FormalPara Take-home message

Among critically ill patients with acute kidney injury, a fluid restrictive management regimen as compared to usual care led to a lower cumulative fluid balance after 72 h from randomization. Adverse events were more frequent in the usual care arm.

 

Introduction

Many critically ill patients develop acute kidney injury (AKI). AKI is associated with an increased risk of adverse outcomes [1, 2]. Current management of AKI remains supportive and includes optimization of hemodynamics and fluid status, avoidance of nephrotoxins [3] and, in severe cases, renal replacement therapy (RRT) [1, 2]. Novel therapies are being evaluated in cardiac surgical [4] and septic patients [5], but no specific pharmacological therapies for general intensive care unit (ICU) patients are currently available.

While fluid therapy is the traditional cornerstone of AKI prevention and treatment [6], patients with AKI are also especially prone to develop fluid overload [7]. Such fluid overload is associated with increased mortality [7,8,9,10,11] and progression of AKI [12, 13]. Thus, a more restrictive fluid management approach among adequately resuscitated patients with AKI might improve outcomes by reducing congestion and improving kidney and other organ function. Notably, a feasibility trial in septic shock patients comparing restrictive fluid resuscitation versus usual care reported a lower risk of worsening of AKI in the restrictive group [14].

Trials among general ICU patients [15], or those with sepsis [14, 16] or acute respiratory distress syndrome (ARDS) [17] have found various fluid restrictive strategies to be safe. Among patients undergoing major abdominal surgery, however, restrictive intra- and post-operative fluid resuscitation and management compared to usual care increased the risk of AKI [18]. Thus, research investigating the role of restrictive fluid management specifically in critically ill patients with AKI has been identified as a research priority [3, 19, 20]. To date, such trials have not been performed.

Accordingly, we conducted a multinational, unblinded, randomized, controlled, feasibility pilot trial comparing a restrictive fluid management (RFM) regimen to usual care among ICU patients with AKI to study the feasibility of such intervention in terms of separation, safety, and compliance. We tested the hypothesis that the proposed RFM regimen would lead to a lower cumulative fluid balance at 72 h post-randomization.

Methods

Trial design

We conducted an investigator-initiated, multicenter, unblinded, randomized, controlled, parallel pilot feasibility study (clinicaltrials.gov identifier NCT03251131) in five European and two Australian ICUs. The trial protocol and statistical analysis plan were published prior to completion of data collection [21]. In each participating site, the trial was approved by the relevant research ethics committee prior to commencing screening. Depending on local practices and ethics approval, we obtained written informed consent either prior to enrollment from patient or proxy, or alternatively a deferred consent with written informed consent obtained from patient or proxy as soon as possible. If consent was not granted or withdrawn, approval of using data collected that far was sought. Trial was performed according to the Declaration of Helsinki and its later amendments.

We randomized patients to either RFM or usual care using an electronic platform with allocation ratio of 1:1. Severity of AKI (stage 1 vs stage 2 or 3 according to the Kidney Diseases: Improving Global Outcomes (KDIGO) criteria [3]) and presence of clinical signs of fluid accumulation (defined by peripheral pitting edema and/or positive fluid balance with P/F ratio less than 200 mmHg) served as stratification variables. We used permuted blocks of varying size (2–4). Due to the nature of intervention, only the statistician conducting the data analysis remained blinded to treatment allocation.

Patients

We included patients who (1) were 18 years or older and admitted to critical care with an arterial line in place, (2) were in critical care for at least 12 but no more than 72 h (3) had AKI but were not receiving RRT, (4) were judged by the treating clinician not to be hypovolemic, and (5) were likely to remain in critical care for 48 h after randomization.

We defined AKI according to the following criteria: (1) increase in plasma creatinine 1.5 times or more above baseline without a decline of 27umol/L or more from the last preceding measurement and/or (2) overall urine output less than 0.5 ml/kg/h (or 6 ml/kg) for the previous 12 h. Exclusion criteria are listed in Fig. 1. The Electronic Supplementary Material (ESM) Tables S1 and S2 present detailed definitions of the eligibility criteria.

Fig. 1
figure 1

Flowchart of trial patients. ICU intensive care unit, RRT renal replacement therapy. a Diabetic ketoacidosis, non-ketotic coma, severe burns or other clinical reason determined by the medical staff. b Glomerulonephritis, vasculitis, acute interstitial nephritis, or post-renal obstruction. c Including two exclusion criteria that were removed in a protocol amendment in April 2018 (1) Metformin-induced lactic acidosis or acute liver failure (n = 8) (2) AKI stage 2 or greater is known to have been present for > 48 h (n = 14). These criteria were amended to clarify the exclusion criteria

Full size image

Interventions

The trial intervention (RFM) consisted of a bundle of treatment recommendations for clinicians treating the trial patients. The overall aim was to reach negative fluid balance on the days following randomization. These recommendations included, first, restricting total daily fluid input only to medications and nutritional fluids (enteral or parenteral) and blood products (if clinically necessary). Second, use of maintenance intravenous fluids was permitted only if enteral nutrition was not tolerated and parenteral nutrition contraindicated. Third, fluid bolus therapy could be given as clinically deemed necessary. Fourth, matching fluid output (with unrestricted use of diuretics) to fluid input whenever possible to achieve a preferably negative cumulative fluid balance but always less than 300 mL/day. Fifth, if the fluid balance target could not be achieved by such means, consideration of RRT to remove the required fluid. Commencing RRT was not mandated in the trial and, if RRT was not considered clinically desirable, the protocol indicated acceptance of a less than targeted fluid balance temporarily (for 24 h) up to 900 mL.

In the usual care group, fluid management was at the discretion of the treating clinical team.

Fluid balance was calculated by subtracting total fluid output (urine output, losses to drains, losses from gastrointestinal tract, ultrafiltration by RRT) from total fluid input (intravenous and oral). Insensible losses were not considered. The intervention period was 7 days from randomization or until ICU discharge whatever occurred first.

Outcomes

The primary outcome was cumulative fluid balance at 72 h after randomization adjusted for stratification variables [21, 22]. Secondary outcomes included duration of AKI in days defined by the KDIGO creatinine and urine output criteria (truncated at ICU discharge or 14 days whichever came first), number of patients requiring RRT (truncated at 14 days, including initiation of RRT after ICU discharge), cumulative fluid balance at 24 h after randomization and at ICU discharge (or truncated at 7 days if ICU stay exceeded 7 days), and cumulative dose of diuretics (furosemide) during the intervention period (while in the ICU, maximum of 7 days) adjusted for the duration of the observation period.

Exploratory outcomes included days free of mechanical ventilation and alive at 14 days, days free of vasopressors and alive at 14 days, days free of ICU and alive at 14 days, days free of RRT and alive (assessed at 90 days), 90-days dialysis dependence and 90-days mortality. The safety and feasibility outcomes included the number of patients with one or more (serious) adverse events (AE) and reactions in both arms (detailed definitions are provided in Table S3), screened versus recruited patients ratio, recruitment rate (patients/center/month), and protocol compliance (number of patients with protocol violations in both arms).

Statistical analysis

Details of statistical analysis plan have been published [21]. We estimated that 50 patients per arm would be required to show a difference of 1200 mL in the primary outcome between treatment arms with a two-sided alpha of 0.05 and power exceeding 80%. Based on unpublished data among AKI patients in the FINNAKI study [1], we estimated that cumulative fluid balance at 72 h would be 2700 mL in the usual care group, and thus 1500 mL in the fluid restrictive group (both with a SD of 2000 mL).

We performed the primary analysis on the intention-to-treat (ITT) population defined as all randomized subjects with consent to use data in the analysis. An additional sensitivity analysis was conducted in the per-protocol population, defined as the ITT population after exclusion of subjects who experienced protocol violation(s) or stayed in the ICU for less than 48 h post-randomization.

The primary, secondary, and exploratory outcomes were adjusted for the stratification variables [22] (mild vs severe AKI and presence/absence of clinical signs of fluid accumulation) using two-tailed logistic regression (dichotomous outcomes) or a linear model (continuous outcome variables, mean or median regression model). We report risk ratios (RR) with 95% confidence intervals (CIs) or difference in means/median with 95% CIs.

We used Shapiro–Wilk test for continuous variables to assess assumption of normality. We set the statistical significance to 0.05 and report two-sided P values. Analyses were conducted using R statistical software version 3.6.2 and SPSS 24.

Results

Patients

Between October 2017 and January 2020, we screened 997 patients and randomized 102 patients. One patient in each group declined consent leaving 100 patients for analysis. Randomization occurred at a median [IQR] 30.7 [20.4–46.2] hours after ICU admission. Forty-nine patients (49%) were assigned to the RFM arm and 51 (51%) to the usual care arm. Figure 1 presents the patient flow. Further, two patients in the RFM arm and three patients in usual care arm withdrew consent and their data obtained until withdrawal were included in the analysis.

Patient baseline characteristics and status at randomization were well-balanced (Table 1). The most frequent reasons for ICU admission were cardiac arrest (8.2%), septic shock (7.2%), and gastrointestinal perforation/rupture (7.2%). Suspected etiology of AKI was multifactorial in 55 (55%) patients. Table S5 presents the observed AKI risk factors.

Table 1 Patient characteristics
Full size table

Primary, secondary, and exploratory outcomes

At 72 h, 66 (66%) patients were still in ICU, while 29 (29%) were discharged, four were deceased (one in RFM arm and three in usual care), and one had withdrawn consent for further data collection. The median length of stay from randomization to discharge among these 34 patients was 49.2 [32.6–51.9] in RFM arm and 51.0 [29.3–60.2] hours in the usual care arm. The primary outcome was assessed after a median [IQR] 72.0 [53.9–72] hours from randomization. Figure 2 presents fluid input and output data along with daily fluid balance in both arms. Table 2 presents the primary and secondary outcomes. The mean difference (95% CI) in cumulative fluid balance at 72 h was − 1148 (− 2200 to − 96) mL, P = 0.033. There was no difference in duration of AKI but fewer patients in the RFM group received RRT compared to the usual care group. Two additional patients in the usual care arm commenced RRT after day 14. Cumulative fluid balance at 24 h and at ICU discharge or on day 7 were significantly lower in the RFM group. Diuretics were administered to 35 (71.4%) in the RFM group and 38 (74.5%) in the usual care group. The median [IQR] day number of first administration of diuretics was 1 [1] in both groups. There were no differences in the cumulative doses of diuretics adjusted for observation days. The crude primary outcome was in line with the adjusted analysis (Table S6). There were no significant differences in the adjusted or unadjusted exploratory outcomes (adjusted in Table 3, crude in Table S7). Figure S1 presents the daily SOFA score according to treatment allocation.

Fig. 2
figure 2

Daily fluid input, fluid output and daily fluid balance from randomization in restrictive versus usual care. Boxplots represent median with IQR and range in mL. Bars present the daily balance (not cumulative) in mL. RFM restrictive fluid management

Full size image
Table 2 Primary and secondary outcomes (adjusted for stratification variables)
Full size table
Table 3 Exploratory outcomes (adjusted for stratification variables)
Full size table

Safety and feasibility outcomes

Eleven (22.4%) patients in the RFM group and 25 (49%) in the usual care group experienced one or more AE (risk ratio 0.46; 95% CI 0.36–0.63, P = 0.001). Serious adverse events (SAE) were observed in six (12.2%) patients randomized to RFM group and in 16 (31.4%) randomized to usual care group (RR 0.39; 95% CI 0.15–0.86, P = 0.031). Table S8 provides a detailed description of AEs and SAEs. Most events were related to disturbances in serum electrolyte concentrations and RRT. The median [IQR] number of patients recruited per center/month was 0.8 [0.5–1.7] (Figure S2).

Fluid protocol and protocol violations

Protocol violations occurred in 18 (36.7%) in the RFM arm and in 5 (9.8%) in the usual care arm. In the RFM arm, use of excess maintenance fluid was the most frequently observed protocol violation in 9 (18.4%) patients (Table S9). Moreover, five patients in the RFM arm had a protocol suspension for the following reasons: hypernatremia (n = 2), need for surgery (n = 1), complete heart block (not considered to be attributable to study intervention) (n = 1), and withdrawal of consent (n = 1). One patient in the usual care arm had protocol suspension as overall ICU care was withdrawn.

Per-protocol analysis

Twenty-three patients (46.9% of 49) in the RFM arm and 37 (72.5% of 51) in the usual care arm were included in the per-protocol analysis. The primary or secondary outcome variables were not different in the per-protocol population except for fluid balance at 24 h (Table S10).

Discussion

Key findings

In this multicenter, multinational, unblinded, randomized, controlled, feasibility pilot study involving 100 critically ill patients with AKI who were considered adequately resuscitated, we assessed whether, compared to usual care, a restrictive fluid management regimen would lead to a meaningful difference in cumulative fluid balance after 72 h from randomization without signals of harm. We found the RFM protocol to be effective in terms of achieving a lower cumulative fluid balance at 72 h compared to usual care. Moreover, patients in the RFM arm received RRT less frequently and had fewer adverse events than patients in the usual care arm.

Relationship with previous studies

No randomized controlled trial has studied a RFM regimen among critically ill patients with AKI.

The FACTT trial, conducted from 2000 to 2005, compared conservative fluid management to liberal therapy in patients with ARDS and found no difference in 60-day mortality, but reported shorter duration of mechanical ventilation and a trend toward less RRT in the conservative arm [17]. Furthermore, a sub-analysis of patients with AKI found an association with high cumulative fluid balance and mortality [9]. In the FACTT, the cumulative fluid balance on day three (corresponding to the primary endpoint in the current analysis) was about 150 mL in the conservative arm and about 5000 mL in the liberal arm, both more positive than in our study [23]. Moreover, we observed a much lower actual fluid balance in the usual care group than we had assumed based on data collected in 2011–2012. Such a shift towards more restrictive fluid management practices among the critically ill has been reported also in another RFM pilot trial [24].

Most interventional fluid trials in the critically ill have concentrated on modifying fluid administration [14, 15, 25,26,27,28] or enhancing fluid output with diuretics and/or ultrafiltration [29]. Our approach of combining the two sides of the coin in one trial gave the treating clinician more tools to individualize fluid therapy while still successfully targeting a clinically meaningful outcome measure. For example, the CLASSIC pilot trial among patients with septic shock achieved a difference in the administered resuscitation fluid but not in the total fluid balance [14]. Notably, fluids other than resuscitation fluid represent the largest proportion of the administered fluids [30, 31] even in patients with shock [32].

A dynamic intervention bundle such as ours may be more suitable for AKI patients who represent a heterogeneous group of patients with some having severely reduced urine output significantly affecting the achievable fluid output while others have a creatinine rise but preserved urine output. A French stepped wedge cluster RCT is assessing another dynamic approach among general ICU patients [33]. The trial intervention involves fluid restriction and/or enhancement of fluid output to achieve a daily weight loss of 0.5 kg in patients with increasing daily weight [33]. However, no results are yet available. A pilot study among septic shock patients reported a rather large but non-significant difference between restrictive and usual care without any signals of harm, but the trial was stopped early as the predefined endpoint of a—500 mL—difference in mean daily fluid balance was not met [24].

Study implications

Our results showed that, in ICU patients with AKI who are considered not to be hypovolemic, using a combined strategy targeting control of daily fluid balance is both feasible and safe. Adverse events were more frequent in the usual care arm, possibly related to more frequent use of RRT in this group. Moreover, the trial provides initial support to the hypothesis that fluid accumulation is linked to important outcomes among AKI patients such as need for RRT, a hypothesis that should be studied in larger clinical trials.

A quarter of excluded patients were ineligible due to being still considered intravascularly hypovolemic albeit having stayed in the ICU for at least 12 h. This is an important patient cohort that should be considered when designing future RFM trials. Potentially, these patients could be enrolled, if the RFM protocol included a well-defined starting trigger after correction of hypovolemia such as using passive leg raising or other dynamic tests to assess fluid responsiveness [34, 35].

Strengths and limitations

Obvious strengths to our study include randomization to minimize selection bias and a multinational approach allowing simultaneous feasibility assessment in multiple locations with potentially different fluid treatment practices and increasing external validity. In addition, we used a treatment bundle targeting both fluid input and output to maximize the efficacy of the intervention, a dynamic two-pronged approach, which increased flexibility and allowed for individualization of treatment according to perceived patients’ needs. Moreover, we collected data on multiple potential adverse events daily to be able to reliably identify any signal of harm potentially associated with RFM.

We acknowledge several limitations. First, blinding of the intervention was not possible due to its nature. Second, calculated fluid balance is not an exact measure, as measurement of fluid output in patients with high insensible losses (such as losses to wound dressing) is not exact. Thus, the actual fluid balance of the included patients in both groups may be lower than reported here. Third, we observed protocol violations in terms of use of excessive maintenance fluid. This may signal the presence of local practices [30, 32] that need to be better addressed in future trials. Fourth, we excluded patients with treatment limitations albeit many from this heterogenous group could potentially have been enrolled. Nonetheless, we observed a signal for positive patient-centered outcomes in the frequency of RRT, which implies that fluid restriction may have an effect on clinical outcomes.

Conclusions

In a pilot multicenter, international, randomized, controlled trial of critically ill patients with AKI who were considered adequately fluid resuscitated, a restrictive fluid management regimen was feasible, resulted in fewer adverse events than usual care, and delivered a significant separation in fluid balance. The results also suggested benefit in terms of patient-centered outcomes implying that this intervention should now be studied in larger trials.