcome of adult patients treated with extracorporeal membrane oxygenation. Intensive Care Med 2014; 40: 1256–1266.

Acheampong

Vincent

. A positive fluid balance is an independent prognostic factor in patients with sepsis. Crit Care 2015; 19: 251.

Boyd

Forbes

Nakada

, et al. Fluid resuscitation in septic shock: a positive fluid balance and elevated central venous pressure are associated with increased mortality. Crit Care Med 2011; 39: 259–265.

Payen

de Pont

Sakr

, et al. A positive fluid balance is associated with a worse outcome in patients with acute renal failure. Crit Care 2008; 12: R74.

Teixeira

Garzotto

Piccinni

, et al. Fluid balance and urine volume are independent predictors of mortality in acute kidney injury. Crit Care 2013; 17: R14.

Wang

Jiang

Zhu

, et al. Fluid balance and mortality in critically ill patients with acute kidney injury: a multicenter prospective epidemiological study. Crit Care 2015; 19: 371.

10.

Ostermann

Straaten

HMO

Forni

. Fluid overload and acute kidney injury: CAUSE or consequence?. Crit Care 2015; 19: 443.

11.

von Elm

Altman

Egger

, et al. The strengthening the reporting of observational studies in epidemiology (strobe) statement: guidelines for reporting observational studies. Epidemiology 2007; 18: 800–804.

12.

Zimmerman

Kramer

McNair

, et al. Acute Physiology and Chronic Health Evaluation (APACHE) IV: hospital mortality assessment for today's critically ill patients. Crit Care Med 2006; 34: 1297–1310.

13.

Vincent

J-L

Moreno

Takala

, et al. The SOFA (Sepsis-related Organ Failure Assessment) score to describe organ dysfunction/failure. Intens Care Med 1996; 22: 707–710.

14.

Bellomo

Ronco

Kellum

, et al. Acute dialysis quality initiative. Acute renal failure – definition, outcome measures, animal models, fluid therapy and information technology needs: The Second International Consensus Conference of the Acute Dialysis Quality Initiative (ADQI) Group. Crit Care 2004; 8: R204–R212.

15.

Pickering

Endre

. Back-calculating baseline creatinine with MDRD misclassifies acute kidney injury in the intensive care unit. Clin J Am Soc Nephrol 2010; 5: 1165–1173.

16.

Schoenfeld

. Survival methods, including those using competing risk analysis, are not appropriate for intensive care unit outcome studies. Crit Care 2005; 10: 103.

17.

Allison PD. Multiple regression: a primer. Thousand Oaks: Pine Forge Press, University of Pennsylvania, 1999.

18.

R Core Team, R. A language and environment for statistical computing, Vienna: R Foundation for Statistical Computing, 2017.

19.

Wickham H. Tidyverse: easily install and load the ‘Tidyverse’. R package version 1.2.1; 2017, https://CRAN.R-project.org/package=tidyverse (28 June 2019).

20.

Fox

Weisberg

. An R companion to applied regression, 2nd edn. Thousand Oaks: Sage, 2011.

21.

Vincent

Sakr

Sprung

, et al. Sepsis in European intensive care units: results of the SOAP study. Crit Care Med 2006; 34: 344–353.

22.

Hoste

Maitland

Brudney

, et al. Four phases of intravenous fluid therapy: a conceptual model. Br J Anaesth 2014; 113: 740–747.

23.

Shum

Lee

FMH

Chan

, et al. Interaction between fluid balance and disease severity on patient outcome in the critically ill. J Crit Care 2011; 26: 613–619.

24.

Kim

Paek

Song

, et al. Permissive fluid volume in adult patients undergoing extracorporeal membrane oxygenation treatment. Crit Care 2018; 22: 270.

25.

Sidebotham

McGeorge

McGuinness

, et al. Extracorporeal membrane oxygenation for treating severe cardiac and respiratory failure in adults: part 2 – technical considerations. J Cardiothorac Vasc Anesth 2010; 24: 164–172.

26.

Claure-Del Granado

Mehta

. Fluid overload in the ICU: evaluation and management. BMC Nephrol 2016; 17: 109.

27.

Singbartl

Kellum

. AKI in the ICU: definition, epidemiology, risk stratification, and outcomes. Kidney Int 2012; 81: 819–825.

28.

Mehta

Bouchard

Soroko

, et al. Program to improve care in acute renal disease (PICARD) study group. Sepsis as a cause and consequence of acute kidney injury: program to improve care in acute renal disease. Intensive Care Med 2011; 37: 241–248.

29.

Butcher

Liu

. Fluid overload in AKI-epiphenomenon or putative effect on mortality?. Curr Opin Crit Care 2012; 18: 593.

30.

Valentova

Von Haehling

Doehner

, et al. Liver dysfunction and its nutritional implications in heart failure. Nutrition 2013; 29: 370–378.

31.

Li Li

Oi Yan

Ming Chit Arthur

, et al. Insensible water loss through adult extracorporeal membrane oxygenation circuit: an in vitro study. ASAIO J 2014; 60: 508–512.