No evidence of excess mortality in patients receiving human albumin: a meta-analysis of randomized controlled trials
© The Author(s) 2001
Received: 15 January 2001
Published: 2 March 2001
A meta-analysis of randomized controlled trials was conducted to test the hypothesis that albumin administration is not associated with excess mortality. Candidate trials must have compared albumin administration with crystalloid, no albumin or a lower dosage of albumin. No restrictions were placed on clinical indication. Trials were identified by computer searches of bibliographic databases, the Cochrane Controlled Trials Register, the Cochrane Medical Editors Trial Amnesty and other Internet-resident resources; hand-searching of general medical journals; contacts with albumin suppliers; and consulting the reference lists of prior meta-analyses, review articles and other publications. A total of 55 trials with 525 deaths among 3504 randomized patients were included in the meta-analysis. Included trials involved six categories of clinical indications: surgery or trauma (n = 27), burns (n = 4), hypoalbuminemia (n = 5), high-risk neonates (n = 6), ascites (n = 5) and other (n = 8). The pooled relative risk of death and mortality risk difference were 1.11 (95% CI, 0.95-1.28) and 1.5% (95% CI, -0.7 to 3.8%), respectively, under a fixed-effects model (P = 0.66 and P = 0.65 for heterogeneity, respectively). Thus, without the application of any exclusion criteria that might have biased the results, there was no evidence of significant excess mortality risk in albumin recipients. In addition, there was no significant excess risk within any of the six categories of clinical indications. For the seven trials incorporating some form of blinding the pooled relative risk was 0.73 (95% CI, 0.48–1.12). The pooled relative risk among the 17 trials with mortality as an endpoint was 1.00 (95% CI, 0.84–1.18). In the 35 trials without crossover of control group patients to the albumin group the pooled relative risk was 1.04 (95% CI, 0.89–1.22). For all trials there was evidence of significant small-trial bias (ie publication or other bias) favoring control (P = 0.03 by Egger's test). The adjusted relative mortality risk taking into account small-trial bias was 0.83 (95% CI, 0.65–1.08) based on regression analysis. Thus, lack of blinding, absence of mortality as a study endpoint, crossover and small-trial bias all consistently biased pooled relative risk in favor of control. This meta-analysis furnished no evidence of excess albumin-associated mortality and strongly suggested that albumin may reduce mortality. These findings support the safety of albumin.