Skip navigation

Albumin in critical illness

Continuing the theme of issues taken up by the media, Bandolier thought it useful to briefly summarise the systematic review on albumin in critically ill patients [1] that hit the headlines recently. As may be expected from a Cochrane review, the searching was meticulous, and great efforts were expended to find every possible trial and to obtain the numbers of deaths from authors if that was not given in the published report.

Reports and outcomes

Reports were those which had randomly administered albumin or plasma protein fraction in critically ill patients with hypovolaemia from trauma or surgery, with burns, or with hypoalbuminaemia. The outcome measure was all-cause mortality over the period of follow-up, which ranged from about 1 day until discharge.


There were 32 randomised trials with 1204 patients with data for analysis (six studies had no deaths and for one no data could be obtained). Most of the trials involved fewer than 40 patients and only two involved more than 100 patients.

The overall mortality rate was 10% in controls (treated mainly with crystalloids) and 17% in those who were given albumin. Although only one trial itself had a significant result, overall the results were statistically significant for patients with hypoalbuminaemia and burns, but not (just) those with hypovolaemia. Overall the NNH for albumin treatment was 14 (95%CI 9 to 32), and the scatter of results by size of trial is shown in the L'Abbé plot.

Sensitivity analysis showed that using only trials with adequate concealment of allocation did not make any difference. We have also analysed the data by trial size, and found that larger trials yielded a larger effect size. For trials with fewer than 50 patients in the comparison 17% of patients died with control and 23% with albumin, and the relative risk was 1.3 (0.9 to 1.9). For trials with at least 50 patients 6% of patients died with control and 14% with albumin, with a relative risk of 2.2 (1.4 to 3.4) and an NNH of 13 (8.5 to 27).


The authors properly acknowledge that because the review was based on small numbers of deaths the results must be interpreted with caution. They also say that a rigorous randomised trial may need to be done to confirm these results. But the size of that trial would need to be large, and the circumstances in which it could be done are not easy to discern. But this is a lovely example of systematic review uncovering a small but important effect not shown by individual studies themselves.


  1. Cochrane Injuries Group Albumin Reviewers. Human albumin administration in critically ill patients: systematic review of randomised controlled trials. BMJ 1998 317: 235-40.

previous story in this issue