Skip navigation

Albumin in critical illness revisited


Clinical bottom line

The use of albumin in critically ill patients seems not to be associated with higher mortality, despite a review originally suggesting this.


An original meta-analysis [1] featured in Bandolier 54 concluded that there was a strong suggestion that albumin may increase mortality in comparison with crystalloids in critically ill patients with hypovolaemia, burns, or hypoalbuminaemia. This was based on 30 randomised trials with 1,419 patients, in trials that were predominantly small and with low death rates. This has caused considerable controversy, and two additional systematic reviews [2, 3] contest the findings. One of them [2] examined mortality, and the other [3] morbidity. Here we examine mainly the additional review looking at mortality.


  1. Cochrane Injuries Group Albumin Reviewers. Human albumin administration in critically ill patients: systematic review of randomised controlled trials. BMJ 1998 317: 235-40.
  2. MM Wilkes, RJ Navickis. Patient survival after human albumin administration: a meta-analysis of randomized, controlled trials. Annals of Internal Medicine 2001 135: 149-164.
  3. GR Haynes et al. Albumin administration: what is the evidence of clinical benefit? A systematic review of randomized controlled trials. European Journal of Anaesthesiology 2003 20: 771-793.

Systematic review

The newer systematic review on mortality used an extensive search strategy with several electronic databases, hand searching journal, and contacted manufacturers for studies. Inclusion criteria were broad, and included randomised comparison of albumin therapy with crystalloid therapy, no therapy, or a lower dose of purified albumin. Mortality data had to be available. No restriction on setting was used. The primary outcome was relative risk for death, together with planned sensitivity analyses for issues like blinding, allocation concealment, presence of mortality as a study endpoint, and size.


There were 42 trials with at least one death and 13 trials with no deaths. In all there were 3,504 patients with 52 deaths. The median number of patients randomised was 52 (range 10 to 300). Most trials involved surgery (27), and others burns (4), high-risk neonates (6), hypoalbinaemia (5), and other indications (8).

For the 42 trials with at least one death the relative risk was 1.1 (95% CI 0.95 to 1.28). Over all trials, mortality was 273/1456 (19%) with albumin and 252/1502 (17%) with crystalloid. No significantly increased risk of mortality was found for patients undergoing surgery, with burns, hypoalbuminaemia, high-risk neonates, or with any other indication.

Sensitivity analysis revealed that smaller trials of lower quality had more mortality with albumin than crystalloid (Table 1).

Table 1: Sensitivity analysis of trials of mortality

Trials with fewer than 100 patients
Trials with 100 patients or more
Trial category
Relative risk (95% CI)
Relative risk (95% CI)
All randomised trials
1.4 (1.1 to 1.7)
0.9 (0.8 to 1.1)
Blind trials
1.2 (0.6 to 2.6)
0.5 (0.3 to 0.9)
Non-blind trials
1.4 (1.1 to 1.7)
1.0 (0.8 to 1.3)
Adequate allocation concealment
1.4 (1.0 to 2.1)
1.0 (0.7 to 1.3)
Unclear allocation concealment
1.3 (1.0 to 1.8)
0.9 (0.7 to 1.2)
Mortality end point
1.2 (0.9 to 1.6)
0.9 (0.8 to 1.4)
No mortality end point
1.5 (1.1 to 2.1)
1.1 (0.3 to 3.7)


Most of the 32 smaller trials were not blind (27/32) and most (23/32) did not have mortality as their end point. Again, most larger trials were not blind (8/10), but most of these (8/10) did have mortality as an end point. The smaller trials gave a different answer to the larger trials.

The second systematic review had more trials and many more patients than the original review, and used sensitivity analysis to explore important methodological issues. We know that smaller, and lower quality trials may mislead us. Here is another example of where that seems to have happened.

The third systematic review [3] examined issues of morbidity as well as mortality, and comes down in support of albumin use, despite a complex variety of settings and outcomes.