Fluid Overload and Precision Net Ultrafiltration in Critically Ill Patients

Fluid overload, defined as an absolute increase in total body volume or a relative increase in percentage of extracellular volume over the isovolemic status of the patient, is a frequent problem in critically ill patients [1]. For epidemiologic and operational purposes, fluid overload in hospitalized patients is typically defined as a positive value of total fluid intake minus total fluid output expressed as a percentage of patient body weight on hospital admission [2, 3] or cumulative positive fluid balance during hospital stay [4]. Except for chronic conditions that cause fluid overload (e.g., congestive heart failure, cirrhosis of the liver, etc.), in most critically ill patients, fluid overload occurs because of iatrogenic administration of intravenous fluids, medications, and nutrition, coupled with decreased elimination of excess fluid due to kidney dysfunction due to either acute kidney injury (AKI), chronic kidney disease or kidney failure (i.e., end-stage renal disease).

Recent systematic review of critically ill patients and epidemiologic studies suggests that fluid overload defined by weight gain >5% of body weight or positive cumulative fluid balance [3, 4] is associated with poor clinical outcomes [3, 5, 6]. Several studies have documented that fluid overload is associated with multiple adverse events including risk of AKI [7], impaired recovery from AKI [8], prolonged need for mechanical ventilation [9], impaired wound healing [10], abdominal compartment syndrome [11], discharge to a healthcare facility [12], and death [3, 5, 6].

Among critically ill patients with fluid overload and kidney dysfunction, diuretics are often used for volume management. However, when fluid overload is refractory to treatment with diuretics, clinicians frequently use extracorporeal fluid removal (i.e., net ultrafiltration [UF_NET]) during kidney replacement therapy (KRT) for treatment of fluid overload. Although observational studies suggest that extracorporeal volume removal is associated with improved outcomes, recent evidence suggests that the rate of volume removal is also associated with clinical outcomes. In this review, we provide an overview of monitoring of fluid overload; principles of fluid management in the ICU; use of extracorporeal ultrafiltration for fluid management; association of UF_NET rate with clinical outcomes; risks associated with slow and fast UF_NET rates; ultrafiltration and organ dysfunction; importance of precision UF_NET; areas of uncertainty and future research; and the need for innovation in ultrafiltration technology.

Daily monitoring of input and output fluids and net patient fluid balance is critical for prevention and early treatment of fluid overload in patients at risk and those with established AKI. The total extent of fluid overload may be best assessed by serial weights, cumulative fluid balance, and clinical examination, possibly aided by specific investigations such as bioelectrical impedance analysis or lung ultrasound [13, 14]. Bioimpedance analysis is a noninvasive technology that assesses the electrical properties of tissues by measuring the reactance and resistance of an alternating current passed through the body through an electrode placed on the skin [15]. Using bioimpedance, total body, extracellular and intracellular water could be measured. Several studies have evaluated the feasibility of bioimpedance in critically ill patients [14, 16‒20], and volume status assessed by bioimpedance has been associated with mortality in patients treated with continuous kidney replacement therapy (CKRT) [21‒23]. Measurement of daily weight coupled with daily net fluid balance and electronic health record alerting can be used to monitor, prevent, and treat fluid overload. Functional hemodynamic monitoring parameters such as pulse pressure variation and stroke volume variation have been used for determining volume responsiveness in hypotensive and mechanically ventilated patients may reduce unnecessary fluid use and risk of fluid overload.

Among critically ill patients who are at risk of fluid overload or have already developed fluid overload, careful attention should be paid to fluid balance and implementation of a restrictive fluid management strategy. Restrictive fluid management strategy includes minimizing the administration of unnecessary fluids such as continuous infusion of maintenance fluids, reducing the volume for carrier fluids for medications and nutrition, and use of bare minimal volume required to administer medication and parenteral and enteral nutrition [2]. Intravascular hypovolemia and hypotension should be treated with fluid bolus only if the patient is volume responsive and preferably guided using functional hemodynamic monitoring parameters such as pulse pressure variation, stroke volume variation, or passive leg raising [24]. Diuretics should be used judiciously to maintain euvolemia in patients who have or are at risk for fluid overload.

A small recently completed multinational randomized trial found that a restrictive fluid management strategy among patients with AKI targeting negative or neutral daily fluid balance by minimizing fluid input and/or enhancing urine output with diuretics administered at the discretion of the clinician was associated with lower cumulative fluid balance at 72 h after randomization, lower duration of AKI, lower need for KRT, and less adverse events [25]. Similarly, the Fluids and Catheters Treatment Trial (FACTT) compared conservative fluid management to liberal therapy in patients with acute respiratory distress syndrome and found no difference in 60-day mortality but reported shorter duration of mechanical ventilation and a trend toward less KRT in the conservative fluid management arm [26]. However, it is important to note that prolonged negative fluid balance and intravascular volume depletion should also be avoided as it is associated with adverse outcomes compared to patients who remain euvolemic throughout ICU stay [3]. Diuretics are frequently used to treat fluid overload among critically ill patients with oliguria and kidney dysfunction who do not have intravascular hypovolemia. However, when oliguria persists and fluid overload is refractory to diuretic treatment or when the patient develops life-threatening complications (e.g., pulmonary edema or severe hypoxemia), clinicians typically initiate extracorporeal UF_NET for volume management [27].

Ultrafiltration has been used in the treatment of patients with oliguric AKI and fluid overload since the inception of hemodialysis more than 70 years ago [28], and the use of ultrafiltration is currently supported by international guidelines that recommend emergency initiation of KRT and ultrafiltration in response to life-threatening changes in fluids, solutes, and electrolytes [29]. In critically ill children with AKI, ultrafiltration is also frequently used to prevent the development or worsening of fluid overload while enabling simultaneous administration of volume associated with medications, blood products, and nutrition [30].

Among critically ill adults, overall volume status and the optimal UF_NET requirements are determined in part by physical examination (e.g., the presence of tissue edema), patient fluid balance (i.e., total fluid intake minus total fluid loss), weight, hemodynamics (mean arterial pressure, heart rate, pulse pressure, or stroke volume variation), evidence of organ edema (e.g., pulmonary edema on chest radiograph), and adequate perfusion (e.g., capillary refill). Ultrafiltration can be used in conjunction with fluid administration to maintain a precise fluid balance in critically ill patients with oliguric AKI.

Prescription of UF_NET requires careful consideration of the extent of fluid overload, the total amount of fluid that would need to be removed to achieve euvolemia, and the appropriate rate at which this goal should be achieved [1, 2, 31, 32]. Although there is not a single fluid overload threshold at which net fluid removal should be started in critically ill patients, most studies evaluated 5–10% fluid overload threshold for initiation of extracorporeal fluid removal in patients with oliguric AKI. However, a major determinant of the total volume and the rate at which fluid is to be removed is limited by patient’s severity of illness and requires careful consideration of expected fluid inputs and losses, the expected speed of vascular refilling during fluid removal, and the patient’s physiological tolerance to transient reduction in intravascular volume.

The decision to use ultrafiltration for volume management requires selection of a modality such as intermittent hemodialysis, prolonged intermittent forms of KRT or CKRT. This decision should be made by considering the total amount of volume that needs to be removed; the ongoing need for fluid administration; hemodynamic status (i.e., blood pressure and need for vasopressor support); the rate at which fluid needs to be removed; the patient’s response to fluid removal; the need for solute removal, electrolyte correction, or control of uremia; and the resources that are available at the institution.

To precisely manage volume using CKRT, it is essential to understand the hourly machine and patient fluid balances. The machine fluid balance is calculated as the difference in the volume of fluids infused hourly through the dialysis machine (e.g., replacement fluids, dialysate, and anticoagulant) and the total hourly effluent volume, whereas the patient fluid balance is calculated as the hourly difference in patient inputs (e.g., blood products, drugs, and nutrition) and outputs (e.g., urine, drains, and insensible losses). The machine fluid balance and patient fluid balance can be used to determine the hourly overall net machine-patient fluid balance (also known as integrated balance), which can be calculated based on the differences in machine and patient fluid balances in the preceding hour [31].

Several observational studies in critically ill patients with AKI have found that use of UF_NET was associated with lower mortality. For instance, one study found that among patients with AKI, fluid overload of >5% of body weight was associated with increased risk of death that persisted for up to 1 year compared to patients who remained in even fluid balance or negative fluid balance [3]. However, initiation of KRT and UF_NET attenuated this risk associated with fluid overload [3].

In a secondary analysis of the Randomized Evaluation of the Normal versus Augmented Level (RENAL) trial of CKRT in patients with severe AKI and treated with UF_NET [33], daily fluid balance among survivors was −234 mL compared with +560 mL among non-survivors. Moreover, a negative mean daily fluid balance was independently associated with a decreased risk of death and increased dialysis-free and hospital-free days compared to patients who remained in positive fluid balance. Moreover, recent mediation analysis has found that the ability to achieve negative fluid balance was also independently associated with lower mortality [34].

In another study that used the Program to Improve Care in Acute Kidney Disease (PICARD) cohort of critically ill patients with AKI and >10% fluid overload who were treated with UF_NET, survivors had less fluid accumulation at the initiation of KRT than non-survivors [35]. Among critically ill children, observational studies have found that earlier use of UF_NET to prevent worsening of fluid overload attenuated the risk of death [30, 36, 37]. Based on these observational studies, the Acute Disease Quality Initiative suggests early initiation of ultrafiltration in patients who are at risk for severe fluid overload to prevent further worsening of fluid overload while enabling the administration of fluid volumes associated with medications, blood products, and nutrition in adults and children [2, 31, 38].

However, it is important to note that randomized clinical trials have not been conducted to examine whether extracorporeal fluid removal is associated with lower morbidity and mortality in critically ill patients compared with no extracorporeal fluid removal. Observational studies showing benefits associated with extracorporeal fluid removal have the potential to be confounded by hemodynamic tolerance to fluid removal as patients who tolerate fluid removal have better physiological reserve that is associated with improved outcomes compared to patients who do not tolerate fluid removal.

Emerging evidence from observational studies suggests that a “J”-shaped association exists between UF_NET rate (i.e., rate of fluid removal) and mortality (shown in Fig. 1). A single-center retrospective study examined the association of UF_NET rate with 1-year mortality among 1,075 critically ill patients with AKI and fluid overload of >5% of body weight who were treated with intermittent hemodialysis and/or CKRT (Table 1) [6]. The UF_NET rate was calculated as the net volume of fluid removed per day from initiation of either CKRT or intermittent hemodialysis until the end of ICU stay, adjusted for body weight at hospital admission. After accounting for confounders, UF_NET rates ≤20 mL/kg/day were associated with higher mortality than UF_NET rates >25 mL/kg/day in the overall study cohort. Moreover, among patients who were treated with CKRT, UF_NET rates <0.5 mL/kg/h were associated with higher odds of death than UF_NET rates >1.0 mL/kg/h. Another study in 1,398 patients with AKI who were treated with CKRT found that UF_NET rates <35 mL/kg/day were associated with higher odds of 30-day mortality than rates ≥35 mL/kg/day [39]. These findings suggest that slower UF_NET rates are associated with an increased risk of death, possibly due to prolonged exposure to fluid overload (shown in Fig. 2) depending upon the comparator UF_NET rate group.

In a secondary analysis of the RENAL cohort, studies found an association between high UF_NET rate with low risk-adjusted 90-day survival and dialysis dependence in 1,434 critically ill patients with AKI treated with CKRT [40, 41]. UF_NET rate was defined as the volume of fluid removed per hour adjusted for patient body weight. The studies found that UF_NET rates >1.75 mL/kg/h were associated with lower risk-adjusted 90-day survival and longer time to independence from KRT than rates of 1.01–1.75 mL/kg/h or rates <1.01 mL/kg/h (shown in Fig. 1). There was no significant difference in 90-day survival between patients treated with UF_NET rates of 1.01–1.75 mL/kg/h and those treated with rates <1.01 mL/kg/h. However, when UF_NET rates were analyzed as a continuous variable, there was a J-shaped relationship between UF_NET rate and 90-day mortality and dialysis dependence after accounting for competing risk of death. Furthermore, the incidence of cardiac arrhythmias requiring treatment was highest among patients with UF_NET rates >1.75 mL/kg/h.

Using cluster and Bayesian heterogeneity of treatment effect analysis, the probability of harm associated with UF_NET rate >1.75 mL/kg/h was 99.6% compared with UF_NET rate group of 1.01–1.75 mL/kg/h and 32.5% compared with UF_NET rate <1.01 mL/kg/h among severely ill patients who have sepsis, metabolic acidosis, organ edema, those treated with mechanical ventilation and vasopressors, and those treated with CKRT [42]. However, the moderate UF_NET rate group between 1.01 and 1.75 mL/kg/h was consistently associated with the lowest risk of death compared with UF_NET rates <1.01 mL/kg/h or rates >1.75 mL/kg/h. These findings suggest that faster UF_NET rates might be associated with organ ischemia due to intravascular hypovolemia (shown in Fig. 2).

Among patients with organ edema, the probability of harm associated with UF_NET rates >1.75 mL/kg/h compared with rates <1.01 mL/kg/h was only 35.4%, suggesting that higher UF_NET rates in patients with isolated edema may be associated with a lower risk of death [42]. However, among hemodynamically unstable patients, both UF_NET rates >1.75 mL/kg/h and rates <1.01 mL/kg/h were associated with mortality compared with rates 1.01–1.75 mL/kg/h [42]. These findings suggest that moderate UF_NET rates between 1.01 and 1.75 mL/kg/h during CKRT are probably safe and associated with lowest risk of death across subgroups of patients with organ edema and hemodynamic instability.

Emerging evidence suggests that an association exists between ultrafiltration in critically ill patients and development of new organ dysfunction. A study in 11 critically ill patients with AKI who were treated with hemodialysis and ultrafiltration rates of 5 ± 3 mL/kg/h found acute deterioration in regional, segmental, and global left ventricular contractility together with an increase in biomarkers of cardiac injury suggestive of myocardial ischemia but no substantial change in mean arterial blood pressure [43]. Another study in patients treated with CKRT found that 10 of 11 patients developed new regional cardiac stunning, including 8 patients who developed stunning within 4 h of starting dialysis despite stable hemodynamics [44]. The stunning occurred in patients with preserved and impaired baseline cardiac function, and 7 patients died. Importantly, cardiac stunning occurred despite a low median ultrafiltration rate of 1.1 mL/kg/h and stable hemodynamics.

Several mechanistic studies in patients with kidney failure treated with hemodialysis also suggest that higher ultrafiltration rates might be associated with increased risk of intradialytic hypotension [45, 46], decreased microcirculatory blood flow [47, 48], and hypoperfusion of the heart [43, 45, 49‒51], brain [52, 53], kidneys [54], and gastrointestinal tract (shown in Fig. 2) [55]. Higher ultrafiltration rates have also been associated with lower oxygen saturation in the superior vena cava, suggesting mismatch between low oxygen delivery and relatively high tissue oxygen consumption [56]. Ultrafiltration volumes and reduction in intradialytic blood pressure are independently associated with myocardial stunning [50, 57], cardiac arrhythmias [49, 58], and increased risk of death [50, 58]. In addition, both high absolute ultrafiltration volume and high ultrafiltration rates have been associated with a decline in global cerebral blood flow [52], white matter changes, and cognitive dysfunction [59, 60]. Higher ultrafiltration rates have also been associated with loss of residual kidney function [61‒63]. Several studies have found hepato-splanchnic vasoconstriction and decreased splanchnic perfusion during hemodialysis with ultrafiltration [64‒66]. In patients with chronic kidney disease, circulating endotoxin levels have been found to increase significantly following initiation of hemodialysis and ultrafiltration as a result of decreased splanchnic blood flow and gut ischemia [55]. These findings suggest that both initiation of ultrafiltration and increasing ultrafiltration rates are associated with decreased organ perfusion and multisystem ischemic organ injury.

During CKRT, the relationship between high UF_NET rate and mortality and renal recovery may depend upon the balance between the possible direct harmful effect of high UF_NET rate and the indirect beneficial effect of more negative daily fluid balance as several observational studies have shown lower mortality with negative fluid balance [33]. In the RENAL cohort, compared with middle UF_NET rates of 1.01–1.75 mL/kg/h, high UF_NET rates >1.75 mL/kg/h were associated with higher mortality among patients treated with CKRT [40]. This appeared to be due to a direct harmful relationship and was not explained by its impact on daily fluid balance [34]. On the contrary, the more negative daily fluid balance achieved with such rates attenuated but did not remove such risk. Moreover, the reduced mortality of middle compared with low UF_NET rates was mediated by the greater ability of moderate UF_NET rates to achieve a more negative daily fluid balance among patients with baseline fluid overload, increasing the protective impact of a more negative daily fluid balance. This effect was significant in interaction testing when comparing moderate with high UF_NET rates. For all comparisons, irrespective of daily fluid balance and baseline fluid overload, the moderate UF_NET rate was also associated with a lowest risk of mortality [34].

Currently, there is a global variation in prescription and practice of UF_NET in critically ill patients [27, 67]. Variations were noted among clinician perceptions toward fluid removal, assessment of prescribed-to-delivered dose of UF_NET, KRT modality used for UF_NET, methods used to achieve fluid removal during CKRT, assessment of fluid balance, and management of complications during fluid removal and perceived barriers to fluid removal. This is partly due to lack of evidence-based guidelines for UF_NET prescription and practice in critically ill patients. Unlike management of solute clearance where currently there is strong evidence for dosing, timing, and modality of KRT, no such evidence exists for UF_NET prescription and practice for critically ill patients [68‒70]. Despite lack of evidence, most practitioners believed in early initiation of UF_NET despite several trials showing no difference in patient outcomes among patients with early initiation of KRT, even though these trials did not evaluate the timing of volume management [70]. Despite variation in attitudes toward the use of protocol and timing of initiation of UF_NET, critical care practitioners were willing to enroll patients in a protocol-based clinical trial of UF_NET.

Currently, there are several areas of uncertainty in volume management. First, randomized trials have not been conducted in critically ill patients despite several observational studies that have found that use of UF_NET is associated with lower mortality [33, 35]. Although it may not be ethical to randomize patients to fluid removal versus no fluid removal, randomizing patients to different strategies of fluid removal could be performed and outcomes compared. Newer and minimally invasive ultrafiltration technologies might facilitate enrollment in clinical trials using newer study designs. For example, various fluid removal strategies could be examined in patients with oliguric AKI and fluid overload who would otherwise not meet criteria for initiation of traditional KRT.

Second, the optimal timing of initiation of UF_NET in relation to fluid overload is unclear. Although fluid overload was the main reason for initiation of KRT among patients randomized to late arms of KRT in several randomized clinical trials [70, 71], none exclusively examined timing of initiation of UF_NET with respect to fluid overload. Currently, initiation of UF_NET is recommended by clinical practice guidelines only when “life-threatening” changes in volume are present. However, no recommendations exist for optimal timing of discontinuation of UF_NET.

Third, the optimal fluid overload threshold for UF_NET initiation is also unclear. For instance, whether UF_NET should be initiated at 5%, 10%, or 15% fluid overload is also unknown, even though these thresholds are associated with morbidity and mortality. Whether preemptive initiation of UF_NET among patients at risk for fluid overload before development of life-threatening features (e.g., severe hypoxemia) is associated with improved outcomes is unclear. This is important because many such patients will not meet criteria for initiation of traditional KRT for volume management since clinical trials indicate that early initiation of KRT is associated with lack of mortality benefit and adverse outcomes [70].

Fourth, finally, technologies that allow uncoupling of solute management and volume management may open therapeutic areas of research. For instance, whether sequential use of UF_NET followed by initiation of traditional KRT for solute management is associated with less complications and improved outcomes in critically ill patients is unclear and needs to be evaluated.

Given that several studies show an association of fluid overload with mortality, it is paramount to prevent iatrogenic fluid overload by fluid restriction, eliminating unnecessary fluids, and judicious use of diuretics. However, many patients with oliguria and fluid overload only have modest response to diuretics, especially when kidney injury is severe or is rapidly progressing. Since duration of fluid overload is associated with morbidity and mortality [3], strategies that prevent fluid overload and reduce the duration of exposure to fluid overload are needed in patients who fail conservative fluid management strategies.

Many patients with kidney dysfunction and fluid overload may also not have a definitive indication for initiation of KRT for management of solute or volume control, given that recent trials show no benefit of early initiation of KRT in critically ill patients [70]. Thus, technologies are needed to effectively manage volume when patients have inadequate response to diuretics and are not candidates for initiation of traditional KRT for volume management for fluid overload.

Minimally invasive and isolated UF_NET devices may address an unmet clinical need in ICU patients with fluid overload. Such a technology that allows volume management independent of solute management may obviate the need for use of traditional dialysis and its associated complications (Table 2). Using smaller catheters inserted in peripheral veins without the need for insertion of larger dialysis catheters may facilitate slow fluid removal and avoid risks associated with rapid UF_NET rates. Catheter-related complications could be minimized, and UF_NET could be performed without the need for specialized training. Blood flow rates could be kept low, and moderate ultrafiltration rates might be associated with better hemodynamic stability and lower risk of intra-dialytic hypotension. Such technologies can also be used in conjunction with diuretics for efficient volume management or as an interim measure before initiation of traditional KRT. Thus, further clinical trials are warranted using miniaturized and novel ultrafiltration technologies.

Among critically ill patients, fluid overload is associated with significant morbidity and mortality. UF_NET is frequently used for volume management among patients who are refractory to diuretic therapy. Recent evidence suggests that moderate UF_NET rates are associated with lower risk of complications and better survival during CKRT. Emerging novel and minimally invasive UF_NET devices may provide an unmet need for patients with acute AKI and fluid overload who otherwise will not meet criteria for initiation of traditional KRT. Thus, further clinical trials are warranted using novel ultrafiltration devices.

Dr. Raghavan Murugan reported receiving grants and personal fees from La Jolla Inc.; grants from Bioporto, Inc., and the National Institute of Diabetes and Digestive and Renal Diseases; personal fees from Beckman Coulter and AM Pharma, Inc.; and consulting fee from Baxter outside the submitted work. Dr. Claudio Ronco reported receiving consulting fees or participating in advisory boards in the last 3 years for ASAHI, Astute, Baxter, bioMerieux, B. Braun, CytoSorbents, ESTOR, FMC, GE, Jafron, Medtronic, and Toray.

Partial support for this work was provided BY the National Institute of Diabetes and Digestive and Renal Diseases (R01DK128100).

Study concept and design: Raghavan Murugan and Claudio Ronco. Drafting of the manuscript: Raghavan Murugan. Critical revision of the manuscript for important intellectual content: Claudio Ronco, Amir Kazory, and Luca Sgarabotto.