Article Text

Restrictive versus Liberal Rate of Extracorporeal Volume Removal Evaluation in Acute Kidney Injury (RELIEVE-AKI): a pilot clinical trial protocol
  1. Raghavan Murugan1,
  2. Chung-Chou H Chang2,
  3. Maham Raza3,
  4. Nasrin Nikravangolsefid4,
  5. David T Huang3,
  6. Paul M Palevsky5,6,
  7. Kianoush Kashani4
  8. for the (RELIEVE-AKI) Study Investigators
    1. 1Program for Critical Care Nephrology, Department of Critical Care Medicine, University of Pittsburgh School of Medicine, Pittsburgh, Pennsylvania, USA
    2. 2Biostatistics and Data Management Core, Department of Critical Care Medicine, University of Pittsburgh School of Medicine, Pittsburgh, Pennsylvania, USA
    3. 3Multidisciplinary Acute Care Research Organization, Department of Critical Care Medicine, University of Pittsburgh School of Medicine, Pittsburgh, Pennsylvania, USA
    4. 4Division of Nephrology and Hypertension, Division of Pulmonary and Critical Care Medicine, Department of Medicine, Mayo Clinic, Rochester, Minnesota, USA
    5. 5Renal and Electrolyte Division, Department of Medicine, University of Pittsburgh School of Medicine, Pittsburgh, Pennsylvania, USA
    6. 6Kidney Medicine Section, VA Pittsburgh Healthcare System, Pittsburgh, Pennsylvania, USA
    1. Correspondence to Dr Raghavan Murugan; muruganr{at}


    Introduction Observational studies have linked slower and faster net ultrafiltration (UFNET) rates during kidney replacement therapy (KRT) with mortality in critically ill patients with acute kidney injury (AKI) and fluid overload. To inform the design of a larger randomised trial of patient-centered outcomes, we conduct a feasibility study to examine restrictive and liberal approaches to UFNET during continuous KRT (CKRT).

    Methods and analysis This study is an investigator-initiated, unblinded, 2-arm, comparative-effectiveness, stepped-wedged, cluster randomised trial among 112 critically ill patients with AKI treated with CKRT in 10 intensive care units (ICUs) across 2 hospital systems. In the first 6 months, all ICUs started with a liberal UFNET rate strategy. Thereafter, one ICU is randomised to the restrictive UFNET rate strategy every 2 months. In the liberal group, the UFNET rate is maintained between 2.0 and 5.0 mL/kg/hour; in the restrictive group, the UFNET rate is maintained between 0.5 and 1.5 mL/kg/hour. The three coprimary feasibility outcomes are (1) between-group separation in mean delivered UFNET rates; (2) protocol adherence; and (3) patient recruitment rate. Secondary outcomes include daily and cumulative fluid balance, KRT and mechanical ventilation duration, organ failure-free days, ICU and hospital length of stay, hospital mortality and KRT dependence at hospital discharge. Safety endpoints include haemodynamics, electrolyte imbalance, CKRT circuit issues, organ dysfunction related to fluid overload, secondary infections and thrombotic and haematological complications.

    Ethics and dissemination The University of Pittsburgh Human Research Protection Office approved the study, and an independent Data and Safety Monitoring Board monitors the study. A grant from the United States National Institute of Diabetes and Digestive and Kidney Diseases sponsors the study. The trial results will be submitted for publication in peer-reviewed journals and presented at scientific conferences.

    Trial registration number This trial has been prospectively registered with (NCT05306964). Protocol version identifier and date: 1.5; 13 June 2023.

    • Clinical Trial
    • Adult intensive & critical care
    • Dialysis
    • Acute renal failure

    This is an open access article distributed in accordance with the Creative Commons Attribution Non Commercial (CC BY-NC 4.0) license, which permits others to distribute, remix, adapt, build upon this work non-commercially, and license their derivative works on different terms, provided the original work is properly cited, appropriate credit is given, any changes made indicated, and the use is non-commercial. See:

    Statistics from

    Request Permissions

    If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.


    • The trial assesses attending physician equipoise for rates of net fluid removal before study enrolment.

    • A web-based net fluid removal rate calculator provides decision support for critical care clinicians to deliver the study intervention precisely.

    • Using predicted body weight to calculate net ultrafiltration rate provides an objective and precise volume dosing.

    • The characteristics of the two strategies under evaluation make blinding impossible.

    • The study is unlikely to have sufficient statistical power to assess all secondary and safety outcomes.


    Background and rationale

    Fluid overload develops in over two-thirds of critically ill patients with acute kidney injury (AKI) who require kidney replacement therapy (KRT) and is independently associated with morbidity and mortality.1 2 International consensus guidelines recommend initiating extracorporeal fluid removal when a life-threatening fluid overload occurs.3–5 However, there is uncertainty about the optimal net fluid removal rate, and there is a global variation in clinical practice.6–9 In addition, complications such as intradialytic hypotension, hypertension and cardiac arrhythmias frequently occur during fluid removal.6 10

    Observational studies have found a ‘J’ shaped association between the rate of net fluid removal (ie, net ultrafiltration (UFNET) rate) and mortality among critically ill patients with AKI undergoing continuous KRT (CKRT). Patients who received UFNET rates of 1.01–1.75 mL/kg/hour had the lowest mortality and KRT dependence compared with patients who received slower rates of less than 1.01 mL/kg/hour or faster rates of greater than 1.75 mL/kg/hour.11–14 In addition, UFNET rates greater than 1.75 mL/kg/hour were associated with an increased risk of cardiac arrhythmias requiring treatment.11 However, the causality between UFNET rate, a process of care variable, and mortality remains unclear, as randomised trials investigating this association have not been conducted.

    Need for a trial

    First, the net risks and benefits of slower and faster UFNET rates on clinical outcomes are unclear due to the limitations in observational studies. For instance, slower UFNET rates are tolerated haemodynamically but expose patients to prolonged fluid overload, organ oedema and increased risk of death.15–17 Moreover, slower UFNET rates may also reflect the underlying severity of illness with greater haemodynamic instability. Whereas faster UFNET rates are associated with earlier attainment of negative fluid balance and less fluid overload, but may also cause cardiac arrhythmias, haemodynamic instability, ischaemic organ injury and mortality.11 However, faster UFNET rates may also reflect more fluid overload in patients with severe illnesses or comorbidities.

    Second, there is significant variation in the current clinical practice of UFNET.6 A clinical trial is essential to develop best practice guidelines for volume dosing in critically ill patients. Third, no clinical trial has investigated the effectiveness of UFNET rates on long-term patient-centred clinical outcomes. Although it may be unethical to randomise patients to UFNET versus no UFNET, clinical trials comparing alternative approaches to UFNET can be performed. Indeed, in surveys, more than two-thirds of clinicians had the equipoise to enrol patients in a clinical trial of protocol-based UFNET.6 7 9

    Study objectives

    The overall objectives of this clinical trial are to (1) determine the feasibility of maintaining the alternative restrictive and liberal UFNET rate strategies in critically ill patients with AKI and treated with CKRT and (2) explore the effects of alternative UFNET rate strategies on patient secondary clinical and safety outcomes. The UFNET rates in alternative strategies are widely used in clinical practice.6 11–13 15

    Methods and analysis

    Study design and setting

    This ongoing clinical trial is an investigator-initiated, prospective, unblinded, parallel-group, 2-arm, comparative-effectiveness, stepped-wedged cluster randomised trial (SW-CRT) including 112 critically ill patients with AKI treated with CKRT conducted in 10 intensive care units (ICUs) across 2 hospital systems (figure 1) in the USA. The trial initially planned to enrol 144 patients in 6 ICUs. Enrolment was slower than expected, so the sample size was decreased to 112 patients, and the number of ICUs was increased to 10. Participants are enrolled from 5 ICUs at the University of Pittsburgh Medical Center, Pittsburgh, Pennsylvania, and 5 ICUs at the Mayo Clinic, Rochester, Minnesota. The study commenced enrolment on 5 July 2022 and is anticipated to end on 12 January 2025. The University of Pittsburgh Human Research Protections Office (HRPO) approved the protocol (No. 21080010), and the trial has been prospectively registered with (NCT05306964). The protocol adhered to the Standard Protocol Items: Recommendations for Interventional Trials checklist.18 19 The study team is shown in the online supplemental figure S1.

    Figure 1

    The stepped-wedged cluster randomised trial design. All ICUs start with the liberal UFNET rate group (light blue cell). After the first 6 months (period 1) and every 2 months thereafter (periods 2–12), one ICU is randomly allocated to the restrictive UFNET rate group (dark blue cell). ICU, intensive care units; UFNET, net ultrafiltration.

    Education and training

    One month before patient recruitment began at each site, we prepared standardised educational materials, including slide presentations, videos, virtual sessions, prominently placed posters and pocket cards. We also provided study orientation using methods including (1) circulated educational materials, including frequently asked questions via newsletters and emails; (2) presented the study at clinician meetings; (3) scheduled webinars before study initiation and at periodic intervals; (4) provided just-in-time training for the ICU nurse before the study intervention; (5) provided impromptu training sessions; and finally (6) embedded an educational training video in the web application that the clinicians watched.


    Research coordinators screen the study ICUs at each site by reviewing electronic medical records fitted with the AKI alert system per the Kidney Disease: Improving Global Outcomes definition5 with screening sweeps in the morning and afternoon. A screening log documents all individuals evaluated for the study eligibility and contains patient age, sex, race, ethnicity, eligibility status and reasons for non-enrolment. Table 1 shows the inclusion and exclusion criteria, and online supplemental figure S2 shows the study flow diagram.

    Table 1

    Inclusion and exclusion criteria

    Assessing attending physician equipoise

    The patient is considered provisionally eligible if all inclusion and none of the exclusion criteria are met (figure 2). Once provisionally eligible, the attending intensivist or nephrologist is asked two times per day (a) if he/she strongly believed that emergent and rapid fluid removal should occur or (b) if he/she strongly believed that deferral of fluid removal should occur. If the answer to both questions is negative, the patient is fully eligible, and efforts to obtain informed consent will commence. If a physician does not have equipoise for the rate of fluid removal, the study team reapproaches the physician in the next 12–24 hours to reassess equipoise.

    Figure 2

    Screening and assessment of attending physician equipoise. Attending physician equipoise for net fluid removal rate was evaluated two times per day for patients meeting inclusion and none of the exclusion criteria. AKI, acute kidney injury; CKRT, continuous kidney replacement therapy; CMO, comfort measures only; ECMO, extracorporeal membrane oxygenation; IABP, intra-aortic balloon pump; KRT, kidney replacement therapy; VAD, ventricular assist device.

    Informed consent

    This trial complies with the declaration of Helsinki and the US federal regulations. Written informed consent is required for all participants or their legally authorised representatives (LAR) before enrolment. Most participants are critically ill with impaired decision-making capacity, so consent is obtained from the LARs after discussing the potential risks and benefits of study participation. Before the consent discussion, participants and LARs receive a study brochure (online supplemental figure S3) and consent form (online supplemental file 2). The consent process is conducted in person or virtually. Appropriate signatures are obtained on the consent form before the study intervention. If the study participant regains decision-making capacity, the participant is being reconsented for ongoing participation and data collection.


    In this SW-CRT, ICUs start in the liberal UFNET rate group and are randomly assigned to the restrictive UFNET rate group every 2 months after the first 6 months (figure 1). The data management team uses a computer-generated randomisation scheme to determine how each ICU would cross from the liberal to the restrictive group. We chose SW-CRT because (1) individual patient-level randomisation may be scientifically problematic due to a high risk of contamination between the two UFNET rate groups (ie, Hawthorne effect), (2) a cluster randomised trial is infeasible because ICUs are unwilling to continue with a single intervention (ie, only restrictive or liberal UFNET rate group) during the entire study period, (3) each ICU act as their control and thus fewer units or clusters are required than a traditional cluster-randomised trial,20 (4) to alleviate logistical challenges associated with introducing the intervention (ie, the restrictive group) in all ICUs at once—an SW-CRT will provide adequate time for training all staff on liberal UFNET rate group before transitioning to the restrictive group, (5) offers the opportunity to evaluate ICU-level effectiveness of a new intervention and (6) to study the effect of time on intervention effectiveness.


    We follow standard start-up procedures, with the intervention initiated within 24 hours of study enrolment (online supplemental box S1). Blinding is not feasible due to the nature of the intervention. To precisely determine and dose the UFNET rate, we use a web-based UFNET rate calculator (ie, a web app). This web app calculates the hourly UFNET rate range to be set on the CKRT machine based on the allocation to the intervention group, the rate of intravenous fluids infused in the current hour, and the patient predicted body weight (PBW). We accounted for the rate of intravenous infusions because the UFNET rate dosing represents the net intravascular volume removed beyond the volume infused in the current hour. We chose predicted PBW,21 22 as it is free of confounding by fluid overload and catabolism due to critical illness and measurement errors. In both groups, the initial UFNET rate is set at 0.5 mL/kg/hour and then increased by 0.5 mL/kg/hour, as tolerated (figure 3).

    Figure 3

    The study intervention. Fluid removal will not be started until the mean arterial pressure (MAP)≥65 mm Hg with or without the need for vasopressors. In either group, the initial UFNET rate will be set at 0.5 mL/kg/hour and then increased 0.5 mL/kg/hour, until the target UFNET rate range is reached. Net fluid removal can be stopped at any time and reinitiated as tolerated by the patient. The UFNET rate is calculated based on the allocation to the intervention group, predicted body weight and rate of fluids infused into the patient. CKRT, continuous kidney replacement therapy; UFNET, net ultrafiltration.

    Restrictive strategy

    The UFNET rate is titrated and maintained throughout the study between 0.5 and 1.5 mL/kg/hour. For instance, in a patient with PBW 70 kg receiving no intravenous infusion, the UFNET is initiated at 35 mL/hour (ie, 70×0.5) and increased by 35 mL/hour to a maximum of 105 mL/hour (ie, 70×1.5). If, however, the patient is also receiving an infusion of 100 mL/hour in intravenous fluids (eg, parenteral nutrition, etc), the initial UFNET rate is set at 135 mL/hour (ie, (70×0.5)+100). The UFNET rate is then increased to a maximum of 205 mL/hour (ie, (70×1.5)+100) and maintained between 135 and 205 mL/hour, as tolerated, for the duration of the patient’s 100 mL/hour volume infusion. If the additional 100 mL/hour fluid is discontinued in the patient, the UFNET rate will be reset between 35 and 105 mL/hour.

    Liberal strategy

    The UFNET rate is titrated and maintained between 2.0 and 5.0 mL/kg/hour throughout the study. For instance, in a patient with PBW 70 kg, the UFNET is initiated at 35 mL/hour (ie, 70×0.5) and increased by 35 mL/hour to a maximum of 350 mL/hour (ie, 70×5.0) as tolerated. The UFNET rate is then titrated between 140 and 350 mL/hour. If the patient also receives an infusion of 100 mL/hour in fluids, the initial UFNET rate will be set at 135 mL/hour. The UFNET rate is then increased and maintained between 240 and 450 mL/hour for the duration the patient receives 100 mL/hour volume infusion. If the additional 100 mL/hour fluid is discontinued in the patient, the UFNET rate will be maintained between 140 and 350 mL/hour.

    Concurrent interventions

    In both study groups, we will follow standard start-up procedures (online supplemental box S1); provide guidelines to the clinicians for assessment and management of haemodynamic instability (online supplemental box S2); recommend conservative fluid management strategy (online supplemental box S3); outline criteria for rescue UFNET procedure for life-threatening fluid overload (online supplemental box S4). The management of CKRT, including the prescription of modality, blood flow, dialysate, replacement fluids, hemofilter and anticoagulation, will be as per the attending nephrologist and intensivist. All patients will receive an effluent flow rate of 20–25 mL/kg/hour.23 24 We will protocolise a low tidal volume ventilation strategy.21

    Discontinuation of study protocol

    The study protocol will be continued until one of the following occurs: (1) the attending physician determines that fluid removal is no longer needed permanently using CKRT; (2) a decision is made to stop CKRT and transition the patient to intermittent haemodialysis; (3) the patient or surrogate decision-makers decide to withdraw life-sustaining treatment; (4) the patient dies; or (5) day 28 after study enrolment, whichever occurs first.

    Facilitating adherence to study protocol

    We use educational programmes and academic detailing,25 automated reminders26 and regular audit with feedback.27 We will assess reasons for non-compliance, focusing on physician rationales, logistics and other factors. The web app UFNET rate calculator will track the recommended and the actual delivered UFNET rate by the ICU nurse. We suggest the clinician enter a comment in the web app when there is a deviation and the reason for deviation (eg, patient intolerance). We check protocol adherence two times per day and provide feedback at both sites. Reasons for non-adherence are recorded using pretested taxonomy distinguishing clinical and research-related reasons.

    Outcome measures

    The three primary feasibility outcomes, secondary and safety outcomes are shown in table 2.

    Table 2

    Study outcomes

    The three coprimary outcomes are:

    The between-group difference in mean delivered UFNET rate

    The primary objective is to achieve a minimum of 0.53 mL/kg/hour separation in the patient-delivered mean UFNET rates between the restrictive and liberal UFNET rate groups. We chose between-group separation as a feasibility metric because it is a robust measure of adherence to complex protocols and has been used in ICU trials assessing the feasibility of frequently titrated interventions.28 29 Specifically, we reasoned that a larger study would not be feasible if the separation in the UFNET rates were less than 0.5 mL/kg/hour. We chose 0.5 mL/kg/hour as a clinically meaningful difference because observational studies indicate that a 0.50 mL/kg/hour increase in UFNET rate is associated with increased mortality.11

    Protocol adherence

    In this trial, we define protocol deviation a priori as a delivered UFNET rate that lies>0.5 mL/kg/hour outside of the target UFNET rate range in the assigned treatment group for greater than six consecutive hours during fluid removal without significant changes in mean arterial pressure (MAP) (ie, MAP<65 mm Hg or ≥90 mm Hg). As such, out-of-range UFNET rates>0.5 mL/kg/hour beyond the target UFNET rate range will not constitute a protocol deviation when the bedside team titrated the UFNET rate as required to manage the patient haemodynamics (ie, when clinicians appropriately decreased rate or stopped fluid removal for hypotension; increased the rate for hypertension or treatment of respiratory distress due to pulmonary oedema).

    Recruitment rate

    A successful recruitment rate will be defined as achieving an enrolment rate of one patient every 2 months per ICU during the trial.

    Sample size calculations

    We initially planned to enrol 126 patients to detect a 0.52 mL/kg/hour difference in delivered UFNET rates between the two groups. After accounting for a 10% attrition rate, we aimed to enrol 144 patients. Subsequently, we estimated that if each of the 10 ICUs enrolled a minimum of 0.93 patients per 2 months for 24 months, we would have 111 patients. Using the sample size calculation for SW-CRT design, we estimated that 111 patients would have 80% power at a two-sided alpha of 0.05 to reject the null hypothesis that the average UFNET rate was at least 0.53–0.57 mL/kg/hour different between the two groups, using intracluster correlation coefficient of 0.01, at a Standard Deviation of 0.75 and assuming mean UFNET rate of 1.0 mL/kg/hour. Thus, we plan to recruit 112 patients or 56 patients per group (online supplemental table S1).

    Analysis plan

    All analyses will be performed intention to treat, and there will be no interim analysis. We will also perform per-protocol analyses where the assigned intervention was followed. Missing data will be imputed after examining the reason for missingness. For data that are missing completely at random or missing related to other collected covariates, we will use the multivariable imputation by chained equation to impute data before fitting models.30 We will use adjusted generalised linear mixed models (GLMM) for all primary and secondary outcomes to account for temporal and clustering effects.20 31 32

    We will adjust all analyses for prespecified variables such as age, baseline estimated glomerular filtration rate (eGFR), the severity of illness as measured by the Acute Physiology and Chronic Health Evaluation—III scores, Elixhauser Comorbidity Index Score, use or non-use of mechanical ventilation, admission source, percentage of fluid overload before study enrolment, presence or absence of sepsis, baseline cardiovascular Sequential Organ Failure Assessment (SOFA) Score and any other variable that is associated with the outcome of UFNET and had a between-group difference (p≤0.2) on univariable analysis. We will report adjusted and unadjusted analyses stratified by restrictive and liberal UFNET groups.

    Analysis methods for primary outcomes

    The between-group difference in mean delivered UFNET rate

    We will report the patient-averaged UFNET rate with a corresponding 95% CI for the two groups. The patient averaged UFNET rate was the average of all hours for days where any UFNET was delivered. We will use GLMM with a two-sided alpha of 0.05 to test the null hypothesis that the mean difference in the patient-averaged delivered UFNET rate was less than 0.53 mL/kg/hour between the liberal and restrictive UFNET groups after adjusting for prespecified variables. We will perform exploratory analysis to examine between-group differences in peak (ie, maximum) delivered UFNET rates and compare the longitudinal UFNET rate trajectories between the two groups. Subgroup analysis will be conducted among those with and without>10% fluid overload at enrolment; eGFR≥ and <60 mL/min/1.73 m2; duration of UFNET before enrolment≥6 and <6 hours; age≥65 and <65 years; with and without sepsis; and baseline cardiovascular SOFA Score≥3 and <3.

    Protocol adherence

    We will report the following stratified by restrictive and liberal UFNET rate groups: (1) no. of occurrences of deviations for 6 consecutive hours divided by UFNET days (mean events per day); (2) no. of days with UFNET out of range for at least 6 consecutive hours; (3) no. of patients with at least one occurrence of UFNET rate out of range for 6 consecutive hours. The proportion of patients with protocol deviation will be compared using the Wald test for GLMM and adjusted for prespecified covariates. As an exploratory analysis, we will also report the proportion of total hours of UFNET rate within, above or below range for each patient weighted equally and proportionally to total hours of fluid removal.

    Recruitment rate

    The recruitment metric will be calculated as the mean number (SD) of recruited patients per active screening month.

    Analysis methods for secondary and safety outcomes

    We will use GLMM regression after adjusting for various prespecified variables. We will consider observed differences between groups statistically significant at a two-sided, nominal alpha of 0.05 for all outcomes. We will use a Wald test for GLMM to compare binary outcomes and zero-truncated negative binomial regression to calculate incident rate ratios for continuous variables, adjusted for prespecified baseline covariates. For independence from KRT and hospital mortality, we will fit GLMM regression models, and report adjusted ORs with corresponding 95% CIs from logistic regression. To safeguard against erroneous type 1 error inflation in secondary outcomes, we will apply the conservative Hochberg procedure for adjustment on multiplicity to two key secondary outcomes of mortality and KRT dependence.33 34 Because of the potential for type 1 error due to multiple comparisons, findings for analyses of the other secondary and safety endpoints will be interpreted as exploratory.

    Data collection and management

    Data are collected from direct observation and electronic health record review. Data quality will be reviewed remotely using front-end range and logic checks during data entry and back-end data monitoring using SAS (Version 9.4) reports. The principal investigators at the two sites will perform random audits of up to 10% of electronic data collection forms and verify source documents. All clinical trial data will be stored for at least 7 years. Online supplemental table S2 shows the scheduled events for data collection.


    We will share our trial findings widely using oral presentations, abstracts and publications. We will publish the study results in a peer-reviewed medical journal with open access within 1 year of the end of the trial. Authorship guidelines will be consistent with those proposed by the International Committee of Medical Journal Editors policy. We will post the protocol outline on and publish our statistical analysis plan. Our final study dataset will be deidentified and not include protected health information or patient identifiers. We will follow National Institutes of Health policy on sharing and releasing data.


    To maintain patient confidentiality, all evaluation forms and reports will be identified only by a coded number. A computer will generate the coded number; only the study team can access the codes. All records will be kept in a locked, password-protected computer. All computer entry and networking programmes will be done only with coded numbers. Study information will not be released without the patient’s permission except as necessary for the Data Safety Monitoring Board (DSMB) monitoring.

    Safety monitoring

    All adverse events (AE) and serious AE (SAE) are reportable to the University of Pittsburgh HRPO if they are unanticipated, related to the study intervention, and place patients or others at a greater risk of harm. Unanticipated AEs must be reported within 10 working days, and SAEs must be reported within 24 hours to HRPO. The DSMB includes members with expertise in critical care medicine, nephrology and biostatistics that provide trial oversight. The DSMB will be notified of all unanticipated SAEs within 24–48 hours and other unanticipated problems within 10 working days. The HRPO and DSMB will also review all AEs and clinical outcomes during regularly scheduled meetings and at the final analyses.

    Patient and public involvement

    Patients and/or the public were not involved in the design, conduct, reporting or dissemination plans of this research.


    We wrestled with a few study design challenges. We made decisions based on previous research, surveys of existing practice and an iterative consensus model when data were less rigorous. We detail these challenges and decisions subsequently here.

    First, our overall objective is to conduct a larger randomised trial to assess the influence of protocol-based UFNET strategy on patient-centred outcomes. Thus, we sought to develop a pilot protocol that is feasible to deploy. Based on observational studies, we hypothesised that a restrictive UFNET rate strategy embracing a ‘slow and steady’ approach to fluid removal is feasible and is associated with fewer complications than a more liberal ‘sprint and pause’ strategy. We chose a UFNET rate of 0.5–1.5 mL/kg/hour in the restrictive arm because these UFNET rates are inclusive of the rates studied in the middle UFNET rate group in observational studies.11 13 We chose 2.0–5.0 mL/kg/hour in the liberal arm because these UFNET rates are also within the UFNET rate ranges studied in observational studies11 13 and surveys.6–9

    Second, we debated including a usual care arm in our trial. However, there are few published protocols for UFNET, and surveys show considerable variation in clinical practice.6–9 This makes it difficult to define what represents usual care for UFNET. A usual care arm was also not feasible due to budgetary constraints. Additionally, the scientific validity of a usual care arm is questionable, given the likelihood of drift in treatment practice over time due to the Hawthorne effect. To ensure the separation of delivered UFNET rates between study groups and to assess clinical outcomes, we decided to protocolise the UFNET. Our study design is similar to previous trials of protocol-driven solute dosing strategies for AKI,23 24 acute respiratory distress syndrome21 35 and sepsis.36 These trials did not include a usual care arm because there was no consensus regarding best practices or precise knowledge of prevailing practice patterns before trial implementation.37–39

    Third, we ensured the treatment arms paralleled current clinical practice in implementing the two intervention strategies. As a preparatory to the trial, a survey of attending intensivists and nephrologists at the two study sites suggested that most clinicians had the equipoise to enrol their patients in this trial. Fourth, we recognised that the protocol might be challenging based on haemodynamic factors, clinician preference, staffing burden and safety. Thus, we assessed attending physician equipoise to the UFNET rate before enrolment; developed protocols for managing haemodynamic instability; recommended conservative fluid management and rescue UFNET procedures; and allowed clinicians to stop or personalise UFNET based on patient needs. Finally, we collect granular safety outcomes data to assess risks to the patients monitored by the DSMB and the University of Pittsburgh HRPO.

    The design of this feasibility trial has limitations. First, caregivers cannot be blinded due to the nature of the intervention; however, awareness of treatment assignment is unlikely to compromise the study’s validity due to the protocolisation of intervention and the objective outcomes. Second, it is possible that attending physicians will have firm beliefs about the rates of UFNET, and awareness of treatment assignment may bias physicians to allow or not allow enrolment of their patients; for this reason, we assessed equipoise every 12–24 hours and are capturing reasons for non-enrolment that will be examined across the two groups. Third, we did not assess circulating intravascular volume to guide UFNET, as no technology has been validated for fluid removal.17

    Extracorporeal net fluid removal is labor-intensive, expensive and associated with complications. The lack of high-quality data makes it challenging for clinicians to determine the optimal approach to UFNET. The Restrictive versus Liberal rate of Extracorporeal Volume removal Evaluation in Acute Kidney Injury study is the first randomised trial to evaluate the feasibility of alternative UFNET approaches and will inform the study design of a larger clinical trial to assess patient-centred outcomes.

    Ethics statements

    Patient consent for publication


    We thank the study participants of the Restrictive versus Liberal rate of Extracorporeal Volume removal Evaluation in Acute Kidney Injury clinical trial and their legally authorised representatives. We also thank the Biostatistics and Data Management Core of the Clinical Research, Investigation, Systems Modelling of Acute Illness Center for data management and the Multidisciplinary Acute Care Research Organization for providing coordinator support.


    Supplementary materials

    • Supplementary Data

      This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.

    • Supplementary Data

      This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.


    • Twitter @PaulPalevsky

    • Correction notice This article has been corrected since it was published. State in affiliation 4 has been corrected.

    • Collaborators Ali Al-Khafaji, Brad Butcher, Philip Lamberty, Firas Abdulmajeed, Scott Gunn, Andrea Kattah, Michele Elder, Denise Scholl, William Sabol, Tina Vita

    • Contributors RM designed the study and wrote the manuscript. KK, DTH, PMP and CCH contributed substantially to the study’s conception and design. RM and CCH wrote the statistical analysis plan and estimated sample size. KK, DTH, PMP, CCH, MR and NN contributed towards drafting the work, revising it critically for important intellectual content and approving the final version of the manuscript. All authors agreed to be accountable for all aspects of the work and ensure the accuracy and integrity of any part of the work.

    • Funding This study was supported by United States National Institute of Diabetes and Digestive and Kidney Diseases (NIDDK), grant number: R01DK128100.

    • Competing interests RM received research grants from NIDDK, consulting fees from Baxter, AM Pharma, Bioporto and La Jolla unrelated to this study; DTH received grants from NIH. KK received research grants NIDDK and from Philips Research North America and Google, speaker honorarium from Nikkiso Critical Care Medical Supplies (Shanghai) and consulting fees to Mayo Clinic and from Baxter; PMP received consulting fees and advisory committee fees from Durect, Health-Span Dx and Novartis; served on a Data and Safety Monitoring Board for Baxter; served as a member of an endpoint adjudication committee for GE Healthcare; and CCH, MR and NN has nothing to disclose.

    • Patient and public involvement Patients and/or the public were not involved in the design, or conduct, or reporting, or dissemination plans of this research.

    • Provenance and peer review Not commissioned; peer reviewed for ethical and funding approval prior to submission.

    • Supplemental material This content has been supplied by the author(s). It has not been vetted by BMJ Publishing Group Limited (BMJ) and may not have been peer-reviewed. Any opinions or recommendations discussed are solely those of the author(s) and are not endorsed by BMJ. BMJ disclaims all liability and responsibility arising from any reliance placed on the content. Where the content includes any translated material, BMJ does not warrant the accuracy and reliability of the translations (including but not limited to local regulations, clinical guidelines, terminology, drug names and drug dosages), and is not responsible for any error and/or omissions arising from translation and adaptation or otherwise.