Colloid or Crystalloid: Pick Your Poison

Abstract & Commentary

Synopsis: A large multicenter prospective randomized double-blind trial found no difference in 28-day outcomes for fluid resuscitation with normal saline vs 4% albumin for a heterogeneous ICU population.

Source: The SAFE Study Investigators. N Engl J Med. 2004; 350(22):2247-2256.

It is unknown whether the choice of resuscitation fluid impacts outcomes of critically ill patients. This study hypothesized there would be no difference in 28-day mortality for patients resuscitated with normal saline (NS) vs 4% albumin.

A 20-month prospective, randomized, double-blind study designed to detect a 3% difference in absolute mortality was performed in 16 closed tertiary care ICUs in Australia and New Zealand. Patients older than age 18 and judged to require fluid resuscitation by the treating physician were eligible. Those admitted after cardiac surgery, liver transplantation or burn injuries were excluded. Patients were randomized to receive either NS or 4% albumin for all fluid resuscitation until death, discharge, or 28 days after randomization. Quantity and rate of study fluid administration and choice of other fluids (maintenance, nutrition, etc) and care were at the discretion of treating physicians. Data collected included demographics, Acute Physiology and Chronic Health Evaluation II (APACHE II) and Sequential Organ-Failure Assessment (SOFA) scores, diagnostic criteria for severe sepsis, acute respiratory distress syndrome (ARDS) and traumatic brain injury, daily vital signs and infused fluid volumes, survival time, organ failures and duration of ventilation, renal-replacement therapy and hospital and ICU stays. Two interim analyses were planned. Intention-to-treat analysis and standard statistical techniques were used.

Seven thousand patients were randomized. The 2 study groups had similar baseline characteristics (mean age, 59 years; mean APACHE II score, 19; 43% surgical, 18% severe sepsis). A difference in total volume of study fluid administered was seen only in the first 3 days after randomization (ratio of volume for albumin group: NS group 1:1.3 to 1:1.6 per day). The albumin group also received more red blood cells (RBC) (mean, 71 mL) during this time and had slightly higher central venous pressures and serum albumin. There was, however, no difference in mean arterial pressure between the groups. There was no mortality difference at 28 days (absolute difference for albumin vs NS was -0.2%; 95% CI, -2.1-1.8%). Length of ventilation, renal-replacement, and ICU and hospital stays, number of organ failures and survival times were also the same. Analysis of mortality per subgroups defined a priori revealed a weak trend towards increased relative risk (RR) of mortality among trauma patients in the albumin group (RR, 1.36; 95% CI, 0.99-1.86): all of the increased risk was in the trauma patients with associated brain injury (RR 1.62; 95% CI, 1.12-2.34). There were no significant differences in mortality of patients with severe sepsis or ARDS receiving albumin vs NS; however, there was suggestion of a trend towards lower mortality in the severe sepsis group receiving albumin compared to those without severe sepsis (RR 0.87; P = 0.06). The authors conclude that NS and 4% albumin are equivalent therapies for fluid resuscitation but suggest further studies in particular subgroups.

Comment by Saadia R. Akhtar, MD, MSc

The ideal choice of fluid for resuscitation has been a subject of discussion for decades. The primary debate focuses on whether colloid (usually albumin) or crystalloid (NS) is preferable. There are several reasons why albumin is theorized to have potential benefits. It may support colloid oncotic pressure and thus be superior for volume replacement. Furthermore, because low serum albumin is associated with increased mortality in a variety of acutely ill patients, albumin supplementation may improve outcomes.

Albumin appears to have benefit in patients with cirrhosis and spontaneous bacterial peritonitis (SBP); administration of a standard replacement on day 1 and 3, along with antibiotics and other supportive care, reduced mortality and risk of renal dysfunction in 1 randomized controlled trial.1 There are multiple small studies (50-100 patients) showing trends towards both reduced and increased mortality with albumin for other indications. Meta-analyses of these suggest there is no role for albumin administration in general ICU populations. In 1998, the Cochrane Group reviewed 30 randomized studies (1419 patients) comparing albumin to placebo or NS or comparing differing levels of albumin supplementation in critically ill patients with hypovolemia (in the setting of trauma or surgery), burn injuries or hypoalbuminemia. There was an increased risk of mortality with albumin administration for all studies; this was persistent even when only studies with adequate blinding were analyzed (RR, 1.61; 95% CI, 1.09-2.38). The increased risk was seen particularly in patients with burn injuries or hypoalbuminemia as the indication for albumin administration.2

A 2000 updated Cochrane review reported the same results. In a similar analysis, Ferguson et al reviewed 24 trials (1204 patients) comparing albumin to no albumin or crystalloid in the same patient populations. No individual study "with a reasonable sample size" (not defined) suggested improved outcomes with albumin. The pooled relative risk of mortality was higher for patients receiving albumin (RR, 1.68; 95% CI, 1.07 to 2.23).3 Finally, most recently, Wilkes and Navickis reviewed 55 trials including > 3500 patients; these compared albumin to placebo or crystalloid or other concentration of albumin for any indication in any patient population. There was a non-statistically significant trend towards reduced mortality with albumin in trials with 2 of the following characteristics: adequate blinding, primary end point of mortality or no crossover. Pooled analysis revealed no difference in mortality between patients receiving albumin vs NS. The authors concluded that albumin was unlikely to be harmful but that is was unclear whether it added benefit.4

These reports have altered practice to some degree, but their findings remain controversial due to the inherent limitations of meta-analyses. The SAFE study, the largest and best-designed trial to date on this subject, finally provides more certain support for the assertion that albumin administration is equivalent to NS for fluid resuscitation for general ICU patients. I would extend this further to state that because of albumin’s lack of benefit, associated increased RBC transfusion in this study, up to 30 to 40 fold higher cost and greatly limited supply (as a human blood product), it should simply not be used for fluid resuscitation for general ICU patients at this time. I suggest that its use be restricted to SBP (as defined by Sort et al,1 until further investigated) and to large well-designed controlled trials for other specific indications. In addition, I wonder whether other products for volume replacement such as hypertonic saline with and without dextran, new starch products and novel blood substitutes may ultimately make albumin obsolete.


1. Sort P, et al. N Engl J Med. 1999:341(6):403-409.

2. Cochrane Injuries Group Albumin Reviewers. BMJ. 1998:317(7153):235-240.

3. Ferguson ND, et al. Intensive Care Med. 1999: 25(3):323-325.

4. Wilkes MM, et al. Ann Intern Med. 2001:135(3):149-164.

Saadia R. Akhtar, MD, MSc, Pulmonary and Critical Care Medicine, Yale University School of Medicine, is Associate Editor of Critical Care Alert.