Editorials

Excess mortality after human albumin administration in critically ill patients

Clinical and pathophysiological evidence suggests albumin is harmful

Papers p 235 Letters p 277 Science commentary p   240

Albumin is a medium weight colloid which plays an essential role in generating the colloid-osmotic pressure. It facilitates fluid retention in the intravascular space. Human albumin is often given to critically ill patients with life threatening hypovolaemia. Low serum albumin concentrations are seen in various disease states and may be due to leakage, increased metabolism, or insufficient synthesis in the liver. The serum albumin concentration in critically ill patients seems to be inversely related to mortality.¹ Yet does this observation imply that hypoalbuminaemia should be treated with albumin? In this week's issue a systematic reviewpublished simultaneously in the Cochrane Library² and a sequel to a paper on the controversy of whether critically ill patients with hypovolaemia should be given colloid or crystalloid fluids³evaluates the use of human albumin in various clinical settings (p 235).⁴

The paper is clinically important because it suggests that a respected and widely used treatment given to neonates, children, and adults with hypovolaemia, burns, or hypoalbuminaemia is associated with increased mortality: on average six extra deaths for every 100 patients treated. The authors conclude that human albumin should not be given any more "outside the context of a rigorously conducted randomised controlled trial."

Can we trust these findings? As systematic overviews of the medical literature are becoming more prevalent, it is important for clinicians to understand how to decide whether an overview is credible and how to interpret its results. Guidelines to help assess the scientific quality of a systematic review focus on the definition of the question, the comprehensiveness of the search strategy, the methods of choosing and assessing the primary studies, and the technique of combining the results and reaching appropriate conclusions. ^{5 6}

The present overview addressed a focused clinical question, the relation between one determinantadministration of human albumin versus no albumin or crystalloidsand a clinically important outcome (death); the criteria used to select articles for inclusion were appropriate; and it is unlikely that relevant studies were missed. The validity of the studies included was appropriately appraised. The authors also make explicit what the data are, and the assessments of studies are reproducible. Patients from many different hospitals were studied, but the results were similar in the three different settings: volume expansion, burns, and treating low serum albumin. The review therefore seems to be scientifically robust. What adds to the credibility of these results is that if results are consistent across studies they are likely to apply to this wide variety of patients.⁵ However, favourable effects of albumin administration in certain patients may have been obscured in the analysis and cannot be totally excluded.

Another requirement is that there is a plausible pathophysiological mechanism to explain the excess mortality. Without one, it is hard to understand and accept study results of this kind. A low serum albumin value is a marker for serious disease associated with high mortality. However, a direct causal relation between low albumin values and mortality has not been established, and it is difficult to justify maintaining serum albumin values within the "normal" range without clinical evidence that this improves patients' outcome. On the contrary, there are several reasons why albumin supplementation might make things worse for critically ill patients.

Firstly, cardiac decompensation may occur after rapid volume replacement with 20% albumin since this leads to an increase in volume retention (of up to fourfold). Indeed, an older study in baboons found that interstitial pulmonary oedema develops after albumin infusion in haemorrhagic shock.⁷ Secondly, in patients with increased capillary permeability or the capillary leak syndrome albumin administration may become detrimental when albumin and water cross the capillary membrane and cause or worsen (pulmonary) oedema, thus compromising tissue oxygenation and finally leading to multiorgan failure. Thirdly, the antihaemostatic and platelet lowering properties of albumin may increase blood loss in postsurgical or trauma patients.⁸ Finally, albumin administration in the resuscitation of hypovolaemic shock may impair sodium and water excretion and worsen renal failure.⁹ Thus, although not fully understood, several potential mechanisms may explain how human albumin administration may worsen the condition of critically ill patients, but they need to be delineated in more detail.

Alternatives to albumin are available for most acute situationshypovolaemic shock, burns, and in postsurgical patients with hypovolaemia. Unfortunately, they are not without drawbacks. The newer synthetic colloids like hydroxyethyl starch (high molecular weight hetastarch) are larger in molecular size and hopefully do not leak into the extravascular spaces. Indeed, in recent studies volume replacement using hetastarch in patients with the capillary leak syndrome led to improved haemodynamics with a lower incidence of pulmonary oedema than with saline solutions.¹⁰ Hetastarch, however, reduces platelet aggregation, prolongs bleeding time, and decreases the levels of circulating factor VIII.¹¹ Gelatin based plasma substitutes (such as Gelofusine) can cause anaphylactic reactions and impair primary haemostasis and thrombin generation. The defect in primary haemostasis seems to be related to a gelatin induced reduction in von Willebrand factor activity, whereas the decreased thrombin generation is due to dilution.¹² Dextran infusion may also lower plasma factor VIII, and it prolongs bleeding time. Although high quality comparative studies in critically ill patients are not yet available, clinicians should be aware of these adverse effects on the haemostatic system. Crystalloids do not influence haemostasis but more volume needs to be infused to reach adequate clinical effects. This is usually unwanted in young children and patients with renal failure, who are at increased risk of volume overload, oedema, and subsequent compromised oxygenation.

How then should we use albumin from now on? Although albumin administration is surely harmful in certain categories of patients, favourable effects in particular patients cannot yet be excluded. An effort must be made to identify these patients. As agreed in the North American consensus conference,¹³ albumin should not be used for the treatment of septic shock. Hypoalbuminaemia in patients without circulatory failure is a symptom that should not be treated: instead the cause should be identified and treated. In other clinical circumstances synthetic colloids and crystalloids may offer an effective,² relatively cheap, and safe (no viral or prion risk) alternative.

After evaluating the evidence that a treatment is not beneficial and may even be harmful, deciding on subsequent actions may not be simple. If one accepts that the results of this systematic review are valid, the differences in mortality are clinically relevant, and plausible mechanisms exist to explain these differences, and if one thinks that the results apply to patients in one's own practice then one has to decide whether to continue to administer human albumin. Given the succession of positive answers to these questions the administration of albumin should be halted until, as the authors suggest, the results of a high quality large clinical trial are available.