SAFE: Saline versus Albumin Fluid Evaluation in Critical Illness (2004)

“In a heterogeneous group of patients in the ICU, the use of either 4 percent albumin or normal saline for fluid resuscitation resulted in similar outcomes at 28 days.”

  • The SAFE Study Investigators

1. Publication Details

  • Trial Title: A Comparison of Albumin and Saline for Fluid Resuscitation in the Intensive Care Unit
  • Citation: The SAFE Study Investigators. A comparison of albumin and saline for fluid resuscitation in the intensive care unit. N Engl J Med. 2004;350(22):2247-2256. DOI: 10.1056/NEJMoa040232
  • Published: May 27, 2004, in The New England Journal of Medicine
  • Author: The Saline versus Albumin Fluid Evaluation (SAFE) Study Investigators
  • Funding: The National Health and Medical Research Council of Australia; and others.

2. Keywords

  • Fluid Resuscitation, Critical Care, Albumin, Saline, Colloids, Crystalloids, Randomized Controlled Trial

3. The Clinical Question

  • In a general population of critically ill adult patients in the ICU (Population), does fluid resuscitation with 4% albumin (Intervention) compared to 0.9% saline (Comparison) affect 28-day all-cause mortality (Outcome)?

4. Background and Rationale

  • Existing Knowledge: The choice between colloid (like albumin) and crystalloid (like saline) solutions for fluid resuscitation was one of the longest-standing and most intense debates in critical care. Proponents of albumin argued that its higher oncotic pressure would keep fluid in the vascular space more effectively, while proponents of saline cited its lower cost and questioned the clinical benefit of albumin.
  • Knowledge Gap: Despite decades of debate, there was no high-quality evidence from a large, multicenter randomized controlled trial to definitively determine if one fluid type was superior to the other in a broad population of critically ill patients.
  • Proposed Hypothesis: The authors hypothesized that there would be no significant difference in 28-day mortality between patients resuscitated with albumin and those resuscitated with saline.

5. Study Design and Methods

  • Design: A very large, multicenter, prospective, randomized, double-blind, controlled trial (used to test the effectiveness of interventions).
  • Setting: 16 intensive care units (ICUs) in Australia and New Zealand.
  • Trial Period: Enrollment ran from November 2001 to June 2003.
  • Population:
    • Inclusion Criteria: Adult patients (≥18 years) admitted to the ICU who were judged by their treating clinician to require fluid resuscitation.
    • Exclusion Criteria: Included patients with traumatic brain injury, burns, or those undergoing liver transplantation.
  • Intervention: Fluid resuscitation with 4% human albumin.
  • Control: Fluid resuscitation with 0.9% saline.
  • Management Common to Both Groups: The assigned study fluid was used for all fluid resuscitation needs throughout the ICU stay. The choice of all other interventions was at the discretion of the treating clinicians.
  • Power and Sample Size: The authors calculated that a sample size of 7000 patients would provide 90% power to detect a 3% absolute risk reduction in 28-day mortality. (Power is a study’s ability to find a real difference between treatments if one truly exists; 90% power means the study had a 90% chance of detecting the specified effect, which is considered very high).
  • Outcomes:
    • Primary Outcome: All-cause mortality at 28 days.
    • Secondary Outcomes: Included duration of mechanical ventilation, duration of renal-replacement therapy (RRT), and length of ICU and hospital stay.

6. Key Results

  • Enrollment and Baseline: 6997 patients were randomized (3497 to albumin and 3500 to saline). The groups were well-matched at baseline.
  • Trial Status: The trial was completed as planned.
  • Primary Outcome: There was no significant difference in 28-day mortality. 726 of 3473 patients (20.9%) in the albumin group died, compared with 729 of 3460 patients (21.1%) in the saline group (p=0.87).
  • Secondary Outcomes: There were no significant differences between the groups in the duration of mechanical ventilation, duration of RRT, or length of ICU and hospital stay.
  • Adverse Events: The incidence of adverse events was similar in both groups.

7. Medical Statistics

  • Analysis Principle: The trial was analyzed using an intention-to-treat principle.
  • Statistical Tests Used: The primary outcome was analyzed using a chi-square test.
  • Primary Outcome Analysis: The primary outcome was a comparison of the proportions of death between the two groups.
  • Key Statistic(s) Reported: Relative Risk (RR) for death at 28 days: 0.99 (95% CI, 0.91 to 1.09; P-value: 0.87).
  • Interpretation of Key Statistic(s):
    • Relative Risk (RR):
      • Formula: Conceptually, RR = (Risk in Intervention Group) / (Risk in Control Group).
      • Calculation: The paper reports the result as 0.99.
      • Clinical Meaning: An RR of 0.99 indicates virtually no difference in the risk of death between the two groups.
    • Confidence Interval (CI):
      • Formula: Conceptually, CI = (Point Estimate) ± (Margin of Error).
      • Calculation: The 95% CI was 0.91 to 1.09.
      • Clinical Meaning: Since this range crosses the line of no effect (1.0), it confirms that the result is not statistically significant. Clinically, this means the true effect could range from a 9% benefit to a 9% harm.
    • P-value: The p-value of 0.87 is much higher than the 0.05 threshold, indicating the result is not statistically significant and very likely due to chance (a result is conventionally considered statistically significant if the p-value is less than 0.05).
  • Clinical Impact Measures: As the trial was neutral, ARR and NNT are not applicable.
  • Subgroup Analyses: In a pre-specified subgroup analysis of patients with severe sepsis, there was a trend towards lower mortality in the albumin group, but this was not statistically significant. Conversely, in the subgroup of patients with traumatic brain injury (who were included in a nested study), there was a trend towards harm with albumin.

8. Strengths of the Study

  • Study Design and Conduct: The very large, multicenter, randomized, double-blind, controlled design is the gold standard and provided a massive amount of high-quality data.
  • Generalizability: The pragmatic design and inclusion of a very large, heterogeneous population of ICU patients make the findings highly generalizable to real-world practice.
  • Statistical Power: The enormous sample size provided definitive power to confidently rule out even a small but clinically important mortality difference.
  • Patient-Centered Outcomes: The primary outcome of 28-day mortality is a robust and patient-centered endpoint.

9. Limitations and Weaknesses

  • Internal Validity (Bias): The study was well-conducted with a very low risk of bias.
  • External Validity (Generalizability): The findings are highly generalizable to the broad population of critically ill patients.
  • Other: The study used 4% albumin, and the results may not apply to other concentrations (e.g., 20% albumin). The comparator was saline, which has its own potential harms.

10. Conclusion of the Authors

  • The authors concluded that in a diverse group of critically ill patients, 4% albumin and normal saline are clinically equivalent fluids for resuscitation.

11. To Summarize

  • Impact on Current Practice: This was a landmark trial that provided definitive evidence to resolve one of the longest-standing debates in critical care. It demonstrated that for the majority of ICU patients, the more expensive colloid (albumin) offered no survival advantage over the less expensive crystalloid (saline).
  • Specific Recommendations:
    • Patient Selection: For the broad population of adult ICU patients requiring fluid resuscitation.
    • Actionable Intervention: The results support the use of crystalloids (like saline) as the first-line fluid for resuscitation.
  • What This Trial Does NOT Mean: This trial does NOT mean that albumin has no role in the ICU. The subgroup analysis in sepsis was hypothesis-generating, and albumin may still be considered in specific situations (e.g., patients with severe hypoalbuminemia or those who have received very large volumes of crystalloids).
  • Implementation Caveats: Given the lack of benefit and significantly higher cost, albumin should not be used as a routine, first-line resuscitation fluid in a general ICU population.

12. Context and Related Studies

  • Building on Previous Evidence: The SAFE trial (2004) was designed to provide a definitive answer to the “colloid vs. crystalloid” debate, which had been fueled by decades of smaller, inconclusive studies and strong opinions.
  • Influence on Subsequent Research: The definitive neutral result of this trial, combined with its intriguing subgroup analysis in sepsis, was highly influential. It led directly to the design of the ALBIOS trial (2014), which specifically tested the role of albumin in patients with severe sepsis.

13. Unresolved Questions & Future Directions

  • Unresolved Questions: The key unresolved question from this trial was whether the signal of benefit for albumin in the severe sepsis subgroup was real.
  • Future Directions: The ALBIOS trial was the direct successor to this study, designed to answer the specific question of albumin’s efficacy in sepsis. The SAFE trial also helped shift the focus of the fluid debate away from “colloid vs. crystalloid” and towards “balanced vs. unbalanced” crystalloids (e.g., the SMART and BaSICS trials).

14. External Links

15. Framework for Critical Appraisal

  • Clinical Question: The research question was of the highest relevance, addressing a fundamental and ubiquitous aspect of critical care for which there was no high-quality evidence.
  • Methods: The very large, multicenter, double-blind RCT design was of the highest quality and was essential for providing a definitive answer. The pragmatic design ensured high external validity.
  • Results: The study reported a clear neutral finding for its primary outcome, with a narrow confidence interval centered on the null value. This provides strong evidence against a meaningful clinical benefit of routine albumin use in this population.
  • Conclusions and Applicability: The authors’ conclusion is a direct and fair reflection of the data. The trial is a landmark in evidence-based critical care and a classic example of a high-quality “negative” trial that was profoundly practice-changing by providing strong evidence that a more expensive intervention is not superior to a simpler, less expensive standard of care.

16. Disclaimer and Contact

  • This summary is provided by the Academic Committee of ESBICM (ACE) to facilitate the understanding of this study; readers are advised to refer to the original trial document for a deeper understanding. If you find any information incorrect, or missing, or it needs an update or have a request for a specific critical care trial summary, kindly write to us at academics[at]esbicm.org.
Scroll to Top