Albumin has no benefit over saline in the critically ill

THE majority of protein that makes up the human body is albumin. It has important physiological functions, such as maintaining the colloidal osmotic pressure and delivering as well as storing various organic substances. A reduced albumin level has been shown to correlate with a worsened prognosis in several studies. Despite the functions described above, albumin treatment has been categorized as inconsequential (1, 2) or even damaging (3) in recent meta-analyses. The problems associated with the methodology of meta-analyses and small sample sizes of the included original studies led an Australian—New Zealander study group (The SAFE Study Investigators) to assess the benefits and safety of administering albumin to intensive care patients. It was conducted with two large groups (albumin group 3497 patients, saline [NaCl 0.9%] group [3500]) as a randomized, double-blind study (4). In this — by far the largest — study of nearly 7000 patients, the treatment for hypovolemia during intensive care was either albumin or physiological saline solution. The effectiveness of the treatment was assessed according to the intention-to-treat principle as day-28 all-cause mortality. The results of the study are unquestionable and obvious: administering albumin (4%) as a fluid treatment for critically ill adults to replace lost blood volume is ineffective for reducing mortality by day 28 of care (death rates 726/3473 in the albumin vs. 729/ 3460 in the saline group; RR (relative risk) for death 0.99, 95% CI 0.91—1.09) for the albumin group. Fluid treatment had no effect on the length of stay either in the ICU or in the hospital. Albumin treatment did not reduce the time spent in mechanical ventilation or on dialysis. Trauma patients were randomized separately. In this predefined subgroup, there was almost a statistically significant increase in mortality (81/596 with albumin, 59/590 with saline, RR 1.36, 95% CI 0.99—1.86). The upper confidence limit means that this study cannot exclude a nearly doubled risk of death in trauma patients when albumin is involved. A retrospective further analysis suggests that albumin may be deleterious, especially when cerebral injury is included (deaths 59/241 in the albumin group vs. 38/251 in the saline group; RR 1.62, 95% CI 1.12—2.34, P 1⁄4 0.009). In severe sepsis and ARDS (predefined subgroup analyses without randomization) no significant differences between groups could be detected. Why did the randomized SAFE study reach a different conclusion than the groups using metaanalysis? The SAFE study did not include patients with postoperative cardiac surgery, with liver transplantation, with burns, or pediatric patients. The previously published meta-analyses (1—3) can be extensively criticised for differences in their followup time for mortality, missing data of original studies, confusing factors such as fresh-frozen plasma, the use of albumin in the control group, and variations in the diagnoses, phase of treatment and age of the patients. The guiding principle in metaanalyses should be combining studies that are as homogenous as possible, using the same treatments and having the same follow-up times. The analyses cited above, or the principal studies therein cited, do not provide a convincing case to support the use of albumin — neither does the large RCT. A small study of patients receiving a liver transplant (5) could not establish any effect on mortality, and no other randomized studies are available. Replacement treatment of ascites fluid with albumin in liver cirrhosis patients has shown no convincing benefit: Wilkes and Navickis describe a RR of 0.93 (CI 95% 0.67—1.28) in their meta-analysis. With burns, the meta-analysis suggests damaging effects of albumin (RR 1.76, CI 95% 0.97—3.17) (1). With neonatal patients the evidence is also limited (1—3). Acta Anaesthesiol Scand 2005; 49: 599—600 Copyright # Acta Anaesthesiol Scand 2005 Printed in UK. All rights reserved ACTA ANAESTHESIOLOGICA SCANDINAVICA