In July 1998, members of the Cochrane Injuries Group published a meta-analy
sis on the risk of death associated with the use of human albumin in patien
ts with hypovolaemia, burns or hypoalbuminaemia. The control groups receive
d crystalloid solutions or no albumin. Data of 1204 patients from 24 studie
s were included in the analysis. In both the entire group and the three sub
groups the relative risk of death was increased for the albumin-treated pat
ients. The pooled absolute difference in the risk of death was 6%, suggesti
ng one additional death in 17 patients treated with albumin. Even if the co
nclusion that albumin increases mortality is rejected on methodological gro
unds, there is no reason to assume that albumin is superior to crystalloids
or artificial colloids. Commonly accepted hut largely unproven indications
for the use of albumin include hypovolaemia, burns, hypoalbuminaemia, abdo
minal paracentesis, subarachnoid haemorrhage, cerebral trauma or ischaemia,
and neonatal care. None of these is supported by randomised clinical trial
s. Therefore, crystalloids and/or colloids (gelatine, hydroxyethyl starch)
should be preferred and albumin avoided in the vast majority of cases.