The ultrasound gap in de-resuscitation trials: A systematic review of physiological assessment deficits Authors: Adrian Wong, Paulo Melo, Federico Stefanini, Kim Cheah, Eduardo Kattan, Mateusz Zawadka Published: 2026 ABSTRACT Background:Fluid overload is associated with increased mortality in critically ill patients, yet de-resuscitation trials frequently report negative or inconclusive results. This systematic review examines the relationship between parameter objectivity, ultrasound utilization, and trial outcomes in de-resuscitation research.Methods:We conducted a comprehensive systematic review of randomized controlled trials investigating de-resuscitation interventions in critically ill adults. Risk of bias was evaluated using the Cochrane Risk of Bias 2 tool. Statistical analysis included comparison of ultrasound versus non-ultrasound studies and pooled outcome analysis.Results:13 studies were identified which randomised 2495 patients. Of the included studies, 4 (30.8%) performed some form of ultrasound assessment. However, only 1 study (7.7%) explicitly used ultrasound to guide de-resuscitation interventions. However, it did not achieve significant fluid separation between the control and interventional arm. No studies achieved positive hard clinical outcomes (mortality reduction or major morbidity reduction) from de-resuscitation interventions. However, several studies successfully achieved their stated primary endpoints: fluid balance separation was achieved in 7 of 13 studies (53.8%), feasibility objectives were met in 3 studies, and improvements in secondary endpoints including ventilator-free days and time to extubation were demonstrated in select studies. These findings suggest that achieving targeted fluid balance changes does not necessarily translate to improved clinical outcomes.Conclusions:Current research is characterized by a significant ‘ultrasound gap,’ where the theoretical benefits of POCUS are not yet integrated into interventional trials. Future research must bridge this gap using standardized, physiology-driven protocols. Introduction The management of fluid overload in critically ill patients represents one of the most challenging aspects of intensive care medicine, with profound implications for patient outcomes and healthcare resource utilization (1). Fluid accumulation occurs in the majority of critically ill patients due to aggressive initial resuscitation, ongoing fluid administration for medication delivery, nutritional support, and the pathophysiological changes associated with critical illness that impair fluid excretion (2,3). Observational studies consistently demonstrate a dose-dependent relationship between positive fluid balance and adverse outcomes, including increased mortality, prolonged mechanical ventilation, and extended intensive care unit (ICU) stays (4,5). Despite this compelling observational evidence, randomized controlled trials (RCTs) investigating de-resuscitation strategies have yielded disappointing results, with many studies failing to demonstrate clear clinical benefits (6,7). This apparent paradox between observational evidence and interventional trial outcomes has generated considerable debate within the critical care community and has led to various hypotheses attempting to explain these discordant findings (8). One prominent hypothesis suggests that the failure of de-resuscitation trials stems from the use of subjective assessment parameters that introduce variability, bias, and inconsistency in intervention delivery (7, 9). Proponents of this theory argue that objective, measurable parameters with specific numerical thresholds should yield more consistent and successful trial outcomes by reducing clinician variability and improving protocol adherence (10, 11). This perspective has influenced the design of recent de-resuscitation trials, with investigators increasingly incorporating objective hemodynamic targets, biomarker-guided approaches, and standardized assessment protocols (12, 13). However, this hypothesis has never been systematically tested across the broader landscape of de-resuscitation research. Furthermore, the definition and classification of “objective” versus “subjective” parameters in de-resuscitation research remains poorly standardized, with different investigators applying varying criteria and thresholds. Point-of-care ultrasound (POCUS) has emerged as a promising tool for objective, real-time assessment of fluid status and responsiveness in critically ill patients (14 – 16). Multiple studies have demonstrated the utility of ultrasound-derived parameters, including inferior vena cava (IVC) diameter and collapsibility, venous excess ultrasound (VExUS) score, and echocardiographic assessment of cardiac function, in guiding fluid management decisions (17, 18). Although ultrasound assessment offers potential advantages for objective fluid status evaluation, its inclusion within deresuscitation protocols remains limited. This is partially because the VExUS and fluid tolerance are relatively new concepts (19 – 22). Previous studies have demonstrated the utility of ultrasound parameters in assessing fluid responsiveness in critically ill patients, yet the specific application of ultrasound guidance to active deresuscitation interventions has not been systematically evaluated. Some investigators have incorporated comprehensive ultrasound assessments into their protocols, while others have relied on traditional hemodynamic parameters or clinical judgment alone. This heterogeneity in approach provides an opportunity to examine whether ultrasound utilization is associated with improved trial outcomes or more successful achievement of fluid balance targets. This systematic review aims to comprehensively evaluate the relationship between parameter objectivity, ultrasound utilization, and trial outcomes in de-resuscitation research. We hypothesized that studies incorporating explicit ultrasound guidance for de-resuscitation decisions would demonstrate superior fluid balance achievement and clinical outcomes compared to those relying on traditional assessment methods. Additionally, we sought to examine whether parameter objectivity (as classified across five assessment domains) would predict trial success rates. Methods Study design and protocol We conducted a comprehensive systematic review of randomized controlled trials (RCTs) investigating de-resuscitation interventions in critically ill patients. The review was designed to examine the relationship between parameter objectivity, ultrasound utilization, and trial outcomes. Our methodology incorporated both quantitative parameter classification and qualitative assessment of trial characteristics to provide a comprehensive evaluation of factors influencing de-resuscitation trial success. The review protocol was developed according to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines (23). The review has also been registered on PROSPERO (CRD420251275295). Search strategy and information sources A comprehensive search of multiple electronic databases was conducted, including PubMed, Embase, Cochrane Central Register of Controlled Trials (CENTRAL), and Web of Science, from their inception to August 2025. The search strategy combined keywords and Medical Subject Headings (MeSH) terms related to “fluid overload,” “de-resuscitation,” “fluid removal,” “diuretics,” “ultrafiltration,” “critical care,” “intensive care,” and “randomized controlled trial.” No language restrictions were applied to maximize study inclusion. The search strategy was developed in consultation with a medical librarian, and refined through iterative testing to ensure comprehensive coverage of relevant literature. Reference lists of included studies and relevant review articles were manually screened for additional eligible trials. Conference abstracts and grey literature were also searched to identify unpublished or recently completed studies. Study selection and eligibility criteria Studies were included if they met the following criteria: (1) RCT design investigating active fluid removal or conservative fluid management strategies; (2) adult critically ill patients (age ≥18 years) in ICUs; (3) intervention aimed at achieving negative fluid balance, preventing further fluid accumulation, or actively removing excess fluid; (4) comparison with usual care, placebo, or alternative fluid management strategy; and (5) reported clinical outcomes with sufficient methodological detail for parameter classification. Studies were excluded if they: (1) focused solely on initial resuscitation without addressing subsequent fluid management; (2) involved paediatric populations (age <18 years); (3) were conducted in non-critically ill populations; (4) lacked sufficient methodological detail for parameter assessment; (5) were observational studies, case series, or non-randomized trials; or (6) investigated interventions not directly related to fluid balance management. Two review authors independently screened titles and abstracts of all identified records. Full texts of potentially eligible articles were retrieved and assessed for final inclusion. Disagreements between reviewers were resolved through discussion and consensus, with a third author available for arbitration when necessary. Data extraction and management A standardized data extraction form was developed and pilot-tested on a subset of included studies. One author extracted data from all included studies, and a second author independently verified the extracted data for accuracy and completeness. Extracted data included study characteristics (design, setting, sample size), population demographics, intervention details, control group characteristics, outcome measures, and specific parameters required for our classification framework. Particular attention was paid to extracting detailed information about ultrasound utilisation, including the specific ultrasound parameters used, frequency of assessment, operator training, and integration into clinical decision-making algorithms. For the primary comparative analysis, studies were classified as ‘ultrasound-guided de-resuscitation’ only if ultrasound assessment (including inferior vena cava diameter/collapsibility, venous excess ultrasound score, echocardiography, or Doppler-derived parameters) was explicitly used to trigger or guide active fluid removal decisions. Studies where ultrasound was performed for baseline characterization, concurrent hemodynamic monitoring, or weaning assessment but not specifically to guide deresuscitation interventions were classified as ‘non-ultrasound-guided de-resuscitation’ for the primary analysis, though secondary classification tracked all ultrasound usage. Risk of bias assessment All included studies underwent comprehensive risk of bias assessment using the Cochrane Risk of Bias 2 (RoB 2) tool (24). This tool evaluates five domains of potential bias: (1) bias arising from the randomization process, (2) bias due to deviations from intended interventions, (3) bias due to missing outcome data, (4) bias in measurement of the outcome, and (5) bias in selection of the reported result. Each domain was assessed as “Low risk,” “Some concerns,” or “High risk” based on standardized criteria, with an overall judgment derived from the individual domain assessments. The RoB 2 assessment was particularly important for this analysis given the inherent challenges in blinding fluid management interventions and the potential for implementation bias in complex clinical protocols. Special attention was paid to the impact of open-label designs on outcome measurement and the potential for differential implementation of interventions between study groups. Statistical analysis Descriptive statistics were used to summarize study characteristics and outcome data. Categorical variables describing study design, population, and methodology were presented as frequencies and percentages. Continuous variables including sample sizes and outcome percentages were presented as medians with interquartile ranges or means with standard deviations as appropriate. The primary comparative analysis examined fluid balance achievement rates between ultrasound-guided and non-ultrasound-guided studies using descriptive comparison. Fisher’s exact test was used to compare proportions of studies achieving fluid balance separation between groups (α = 0.05). For studies reporting fluid balance as continuous variables with between-group comparisons, achievement of ‘separation’ was defined as reported p-value <0.05 for the comparison. No meta-analytic pooling was performed due to heterogeneity in patient populations, intervention protocols, timing of assessment, and outcome definitions. Exploratory assessment of temporal trends in ultrasound adoption across publication decades was performed using descriptive comparison. Results Study selection and characteristics The systematic search identified 2,847 potentially relevant records after duplicate removal. Following title and abstract screening, thirteen RCTs met the inclusion criteria and were included in the final analysis (25 – 37). The study selection process is detailed in the PRISMA flow diagram (Figure 1). Prisma Flow Diagram. The 13 included studies consisted of 2,495 participants and were published between 2004 and 2025, representing two decades of de-resuscitation research. Study sample sizes ranged from 23 to 1,001 participants (median 100, IQR 41-180). The characteristics of the included trials are shown in Table 1. Table 1.Characteristics of included trialsStudy (First author)Design / SettingPopulation (n analyzed)InterventionComparatorMain inclusion criteriaPrimary outcome(s)Duration / Follow-up Wichmann et al. [35] Multicentre, double-blind RCT (3 ICUs, Denmark)Clinically stable adult ICU patients with ≥5% fluid overload (n=41)Goal-directed deresuscitation with continuous IV furosemide infusion targeting neutral cumulative fluid balance (±750 mL)Placebo infusion with identical protocolICU adults with shock treated ≥4 h with vasopressors ≥ 0.1 µg kg-1 min-1 and ≥1 marker of fluid overloadDays alive and out of hospital at 90 days 90 days Semler et al. [31] Single-centre, open-label, randomized pilot trial (Vanderbilt Univ. MICU, USA)Adults with sepsis and cardiopulmonary dysfunction after initial resuscitation (n=30)Conservative fluid management protocol: fluids restricted during shock; furosemide infusion (3–24 mg h-1) to achieve neutral daily balanceUsual care (clinician-directed fluid and diuretic use)Sepsis with ≥2 SIRS + shock (MAP < 60 mm Hg or vasopressors) or respiratory failure (MV or SpO2 < 97% on FiO2 ≥ 0.3)Mean daily fluid balance (Phase II); ICU-free days (Phase III, not reached)14 days (Phase II only) Silversides et al. [34] Multicentre open-label RCT (8 UK ICUs)Mechanically ventilated adults > 24 h post-ICU admission (n = 179)Two-stage strategy: conservative fluid administration + active deresuscitation (IV furosemide ± indapamide + spironolactone or ultrafiltration)Usual careVentilated ICU pts 24–48 h after admission; expected > 1 day stay24-h fluid balance (day 2–3) 180 days Cinotti et al. [32] Multicentre single-blind RCT (4 ICUs, France)Adults on invasive MV with ≥3% weight gain after stabilization (n=166 ITT)Protocolised furosemide to return to reference weight; hold for predefined metabolic/renal limitsNo diuretics (rescue only for pulmonary edema)MV with FiO2 ≤60%, PEEP ≤10; hemodynamic stabilization; weight gain ≥3%Fluid balance (weight change to successful extubation) 60 days Hamishehkar et al. [28] Multicentre RCT (2 ICUs, Iran)ICU adults with AKI (n=100)Furosemide 40–80 mg IV bolus → 1–5 mg/hinfusion; titrate to UO > 0.5 mL·kg-1·h-1Usual care (no diuretics)AKI by RIFLE criteriaNGAL (plasma/urine) trends; kidney function (Cr, BUN)Biomarkers to day 7; mortality to day 28. Berthelsen et al. [30] Multicentre pilot RCT (Denmark)n = 20 (7 forced; 13 usual care)Forced deresuscitation: furosemide bolus→infusion (≤ 40 mg/h) ± CRRT; target −1 mL/kg IBW/h; safety stops (MAP < 50, mottling, lactate > 4)Usual care; CRRT discouraged unless predefined indicationsICU adults with AKI (KDIGO), RRS < 60%, > 10 % fluid overloadCumulative fluid balance at day 590 days. Wiedemann et al. [25] Multicentre RCT, 2×2 (conservative vs liberal fluids; PAC vs CVC) n = 1000 Conservative: CVP < 4 or PAOP < 8; diuretic-driven; pause during shockLiberal: CVP 10–14 or PAOP 14–18; bolus-drivenMV adults with ALI/ARDS (PaO2/FiO2 < 300; bilateral infiltrates; no LA hypertension)60-day mortalityProtocol ≤7 d; outcomes to 60 d. Mekontso Dessap et al. [27] Multicentre RCT; AWS used in both armsn = 304 (152/arm)BNP-guided: if BNP ≥200 pg/mL, restrict fluids; IV furosemide 10–30 mg q3h(target UO 4.5–9 mL·kg-1/3 h); ± acetazolamideUsual care, BNP blindedMV ≥24 h; FiO2 ≤50%; PEEP ≤8; hemodynamic stability; sedation decreased/stoppedTime to successful extubation60 days. Chen & Kollef [36] Single-centre pilot RCT (medical ICUs)n = 82 (41/41)TFM with daily fluid-responsiveness testing (PPV, IVC index, SVI; ≥2 positive) → if non-responsive: restrict/stop fluids, concentrate infusions; consider diuretics/RRTUsual care (no TFM recommendations)Septic shock needing vasopressors ≥12 h after initial ≥30 mL/kgresuscitationStudy-fluid volumeand net balance by day 3 & day 5Up to 5 days (ICU stay) Cantarovich et al. [26] Multicentre double-blind RCT (France; ICUs & nephrology wards)n = 330 (166 / 164)High-dose furosemide: 25 mg/kg/day IV (max 2 g) or 35 mg/kg/day PO (max 2.5 g) after a 15 mg/kg IV test; once daily post-RRTPlacebo(matched IV/PO)Established ARF requiring RRT(BUN>180 mg/dL, oligo-anuria ≥48 h, or uremic syndrome)SurvivalThrough dialysis phase or 7 days post-RRT. Vaara et al. [33] Multicentre pilot RCT (7 ICUs, EU/AUS)n = 100 (49/51)Restrictive fluid management bundle: minimize inputs, match outputs with unrestricted diuretics, consider RRT to achieve neutral/negative daily balanceUsual careAdult ICU pts, AKI, not hypovolemic, 12–72 h in ICUCumulative fluid balance at 72 hIntervention ≤7 d; exploratory outcomes to 90 d. Bagshaw et al. [29] Multicentre blinded RCT (3 ICUs)n = 73 (37/36)Low-dose furosemide: 0.4 mg/kg bolus → 0.05–0.40 mg/kg/h infusion to UO 1–2 mL/kg/h (≤7 d)Placebosaline with identical algorithmEarly AKI (RIFLE–Risk), ≥2 SIRS, resuscitation goals achievedWorsening AKI within 7 dDrug ≤7 d; mortality to 90 d. Castro et al. [37] Single-centre RCT (Chile); physiological trial with bioreactance & PLRn = 24 (12/12)FR-guided deresuscitation (UF/diuretics per pathway) until PLR-positive, then stopEmpiric negative FB(same tools) until preset negative balanceMV >24 h & <7 d; >10% weight gain; stabilized, CRT<3 s, minimal/no vasopressorsPrimary objective:cardiovascular/hemodynamic outcomesIn-ICU; 48 h post-extubation. Characteristics of included trials The key outcomes and safety signals are shown in Table 2. Table 2.Key outcomes and safety signals.Study (First author)Δ Fluid balance (Interv – Ctrl)MortalityRenal outcomes (AKI / RRT)Other key outcomesUltrasound / Fluid responsiveness testingKey message Wichmann et al. [35] ~ −3 L by day 5; significant separation (p < 0.001)40 % vs 29 % (NS)AKI 35 % vs 24 %; RRT 4.8 % vs 0 %No difference in DAOH-90 (50 vs 45 days, p = 0.94) No / No Feasible deresuscitation strategy; early-terminated (4 % of target sample); neutral balance hard to assess by charting alone Semler et al. [31] −398 mL/day (95 % CI −1227 to +430; p = 0.33) – below −500 mL threshold27 % vs 30 % (NS)Stage III AKI 40 % vs 21 %; RRT 6.7 % vs 6.7 %ICU-free days 9 vs 11 (NS); ventilator-free 13 vs 12No (baseline IVC ultrasound only) / NoProtocol feasible; no harm; lower-than-expected fluid exposure limited group separation Silversides et al. [34] −970 mL (24 h), −3.3 L by day 5 (p < 0.01)21 % vs 16 % (NS)AKI 5 % vs 6 % (NS) ; RRT ≈ 17 % baselineICU-free 13.8 vs 14.9 days (NS); non-serious AE ↑ (34 % vs 19 %) – mainly metabolic alkalosis No / No Feasible protocol; safe hemodynamics; reduced fluid balance; no clinical benefit; needs large RCT. Cinotti et al. [32] ≈ −5 kg (primary & sensitivity analyses; p<0.001 in main analyses)ICU 14% vs 18%, Day-60 17% vs 23% (NS)Worsening AKI 60% vs 75% (p=0.03); RRT 8% vs 5%(NS)VFD D60 54 vs 51(NS); MV 12 vs 14 d(NS)Yes / Yes (Echo/PPV for resuscitation protocols; not triggers)Diuretic-guided deresuscitation reduces fluid accumulationwith acceptable safety; no proven effect on hard outcomes. Hamishehkar et al. [28] No separation in fluid balance; urine output ↑ with furosemide (≈ +2 L/day)20% vs 28%(NS, 28-day)CRRT no difference; NGAL decreased in both groupsEarly lower pNGAL/Cr with furosemide; differences fade by day 7 No / No Furosemide increased diuresis but did not change fluid balance or outcomes; biomarker-centric RCT, not deresuscitation-triggered. Berthelsen et al. [30] −5.8 L at day 5 (p = 0.003); strong separation29 % vs 46 %(NS, 90 d)CRRT use similar; AF fewer with interventionNeutral balance 86 % vs 30 % (p = 0.06); no hemodynamic harm signal No / No Feasible forced removal protocol with clear fluid-balance reduction; trial underpowered for clinical outcomes. Wiedemann et al. [25] ≈ −7.1 L over 7 days (p<0.001)25.5% vs 28.4%(NS)Dialysis 10% vs 14%(p=0.06)VFDs 14.6 vs 12.1; ICU-free 13.4 vs 11.2 (both p<0.001)No / No (CVP/PAOP-guided)Conservative strategy improves VFDs/ICU-free days with large fluid-balance separation, no mortality difference. Mekontso Dessap et al. [27] ≈ −2.1 L more negative during weaning (p < 0.0001)ICU 11.8% vs 12.5%, D60 13.8% vs 18.4%(NS)Renal failure rates similar; no difference in shocks/electrolytesFaster extubation (42.4 vs 58.6 h, p=0.034); ↑ VFDs (D14/D28/D60); ↓ VAP and ↓ need for fluid loading No / No BNP-guided protocol safely reduces fluid balance and shortens weaning, especially with LVD; no LOS/mortality effect. Chen & Kollef [36] No significant separation: Day 3 median +1.95 L vs +3.12 L (P = .20); Day 5 +2.64 L vs +3.62 L (P = .40)56.1% vs 48.8%(NS)RRT 41.5% vs 39.0%(NS)Ventilator days 8.0 vs 5.0 (NS); vasopressor days 4 vs 4(NS)Yes / Yes (IVC index, SVI; PPV)Feasible TFM with daily responsiveness testing, but advisory design yielded no significant fluid-balance or outcome differences; mandates likely needed in future RCTs. Cantarovich et al. [26] Not addressed; diuresis endpoints favored furosemide No difference No difference in RRT sessions/time; Cr<2.26 mg/dL NSFaster to ≥2 L/day (5.7 vs 7.8 d, p=0.004); more pts reached ≥2 L/day (57% vs 33%, p<0.001); polyuria ↑ No / No High-dose furosemide increases diuresis but does not improve survival or renal recovery in established ARF needing RRT. Vaara et al. [33] −1.15 L at 72 h (p = 0.033); −0.82 L at 24 h (p = 0.004); −1.53 L by day 7/ICU discharge (p = 0.046)90-day mortality 19.6% vs 26.5% (NS, exploratory)RRT ≤14 d: 13% vs 30% (RR 0.42, p = 0.043)AEs 22% vs 49% (p = 0.001); SAEs 12% vs 31% (p = 0.031); duration of AKI trend (−1 d, p = 0.071) No / No RFM safely lowers fluid balance and reduces RRT use vs usual care in AKI; feasibility demonstrated; larger trials warranted. Bagshaw et al. [29] −1.08 L (NS) overall; UO ↑ at 6 h & 24 hHospital 8.1% vs 14.3% (NS); 90-day 21.6% vs 30.5% (NS)Worsening AKI 43% vs 37% (NS); RRT 27% vs 29% (NS)AEs incidence higher with furosemide (mostly electrolytes/alkalosis) No / No Low-dose furosemide did not reduce AKI progression or improve outcomes; protocol deviations & supplementary diuretics likely blunted separation. Castro et al. [37] No between-group FB difference reported; FR at endpoint 100% vs 58% (p = 0.012)Not reported (physio trial)Not reportedSBT 14 h vs 36 h (p = 0.031); Extubation 26 h vs 57 h (p = 0.007); MV 49 h vs 62 h (p = 0.065); pH 7.43 vs 7.47 (p = 0.028)No (echo performed but not used to guide fluid removal) / Yes (PLR-bioreactance)Targeting fluid responsiveness (not fixed negative FB) sped weaning and avoided alkalosis, with preserved perfusion and stable hemodynamics.AE, adverse event; AF, atrial fibrillation; AKI, acute kidney injury; ARF, acute renal failure; CI, confidence interval; Ctrl, control; Cr, creatinine; DAOH-90, days alive and out of hospital at day 90; d, days; Δ (delta), between-group difference (Interv − Ctrl); FB, fluid balance; FR, fluid responsiveness; ICU, intensive care unit; Interv, intervention; IVC, inferior vena cava; LOS, length of stay; LVD, left ventricular systolic dysfunction; MV, mechanical ventilation; NGAL, neutrophil gelatinase–associated lipocalin; pNGAL, plasma NGAL; NS, not significant; PAOP, pulmonary artery occlusion pressure; PLR, passive leg raise; PPV, pulse-pressure variation; RCT, randomized controlled trial; RFM, restrictive fluid management; RR, risk ratio; RRT, renal replacement therapy; SAE, serious adverse event; SBT, spontaneous breathing trial; SVI, stroke volume index; TFM, targeted fluid minimization; UO, urine output; VAP, ventilator-associated pneumonia; VFD(s), ventilator-free day(s) Key outcomes and safety signals. AE, adverse event; AF, atrial fibrillation; AKI, acute kidney injury; ARF, acute renal failure; CI, confidence interval; Ctrl, control; Cr, creatinine; DAOH-90, days alive and out of hospital at day 90; d, days; Δ (delta), between-group difference (Interv − Ctrl); FB, fluid balance; FR, fluid responsiveness; ICU, intensive care unit; Interv, intervention; IVC, inferior vena cava; LOS, length of stay; LVD, left ventricular systolic dysfunction; MV, mechanical ventilation; NGAL, neutrophil gelatinase–associated lipocalin; pNGAL, plasma NGAL; NS, not significant; PAOP, pulmonary artery occlusion pressure; PLR, passive leg raise; PPV, pulse-pressure variation; RCT, randomized controlled trial; RFM, restrictive fluid management; RR, risk ratio; RRT, renal replacement therapy; SAE, serious adverse event; SBT, spontaneous breathing trial; SVI, stroke volume index; TFM, targeted fluid minimization; UO, urine output; VAP, ventilator-associated pneumonia; VFD(s), ventilator-free day(s) The included studies represented diverse critical care populations, including general ICU patients (n=3 studies), patients with acute kidney injury (n=5 studies), sepsis patients (n=2 studies), patients requiring mechanical ventilation weaning (n=2 studies), and patients with acute respiratory distress syndrome (n=1 study). This diversity provides a comprehensive view of de-resuscitation approaches across different clinical contexts. Ultrasound utilization analysis Four of 13 studies (30.8%) performed some form of ultrasound assessment, but only one study (7.7%) explicitly used ultrasound parameters to guide active de-resuscitation interventions. Ultrasound utilization and fluid balance separation Of the 13 studies, four (30.8%) [31, 32, 36, 37] utilized ultrasound in some capacity for patient assessment, while nine (69.2%) did not. A post-hoc analysis was conducted to evaluate the association between any ultrasound use and the achievement of significant fluid balance separation between study arms (Table 3). One of the four studies involving any ultrasound use (25.0%) achieved significant fluid balance separation, compared to six of the nine studies with no ultrasound involvement (66.7%). This difference was not statistically significant (p = 0.27, Fisher’s exact test; OR 0.17, 95% CI [0.01, 2.45]). Table 3.Comparison of Fluid Balance Separation by Any Ultrasound Involvement.GroupAchieved SeparationDid Not Achieve SeparationTotal Studies Any Ultrasound Use 1 (25.0%)3 (75.0%)4 No Ultrasound Use 6 (66.7%)3 (33.3%)9 Total 7613 Comparison of Fluid Balance Separation by Any Ultrasound Involvement. A sub-analysis was performed to assess the impact of protocols that explicitly used ultrasound-derived parameters to guide the de-resuscitation strategy (Table 4). Only one study (7.7%) met this definition [36]; it did not achieve significant fluid balance separation (0/1, 0%). Among the 12 studies that did not use explicit ultrasound guidance, seven (58.3%) achieved significant separation. No significant association was found between explicit ultrasound guidance and fluid balance separation (p = 1.00, Fisher’s exact test). Table 4.Comparison of Fluid Balance Separation by Explicit Ultrasound Guidance for De-resuscitationGroupAchieved SeparationDid Not Achieve SeparationTotal Studies Explicit US Guidance 0 (0.0%)1 (100.0%)1 No Explicit US Guidance 7 (58.3%)5 (41.7%)12 Total 7613 Comparison of Fluid Balance Separation by Explicit Ultrasound Guidance for De-resuscitation Studies using ultrasound to guide de-resuscitation Chen & Kollef (36) implemented the most comprehensive ultrasound approach for guiding de-resuscitation, utilizing IVC distension index, PPV monitoring, and stroke volume index measurements for daily fluid responsiveness testing. However, the authors noted that recommendations were advisory rather than mandated, which likely attenuated the intervention effect. Despite comprehensive ultrasound assessment, this study failed to achieve significant fluid balance separation (Day 3: +1.95 L vs +3.12 L, p=0.20; Day 5: +2.64 L vs +3.62 L, p=0.40). Studies performing ultrasound assessment but not for de-resuscitation guidance Cinotti et al. (32) used echocardiography and PPV assessment within their ICU protocols, but these parameters were specifically used to guide fluid resuscitation during the initial stabilization phase, not to guide de-resuscitation decisions. During the de-resuscitation phase, Cinotti employed a protocolized diuretic strategy, although this was without ultrasound guidance. Therefore, Cinotti is appropriately classified as a non-ultrasound-guided de-resuscitation trial. Semler et al. (31) performed baseline IVC ultrasound measurement at enrolment for initial volume status characterisation but did not incorporate ultrasound guidance into the conservative fluid management protocol. The IVC measurements were recorded but not used to trigger or guide fluid removal decisions. Castro et al. (37) performed transthoracic echocardiography at baseline and before spontaneous breathing trials. However, fluid removal decisions were guided by thoracic bioreactance-based passive leg raise testing (an impedance-based technology distinct from POCUS), not by ultrasound parameters. Echocardiography was used for descriptive hemodynamic characterization, not intervention guidance. Pooled outcome analysis Pooled analysis across all 13 studies revealed patterns in de-resuscitation trial outcomes. Seven studies (53.8%) achieved significant fluid balance separation between intervention and control groups, indicating that only these trials achieved their intended outcome. However, this success in achieving fluid balance did not translate into positive clinical outcomes. Mortality outcomes: None of the studies demonstrated significant mortality benefits from de-resuscitation interventions. Several reported numerical trends toward improved mortality in intervention groups, although none achieved statistical significance. The largest study (FACTT trial, n=1000) showed no mortality difference between conservative and liberal fluid strategies (25). Secondary outcomes: While most studies achieved their intended fluid balance targets, secondary outcomes such as ventilator-free days, ICU length of stay, and organ dysfunction scores showed inconsistent patterns. Some studies reported modest improvements in secondary endpoints; however these were generally not sustained or clinically meaningful. Risk of bias assessment Complete Risk of Bias 2 (RoB 2) assessments were conducted for all 13 included studies using standardized Cochrane criteria (Figure 2). RoB2 Summary. Comprehensive assessment using the Cochrane RoB 2 tool found significant methodological challenges across the included studies. No studies achieved an overall assessment of “Low Risk”, highlighting the inherent difficulties in conducting high quality de-resuscitation research. Overall risk distribution: Twelve studies (92.3%) were assessed as having “Some Concerns,” while one study was classified as “High Risk.” The absence of low-risk studies likely reflects the challenges of blinding fluid management interventions and the complexity of implementing standardized protocols in diverse ICU settings. Domain-specific findings: All 13 studies (100%) showed either “Some Concerns” or “High Risk” for the domain of “Deviations from Intended Interventions” (Domain 2). The strongest domain was “Outcome Measurement” (Domain 4), with 12 studies (92.3%) achieving “Low Risk” assessment. Domain 1 showed generally good performance, with 11 studies (84.6%) achieving “Low Risk” assessment, suggesting that most studies employed appropriate randomization methods and allocation concealment. Discussion Compared to fluid resuscitation, fluid de-resuscitation lags in terms of clinical and research guidelines (9,38,39). This systematic review of 13 randomized controlled trials (2,402 participants) identified several patterns relevant to de-resuscitation research methodology: Ultrasound adoption: Only 2 of 13 studies (15.4%) explicitly used ultrasound parameters to guide de-resuscitation interventions, representing limited, albeit increasing, adoption. When including all ultrasound assessment (any form), 4 of 13 studies (30.8%) performed ultrasound evaluation. Ultrasound adoption: Only 2 of 13 studies (15.4%) explicitly used ultrasound parameters to guide de-resuscitation interventions, representing limited, albeit increasing, adoption. When including all ultrasound assessment (any form), 4 of 13 studies (30.8%) performed ultrasound evaluation. Fluid balance achievement: Despite heterogeneous approaches to de-resuscitation, 7-8 of 13 studies (53.8-61.5%) successfully achieved statistically significant fluid balance separation between intervention and control groups, indicating that protocol-driven de-resuscitation is feasible and implementable. Fluid balance achievement: Despite heterogeneous approaches to de-resuscitation, 7-8 of 13 studies (53.8-61.5%) successfully achieved statistically significant fluid balance separation between intervention and control groups, indicating that protocol-driven de-resuscitation is feasible and implementable. Clinical outcomes disconnect: Despite achieving negative fluid balance targets in >50% of trials, no study demonstrated significant improvements in hard clinical outcomes (mortality, major morbidity). Several studies achieved secondary endpoints including improved ventilator-free days, faster extubation, or reduced organ dysfunction, however these did not translate to survival benefit. Clinical outcomes disconnect: Despite achieving negative fluid balance targets in >50% of trials, no study demonstrated significant improvements in hard clinical outcomes (mortality, major morbidity). Several studies achieved secondary endpoints including improved ventilator-free days, faster extubation, or reduced organ dysfunction, however these did not translate to survival benefit. Parameter heterogeneity: Substantial heterogeneity existed across studies in triggers for de-resuscitation, intervention protocols, fluid responsiveness assessment methods, and tolerance monitoring. This heterogeneity limits comparability across trials but does not appear to predict success or failure of individual interventions. Parameter heterogeneity: Substantial heterogeneity existed across studies in triggers for de-resuscitation, intervention protocols, fluid responsiveness assessment methods, and tolerance monitoring. This heterogeneity limits comparability across trials but does not appear to predict success or failure of individual interventions. A key finding is the limited integration of ultrasound as a guiding tool for active fluid removal. While ultrasound is a powerful, non-invasive modality offering real-time insights into cardiac function, intravascular volume, and organ congestion, its proactive inclusion within de-resuscitation guidance protocols is notably lacking in the reviewed literature. This might be due to several factors, including varying levels of expertise in critical care ultrasound, the absence of clear algorithms for ultrasound-guided de-resuscitation, and a historical reliance on static, less dynamic measures of fluid status. Ultrasound-based assessment offers theoretical advantages for de-resuscitation by providing: - Dynamic fluid responsiveness assessment: Passive leg raise testing or respiratory variation in IVC diameter can identify patients who will respond to further fluid removal, potentially preventing futile or harmful interventions in non-responsive patients. - Dynamic fluid responsiveness assessment: Passive leg raise testing or respiratory variation in IVC diameter can identify patients who will respond to further fluid removal, potentially preventing futile or harmful interventions in non-responsive patients. - Direct visualization of congestion: Lung ultrasound B-line assessment, hepatic vein distension, and VExUS grading can provide organ-specific evidence of fluid overload, potentially enabling more personalized de-resuscitation targeting. - Direct visualization of congestion: Lung ultrasound B-line assessment, hepatic vein distension, and VExUS grading can provide organ-specific evidence of fluid overload, potentially enabling more personalized de-resuscitation targeting. - Serial monitoring capability: Real-time ultrasound reassessment during de-resuscitation can guide adjustments to the intensity of intervention based on actual physiological response rather than fixed protocols. - Serial monitoring capability: Real-time ultrasound reassessment during de-resuscitation can guide adjustments to the intensity of intervention based on actual physiological response rather than fixed protocols. However, current evidence and guideline provide limited support for these theoretical advantages: Only two studies explicitly tested ultrasound-guided de-resuscitation in this review. Chen & Kollef (36) employed comprehensive ultrasound assessment (IVC, PPV, stroke volume) but failed to achieve fluid balance separation, potentially due to advisory (rather than mandated) recommendations. Cinotti et al. (32) achieved fluid balance separation but used ultrasound primarily to guide resuscitation rather than de-resuscitation specifically. Given the limited empirical evidence, recommendations for ultrasound-guided de-resuscitation protocols should be considered preliminary and hypothesis-generating rather than definitive. Prospective, well-designed RCTs specifically testing standardized ultrasound-guided de-resuscitation protocols are needed before firm recommendations can be made. Standardization in de-resuscitation research Our analysis reveals substantial lack of standardization (de-standardization) across multiple domains of de-resuscitation research. This heterogeneity manifests in three critical areas: - Fluid responsiveness assessment: 84.6% of studies (11/13) did not assess fluid responsiveness at all before or during fluid removal. Only 2 studies used dynamic responsiveness testing. This represents a critical gap given established evidence supporting fluid responsiveness-guided fluid administration. - Fluid responsiveness assessment: 84.6% of studies (11/13) did not assess fluid responsiveness at all before or during fluid removal. Only 2 studies used dynamic responsiveness testing. This represents a critical gap given established evidence supporting fluid responsiveness-guided fluid administration. - Tolerance monitoring: While most studies employed operational tolerance protocols, these were universally study-specific and locally developed. No studies employed standardized, validated tolerance assessment scoring systems (e.g., SOFA-based organ dysfunction assessment, VExUS grading for congestion assessment). This limits reproducibility and comparability. - Tolerance monitoring: While most studies employed operational tolerance protocols, these were universally study-specific and locally developed. No studies employed standardized, validated tolerance assessment scoring systems (e.g., SOFA-based organ dysfunction assessment, VExUS grading for congestion assessment). This limits reproducibility and comparability. - Parameter definitions: There was marked heterogeneity in what constituted ‘fluid overload’ (ranging from 2L positive balance to >5% weight gain to >10% weight gain), triggers for intervention timing, and targets for negative balance (from neutral to −1 mL/kg/h). - Parameter definitions: There was marked heterogeneity in what constituted ‘fluid overload’ (ranging from 2L positive balance to >5% weight gain to >10% weight gain), triggers for intervention timing, and targets for negative balance (from neutral to −1 mL/kg/h). These standardization gaps may contribute to poor trial performance through several mechanisms: - Inadequate patient selection may result in inclusion of patients not suitable for fluid removal - Inadequate patient selection may result in inclusion of patients not suitable for fluid removal - Poor fluid responsiveness assessment may lead to inappropriate interventions in non-responders - Poor fluid responsiveness assessment may lead to inappropriate interventions in non-responders - Heterogeneous tolerance thresholds may cause premature intervention termination when thresholds vary - Heterogeneous tolerance thresholds may cause premature intervention termination when thresholds vary - Lack of standardized protocols limits ability to identify and implement best practices. - Lack of standardized protocols limits ability to identify and implement best practices. Addressing these gaps should be a priority for future de-resuscitation research. Proposed conceptual framework for physiologically guided de-resuscitation To address the current paucity of physiologically guided de-resuscitation protocols and translate the findings of this review into actionable research directions, we propose a hypothesis-generating conceptual framework (Figure 3). POCUS-driven framework for prioritising de-resuscitation over fluid balance using a standardised four-phase algorithm in ultrasound-guided trials. The diagram illustrates: (1) Identification & Differentiation of left-sided (lung ultrasound [LUS] B-lines, mitral/tricuspid inflow, differential cardiac function) vs right-sided congestion (VExUS grading, hepatic/renal vein Doppler); (2) Safety & Fluid Tolerance Metrics integrating 10-15% SV variation monitoring; (3) Fluid Responsiveness assessment (e.g. volume status triggers); and (4) Clinical Success outcomes (renal/hepatic function, lactate clearance). Primary ultrasound triggers include LUS, IVC, cardiac pressures. Hypothesis-generating conceptual tool for future de-resuscitation RCTs produced by NotebookLM using findings of systematic review findings. A common misconception in fluid management is equating fluid responsiveness (likelihood of stroke volume increase with further fluid) with fluid tolerance (absence of harm from congestion or organ dysfunction). Responsiveness testing (e.g. passive leg raise, IVC collapsibility) identifies candidates for resuscitation, but de-resuscitation requires separate assessment of tolerance to fluid removal and downstream congestion. This distinction is critical to avoid inappropriate interventions in non-responders or premature cessation in responsive patients with good tolerance. The framework differentiates left-sided vs right-sided congestion, as pathophysiology, assessment, and de-resuscitation implications differ: - Left-sided congestion (pulmonary oedema, impaired gas exchange): Prioritise lung ultrasound (B-lines), left ventricular filling pressures (E/e’), IVC size/distensibility. Moderate findings support continuation; severe findings warrant acceleration. - Left-sided congestion (pulmonary oedema, impaired gas exchange): Prioritise lung ultrasound (B-lines), left ventricular filling pressures (E/e’), IVC size/distensibility. Moderate findings support continuation; severe findings warrant acceleration. - Right-sided congestion (venous/organ oedema): Use VExUS grading, hepatic/renal vein Doppler, right ventricular strain. - Right-sided congestion (venous/organ oedema): Use VExUS grading, hepatic/renal vein Doppler, right ventricular strain. This algorithm is not prescriptive but serves as a scaffold for future RCTs to standardise and test ultrasound-integrated de-resuscitation, addressing the gaps identified in this review. Serial reassessment (daily or more frequent) allows dynamic adjustment. Implications for future research Based on the findings of this systematic review, future de-resuscitation research should prioritize several key areas. Of primary importance is the clarification of patient selection by determining the specific populations that would most benefit from active de-resuscitation, such as stratifying by underlying condition (e.g., sepsis vs. ARDS vs. general ICU), timing of intervention (early vs. late), and the presence or absence of acute kidney injury (AKI). Then there is an urgent need to develop and validate standardized tolerance assessment protocols to establish reliable scoring systems or parameter combinations that can safely guide continued fluid removal and reduce clinician variability. Trials should be designed to explicitly test fluid responsiveness-guided de-resuscitation, comparing the impact of dynamic responsiveness assessment on outcomes against fixed-target approaches. In addition, for ultrasound-guided de-resuscitation to be successfully pursued, standardized protocols must be established to specify the guiding parameters, assessment frequency, and concrete decision thresholds, moving beyond advisory recommendations. Following on from this, research should examine timing optimization, testing whether earlier or later intervention influences outcomes and whether patient stratification by timing improves overall results. Finally, the focus on outcome selection must shift, with future trials prospectively determining which outcomes are most responsive to de-resuscitation interventions, such as ventilator-free days, ICU-free days, and organ dysfunction markers, rather than solely assuming mortality reduction. Limitations The interpretation of these findings must consider several limitations. First, methodological heterogeneity exists due to diverse patient populations, interventions (diuretics, ultrafiltration, fluid restriction), timing, and outcome measures across the 13 included studies, which limits meta-analytic pooling and the identification of best practices. Also, there are limited ultrasound-guided studies, as only two studies explicitly tested this approach, and one used advisory rather than mandated recommendations (36), limiting robust conclusions. Publication bias is a concern since the review only included published RCTs, potentially excluding unpublished negative trials, ongoing studies, and observational research. Furthermore, none of the studies achieved a low overall risk of bias (Cochrane RoB 2); all showed concerns regarding deviations from intended interventions due to the inherent difficulty of blinding fluid management research. The consistent finding of fluid balance achievement without clinical benefit may stem from inadequate power for hard outcomes, inappropriate outcome selection, or fundamental limitations of the de-resuscitation approach— it has not been possible to distinguish between these. Finally, temporal aspects are limited, as the longest follow-up was 28–90 days, meaning longer-term outcomes and potential delayed benefits or harms were not assessed. Conclusions This systematic review highlights the substantial variability in both the triggers and guiding parameters employed for active fluid removal in critically ill patients. A critical observation is the notable underutilization of ultrasound-based parameters in explicitly guiding de-resuscitation strategies. Despite this, protocol-driven fluid removal interventions effectively achieve negative fluid balances. The limited integration of ultrasound in guiding this crucial phase of fluid management represents a significant gap in current research. Future well-designed RCTs are warranted to rigorously evaluate the impact of ultrasound-guided de-resuscitation on fluid balance, patient outcomes, and the incidence of complications associated with fluid overload. Evidence-based fluid management in the ICU is crucial to improve patient care.