The New England Journal of Medicine May 27, 2004

A Comparison of Albumin and Saline for Fluid Resuscitation in the Intensive Care Unit (The SAFE Study)

The SAFE Study Investigators (Finfer S, Bellomo R, Boyce N, French J, Myburgh J, Norton R)

Bottom Line

In a large, multicenter, randomized trial of critically ill adults, fluid resuscitation with 4% albumin resulted in similar 28-day all-cause mortality compared to 0.9% normal saline, establishing the safety and equivalence of both fluids for general ICU resuscitation.

Key Findings

1. Overall 28-day all-cause mortality was virtually identical between the two groups: 726 deaths in the albumin group (20.9%) compared to 729 deaths in the saline group (21.1%) (Relative Risk [RR], 0.99; 95% CI, 0.91 to 1.09; P=0.87).

2. There was no significant difference in the incidence of new single-organ or multiple-organ failure between the albumin and saline cohorts (P=0.85).

3. No significant differences were observed in mean days spent in the ICU (6.5 vs. 6.2 days; P=0.44), hospital length of stay (15.3 vs. 15.6 days; P=0.30), days of mechanical ventilation (4.5 vs. 4.3 days; P=0.74), or days requiring renal-replacement therapy (0.5 vs. 0.4 days; P=0.41).

4. In a predefined subgroup of patients with severe sepsis, 28-day mortality was 30.7% in the albumin group versus 35.3% in the saline group, a trend that did not reach statistical significance (RR, 0.87; 95% CI, 0.74 to 1.02; P=0.09).

5. In patients with trauma, mortality trended higher with albumin (13.6% vs. 10.0%; RR, 1.36; 95% CI, 0.99 to 1.86; P=0.06). A post-hoc analysis identified that this was driven by patients with traumatic brain injury (TBI), where mortality was significantly higher with albumin (24.5% vs. 15.1%; RR, 1.62; 95% CI, 1.12 to 2.34; P=0.009).

Study Design

Design

RCT

Double-Blind

Sample

6,997

Patients

Duration

28 days

Median

Setting

Multicenter, Australasia

Population Adult patients (aged >=18 years) admitted to 16 academic tertiary intensive care units who were judged by treating clinicians to require fluid administration to maintain or increase intravascular volume. Patients admitted after cardiac surgery, liver transplantation, or for the treatment of burns were excluded.

Intervention Intravascular-fluid resuscitation using 4% human albumin, administered in 500-ml specifically blinded bottles, given at the discretion of the treating clinician for up to 28 days or until ICU discharge.

Comparator Intravascular-fluid resuscitation using 0.9% sodium chloride (normal saline), administered in identical 500-ml blinded bottles under the same clinical parameters.

Outcome Death from any cause during the 28-day period after randomization.

Study Limitations

• Subgroup analyses (including those for trauma and severe sepsis) were insufficiently powered in the primary study, serving primarily as hypothesis-generating signals rather than definitive conclusions.

• The primary endpoint was limited to 28-day mortality, which may miss longer-term survival differences or functional neurological outcomes, particularly in traumatic brain injury patients.

• The study protocol allowed clinicians to dictate the rate and volume of fluid administration based on clinical judgment, which may have introduced practice variations, although randomization likely balanced these effects.

• Specific populations such as patients undergoing liver transplantation, elective cardiac surgery, or treatment for burns were excluded, meaning results cannot be generalized to those groups.

Clinical Significance

The SAFE trial was a landmark study that definitively resolved a major medical controversy by proving that 4% albumin and 0.9% saline are equally safe and effective for fluid resuscitation in a heterogeneous ICU population. Because albumin is a blood product and significantly more expensive than crystalloids, these findings cemented normal saline (and later, balanced crystalloids) as the default first-line resuscitation fluid globally. Crucially, the trial also generated two practice-changing subgroup hypotheses: it identified a clear harm signal for albumin in traumatic brain injury (which became a firm contraindication) and suggested a potential benefit in sepsis, which directly inspired subsequent major trials like the ALBIOS study.

Historical Context

During the late 1990s, fluid resuscitation practices were thrown into disarray after a 1998 Cochrane meta-analysis by the Injuries Group Albumin Reviewers suggested that human albumin administration increased the absolute risk of death in critically ill patients by 6%. This meta-analysis, based on small and heterogeneous trials, prompted widespread panic, regulatory warnings, and a massive drop in albumin use. The critical care community recognized the urgent need for a massive, high-quality, blinded randomized controlled trial to determine the true safety profile of albumin. The SAFE trial was funded and conducted in Australia and New Zealand to provide a definitive answer, ultimately rescuing albumin from claims of universal harm while concurrently demonstrating it offered no survival superiority over cheap, widely available saline.

Guided Discussion

High-yield insights from every perspective

Med Student

Medical Student

Based on Starling's forces, why was 4% albumin theoretically expected to be a superior volume expander compared to 0.9% normal saline, and why might the pathophysiological changes of critical illness explain the lack of clinical mortality benefit seen in the SAFE study?

Key Response

This tests foundational physiology. Albumin is a colloid that provides oncotic pressure, theoretically keeping fluid in the intravascular space longer than crystalloids like saline. However, in critical illness (like sepsis), systemic inflammation causes endothelial glycocalyx degradation and capillary leak, allowing albumin to escape into the interstitium, neutralizing its theoretical oncotic advantage and leading to similar clinical outcomes.

Resident

Although the SAFE study demonstrated overall mortality equivalence between albumin and saline, there was a specific patient population where albumin was associated with increased mortality. Which subgroup was this, and how should this finding directly alter your fluid resuscitation management in the ER or ICU?

Key Response

This focuses on clinical application. The post-hoc and subsequent follow-up analyses showed increased mortality in patients with traumatic brain injury (TBI) who received albumin, likely because the 4% albumin used is slightly hypotonic compared to plasma, worsening cerebral edema and intracranial pressure. Residents must know to strictly avoid albumin in TBI patients.

Fellow

The SAFE trial compared albumin specifically to 0.9% normal saline. In the context of modern fluid resuscitation literature (e.g., SMART and SALT-ED trials), how does the risk of hyperchloremic metabolic acidosis and acute kidney injury associated with normal saline complicate our retrospective interpretation of the SAFE trial's 'control' group?

Key Response

Fellows must integrate historical and modern evidence. Since large volumes of 0.9% saline can cause hyperchloremic non-anion gap metabolic acidosis and potentially worsen renal perfusion, the saline 'control' group in SAFE may have experienced iatrogenic harm. This raises the nuance that albumin might not be equivalent to an optimal fluid, but merely equivalent to a flawed crystalloid, shifting modern debates toward balanced crystalloids (like Lactated Ringer's) instead.

Attending

When evaluating a patient with severe sepsis requiring massive volume resuscitation, how do you balance the physiological rationale for colloid use, the neutral overall mortality outcome of the SAFE study, and the principles of cost-conscious care when teaching trainees about fluid stewardship?

Key Response

This targets wise, practice-changing teaching points. Attendings must guide trainees to understand that because efficacy (mortality) is equivalent for the general ICU population, the massive cost difference between albumin and saline makes crystalloids the definitive first-line choice. It emphasizes value-based care, reserving expensive albumin only for specific salvage scenarios rather than routine resuscitation.

Scholarly Review

Critical appraisal through the lens of expert reviewers and guideline development

PhD

The SAFE trial was a massive, pragmatically designed, double-blind randomized controlled trial. Given the physical characteristics of albumin (e.g., foaming upon agitation) versus saline, what methodological challenges exist in truly maintaining the blind in such trials, and how might unrecognized unblinding bias the administration of non-protocolized co-interventions?

Key Response

This evaluates study design critique. True blinding of intravenous fluids is notoriously difficult because bedside nurses and physicians may notice visual cues (like foaming). If unblinded, clinicians might alter their threshold for administering vasopressors, diuretics, or other supportive care (co-interventions). PhDs must critically assess how such biases are handled, even when the primary endpoint (28-day mortality) is robust and objective.

Journal Editor

As an editor reviewing the original SAFE manuscript, how would you scrutinize the authors' reporting of subgroup outcomes, particularly the trend toward benefit in severe sepsis and the signal for harm in trauma/TBI, regarding the risks of multiplicity and statistical power?

Key Response

This focuses on critical appraisal. A seasoned reviewer would flag that while the overall trial was well-powered, subgroup analyses are often underpowered and susceptible to Type I errors (false positives) due to multiple testing. The editor must ensure that the manuscript's conclusions and abstract accurately reflect the overall neutral primary outcome, relegating subgroup findings to hypothesis-generating observations rather than definitive conclusions.

Guideline Committee

How does the foundational equivalence demonstrated in the SAFE study influence the Surviving Sepsis Campaign (SSC) guidelines' recommendations on the initial choice of resuscitation fluid, and what specific recommendation does the SSC make regarding the addition of albumin based on this and subsequent evidence?

Key Response

This addresses guideline formulation. The Surviving Sepsis Campaign strongly recommends crystalloids as the first-line fluid for initial resuscitation due to equivalent efficacy and lower cost compared to colloids (Level of Evidence: Strong, based heavily on SAFE). The guidelines suggest using albumin only in patients who require substantial amounts of crystalloids to maintain MAP, balancing the lack of overall mortality benefit with the practical need to limit massive crystalloid overload.

Clinical Landscape

Noteworthy Related Trials

2012

CHEST Trial

n = 7,000 · NEJM

Tested

6% Hydroxyethyl starch (HES) 130/0.4

Population

ICU patients requiring fluid resuscitation

Comparator

0.9% normal saline

Endpoint

90-day all-cause mortality

Key result: HES did not reduce 90-day mortality compared to saline but significantly increased the need for renal-replacement therapy.

2014

ALBIOS Trial

n = 1,818 · NEJM

Tested

20% albumin and crystalloid to maintain serum albumin >= 30 g/L

Population

ICU patients with severe sepsis or septic shock

Comparator

Crystalloid alone

Endpoint

28-day all-cause mortality

Key result: Albumin replacement did not improve 28-day or 90-day survival in patients with severe sepsis, though a post-hoc analysis suggested potential benefit in septic shock.

2018

SMART Trial

n = 15,802 · NEJM

Tested

Balanced crystalloids (Lactated Ringer's or Plasma-Lyte A)

Population

Critically ill adults in the ICU requiring fluid resuscitation

Comparator

0.9% normal saline

Endpoint

Major Adverse Kidney Events within 30 days (MAKE30)

Key result: Balanced crystalloids resulted in a lower rate of the composite outcome of death, new renal-replacement therapy, or persistent renal dysfunction compared to normal saline.

Tailored to your role

Want this tailored to you?

Add your specialty or training stage to get role-specific takeaways and more questions.

Personalize this analysis