Skip to main content

Foundational Trials

ALBIOS

Image: Shutterstock / ChaNaWiT

Publication

  • Title: Albumin replacement in patients with severe sepsis or septic shock
  • Acronym: ALBIOS (Albumin Italian Outcome Sepsis)
  • Year: 2014
  • Journal published in: New England Journal of Medicine
  • Citation: Caironi P, Tognoni G, Masson S, Fumagalli R, Pesenti A, Romero M, et al. Albumin replacement in patients with severe sepsis or septic shock. N Engl J Med. 2014;370(15):1412-1421.

Context & Rationale

  • Background
    Colloids (including human albumin) were widely used in sepsis for presumed intravascular volume expansion, yet clinically important outcome benefit over crystalloids remained uncertain.
    Observational and mechanistic work suggested potential non-oncotic effects of albumin (endothelial/glycocalyx interactions, binding/transport, antioxidant capacity), but translation to patient-centred benefit was unproven.
    The SAFE trial (ICU-wide) found no overall mortality difference between albumin and saline; its pre-specified severe sepsis subgroup suggested possible benefit, but this was hypothesis-generating rather than definitive.
  • Research Question/Hypothesis
    In adults with severe sepsis (including septic shock), does adding 20% albumin (with a protocolised target serum albumin ≥30 g/L) to crystalloids reduce 28-day mortality compared with crystalloids alone?
  • Why This Matters
    Albumin is costly and resource-limited; a reliable mortality or organ-failure signal would justify targeted use, while null results would support crystalloids as default and constrain colloid practice.
    The trial also tested a conceptually distinct strategy: sustained correction of hypoalbuminaemia (hyperoncotic 20% albumin) rather than a single “resuscitation fluid” substitution.

Design & Methods

  • Research Question: In adults with severe sepsis/septic shock, does protocolised 20% albumin replacement to maintain serum albumin ≥30 g/L (in addition to crystalloids) reduce 28-day mortality versus crystalloids alone?
  • Study Type: Multicentre, randomised, open-label, controlled trial in 100 ICUs (Italy); central computer-generated allocation; stratified by ICU and by time from severe sepsis criteria to randomisation (<6 hours vs 6–24 hours).
  • Population:
    • Adults (≥18 years) admitted to ICU with severe sepsis (Sepsis-1 era) within the prior 24 hours, including patients with septic shock; enrolment and protocol procedures per trial protocol1.
    • Key exclusions included clinical conditions mandating albumin, contraindication/allergy, and other protocol-specified exclusions (e.g., consent constraints), as detailed in the protocol1.
  • Intervention:
    • 20% albumin plus crystalloids, with albumin dosing guided by daily serum albumin to target ≥30 g/L for up to 28 days or ICU discharge (whichever occurred first), alongside haemodynamic resuscitation using standard targets; protocolised dosing thresholds are specified in the trial protocol1.
  • Comparison:
    • Crystalloids alone for volume replacement, using the same haemodynamic targets; albumin and other colloids were discouraged except under defined exceptions/violations documented in trial materials12.
  • Blinding: Open-label (no blinding of treating clinicians); mortality outcomes are objective but co-intervention and performance bias are plausible.
  • Statistics: A total of 1350 patients were planned to detect a 17% relative reduction in 28-day mortality (from 45% to 37.4%) with 80% power at the 5% significance level; an interim assessment allowed sample-size re-estimation up to ~1800 if needed; primary analyses were intention-to-treat1.
  • Follow-Up Period: 28 days (primary) and 90 days (key secondary), with additional in-ICU and in-hospital outcomes.

Key Results

This trial was not stopped early. A planned interim analysis occurred after enrolment of 700 patients; recruitment continued to 1818 randomised participants.

Outcome 20% albumin + crystalloids Crystalloids Effect p value / 95% CI Notes
Death at 28 days (primary) 285/895 (31.8%) 288/900 (32.0%) RR 1.00 95% CI 0.87 to 1.14; P=0.94 Primary endpoint; intention-to-treat denominators as reported.
Death at 90 days 365/888 (41.1%) 389/893 (43.6%) RR 0.94 95% CI 0.85 to 1.05; P=0.29 No overall mortality benefit.
Death at 90 days in septic shock at enrolment (subgroup) 243/558 (43.6%) 281/563 (49.9%) RR 0.87 95% CI 0.77 to 0.99; P=0.03 Heterogeneity (shock vs no shock) P=0.03; adjusted analyses in supplement attenuated with fuller covariate adjustment (RR 0.88; 95% CI 0.77 to 1.01; P=0.07)2.
Death at 90 days without septic shock at enrolment (subgroup) 122/330 (37.0%) 108/330 (32.7%) RR 1.13 95% CI 0.92 to 1.39; P=0.25 Directionally opposite to septic shock subgroup; interaction caution warranted.
SOFA score (days 1–7; median) 6.00 (IQR 4.00–8.50) 5.62 (IQR 3.92–8.28) Not reported P=0.23 No signal for global organ-dysfunction reduction.
Cardiovascular SOFA subscore (median) 1.20 (IQR 0.46–2.31) 1.42 (IQR 0.60–2.50) Not reported P=0.03 Statistical difference; clinical magnitude modest.
Time to suspension of vasopressor/inotropic agents (days; median) 3 (IQR 1–6) 4 (IQR 2–7) Not reported P=0.007 Tertiary outcome; susceptible to practice variation in an open-label design.
Renal-replacement therapy 222/903 (24.6%) 194/907 (21.4%) Not reported P=0.11 No statistically significant difference; direction towards more RRT with albumin.
Acute kidney injury 183/834 (21.9%) 190/837 (22.7%) Not reported P=0.71 No difference in AKI incidence as reported.
Serum albumin at day 1 (g/L; mean ± SD) 28.6 ± 5.4 24.0 ± 5.2 Not reported P<0.001 Demonstrates biological separation (supplement)2.
Mean arterial pressure at 6 hours (mmHg; mean ± SD) 79 ± 14 77 ± 13 Not reported P<0.001 Physiological effect (supplement)2.
Net fluid balance at day 2 (mL; median [IQR]) 350 [−705 to 1607] 620 [−480 to 1945] Not reported P=0.002 Daily net balance differences were statistically significant on selected days; total 7-day net balance did not differ (P=0.13) (supplement)2.
  • Protocolised albumin replacement increased serum albumin early (day 1: 28.6 ± 5.4 g/L vs 24.0 ± 5.2 g/L; P<0.001) and produced small but measurable haemodynamic differences (6-hour MAP: 79 ± 14 vs 77 ± 13 mmHg; P<0.001)2.
  • Despite biological and physiological separation, there was no reduction in 28-day or 90-day all-cause mortality.
  • A septic shock subgroup signal at 90 days (RR 0.87; 95% CI 0.77 to 0.99) was present but attenuated with covariate adjustment and sits within a multiple-comparisons context2.

Internal Validity

  • Randomisation and Allocation: Central, computer-generated assignment sequence; stratified by ICU and enrolment time stratum; allocation concealment was likely adequate up to assignment.
  • Drop out / exclusions: 1818 randomised; outcome denominators indicate limited missing follow-up for mortality (e.g., 28-day mortality assessed in 895/903 vs 900/907). The supplementary CONSORT flow diagram documents post-randomisation exclusions and missingness with reasons2.
  • Performance/Detection Bias: Open-label design risks co-intervention differences (e.g., fluid choices, transfusion thresholds, vasopressor weaning practices); mortality is objective, but tertiary outcomes (vasopressor duration) are more vulnerable to bias.
  • Protocol Adherence: Protocol violations were common: 377/903 (41.8%) in the albumin arm were noncompliant with albumin administration at least once; 336/907 (37.1%) in the crystalloid arm received albumin at least once; synthetic colloids were administered as violations in 205/903 (22.7%) vs 214/907 (23.6%)2.
  • Baseline Characteristics: Broadly similar at baseline for key reported variables (e.g., median age 70 vs 69 years; female sex 39.9% vs 39.4%; SAPS II median 48 vs 48)2.
  • Heterogeneity: Stratification by centre and enrolment timing addressed some clinical variability; however, sepsis management practices across 100 ICUs and the allowance of protocol “exceptions” could dilute effects.
  • Timing: Enrolment within 24 hours of severe sepsis criteria, with pre-specified time strata (<6 hours vs 6–24 hours) supporting early delivery of the intervention (important given time-sensitive resuscitation physiology)1.
  • Dose: Albumin strategy was dose-to-target (serum albumin ≥30 g/L), rather than fixed-volume resuscitation; this strengthens internal validity for the “replacement” hypothesis but may weaken inference about albumin as a primary resuscitation fluid.
  • Separation of the Variable of Interest: Clear biochemical separation and modest haemodynamic separation: day 1 serum albumin 28.6 ± 5.4 vs 24.0 ± 5.2 g/L (P<0.001); day 7 serum albumin 29.4 ± 3.3 vs 23.1 ± 4.7 g/L (P<0.001); 6-hour MAP 79 ± 14 vs 77 ± 13 mmHg (P<0.001)2.
  • Key Delivery Aspects: Total administered fluids over 7 days were similar (median 31,867 vs 31,970 mL; P=0.51), but crystalloids were lower on several days in the albumin arm; total crystalloids over 7 days: 14,150 vs 16,160 mL (P=0.07)2.
  • Crossover: Meaningful contamination occurred (37.1% of crystalloid arm received albumin at least once), plausibly biasing towards the null for the primary endpoint2.
  • Adjunctive therapy use: Some co-intervention imbalances were present in patient-days (e.g., packed red blood cell volumes: median 600 [0–1200] mL vs 300 [0–950] mL; P<0.001) and minor differences in corticosteroid and glycaemic control use (supplement)2.
  • Outcome Assessment: Mortality endpoints are objective; organ failure outcomes were standardised using SOFA scoring; repeated measures used for physiological variables.
  • Statistical Rigor: Primary effect estimates were reported with relative risks and 95% CIs; the observed 28-day mortality (~32%) was lower than the assumed baseline (45%) used for powering, reducing effective power for modest benefits1.

Conclusion on Internal Validity: Moderate: allocation procedures were robust and endpoints objective, but open-label conduct with substantial protocol deviations/crossover and potential co-intervention imbalances likely diluted any true treatment effect and complicate interpretation of tertiary outcomes.

External Validity

  • Population Representativeness: Broad adult ICU severe sepsis cohort (median age ~70 years) across 100 ICUs supports generalisability within high-income ICU systems; however, definitions and care bundles reflect the Sepsis-1/EGDT era, not Sepsis-3 or contemporary resuscitation paradigms.
  • Applicability: The tested strategy requires serial albumin measurement and access to 20% albumin; practice environments with limited albumin supply or cost constraints may not adopt a “target ≥30 g/L” approach.
  • Intervention Specificity: Findings apply to sustained albumin replacement (hyperoncotic 20%) in addition to crystalloids, rather than a simple substitution of crystalloids with iso-oncotic albumin as an initial resuscitation bolus.

Conclusion on External Validity: Generalisability is reasonable for similar ICUs with access to albumin and serial labs, but translation to contemporary sepsis definitions and to resource-limited settings is constrained by changing standard care and the intervention’s resource intensity.

Strengths & Limitations

  • Strengths: Large multicentre ICU trial; central randomisation with stratification; pragmatic severe sepsis population including septic shock; objective primary outcome; demonstrated biochemical and physiological separation with detailed supplemental reporting2.
  • Limitations: Open-label design; high protocol nonadherence and clinically meaningful contamination (albumin exposure in control); power assumptions not met due to lower-than-expected baseline mortality; subgroup findings vulnerable to multiplicity and residual confounding in adjusted analyses2.

Interpretation & Why It Matters

  • Routine use in severe sepsis
    A protocolised strategy of maintaining serum albumin ≥30 g/L using 20% albumin did not improve 28-day or 90-day survival versus crystalloids alone.
  • Physiological effects
    Albumin replacement produced modest haemodynamic changes (e.g., 6-hour MAP 79 ± 14 vs 77 ± 13 mmHg; P<0.001) and altered fluid composition without reducing total fluids (7-day total administered fluids: 31,867 vs 31,970 mL; P=0.51)2.
  • Where albumin might still fit
    The septic shock subgroup signal at 90 days (RR 0.87; 95% CI 0.77 to 0.99) is hypothesis-generating and does not justify routine adoption without careful attention to guideline framing and patient selection2.

Controversies & Subsequent Evidence

  • What ALBIOS actually tested: The intervention was a sustained “albumin replacement to target” strategy rather than albumin as the primary initial resuscitation fluid; correspondence and editorials emphasised that extrapolation to early resuscitation fluid choice is limited345.
  • Protocol violations and contamination: The high rate of deviations (including albumin exposure in 37.1% of controls) risks underestimating any true effect and complicates causal interpretation of secondary/tertiary outcomes2.
  • Septic shock subgroup: The apparent 90-day benefit in septic shock was accompanied by opposite direction in non-shock and was attenuated in covariate-adjusted analyses; this creates a credible “multiplicity + residual confounding” explanation for a borderline P=0.03 interaction finding26.
  • Power and event-rate assumptions: Editorial commentary highlighted that lower-than-anticipated mortality reduced the ability to detect modest absolute benefits, particularly when contamination and practice heterogeneity are present57.
  • Meta-analyses: Post-ALBIOS systematic reviews/meta-analyses have been mixed, often suggesting no clear all-cause mortality benefit overall, while some report possible benefit signals in septic shock or specific albumin concentrations; conclusions are sensitive to inclusion criteria (sepsis definitions, albumin concentration, timing, and co-interventions)89101112.
  • Guidelines: Surviving Sepsis Campaign guidelines recommend crystalloids as first-line and suggest considering albumin when substantial amounts of crystalloids are required (weak/conditional recommendation, low certainty), whereas newer albumin-specific guidance generally discourages routine albumin use in many critical care indications13161514.

Summary

  • In 1818 ICU patients with severe sepsis/septic shock, protocolised 20% albumin replacement (target serum albumin ≥30 g/L) plus crystalloids did not reduce 28-day mortality versus crystalloids alone (31.8% vs 32.0%; RR 1.00; 95% CI 0.87 to 1.14; P=0.94).
  • No overall 90-day mortality benefit was observed (41.1% vs 43.6%; RR 0.94; 95% CI 0.85 to 1.05; P=0.29).
  • Albumin achieved clear biochemical separation and modest haemodynamic effects (e.g., higher early MAP; lower cardiovascular SOFA subscore), without translating into improved global organ dysfunction or length of stay.
  • A septic shock subgroup signal at 90 days (RR 0.87; 95% CI 0.77 to 0.99) was offset by opposite direction in non-shock and attenuated after covariate adjustment.
  • Open-label conduct with high protocol deviations and albumin contamination in controls limits the certainty of causal inference for secondary/tertiary endpoints and likely biased effects towards the null.

Further Reading

Other Trials

Systematic Review & Meta Analysis

Observational Studies

Guidelines

Notes

  • ALBIOS tests a sustained albumin-replacement strategy (20% albumin to target serum albumin ≥30 g/L) rather than “albumin vs crystalloid” as a simple initial resuscitation-fluid substitution; apply conclusions accordingly.
  • The septic shock subgroup signal is not definitive and should be interpreted alongside guideline recommendations and the trial’s contamination/protocol deviations.

Overall Takeaway

ALBIOS is a landmark sepsis fluid trial because it rigorously tested a biologically plausible, protocolised albumin-replacement strategy at scale and demonstrated that achieving biochemical and modest haemodynamic effects does not necessarily translate into improved survival. Its neutral primary outcome underpins modern guideline preference for crystalloids as first-line resuscitation fluid, while leaving only a cautious, conditional space for albumin in selected septic shock patients requiring large crystalloid volumes.

Overall Summary

  • 20% albumin replacement to maintain serum albumin ≥30 g/L did not reduce 28- or 90-day mortality versus crystalloids alone.
  • Clear biochemical separation and small haemodynamic differences occurred, without improvement in global organ dysfunction.
  • Septic shock subgroup benefit is hypothesis-generating and attenuates with adjustment; contamination and protocol violations likely biased towards the null.

Bibliography