Veno-arterial Extra Corporeal Membrane Oxygenation (VA-ECMO) is a salvage therapy used in severe heart failure such as cardiogenic shock or cardiac arrest (1, 2). Patients with acute heart failure often exhibit an inflammatory process with endothelial dysfunction (3–5), and therefore may develop systemic capillary leakage. Moreover, VA-ECMO itself is responsible for inflammation due to the continuous nature of the flow generated by the centrifugal pump, and extensive blood contact with foreign non-endothelialized material (membrane and cannula) (6). Because inflammation can lead to intense vasoplegia (7), these patients most often require large amounts of fluids to maintain an adequate intravascular volume, necessary for the proper functioning of the pump and therefore the blood flow. Moreover, no criteria for fluid responsiveness have been described yet in VA-ECMO patients to optimize vascular filling, leading to approximate and often liberal fluid management. The administration of large amounts of fluids may worsen the fluid shift into interstitial spaces, leading to edema and organ dysfunction (8). In addition, administration of blood products for bleeding events and reduced urine volumes caused by concomitant acute kidney injury play a role in the aggravation of fluid overload. Fluid balance has been extensively studied in the literature as the difference between fluid inputs and outputs. Even if the calculation can be time-consuming, it may be a clinical reflection of the accumulation and overload of fluids in the body. Excess fluid balance has actually been demonstrated to be associated with mortality and morbidity during septic shock (9–11). Nevertheless, the impact of fluid balance has been little studied in patients treated with VA-ECMO, notably during the first hours of device implementation, where the risk of inflammatory process is probably the highest. Thus, we hypothesized that patients with a positive fluid-balance during the very early phase of VA-ECMO therapy would present the highest mortality rate. The main objective of our study was to evaluate the association of early fluid balance and mortality during the first day of VA-ECMO. Secondary objectives were to evaluate the association of fluid balance and adverse outcomes during the first 5 days.
PATIENT AND METHODS
We conducted a retrospective single-center study in the cardiac surgical intensive care unit (ICU) of Rouen University Hospital. Inclusion criteria were patients admitted for mechanical circulatory support with VA-ECMO and surviving at least 12 h after the admission. Exclusion criteria were patients under 18 years old, pregnancy, VA-ECMO therapy less than 12 h, VA-ECMO for septic shock or Acute Respiratory Distress Syndrome, and hemodialysis prior to admission. Indications for VA-ECMO were refractory cardiogenic shock or cardiac arrest refractory to standard resuscitation care but with no-flow<5 min, low-flow<90 min, and ETCO2>10 cm H2O during resuscitation. The refractory nature of the shock was attested by the intensivist or the surgeon. It is usually defined in our daily practice as a low left ventricular ejection fraction and cardiac index (CI <2.2 L/min/m2, evaluated by either echocardiography or invasive measure of blood flow) in association with peripheral signs of shock (systolic blood pressure under 90 mm Hg, hyperlactatemia, organ failure, etc.). The shock was defined as refractory in cases of persistent hypotension despite catecholamines and optimization of volemia if required. A trained surgical team of five surgeons implanted the VA-ECMO device. Briefly, our local protocol for VA-ECMO implementation consists in a small incision of the right or left inguinal region followed by the introduction of a guidewire in a superficial femoral artery and vein using the Seldinger technique, and the insertion of a 17 to 19F arterial cannula and 23 to 25F venous cannula. An additional 7F catheter was inserted in the femoral artery, downstream of the arterial cannula, to prevent leg ischemia. Cannula positions were controlled using X-ray and, if necessary, trans-thoracic or trans-esophageal echocardiography, to place the venous cannula at the atriocaval junction and the arterial cannula in the iliac artery. Blood flow was 2.5 L/min to 5 L/min and the sweep gas flow was set to maintain an arterial carbon dioxide of 4 kPa to 6 kPa (Rotaflow or Cardiohelp system, Maquet, France). Anticoagulation was obtained according to the ELSO international guidelines (12): a bolus of 50 UI/kg to 100 UI/kg of unfractionned heparin was administrated at the time of ECMO cannulation, followed by a continuous infusion of 10 UI/kg/h adjusted for an activated partial thromboplastin time ratio (aPTTr) of [1.8–2.2]. The dose and the aPTTr could be adjusted to the clinical factors (active bleeding, pre-existing risk factors, post-surgical setting, etc.).
The choice of fluids was left to the discretion of the intensivist and it was not possible to standardize the timing of administration and the amount of fluids because of a lack of fluid-therapy tool predicting efficiency. Nevertheless, fluids were usually administered in cases of mean arterial pressure < 65 mm Hg in association with a reduction in pump flow and exaggerated chattering of the venous cannula.
The study protocol was approved by the local institutional ethics board of Rouen University Hospital (Approval number E2016-57) and the need for written consent was waived because of the retrospective design of the study. Data were collected retrospectively for the first 5 days following the implementation of VA-ECMO, using the patient's medical records during the 2013 to 2016 period. Demographic data obtained at admission were age, sex, body mass index (BMI), Simplified Acute Physiologic Score 2 (SAPS-2) and past history of cardiac, respiratory or metabolic chronic diseases. Biologic blood analyses were performed as usual at admission to the unit: blood count, electrolytes and gases, creatinine, alanine (ALAT) and aspartate (ASAT) amino-transferases, bilirubin, arterial lactate, and sensitive cardiac troponin. The biologic analyses were usually repeated daily. Renal failure was classified using the KDIGO classification (13) based on the worst value during the first 5 days after ECMO implementation. The occurrence of major bleeding events was recorded as defined by ELSO guidelines as more than 20 mL/kg/d or the need of more than 10 mL/kg/d of red blood packs or the need for surgical intervention (12, 14). Moreover, because fluid balance is not always easy to calculate in daily practice, we studied variations in weight during ICU stay as another easily attainable criterion of overload. Fluid balance was calculated daily by the difference between fluid inputs (continuous infusion and bolus; fluids used for the administration of drugs also called “creep fluids”; enteral or parenteral nutrition; transfusion (red blood cells, fresh frozen plasmas, platelets)) and outputs (diuresis; ultrafiltration in cases of renal replacement therapy; bleeding; aspiration of gastric content; ascites) and was divided by the patient's body weight at admission. Fluid balance during surgery was also included in the calculation if it occurred during the VA-ECMO therapy. The cumulative fluid balance was obtained by the addition of each daily fluid balance from the starting of VA-ECMO therapy until the day of evaluation. The amount of fluid and its nature was decided by the intensivist to maintain adequate volemia but the type of fluids was recorded (crystalloids including 0.9% saline or other balanced crystalloids; colloids including hydroxyethyl starch 130 kDa 6% or gelatin 4% or albumin 4% or albumin 20%; dextrose). Weight was measured using the same automated weighting bed system at admission and then daily at 8 am, and variations were calculated as difference between admission and daily measures and expressed as percentages. Survival was observed at day-28 and patients were classified as “survivors” or non-survivors.”
The primary objective was to compare the day-1 fluid balance between survivors and non-survivors. Secondary objectives concerned the daily fluid balance during the first 5 days, the cumulative fluid balance over the first 5 days and the main biological and clinical criteria between the two groups.
Because of the absence of normality of the population as observed using a D’Agostino test concerning the primary objective, the results are presented as medians with first and third quartiles. Comparisons between survivors and non-survivors were realized using a two-tailed Mann and Whitney test for continuous values or chi-square test for ordinal values. P < 0.05 was considered statistically significant. Because of the repeated comparisons over the first 5 days for daily fluid balance, cumulative fluid balance, difference in inputs and outputs, and variations in weight, the P value was adjusted according to the number of comparisons in each set of analyses. Survival was then analyzed according to the amount of fluid balance: patients were categorized into quartiles of fluid balance, and survival in the four quartiles was represented using Kaplan–Meier curves and compared using a Log-rank test, with a P value adjusted for the number of comparisons. Univariable and multivariable analyses were then performed to identify factors associated with mortality at day-28. Only data presenting statistical difference in the previous comparisons were included in the logistic regression model using a backward-step process. Age, sex, SAPS-2, indications for VA-ECMO, and year of implementation were included in the model for adjustment on the confounding covariables. Due to the necessary dependence between daily fluid balance, cumulative fluid balance, and weight variations, we only included the results of daily fluid balances in the model. Concerning fluid balance, results were classified as positive (> 0 mL/kg) or negative (≤ 0 mL/kg) because this cut-off value was identified as discriminating for mortality in previous studies (9, 10). Results are presented as odds-ratio (OR) with 95% confidence intervals. An additional survival analysis was realized using a Cox Hazard Proportional modeling with a backward-step process, including the same variables and the time to the censored event (time to death). Results are presented as Hazard-Ratio (HR) with 95% confidence intervals. We then realized receiver operating characteristic (ROC) curves for the clinically significant factors identified with the multivariable analysis. Optimal cut-off values, defined as the best combination of sensitivity and specificity, were determined using the Youden index. To explore alternative methods for the assessment of fluid overload, we correlated weight variations with the cumulative fluid balance each day, using a Pearson correlation test, and present r values with 95% confidence index. A correlation was considered weak for r value < 0.3, moderate for r value [0.3–0.5], and strong for r value > 0.5 (15). These analyses were performed with GraphPad Prism v6.0 (GraphPad, USA) and MedCalc v17.5 (Medcalc software, Belgium).
Over the 2013 to 2016 period, 170 patients were treated with VA-ECMO. After exclusion of noneligible patients (69 patients: 38 patients with VA-ECMO duration < 12 h, 14 incomplete or missing files, eight patients < 18 years, five patients with septic shock, and four pregnant women), data from 101 patients were analyzed and included in the study (Fig. 1). The baseline characteristics of the patients are presented in Table 1. Patients’ median age was 53 [44–61] years and 68.3% were male. Median duration of ECMO support was 4 [3–7] days with no difference between non-survivors and survivors (4 [3–9] vs. 4 [3–6] days, P = 0.39).
There was no difference concerning sex, age, BMI, or pre-existing cardiac diseases between survivors and non-survivors (Table 1). Main etiologies for cardiogenic shock were acute coronary syndrome, poisoning, myocarditis, post-cardiotomy dysfunction, decompensating chronic heart failure or cardiac arrest, and post-cardiac arrest syndrome. A significant difference was observed concerning the different etiologies in the two groups, notably with a higher proportion of cardiac arrest in non-survivors (37.5 vs. 13.2%). SAPS-2 scores before ECMO implementation were higher in non-survivors (71 [55–89] vs. 60 [47–70], P = 0.0008). There was no difference at admission for plasma levels of lactate, pH value, troponin, bilirubin, creatinine, alanine transaminase, or aspartate transaminase. There also no difference between groups concerning the occurrence of a major bleeding (45% in survivors vs. 48% in non-survivors, P = 0.7) and for the total amount of bleeding on the 5 first days after ECMO implementation (2.6 [0–18.5] vs. 3.8 [0–15.9] mL/kg, P = 0.8).
Results are presented in Table 2. At day-1, non-survivors presented a greater positive fluid balance than survivors (47.3 [18.1–71.9] vs. 19.3 [1.5–36.2] mL/kg, P < 0.0001, Fig. 2A). Fluid balance at day-2 was significantly higher in non-survivors (30.6 [14.8–71.0] vs. 10.1 [−9.8–34.7], P = 0.025). No difference was observed from day-3 to day-5 (Fig. 2A).
The cumulative fluid balance increased progressively over time with a significant difference between survivors and non-survivors each day, reaching a cumulation of 107.3 [40.5–146.2] mL/kg in non-survivors versus 53.0 [7.5–74.3] mL/kg in survivors at day-5 (P = 0.038, Fig. 2B).
The analysis of daily total inputs and outputs did not reveal any significant difference except for higher fluid inputs at day-2 for non-survivors (67 [47–102] vs. 49 [35–73] mL/kg, P = 0.045, Supplemental Digital Content 1, https://links.lww.com/SHK/A885). Similarly, no difference was reported for acute renal failure defined by the KDIGO classification over the first 5 days of ECMO therapy, with 60% (n = 29) of stage 3 for non-survivors versus 66% (n = 35) for survivors (P = 0.35), and no difference for the use of Continuous Renal Replacement Therapy during ICU stay (31% (n = 15) in non-survivors vs. 38% (n = 20) in survivors, P = 0.54). At day-1 and day-2, the volume of fluids administered for continuous perfusion and/or resuscitation (association of crystalloids, colloids, and dextrose administration) was significantly higher in non-survivors than in survivors (respectively, 45.7 [30.1–63.8] vs. 34.6 [21.4–51.2] mL/kg, P = 0.02 at day-1 and 36.4 [20.3–52.9] vs. 25.47 [13.99–37.16], P = 0.013 at day-2). The administration of creep fluids (fluids administrated for drugs preparation and infusion) was significantly higher in non-survivors from day-2 to day-5 (Supplemental Digital Content 1, https://links.lww.com/SHK/A885).
Overall 28-day survival was 53% with a median death of 4 [3–9] days (Fig. 3A) and 60% of the non-survivors died within the first 5 days. Deaths were secondary to multiple organ dysfunction in 63% (30/48), brain death in 23% (11/48), limitation of therapeutic efforts in 8% (4/48), major bleeding in 4% (2/48), and thrombosis of ECMO circuit in 2% (1/48). Patients in the 4th quartile of day-1 fluid balance presented reduced survival compared with the 1st and 2nd quartiles (17.4% vs. respectively, 74.1%, P < 0.001 and 66.7%, P = 0.01). Survival in the 3rd quartile was 48.2% (Fig. 3B).
Non-survivors had greater weight gain at day-1 compared with survivors (2.3 [0–5.2] vs. 0.5 [−1.4 to 2.2] %, P = 0.012) and a significant and positive correlation was observed with fluid balance at day-1, but of moderate strength (r = 0.36 [0.17–0.53], P = 0.0003). In the following days, no difference between non-survivors and survivors was observed for weight variations (Table 2). A significant and positive correlation with cumulative fluid balance was observed for each day, but of moderate strength (respectively: r = 0.36 [0.17–0.53], P = 0.0003 at day-1; r = 0.47 [0.28–0.63], P < 0.0001 at day-3; r = 0.50 [0.29–0.66], P < 0.0001 at day-4, and r = 0.41 [0.17–0.60], P = 0.0013 at day-5). A strong correlation was observed between day-2 weight variations and cumulative fluid balance with r = 0.54 [0.37–0.68], P < 0.0001. Correlations are presented in Supplemental Digital Content 2, https://links.lww.com/SHK/A886.
Multivariable logistic regression
After adjustment for confounding covariables (age, sex, VA-ECMO for cardiac arrest, SAPS-2 score at admission and year of VA-ECMO implementation), the occurrence of a positive fluid balance (>0 mL/kg) at day-1 and day-2 was included as categorial variables in logistic regression and Cox hazard-proportional models. The Hosmer and Lemeshow test presented a P value of 0.7, assuming the logistic regression model satisfactorily fitted with data, and chi-square for the Cox modeling was 36 (P < 0.0001), assuming the time to death modeling also satisfactorily fitted with data. Weight variations were not included in the model because of the positive correlation with fluid-balance. Results are presented in Table 3. There were significant and independent associations between 28 days mortality and a positive fluid balance at day-1 (OR = 14.34 [1.58–129.79], P = 0.02) and SAPS-2 score at admission (OR = 1.04 [1.01–1.07], P = 0.02). A positive fluid balance at day-2 was not an independent factor for mortality (2.35 [0.78–7.06], P = 0.1) in this model. No association was observed for age, sex, VA-ECMO implemented for cardiac arrest or the year of implementation in this model. The Cox modeling highlighted an independent association between day-1 and day-2 positive fluid balance and time to death (respectively HR = 8.26 [1.12–60.98], P = 0.04 and HR = 2.89 [1.26–6.65], P = 0.01), but the implementation of ECMO for cardiac arrest was also associated with time to death (HR = 2.38 [1.12–5.08], P = 0.03) whereas SAPS-2 score was not associated.
ROC analyses and determination of cut-off values
ROC curve was performed for day-1 fluid balance at admission with an AUC of 0.749 [0.653–0.843] (Fig. 4). The Youden index identified a cut-off value for the day-1 fluid balance of more than 38.8 mL/kg with a sensitivity of 60% and a specificity of 83%. The positive predictive value for mortality was 76% and the negative predictive value was 70%.
This retrospective study showed that an early and positive fluid balance was associated with worse outcomes in patients treated with VA-ECMO.
Fluid balance and mortality
A positive cumulative fluid balance was identified years ago as a marker of poor outcome in patients with sepsis (9, 10) but few studies have evaluated the impact of fluid balance in VA-ECMO therapy. Schmidt et al. (16) reported a relationship between fluid balance at day-3 and mortality in mixed populations of veno-arterial and veno-venous ECMO but, contrary to our results, the authors did not observe any impact of early fluid balance (day-1 and 2). Nevertheless, they only included patients who survived to day-3, thus excluding patients not requiring short duration support or who early died. Another study on 195 patients observed an association between an early and very high fluid balance and mortality, but most of the patients (76%) were included after a cardiac arrest, limiting the generalizability of the results (17). More recently, a retrospective analysis in 406 patients showed a significant association between mortality and cumulative fluid balance within the first 3 days (18). Cumulative fluid balance probably better reflects the body's accumulation of fluids. The authors observed a median cumulative fluid balance of 64.7 mL/kg at day-3, whereas we observed a slightly lower accumulation (45.0 [11.9–92.9] mL/kg) in our cohort at the same time-point. This study included many patients but to the detriment of a long inclusion period (11 years), during a decade in which fluid-management practices have greatly evolved, whereas we included fewer patients but over a shorter and more recent period, which may explain our lower fluid balance.
The major difference with the previous studies is that we focused on early fluid balance, within the first hours after VA-ECMO implementation. Moreover, we identified a 38.8 mL/kg threshold for day-1 fluid balance as predictive of mortality with good specificity, positive and negative predictive values and an acceptable AUC in our cohort. Indeed, most of ICU scores such as SOFA, SAPS-2, or EuroScore have relatively low AUC (<0.6) in patients under VA-ECMO (19). Some scores specific to this population have been developed such as the SAVE score, whose performance varies according to the studies (AUC between 0.6 and 0.7) (20), or the REMEMBER score, which had an AUC of 0.85 (21). Thus, our value of 0.749 seems relevant in this population and day-1 fluid balance could be of interest as a prognostic marker.
Moreover, despite the obvious association between poor prognosis and fluid overload, it is currently difficult to establish a causal relationship between excessive fluid administration and poor outcomes. Indeed, several studies comparing liberal or restrictive strategies during sepsis are in progress or have not shown any difference (22, 23). Thus, fluid accumulation may reflect the patient's severity as well as the consequence of inadequate fluid management. Prospective randomized trials comparing fluid strategies are mandatory in VA-ECMO patients to assess this question.
We detailed precisely the composition of fluid balance and showed that creep fluids represent an important part of daily inputs, with significantly higher quantities in non-survivors. Creep fluids represented fluids necessary to dissolve and/or administrate drugs. These unintentional fluids were reported in similar proportions in a recent cohort study in critically ill patients (24) and can greatly contribute to positive fluid balance, thus improvements could be proposed to reduce their amount, for example by using enteral route as soon as possible or by reducing the volume used to a maximum.
Moreover, ECMO patients most often require large volumes of fluids, as presented in our results, but optimization of fluid therapy is a major issue. Since almost the entire blood flow is handled by the centrifugal pump, there is little opportunity for the patient to adjust cardiac output to hypovolemia. Therefore, the many tools usually used to guide fluid therapy in conventional ICU patients are not suitable (25).
Although widely studied in the literature, the calculation of fluid balance can be laborious and sometimes inaccurate. We explored the interest of weight variation as an indirect reflection of fluid balance. In our patients, there was a positive correlation but not enough strong to make it a reliable marker (r < 0.6). A previous study observed similar results during septic shock in ICU patients with r value in the same ranges (26). Thus, clinical tools to measure fluid accumulation and edema, and therefore therapeutics, are still lacking in critically ill patients.
Strengths and limitations
Our work has several strengths and limitations that need to be addressed. First, we adjusted the logistic regression model on various factors that could have influenced results. Unlike other studies on the subject, we chose to include patients with a short duration of VA-ECMO, of less than 3 days, allowing us to explore the very early impact of fluid balance whereas most studies explored the effect at day-3. This is of interest because the median duration of VA-ECMO therapy in our cohort was only 4 days and the inflammatory process is probably higher in the first hours. On the other hand, this work is limited by its retrospective and single-center character, but no prospective trial has been designed yet on this topic. The retrospective design of the study limited the number of variables obtained from the medical records such as medications, notably diuretics. We deliberately chose to focus on a recent and short period to limit inclusion bias, particularly with regard to changes in fluid therapy practices over the past two decades. This resulted in a smaller number of patients than in the recent study by Kim et al. and a similar number to Schmidt et al. (16, 18), but over a 4-year period versus 12- and 8-year periods, respectively. Moreover, we adjusted the regression model on the year of implementation of the VA-ECMO to limit bias related to a change in practice related to time. Finally, the inclusion of a broad panel of EMCO indications facilitates the generalizability of data to a current practice. However, the retrospective nature of the work does not allow us to define whether the fluid balance is a reflection of the severity or a causality of the poor prognosis, and therefore we cannot recommend a strategy for fluid administration in the absence of a clinical trial.
In conclusion, an early positive fluid balance is associated with mortality and may be an important and practice-dependent risk factor in VA-ECMO patients, suggesting the need to develop appropriate tools to optimize fluid therapy and to limit the influence of creep fluids. If fluid balance calculation appears to be accurate in identifying fluid overload, variations in weight do not seem accurate enough in patients treated with VA-ECMO.
The authors are grateful to Nikki Sabourin-Gibbs for her help in editing the manuscript.
1. Rastan AJ, Dege A, Mohr M, Doll N, Falk V, Walther T, Mohr FW. Early and late outcomes of 517 consecutive adult patients treated with extracorporeal membrane oxygenation for refractory postcardiotomy cardiogenic shock. J Thorac Cardiovasc Surg
139 (2):302–311, 2010.
2. Shekar K, Mullany DV, Thomson B, Ziegenfuss M, Platts DG, Fraser JF. Extracorporeal life support devices and strategies for management of acute cardiorespiratory failure in adult patients: a comprehensive review. Crit Care
18 (3):219, 2014.
3. Théroux P, Armstrong PW, Mahaffey KW, Hochman JS, Malloy KJ, Rollins S, Nicolau JC, Lavoie J, Luong TM, Burchenal J, et al. Prognostic significance of blood markers of inflammation in patients with ST-segment elevation myocardial infarction undergoing primary angioplasty and effects of pexelizumab, a C5 inhibitor: a substudy of the COMMA trial. Eur Heart
26 (19):1964–1970, 2005.
4. Frydland M, Ostrowski SR, Møller JE, Hadziselimovic E, Holmvang L, Ravn HB, Jensen LO, Pettersson AS, Kjaegaard J, Lindholm MG, et al. Plasma concentration of biomarkers reflecting endothelial cell- and glycocalyx damage are increased in patients with suspected ST-elevation myocardial infarction complicated by cardiogenic shock. Shock
50 (5):538–544, 2018.
5. Kohsaka S, Menon V, Lowe AM, Lange M, Dzavik V, Sleeper LA, Hochman JS. SHOCK Investigators: systemic inflammatory response syndrome after acute myocardial infarction complicated by cardiogenic shock. Arch Intern Med
165 (14):1643–1650, 2005.
6. Millar JE, Fanning JP, McDonald CI, McAuley DF, Fraser JF. The inflammatory response to extracorporeal membrane oxygenation (ECMO): a review of the pathophysiology. Crit Care
20 (1):387, 2016.
7. Burgdorff AM, Bucher M, Schumann J. Vasoplegia in patients with sepsis and septic shock: pathways and mechanisms. J Int Med Res
46 (4):1303–1310, 2018.
8. Arlati S, Storti E, Pradella V, Bucci L, Vitolo A, Pulici M. Decreased fluid volume to reduce organ damage: a new approach to burn shock resuscitation? A preliminary study. Resuscitation
72 (3):371–378, 2007.
9. Payen D, Pont AC, de Sakr Y, Spies C, Reinhart K, Vincent JL. Sepsis Occurrence in Acutely Ill Patients (SOAP) Investigators: a positive fluid balance is associated with a worse outcome in patients with acute renal failure. Crit Care
12 (3):R74, 2008.
10. Alsous F, Khamiees M, DeGirolamo A, Amoateng-Adjepong Y, Manthous CA. Negative fluid balance predicts survival in patients with septic shock: a retrospective pilot study. Chest
117 (6):1749–1754, 2000.
11. Sirvent J-M, Ferri C, Baró A, Murcia C, Lorencio C. Fluid balance in sepsis and septic shock as a determining factor of mortality. Am J Emerg Med
33 (2):186–189, 2015.
12. Extracorporeal Life Support Organization. ELSO Anticoagulation guideline. Ann Arbor, MI USA
13. Kellum JA, Lameire N. KDIGO AKI Guideline Work Group: diagnosis, evaluation, and management of acute kidney injury: a KDIGO summary (Part 1). Crit Care
17 (1):204, 2013.
14. Aubron C, DePuyt J, Belon F, Bailey M, Schmidt M, Sheldrake J, Murphy D, Scheinkestel C, Cooper DJ, Capellier G, et al. Predictive factors of bleeding events in adults undergoing extracorporeal membrane oxygenation. Ann Intensive Care
6 (12):97, 2016.
15. Cohen J. Statistical power analysis for the behavioral sciences. Chapter 4 difference between correlation coefficients
2nd ed.USA: Lawrence Erlabum Associated Publishers; 1988.
16. Schmidt M, Bailey M, Kelly J, Hodgson C, Cooper DJ, Scheinkestel C, Pellegrino V, Bellomo R, Pilcher D. Impact of fluid balance on outcome of adult patients treated with extracorporeal membrane oxygenation. Intensive Care Med
40 (9):1256–1266, 2014.
17. Staudacher DL, Gold W, Biever PM, Bode C, Wengenmayer T. Early fluid resuscitation and volume therapy in venoarterial extracorporeal membrane oxygenation. J Crit Care
18. Kim H, Paek JH, Song JH, Lee H, Jhee JH, Park S, Yun HR, Kee YK, Han SH, Yoo TH, et al. Permissive fluid volume in adult patients undergoing extracorporeal membrane oxygenation treatment. Crit Care
22 (1):270, 2018.
19. Wengenmayer T, Duerschmied D, Graf E, Chiabudini M, Benk C, Mühlschlegel S, Philipp A, Lubnow M, Bode C, Staudacher DL. Development and validation of a prognostic model for survival in patients treated with venoarterial extracorporeal membrane oxygenation: the PREDICT VA-ECMO score. Eur Heart J Acute Cardiovasc Care
20. Schmidt M, Burrell A, Roberts L, Bailey M, Sheldrake J, Rycus PT, Hodgson C, Scheinkestel C, Cooper DJ, Thiagarajan RR, et al. Predicting survival after ECMO for refractory cardiogenic shock: the survival after veno-arterial-ECMO (SAVE)-score. Eur Heart J
36 (33):2246–2256, 2015.
21. Wang L, Yang F, Wang X, Xie H, Fan E, Ogino M, Brodie D, Wang H, Hou X. Predicting mortality in patients undergoing VA-ECMO after coronary artery bypass grafting: the REMEMBER score. Crit Care
23 (1):11, 2019.
22. Self WH, Semler MW, Bellomo R, Brown SM, deBoisblanc BP, Exline MC, Ginde AA, Grissom CK, Janz DR, Jones AE, et al. Liberal versus restrictive intravenous fluid therapy for early septic shock: rationale for a randomized trial. Ann Emerg Med
72 (4):457–466, 2018.
23. Yealy DM, Kellum JA, Huang DT, Barnato AE, Weissfeld LA, Pike F, Terndrup T, Wang HE, Hou PC, LoVecchio F, et al. ProCESS Investigators. A randomized trial of protocol-based care for early septic shock. N Engl J Med
370 (18):1683–1693, 2014.
24. Van Regenmortel N, Verbrugghe W, Roelant E, Van den Wyngaert T, Jorens PG. Maintenance fluid therapy and fluid creep impose more significant fluid, sodium, and chloride burdens than resuscitation fluids in critically ill patients: a retrospective study in a tertiary mixed ICU population. Intensive Care Med
44 (4):409–417, 2018.
25. Marik PE, Monnet X, Teboul J-L. Hemodynamic parameters to guide fluid therapy. Ann Intensive Care
1 (1):1, 2011.
26. You J-W, Lee SJ, Kim YE, Cho YJ, Jeong YY, Kim HC, Lee JD, Kim JR, Hwang YS. Association between weight change and clinical outcomes in critically ill patients. J Crit Care
28 (6):923–927, 2013.