Skip to main content

Foundational Trials

CLOVERS

Image: Shutterstock / ChaNaWiT

Publication

  • Title: Early Restrictive or Liberal Fluid Management for Sepsis-Induced Hypotension
  • Acronym: CLOVERS (Crystalloid Liberal or Vasopressors Early Resuscitation in Sepsis)
  • Year: 2023
  • Journal published in: The New England Journal of Medicine
  • Citation: Shapiro NI, Douglas IS, Brower RG, Brown SM, Exline MC, Ginde AA, et al. Early restrictive or liberal fluid management for sepsis-induced hypotension. N Engl J Med. 2023;388:499-510.

Context & Rationale

  • Background
    • Adult sepsis guidance has historically promoted early intravenous crystalloids (commonly framed as 30 mL/kg within the first hours), despite low/very low-certainty evidence and long-standing concerns about fluid accumulation and downstream organ oedema. 1
    • Clinicians frequently face an early “after the first 1–3 L” decision: continue fluids to augment preload versus initiate vasopressors to restore perfusion pressure and potentially avoid excessive positive fluid balance.
    • Adjacent trial evidence informed parts of this space but left the early hypotension decision point unresolved: ICU-phase fluid restriction in septic shock (CLASSIC) 2 and physiology-guided resuscitation approaches (FRESH; ANDROMEDA-SHOCK). 34
  • Research Question/Hypothesis
    • In adults with sepsis-induced hypotension after initial fluid resuscitation, does an early restrictive fluid strategy (prioritising earlier vasopressors and less additional crystalloid) improve 90-day outcomes compared with an early liberal fluid strategy (prioritising additional crystalloid before vasopressors)?
    • Pragmatic intent: test two implementable resuscitation “bundles” rather than a single physiological variable.
  • Why This Matters
    • Fluid and vasopressor sequencing is a high-frequency, high-variance clinical decision in emergency and early critical care, with plausible competing mechanisms of benefit and harm.
    • Demonstrating non-inferiority/superiority of a vasopressor-prioritised strategy could legitimise earlier vasopressors (including peripheral administration) and reduce compulsory “one-size-fits-all” fluid targets.
    • Conversely, showing benefit of liberal fluids could reinforce continued fluid-first resuscitation after the initial bolus in sepsis-induced hypotension.

Design & Methods

  • Research Question: In adults with sepsis-induced hypotension after receiving 1–3 L of crystalloid, does an early restrictive fluid / vasopressor-prioritised strategy reduce death before discharge home by day 90 compared with an early liberal fluid strategy?
  • Study Type: Multicentre, randomised, open-label, pragmatic superiority trial (60 US centres; emergency department predominant; ICU/ward permitted); allocation stratified by site; protocolised treatment guidance for 24 hours. 5
  • Population:
    • Adults (≥18 years) with suspected/confirmed infection and sepsis-induced hypotension.
    • Required pre-randomisation resuscitation: receipt of 1–3 L intravenous isotonic crystalloid (minimum 1 L; maximum 3 L) before randomisation.
    • Hypotension criterion: systolic blood pressure <100 mmHg or mean arterial pressure <65 mmHg after initial fluids, or already receiving a vasopressor infusion.
    • Randomisation within 4 hours of meeting eligibility criteria; majority enrolled in the emergency department.
    • Key exclusions (high-level): >3 L crystalloid before randomisation, clinician-determined need for immediate alternative management pathway, or clinical states in which protocolised liberal fluids or protocolised fluid restriction would be unsafe (e.g., overt pulmonary oedema / severe volume overload). 5
  • Intervention:
    • Restrictive strategy: prioritised earlier vasopressors (typically norepinephrine) to achieve a target perfusion pressure (MAP target used clinically; protocol guidance centred around restoring MAP ≥65 mmHg) with minimal additional crystalloid during a 24-hour protocol period.
    • Additional fluid boluses were discouraged and permitted as “rescue” only for prespecified indications (e.g., severe hypotension, evidence of hypoperfusion not responsive to vasopressors, or clinician-determined need consistent with protocol guidance).
    • Peripheral vasopressor initiation was allowed and operationally enabled by protocol guidance (with escalation to central access per clinical judgement). 5
  • Comparison:
    • Liberal strategy: prioritised additional isotonic crystalloid boluses early during the 24-hour protocol period, with vasopressors generally reserved for persistent hypotension after additional fluids or if fluids were limited by clinical concerns (e.g., emerging fluid overload).
    • The liberal arm’s early bolus recommendation was amended during the trial (second litre became optional) without evidence of differential effect by epoch in prespecified interaction testing. 6
  • Blinding: Open-label treatment delivery (no blinding of clinicians or participants); primary outcome was objective mortality-based, but some secondary outcomes (e.g., ICU/hospital-free days) could be susceptible to performance bias.
  • Statistics: Planned sample size 2320 to detect a 4.5% absolute reduction in 90-day mortality (from 15.0% to 10.5%) with 90% power at a two-sided 5% significance level; primary analysis by intention-to-treat. 5
  • Follow-Up Period: Primary endpoint assessed through day 90; key secondary outcomes included 28-day organ support–free days; selected physiological/organ dysfunction metrics assessed at 72 hours and complications assessed to day 7.

Key Results

This trial was stopped early. Enrolment stopped after the second interim analysis on the recommendation of the data and safety monitoring board for futility (1563 of the planned 2320 participants were randomised).

Outcome Restrictive strategy Liberal strategy Effect p value / 95% CI Notes
Death before discharge home by day 90 (primary) 109/782 (13.9%) 116/781 (14.9%) Risk difference −0.9 pp 95% CI −4.4 to 2.6; P=0.61 Primary endpoint (mortality with “home discharge” framing)
Death from any cause at any location by day 90 175/782 (22.4%) 171/781 (21.9%) Risk difference 0.5 pp 95% CI −3.6 to 4.7; P not reported Supports interpretation that the primary endpoint framing did not conceal a mortality signal
Ventilator-free days to day 28 Mean 24.5 (95% CI 24.0 to 25.0) Mean 24.7 (95% CI 24.3 to 25.2) Mean difference −0.2 days 95% CI −0.9 to 0.5; P not reported No clinically meaningful separation
Renal-replacement-therapy–free days to day 28 Mean 26.6 (95% CI 26.2 to 27.0) Mean 26.8 (95% CI 26.4 to 27.2) Mean difference −0.2 days 95% CI −0.7 to 0.3; P not reported Consistent with neutral renal signal
Vasopressor-free days to day 28 Mean 23.5 (95% CI 22.9 to 24.1) Mean 24.2 (95% CI 23.7 to 24.8) Mean difference −0.7 days 95% CI −1.5 to 0.1; P not reported Numerically fewer vasopressor-free days with vasopressor-prioritised strategy (expected direction)
ICU-free days to day 28 Mean 21.4 (95% CI 20.7 to 22.0) Mean 22.1 (95% CI 21.5 to 22.8) Mean difference −0.7 days 95% CI −1.6 to 0.2; P not reported Small, imprecise difference
Hospital-free days to day 28 Mean 19.1 (95% CI 18.3 to 19.8) Mean 20.0 (95% CI 19.3 to 20.8) Mean difference −0.9 days 95% CI −1.9 to 0.1; P not reported Directionally favoured liberal strategy, but CI includes no difference
Initiation of mechanical ventilation by day 28 85/782 (10.9%) 68/781 (8.7%) Risk difference 2.2 pp 95% CI −0.7 to 5.1; P not reported Imprecise; does not support a clear harm signal
Initiation of renal-replacement therapy by day 28 30/782 (3.8%) 25/781 (3.2%) Risk difference 0.6 pp 95% CI −1.2 to 2.4; P not reported Low event rate
ARDS within 7 days 11/782 (1.4%) 9/781 (1.2%) Risk difference 0.3 pp 95% CI −0.8 to 1.3; P not reported Very low event rate
Serious adverse events (number of events) 21 events 19 events Difference 2 events 95% CI −10 to 14; P not reported Specific SAE: fluid overload/pulmonary oedema 0 vs 3 events
  • The intervention achieved substantial treatment separation (fluid volumes and vasopressor timing), yet outcomes were tightly clustered around the null across mortality and organ support endpoints.
  • The 90-day mortality point estimates were near-identical, and the confidence interval for the primary endpoint excluded large absolute benefit for either strategy (but did not exclude smaller clinically relevant effects).
  • Signals of increased invasiveness (e.g., more vasopressor exposure and line use) accompanied the restrictive strategy, without offsetting mortality benefit.

Internal Validity

  • Randomisation and allocation: Centralised randomisation stratified by institution; allocation concealment before assignment was feasible; treatment delivery was unblinded by design.
  • Dropout / exclusions: 1563 randomised (782 restrictive; 781 liberal); primary endpoint status was largely complete; censoring for the primary endpoint was rare (0.6% vs 0.5% per supplementary endpoint accounting). 6
  • Performance/detection bias: Open-label delivery could influence co-interventions and discharge decisions; primary endpoint was mortality-based (objective), but ICU/hospital-free days and admission decisions could be affected by clinician behaviour.
  • Protocol adherence: Assigned protocol guidance was reported as used in 761/782 (97.3%) restrictive vs 750/781 (96.0%) liberal; separation was therefore not driven by widespread non-adherence. 6
  • Baseline comparability: Groups were highly similar at baseline (e.g., mean age ~62 years; median pre-randomisation crystalloid 2050 mL in both arms; mean lactate 2.9 mmol/L in both arms; ED enrolment 93.2% vs 90.7%).
  • Heterogeneity: 60-site pragmatic design increases clinical heterogeneity; subgroup analyses did not identify compelling, consistent effect modification (CIs for many subgroups were wide, reflecting limited power after early stopping).
  • Timing: Randomisation was early relative to meeting criteria (median 61 minutes in both arms), consistent with the intervention’s hypothesised window of effect.
  • Dose / separation of variable of interest:
    • Crystalloid in first 6 hours after randomisation: median 500 mL (IQR 0–1000) restrictive vs 2300 mL (IQR 1600–3000) liberal.
    • Total crystalloid from pre-enrolment through 24 hours: median 3300 mL (IQR 2550–4350) restrictive vs 5400 mL (IQR 4400–6575) liberal; difference of medians −2134 mL (95% CI −2318 to −1949).
    • Vasopressor use: 59% restrictive vs 37% liberal; median time to vasopressor initiation 2.4 hours vs 3.9 hours (difference −1.4 hours; 95% CI −1.8 to −1.1); mean duration 14.4 hours vs 10.2 hours (difference 4.2 hours; 95% CI 2.5 to 5.9).
    • Central venous catheter insertion between randomisation and 72 hours: 42.7% vs 33.6% (difference 9.1 pp; 95% CI 4.4 to 13.8); peripheral vasopressor infusion: 36.9% vs 27.4% (difference 9.5 pp; 95% CI 5.0 to 14.0). 6
  • Crossover: Not reported as a discrete crossover metric; both protocols permitted “rescue” actions, so some bidirectional contamination is expected in a strategy trial, but observed separation suggests limited dilution of the intended contrast.
  • Outcome assessment: Mortality and organ-support endpoints were prespecified and largely objective; the “discharge home” framing introduces a health-system and social component, mitigated by reporting of all-cause mortality at any location by day 90.
  • Statistical rigour: Stopping early for futility reduced attained power versus the planned design, widening uncertainty for smaller absolute effects; analyses remained intention-to-treat with prespecified endpoints.

Conclusion on Internal Validity: Overall, internal validity appears moderate-to-strong: randomisation and protocol fidelity were robust with clear treatment separation, and the primary endpoint was objective; the principal threats are open-label performance effects on secondary endpoints and reduced precision from early stopping.

External Validity

  • Population representativeness: Enrolled a broad emergency-care sepsis-induced hypotension population after initial crystalloid (median 2050 mL pre-randomisation) with moderate severity (mean lactate 2.9 mmol/L), rather than a narrowly defined, ICU-only refractory septic shock cohort.
  • Important exclusions: Patients who had already received >3 L crystalloid and those with overt fluid overload physiology were excluded, limiting direct applicability to very late presenters and to patients in whom clinicians have already committed to large-volume resuscitation.
  • Applicability across systems: Conducted in 60 US centres with capability for early vasopressors and close monitoring; generalisability may be reduced in low-resource settings where vasopressor delivery (especially peripherally) and monitoring infrastructure differ substantially.
  • Clinical translation: High relevance to contemporary ED/ICU interfaces where clinicians decide between additional fluid and earlier vasopressors after the first bolus.

Conclusion on External Validity: Findings are highly generalisable to high-income emergency and early critical care settings managing sepsis-induced hypotension after initial crystalloid, but less directly applicable to extreme shock phenotypes, late resuscitation after very large fluid volumes, and resource-limited contexts.

Strengths & Limitations

  • Strengths: Large, pragmatic multicentre design; early randomisation in the clinically relevant decision window; substantial achieved treatment separation; high protocol adherence; patient-important 90-day endpoint with complementary all-location mortality reporting; detailed supplementary reporting on protocol epochs and vascular access consequences.
  • Limitations: Open-label strategy trial (risk of performance effects on ICU/hospital utilisation outcomes); early stopping for futility reduced precision for modest effect sizes; intervention bundled multiple linked decisions (fluid volume, vasopressor timing, line strategy), limiting mechanistic attribution; protocol amendment in the liberal arm reflects real-world practice but complicates a “single fixed-dose” interpretation.

Interpretation & Why It Matters

  • Clinical practice
    An early vasopressor-prioritised, fluid-restrictive approach after the first 1–3 L is a defensible alternative to continued liberal fluids in sepsis-induced hypotension, as neither strategy improved 90-day mortality; choice should be individualised to phenotype (fluid responsiveness, cardiac/renal reserve, pulmonary oedema risk, and perfusion markers).
  • What it does not test
    CLOVERS does not adjudicate the value of giving the first litre(s) of fluid in suspected sepsis, nor does it directly test a strict 30 mL/kg mandate versus a much smaller initial bolus; it tests what to do after initial crystalloid when hypotension persists.
  • Systems implications
    Restrictive strategy implementation increased vasopressor exposure and invasive line use (central and peripheral), highlighting that “less fluid” may shift rather than eliminate resuscitation-related burdens (monitoring intensity, ICU admission thresholds, and procedural risk).

Controversies & Subsequent Evidence

  • Endpoint framing (“death before discharge home”): The primary endpoint embeds discharge destination; while patient-centred, it is partially contingent on health-system and social factors; the near-identical all-location 90-day mortality supports a genuinely neutral mortality result rather than an artefact of discharge patterns.
  • Early stopping for futility: Stopping early improved efficiency but reduced precision for smaller effects; the primary outcome CI (−4.4 to 2.6 percentage points) leaves residual uncertainty about modest benefit/harm.
  • Liberal strategy amendment: The liberal arm’s early bolus recommendation was amended (second litre optional), yet an epoch-by-treatment interaction test did not indicate differential mortality effect by epoch (interaction P=0.855), supporting stability of the neutral conclusion across implementation eras. 6
  • Relationship to other RCTs: CLASSIC, conducted later (ICU phase) in septic shock, similarly did not show a mortality benefit from a restrictive fluid strategy, reinforcing that “less fluid” is not automatically superior across phases of sepsis care. 2
  • Guideline evolution: Contemporary guidance increasingly emphasises individualised fluid resuscitation, dynamic assessment, and early vasopressors when appropriate; recent ESICM recommendations address both fluid choice and fluid volume as separable decisions that should be tailored to patient context. 78
  • Meta-analytic context: Systematic reviews incorporating modern trials highlight heterogeneity by setting and economic context and do not demonstrate a consistent, large mortality advantage of higher versus lower resuscitation volumes; separate evidence supports the importance of vasopressor timing as a potentially modifiable early factor, albeit with residual confounding and varying definitions of “early”. 910

Summary

  • CLOVERS compared two early resuscitation strategies after 1–3 L crystalloid in sepsis-induced hypotension: a restrictive fluid/earlier vasopressor strategy versus a liberal fluid strategy.
  • The trial was stopped early for futility (1563 participants), with no difference in death before discharge home by day 90 (13.9% vs 14.9%; difference −0.9 pp; 95% CI −4.4 to 2.6; P=0.61).
  • All-location 90-day mortality and major organ support outcomes were similarly neutral, with small, imprecise differences.
  • Treatment separation was substantial (6-hour fluids 500 mL vs 2300 mL; 24-hour cumulative fluids 3300 mL vs 5400 mL; vasopressor use 59% vs 37% with earlier initiation), supporting that the null result is not explained by lack of separation.
  • The restrictive strategy shifted resuscitation burden towards earlier/more vasopressor exposure and greater line use, without demonstrated mortality benefit.

Further Reading

Other Trials

Systematic Review & Meta Analysis

Observational Studies

Guidelines

Notes

  • CLOVERS primarily informs the post–initial bolus decision (additional litres versus earlier vasopressors) rather than the value of the very first litre of crystalloid in suspected sepsis.

Overall Takeaway

CLOVERS is “landmark” because it directly tested, at scale and early in the emergency care pathway, two competing real-world resuscitation strategies after initial crystalloid in sepsis-induced hypotension. It demonstrated substantial physiological and process separation (less fluid and earlier/more vasopressors) without mortality benefit, thereby legitimising individualised fluid–vasopressor sequencing rather than enforcing a single default resuscitation pathway.

Overall Summary

  • After 1–3 L crystalloid in sepsis-induced hypotension, restrictive (earlier vasopressors) and liberal (more crystalloid) strategies produced similar 90-day outcomes.
  • Restriction reduced fluid exposure by ~2.1 L over 24 hours but increased vasopressor use and invasive access without a mortality trade-off.
  • The trial supports phenotype-driven resuscitation rather than universal “fluid-first” or “vasopressor-first” dogma.

Bibliography