Skip to main content

Foundational Trials

CLASSIC

Image: Shutterstock / ChaNaWiT

Publication

  • Title: Restriction of Intravenous Fluid in ICU Patients with Septic Shock
  • Acronym: CLASSIC
  • Year: 2022
  • Journal published in: The New England Journal of Medicine
  • Citation: Meyhoff TS, Hjortrup PB, Wetterslev J, et al. Restriction of Intravenous Fluid in ICU Patients with Septic Shock. N Engl J Med. 2022;386(26):2459-2470.

Context & Rationale

  • Background
    Intravenous crystalloids are foundational in septic shock resuscitation, yet observational data consistently associate higher cumulative fluid exposure and fluid overload with worse outcomes.
    International guidance has promoted early fixed-volume fluid resuscitation (e.g., 30 mL/kg), but acknowledges low-certainty evidence for volume targets and emphasises ongoing reassessment of fluid responsiveness and tolerance1.
    Prior to CLASSIC, high-quality randomised evidence specifically testing a restrictive versus usual-care intravenous fluid strategy in established ICU septic shock (after initial resuscitation) was limited.
  • Research Question/Hypothesis
    In adult ICU patients with septic shock who had already received initial resuscitation (≥1 L IV fluid), does a protocolised restrictive IV fluid strategy (compared with standard IV fluid therapy) reduce 90-day all-cause mortality?
  • Why This Matters
    Septic shock mortality remains substantial, and fluid therapy is ubiquitous, modifiable, and plausibly causal in downstream organ dysfunction (lung oedema, venous congestion, renal injury).
    If lower IV fluid volumes are non-inferior (or superior), practice could shift towards earlier de-escalation of fluids and greater reliance on vasopressors/adjuncts after initial resuscitation.

Design & Methods

  • Research Question: Among adult ICU patients with septic shock, does IV fluid restriction during ICU stay (vs standard IV fluid therapy) reduce death by day 90?
  • Study Type: Investigator-initiated, international, multicentre, randomised, parallel-group, open-label trial with concealed allocation and prespecified analyses (including stratification and multiplicity control).
  • Population:
    • Setting: 31 ICUs (predominantly Scandinavia; also Switzerland, Italy, Czech Republic, United Kingdom, Belgium).
    • Key inclusion: Adults (≥18 years) with septic shock; received ≥1 L IV fluid within 24 hours; onset of shock within 12 hours before screening; planned ICU admission or already in ICU.
    • Key exclusions: Conditions mandating specific fluid strategies or with expected large obligatory fluid loads (e.g., major haemorrhage/burns or other clinician-determined contra-indications to the trial strategy); additional exclusions were protocol-defined.
  • Intervention:
    • Strategy: Restrictive IV fluid therapy during ICU stay (max 90 days) aiming to avoid IV crystalloid unless clinically indicated.
    • Permitted indications for IV fluids: IV fluids permitted (including bolus/continuous) if ≥1 predefined criterion was met on that day: plasma lactate ≥4 mmol/L; MAP <50 mmHg; mottling score >2; urine output <0.1 mL/kg/h for ≥2 consecutive hours.
    • Protocol focus: Restriction targeted resuscitation-type fluid administration; unavoidable fluids with medications/nutrition still contributed to total IV exposure but were not the primary lever of the intervention.
  • Comparison:
    • Strategy: Standard IV fluid therapy (no protocol upper limits), reflecting usual care at participating sites.
    • Permitted indications: IV fluids could be given for any clinical indication (including haemodynamic optimisation), as judged by the treating team; protocol violations in this arm were defined as “no IV fluids given on any single day in ICU”.
  • Blinding: Open-label (clinicians and bedside staff not blinded); outcomes were largely objective (mortality; life-support use) and analyses were prespecified with stratification adjustment and multiplicity control.
  • Statistics: A total of 1554 patients were required to detect a 7-percentage-point absolute mortality reduction (from 45% to 38%) with 80% power (β=0.20) at a two-sided 5% significance level (α=0.05); primary analysis by intention-to-treat using logistic regression adjusted for stratification variables; secondary outcomes assessed with multiplicity control (99% CIs and P<0.01 threshold for statistical significance).
  • Follow-Up Period: Primary follow-up to day 90 for mortality and key patient-centred outcomes; longer-term outcomes were prespecified and later reported separately.

Key Results

This trial was not stopped early. The trial completed enrolment to the prespecified sample size (n=1554) with prespecified interim monitoring rules.

Outcome Restrictive-fluid Standard-fluid Effect p value / 95% CI Notes
Death by day 90 323/764 (42.3%) 329/781 (42.1%) Adjusted RR 1.00 95% CI 0.89 to 1.13; P=0.96 Adjusted absolute difference 0.1 percentage points; 95% CI −4.7 to 4.9.
Serious adverse events (composite, ICU) 221/751 (29.4%) 238/772 (30.8%) Adjusted RR 0.95 99% CI 0.77 to 1.15; P=0.46 Adjusted absolute difference −1.7 percentage points; 99% CI −7.7 to 4.3.
Serious adverse reactions (after IV crystalloids) 31/755 (4.1%) 32/776 (4.1%) Adjusted RR 0.99 99% CI 0.50 to 1.93; P=0.95 Adjusted absolute difference −0.1 percentage points; 99% CI −2.8 to 2.6.
Days alive without life support (to day 90) Median 77 (IQR 1 to 87) Median 77 (IQR 1 to 87) Adjusted difference 0 days 95% CI −11 to 11; P=0.84 Analysed with van Elteren test (site-adjusted) due to distributional assumptions.
Days alive and out of hospital (to day 90) Median 21 (IQR 0 to 69) Median 33 (IQR 0 to 70) Adjusted difference −12 days 95% CI −30 to 6; P=0.84 Interpreted cautiously given skew and competing risks (death/hospitalisation).
Cumulative IV fluids in ICU (excluding fluids with medication/nutrition; during 90-day trial period) Median 1798 mL Median 3811 mL Difference −2013 mL Not reported Protocol achieved separation in resuscitation-type IV fluids; total fluid exposure includes non-protocolised sources.
  • Mortality was essentially identical: 42.3% vs 42.1% (adjusted RR 1.00; 95% CI 0.89 to 1.13; P=0.96), excluding very large benefit or harm.
  • Safety endpoints were similar overall (SAEs 29.4% vs 30.8%; SARs 4.1% vs 4.1%), with no statistically significant differences using multiplicity-adjusted inference.
  • Protocol separation was clinically meaningful but not extreme: ICU IV fluids excluding medication/nutrition 1798 mL vs 3811 mL (difference −2013 mL), with median relative fluid balance 1645 mL vs 2368 mL.

Internal Validity

  • Randomisation and allocation concealment: Central allocation with stratification by site and haematologic malignancy/metastatic cancer; allocation concealment reduces selection bias.
  • Attrition / missingness: 90-day vital status obtained in 1545/1554 (99.4%); primary outcome missing in 6 restrictive and 3 standard patients; all secondary outcome data missing for 23 patients (15 restrictive; 8 standard).
  • Performance / detection bias: Open-label delivery risks co-intervention and clinician behaviour effects; however, primary outcome (mortality) and key secondary outcomes (life-support days) are relatively objective.
  • Protocol adherence and discontinuation: Protocol discontinued in ICU for 80/770 (10.4%) restrictive vs 51/784 (6.5%) standard; protocol violations occurred in 162/755 (21.5%) vs 101/776 (13.0%).
  • Baseline comparability: Groups were generally well balanced; examples—age median 70 vs 70 years; SOFA median 10 vs 10; lactate median 3.8 vs 3.9 mmol/L; respiratory support within prior 24h 52.6% vs 48.6%; IV fluid in 24h pre-randomisation median 3200 mL vs 3000 mL.
  • Timing: Enrolment required shock onset within 12 hours and ≥1 L prior IV fluid; intervention therefore tests “post-initial-resuscitation” ICU fluid strategy rather than very early ED resuscitation.
  • Dose / separation of the variable of interest: Median cumulative ICU IV fluids excluding medication/nutrition 1798 mL vs 3811 mL (difference −2013 mL); cumulative all fluids in ICU 10,433 mL vs 12,747 mL (difference −2314 mL); relative fluid balance 1645 mL vs 2368 mL (difference −723 mL).
  • Heterogeneity and subgroup inference: Prespecified subgroup analyses used a stricter threshold (P<0.01); respiratory-support subgroup heterogeneity P=0.03 did not meet this criterion and should be considered hypothesis-generating.
  • Outcome assessment and statistical rigour: Prespecified intention-to-treat primary analysis adjusted for stratification; multiplicity handling for secondary outcomes (99% CIs; P<0.01) strengthens inferential discipline; no imputation because missingness <5% for all outcomes.

Conclusion on Internal Validity: Strong overall: robust randomisation/allocation procedures and near-complete follow-up support unbiased estimation, though open-label delivery and notable protocol violations/discontinuations plausibly attenuated separation and could bias effects towards the null.

External Validity

  • Population representativeness: Adult ICU septic shock population with high severity (SOFA ~10) and high 90-day mortality (~42%) typical of contemporary European ICU septic shock cohorts; authors noted trial participants were broadly representative of participating ICUs, except fewer pulmonary infections.
  • Practice context: High-resource ICUs with established vasopressor, ventilation, and renal replacement therapy capabilities; findings most applicable where close haemodynamic monitoring and early vasopressor escalation are feasible.
  • Intervention context: Strategy implemented after initial fluid resuscitation (≥1 L and within 12 hours of shock onset) and primarily within ICU; generalisability to earlier ED/prehospital fluid decisions is limited.
  • Fluid type and co-interventions: Predominant use of crystalloids; results inform “how much” IV fluid rather than “which fluid”, and should be interpreted alongside fluid-type evidence and local practice.

Conclusion on External Validity: Good for adult ICU septic shock in high-resource settings after initial resuscitation; limited for early sepsis management in the ED/prehospital phase and for settings where fluid restriction may be constrained by limited vasoactive support or monitoring.

Strengths & Limitations

  • Strengths: Large, international, pragmatic ICU trial; objective primary endpoint with near-complete ascertainment; prespecified protocol and statistical approach; clinically meaningful separation in IV fluid exposure; rigorous multiplicity handling for secondary outcomes.
  • Limitations: Open-label delivery with potential co-intervention and contamination; protocol violations/discontinuations were common enough to dilute effects; intervention tested “late restrictive” ICU strategy rather than initial resuscitation; net fluid balance separation was modest relative to total fluid exposure, potentially limiting biological effect size.

Interpretation & Why It Matters

  • Clinical implication
    In established ICU septic shock after initial resuscitation, a restrictive IV fluid strategy (using explicit hypoperfusion triggers) achieved lower IV fluid volumes without increasing mortality or major safety events at 90 days, supporting safety of “less fluid” approaches in appropriately monitored ICUs.
  • What it does not show
    CLASSIC does not resolve optimal early ED fluid bolus volume or the timing of vasopressors versus fluids; it primarily informs ongoing fluid administration (and avoidance) after initial resuscitation once ICU-level support is in place.
  • Methodological inference
    The narrow primary-outcome CI suggests large mortality effects (benefit or harm) are unlikely under these conditions, but moderate clinically important effects remain compatible with the data, particularly given adherence and separation constraints.

Controversies & Subsequent Evidence

  • What exactly was “restricted”? The intervention meaningfully reduced ICU IV fluids excluding medication/nutrition (median 1798 mL vs 3811 mL) and net balance (1645 mL vs 2368 mL), but total fluid exposure remained substantial; debate persists as to whether the achieved separation is sufficient to alter hard outcomes in an already-resuscitated population.
  • Open-label delivery and adherence: Protocol violations (21.5% vs 13.0%) and protocol discontinuations (10.4% vs 6.5%) plausibly biased effects towards null by reducing separation and introducing clinician-driven cross-strategy behaviour.
  • Power and effect-size assumptions: The published protocol/statistical plan specified an effect size (7% absolute mortality reduction) that may have been optimistic for a “post-initial-resuscitation” intervention, raising the possibility that clinically meaningful but smaller benefits (or harms) would be missed2.
  • Subgroup signals: Respiratory-support subgroup heterogeneity (P=0.03) did not meet the prespecified threshold (P<0.01) and should not be used to guide practice without corroboration.
  • Published correspondence: Correspondence and author reply were published in NEJM; no abstract is available, but their existence reflects active debate on interpretation and generalisability3.
  • Longer-term outcomes: Prespecified 1-year follow-up reported similar survival, health-related quality of life, and cognitive outcomes between strategies, while noting that clinically important differences could not be fully excluded4.
  • Bayesian re-analyses: Prespecified Bayesian analyses of CLASSIC provided probabilistic estimates and suggested substantial residual uncertainty for clinically important effects on mortality and some secondary endpoints, while making large effects on serious adverse reactions unlikely5.
  • Related randomised evidence: CLOVERS (ED sepsis-induced hypotension) also found no mortality difference between early restrictive versus liberal fluid strategies, reinforcing that “more fluid” is not clearly beneficial across phases of sepsis care, while highlighting different timing/setting compared with CLASSIC6.
  • Synthesis and guidelines: Meta-analytic work suggests the balance of benefit/harm of higher-volume fluid strategies may vary by setting/resources (e.g., greater harm signals in lower-resource contexts), cautioning against naïve generalisation from high-income ICU trials7.
    Recent ESICM guideline efforts have formalised evidence-graded recommendations for fluid choice and resuscitation volume across critical illness, within which CLASSIC informs the evidence base for “how much fluid” decisions after initial resuscitation89.

Summary

  • CLASSIC randomised 1554 ICU patients with septic shock (after ≥1 L initial IV fluid) to restrictive vs standard IV fluid therapy during ICU stay.
  • 90-day mortality was identical (42.3% vs 42.1%; adjusted RR 1.00; 95% CI 0.89 to 1.13; P=0.96).
  • Restrictive strategy reduced ICU IV fluids excluding medication/nutrition (1798 mL vs 3811 mL) without increasing serious adverse events or serious adverse reactions.
  • Patient-centred secondary outcomes (days alive without life support; days alive and out of hospital) were similar between groups.
  • Open-label delivery and non-trivial protocol violations/discontinuations likely diluted separation and limit conclusions about smaller effect sizes.

Further Reading

Other Trials

Systematic Review & Meta Analysis

Observational Studies

Guidelines

Notes

  • CLASSIC addresses ongoing IV fluid administration after initial resuscitation in ICU septic shock; it should not be over-interpreted as evidence against early fluid bolus resuscitation in undifferentiated shock.
  • When applying “restrictive” strategies, clinicians must ensure timely vasopressor escalation, source control, and careful reassessment of perfusion markers rather than substituting fluid avoidance for active resuscitation.

Overall Takeaway

CLASSIC is a landmark ICU septic shock trial because it rigorously tested a pragmatic, protocolised restriction of IV fluids after initial resuscitation and demonstrated similar 90-day mortality and safety compared with standard practice, while achieving clear separation in fluid exposure. It shifted the evidentiary baseline from “more fluid is presumed beneficial” to “less fluid appears safe in monitored ICU septic shock”, and helped catalyse subsequent trials and guideline efforts to define phase-specific, context-sensitive fluid strategies.

Overall Summary

  • Restrictive IV fluid strategy in ICU septic shock reduced IV fluid exposure but did not change 90-day mortality (42.3% vs 42.1%).
  • No clear safety penalty: SAEs and SARs were similar between groups.
  • Best interpreted as evidence supporting post-resuscitation fluid stewardship rather than as a test of initial ED resuscitation volumes.

Bibliography