The REDUSE Feasibility Trial was an investigator-initiated, multicenter, parallel-group randomized trial. The trial protocol and the statistical analysis plan have been published previously [8, 9]. All procedures followed the ethical standards of the 1964 Helsinki declaration and its later amendments. The trial protocol was approved by the Swedish ethics review authority (#2020-06594, 08 August 2021) and registered at clinicaltrials.gov (NCT05249088, 18 February 2022) prior to trial initiation. We primarily used a deferred consent process and obtained informed consent from all surviving participants. All authors vouch for the completeness of the data and the fidelity of the trial to the protocol.

Adult patients (≥ 18 years of age) with septic shock (suspected/confirmed infection, plasma lactate > 2 mmol/l and infusion of vasopressor to maintain MAP > 65 mmHg after adequate fluid resuscitation) within 12 h of admission to the ICU and ongoing vasopressor therapy at the time of inclusion were eligible for inclusion [10]. Initially, we also mandated that patients should have fulfilled criteria for septic shock within the three hours preceding inclusion. To promote patient inclusion, the latter criterion was dropped after the inclusion of seven patients. The exclusion criterion was suspected or confirmed pregnancy.

Eligible patients were identified by the clinician caring for the patient and were randomized using a centralized, internet-based computer-generated allocation tool, stratified by site. Allocation was performed in a 1:1 ratio in permuted blocks of varying size to either protocolized restriction of non-resuscitation fluids or usual care. The clinical team caring for the patient was not blinded to the assigned intervention due to the nature of the intervention. Study participants, their relatives, trial statisticians, and outcome assessors at 90-days and six months were blinded to the assigned intervention.

In the restrictive fluid group, maintenance fluid was discontinued in participants who were in a positive cumulative balance and judged not to be dehydrated by the treating physician. The time interval for assessment of fluid balance was left at the discretion of the treating physician. Participants deemed to have a neutral or negative cumulative balance received fluid to cover their daily need for water (1 ml/kg/h) and ongoing losses. Enteral nutrition was administered according to local practice but always at an energy density of at least 2 kcal/ml. Glucose could be used at a maximum dose of 1 g/kg/day and a concentration of 20% or greater as nutrition, starting at the earliest at 72 h from inclusion, if enteral nutrition was not tolerated. In participants with insulin-dependent diabetes mellitus, glucose solutions could be started earlier if enteral feeding was not tolerated and if mandated by local protocol. Parenteral nutrition was administered according to local routines, and enteral water/intravenous fluid could be given as needed to correct electrolyte imbalances. Intravenous medications were concentrated according to a trial-specific protocol developed by the investigators in collaboration with two pharmacists and was based on the most concentrated solutions found in the literature (please see Supplementary Material, Appendix A). Volume used to flush lines after injections were administered according to local practice.

Participants pragmatically received non-resuscitation fluids in the usual care group according to local practice. Unless local protocols stated otherwise, maintenance fluid (crystalloids and/or glucose solutions and/or enteral water) was given at a dose of 1 ml/kg/h. Glucose solutions were used at a maximum concentration of 10%, medications, enteral and parenteral nutrition was administered according to local guidelines.

In both groups, resuscitation fluids were administered according to the Surviving Sepsis Campaign (SSC) guidelines in the salvage and optimization phase and according to local protocol in the stabilization and de-escalation phases [11, 12]. In participants who required surgery, fluids were administered at the discretion of the anesthetist. The assigned interventions continued for the duration of the ICU stay, up to a maximum of 90 days.

The primary feasibility outcome was total volume of fluid administered within three days of randomization (D0–D3). The secondary feasibility outcomes were (a) proportion of participants with outcome data on all-cause mortality, days alive and free of mechanical ventilation, acute kidney injury and ischemic events in the ICU (cerebral, cardiac, intestinal or limb ischemia) within 90 days of inclusion (please see Supplementary Material for definitions), (b) proportion of surviving participants assessed for health-related quality of life (HRQoL) by the EQ-5D five level version (EQ-5D-5L) and the Montreal Cognitive Assessment (MoCA) at six months after inclusion, (c) proportion of eligible patients who were randomized and consented, (d) proportion of participants experiencing at least one protocol deviation [13,14,15,16].

Primary exploratory clinical outcomes were all-cause mortality at 90 days from inclusion, ≥ 1 complication in the ICU (acute kidney injury or ischemic events), days alive and free of mechanical ventilation within 90 days of inclusion, cognitive function measured by MoCA, and HRQoL measured by EQ-5D-5L visual analog scale (VAS) at six months. Secondary exploratory clinical outcomes include volumes of resuscitation and non-resuscitation fluids and cumulative fluid balance at day three and five after inclusion, hemodynamic stability (highest lactate, highest dose of noradrenaline and highest cardiovascular SOFA-score) and use of diuretics during the first five days from inclusion, renal function within 90 days from inclusion and functional outcome as measured by the Glasgow Outcome Scale Extended (GOSE) at six months from inclusion [17]]. In addition, we registered incidence of hypoglycemia (≤ 3.9 mmol/l), electrolyte/metabolic disturbances and suspected unexpected serious adverse complications in the ICU (see Supplementary Material for details).

Our previous study suggested that the total volume of fluid may be reduced by a median of 3.1 (interquartile range [IQR] 1.5–5.0 and standard deviation of 2.8) liters in the first 3 days after ICU admission by restrictive administration of non-resuscitation fluids [1]. Based on the standard deviation derived above we needed 42 participants in each group for the trial to have a power of 90% at an alpha level of 0.05 to detect a difference of 2 L between the groups. To account for data not being normally distributed, we included 15% more participants than the calculated sample size, resulting in a total sample size of 98 participants in the trial [18].

Thresholds for the secondary feasibility outcomes were: proportion of participants with clinical outcome data on all-cause mortality, days alive and free of mechanical ventilation and acute kidney injury and ischemic events in the ICU: 95% (95% confidence interval [CI] 89–98); proportion of surviving participants assessed by the EQ-5D-5L and MoCA at six months after inclusion: 85% (95% CI 73–92); proportion of eligible patients who were randomized and consented: 75% (95% CI 67–81); and proportion of participants experiencing at least one protocol deviation: 10% (95% CI 6–18).

Two statisticians performed all analyses independently according to the statistical analysis plan before unblinding data. The analyses were performed according to an intention-to-treat principle and adjusted for participating site [9]. The primary feasibility outcome was analyzed using the van Elteren test and is presented with median differences with 95% Hodges–Lehman confidence intervals (95% HLCI). All available data was used in the primary analysis. The secondary feasibility outcomes are presented as percentages with confidence intervals calculated using the 1-sample proportions test without continuity correction. For the exploratory clinical outcomes, we analyzed count data outcomes using the van Elteren test with adjustment for site, continuous outcomes using mixed effects linear regression with site as random intercept, and dichotomous outcomes using mixed effect logistic regression with site as random intercept and relative risks obtained using the ‘nlcom’ Stata command and/or by G-computation in R. Due to the exploratory nature of the trial, p values were not adjusted for multiple comparisons. All statistical analyses were performed using Stata v. 17 (StataCorp LLC, Texas, USA) and/or R v. 4.2.1 (R Core Team, Vienna Austria). A p value of 0.05 or less is considered statistically significant.