This systematic review of randomized clinical trials was undertaken to determine whether hyperoncotic albumin consistently differs from control regimens in its effects upon clinically relevant endpoints such as morbidity, major organ function, length of stay and cost of care in acutely ill patients. Such endpoints are often not defined, assessed and reported in a consistent and standardized manner, so qualitative summarization is more appropriate than quantitative combination of results across trials. A secondary objective was to evaluate the effect of hyperoncotic albumin on survival by quantitative meta-analysis.

All randomized clinical trials comparing hyperoncotic albumin with a control regimen for volume expansion in acutely ill patients were eligible for inclusion. Both parallel-group and crossover study designs were acceptable. Randomized trials focusing on other uses of albumin such as treatment of hyperbilirubinemia, extracorporeal albumin dialysis or prevention of ovarian hyperstimulation syndrome were excluded. Also excluded were randomized trials of albumin as an adjunct to paracentesis or for correction of hypoalbuminemia, both of which have been evaluated in previous systematic reviews 2829.

Randomized clinical trials fulfilling the inclusion criteria for the systematic review were identified by multiple methods, including computer searches of the Medline and EMBASE bibliographic databases and the Cochrane Library. Search terms included hyperoncotic albumin, resuscitation, hypovolemia, surgery, trauma, sepsis, liver diseases, intensive care, neonatal, brain injuries, nephrotic syndrome and randomized controlled trials. Studies were also sought through examination of reference lists and manual searching of Index Medicus and specialty journals. No restrictions on time period or language of publication were applied.

Data on numbers of patients randomized, clinical indication, hyperoncotic albumin and control regimen, major findings and deaths were extracted from the randomized clinical trial reports and used to populate a relational database. The participating investigators, time periods and study subject characteristics were closely examined to avoid inclusion of duplicate datasets appearing in multiple publications and to ensure the most complete possible dataset. Deaths were recorded on an intent-to-treat basis. Unpublished survival data and clarification regarding study design features were sought as needed via direct inquiries with the randomized trial investigators.

Trial quality was evaluated by the criteria of blinding and allocation concealment. On the basis of the procedures employed in the study, allocation concealment was classified as adequate, inadequate or unclear 32.

R version 2.4.1 (The R Foundation for Statistical Computing, Vienna, Austria), Stata 9.1 (Stata Corp., College Station, TX, USA) and SPSS 11.5 (SPSS Inc., Chicago, IL, USA) statistical software was used for analyses. The attributes of the included trials were summarized by descriptive statistics, i.e., the median and interquartile range (IQR).

The primary endpoint for the quantitative meta-analysis of survival was the relative risk (RR) of death. The 95% confidence interval (CI) of RR was also calculated. RR can only be calculated when at least one death has been observed; hence, trials with no deaths did not enter the survival meta-analysis. Deaths in crossover trials were also excluded from the meta-analysis, since in such trials the same patients are exposed to both the albumin and control regimens and the attributability of death to one regimen or the other is uncertain. Pooled RR was estimated under a fixed-effects model. Publication bias was assessed by the method of Egger et al. 34. The statistical power of the meta-analysis was calculated as 1 - Φ (C_α/2 - λ) + Φ (-C_α/2 - λ), where Φ = the standard normal cumulative distribution function, C_α/2 = the standard normal critical value for a two-sided test at the 0.05 α level, and λ = (ln(pooled RR) - 1)/(pooled standard error) 34.

A total of 25 randomized clinical trials 35363738394041424344454647484950515253545556575859 with an aggregate of 1,485 patients were included in the review (Figure 1). The median number of patients per trial was 30 (IQR, 18–58). In all, 21 of the trials (84%) had been published from 1990 onwards. Parallel groups were compared in 21 trials, and a crossover design was employed in the remaining 4. The control regimen consisted of another colloid, predominantly hydroxyethyl starch (HES), in nine trials (36%), crystalloid in four (16%), no albumin in six (24%) and lower-dose albumin in one (4%). Two separate control regimens composed of another colloid and either crystalloid or no albumin were administered in five trials (20%).

In five trials, the effects of hyperoncotic albumin were separately investigated among patients with either trauma or sepsis 4445464849. Randomized comparisons involving surgery and trauma are summarized in Table 1, sepsis in Table 2 and other indications in Table 3. The other indications were comprised of liver disease, high-risk neonates, brain injury, intradialytic hypotension and nephrotic syndrome. The trials of high-risk neonates involved premature infants with risk factors such as low birth weight, respiratory dysfunction or brain injury.

Evaluated study endpoints included survival in three trials 405153, morbidity in six 354043535658, major organ function in six 414249565859, length of stay in two 5156, and costs of care in one 51. Among additional endpoints assessed were hemodynamics in seven trials 45484952545559, major organ edema in two 3956, whole body edema in two 3842, coagulation function in four 39444649, colloid oncotic pressure (COP) in two 3637, diuretic responsiveness in two 5057, and inflammatory markers in one 47.

Four trials (16%) were blinded 41505356, and the remainder unblinded. Allocation concealment was adequate in four trials (16%) 50515358 and inadequate or unclear in the rest.

Based on randomized clinical trial data from surgical patients summarized in Table 1, hyperoncotic albumin more effectively maintained colloid oncotic pressure than control regimens in both cardiac 37 and non-cardiac surgery 36. Compared with HES, hyperoncotic albumin better preserved renal function 41 and coagulation 39. A greater increase in cardiac output was demonstrated after administration of hyperoncotic albumin than saline to the same right atrial pressure target 52. Hyperoncotic albumin was superior to both HES and crystalloid in preventing intestinal edema during abdominal surgery 39.

In contrast, among the five randomized comparisons in trauma described by Boldt et al. (Table 1) no apparent clinical advantages of hyperoncotic albumin over HES were apparent in three reports 444648. In the other two reports 4549, cardiac index and oxygenation were increased by HES relative to albumin.

Similarly, in sepsis (Table 2) higher cardiac index and oxygenation after HES than albumin administration were observed in several trials 454959. Additionally, HES maintained higher gastric intramucosal pH than hyperoncotic albumin in two trials 4548 and improved Acute Physiology and Chronic Health Evaluation II score in one 59.

In cirrhotic patients with refractory ascites (Table 3), hyperoncotic albumin increased the treatment response rate, shortened hospital stay and reduced costs of care 51. In patients developing spontaneous bacterial peritonitis, hyperoncotic albumin diminished the incidence of renal impairment 53. Unlike HES 200/0.5 (molecular weight/molar substitution ratio), hyperoncotic albumin improved circulatory function of patients with spontaneous bacterial peritonitis 58.

In randomized trials of high-risk premature infants (Table 3), hyperoncotic albumin reduced the frequency of illnesses, lessened whole body edema and improved respiratory function 3542. Apgar scores were higher, the frequency of cerebral edema lower and hospital stay shorter after hyperoncotic albumin administration in newborns with asphyxia and brain edema 56.

Hyperoncotic albumin reduced disability at 3 months in patients with acute ischemic stroke and normal hematocrit (Table 3) 40. Use of hyperoncotic albumin to maintain high oncotic pressure also led to more favorable outcomes in a randomized trial of patients with closed head injury 43. All patients receiving high-oncotic-pressure therapy recovered with minimal or no neurological deficit, whereas 30% of the control group remained in a vegetative state or died.

In two randomized crossover trials (Table 3), hyperoncotic albumin was more effective than either saline 54 or hypertonic saline 55 in averting blood volume declines. Systolic hypotension was also reduced by hyperoncotic albumin compared with saline 55.

Hyperoncotic albumin was investigated in three randomized trials of nephrotic syndrome (Table 3). Hyperoncotic albumin with concomitant furosemide accelerated diuresis and promoted body weight loss 38. In addition, hyperoncotic albumin potentiated both the diuretic and natriuretic effects of furosemide 5057.

Mortality data were available for 24 randomized comparisons. The median duration of follow-up was 5 days (IQR, 5 to 13 days). Three trials 405153 were specifically designed to assess survival as a primary study endpoint. No patient died in 4 of the 24 comparisons 37394156. The remaining 20 comparisons with at least one death were included in a quantitative meta-analysis of survival (Figure 2). Total mortality in the 20 comparisons (Figure 2) was 266/1,287 (21%). In the meta-analysis, there was no evidence of either heterogeneity (p = 0.86) or publication bias (p = 0.87). The observed statistical power of the meta-analysis was 98% to detect a 35% reduction in relative mortality risk by hyperoncotic albumin (RR 0.65) and 82% to demonstrate a 35% relative mortality risk increase (RR 1.35). No significant overall effect of hyperoncotic albumin on survival was detected (RR 0.95; CI 0.78 to 1.17).

The included randomized trials addressed diverse clinical indications for hyperoncotic albumin, and the number of trials focusing on any single indication was, in most instances, relatively small. Within any particular indication, furthermore, no more than two trials evaluated clinically relevant endpoints of the same type. Consequently, firm conclusions about clinical benefit in defined indications are difficult to draw. Another limitation was the variety of control regimens. Treatment effects of hyperoncotic albumin might differ, for instance, in comparison to crystalloid vs colloid control fluid.

The available data in the categories of trauma and of sepsis were predominantly furnished by a single group of investigators, and 10 of the 11 randomized comparisons from this group were performed using the same trial design (Tables 1 and 2). The generalizability of those results is uncertain, and further trials are needed in those categories.

In the survival meta-analysis, the short median follow-up of 5 days may account partly or entirely for the lack of overall effect. Follow-up of 30 days or longer is typical for outcomes trials. Even at 30 days survival, differences may be substantially underestimated compared with those observed at 90 days 60. In all three included trials designed to assess survival, the follow-up period was at least 90 days 405153.

As in previous systematic reviews of albumin administration 2627, one of the control regimens represented among the trials in this review was no albumin, accounting for six of the included trials (24%). No albumin as the designated control regimen did not necessarily mean no volume expansion, however. In one surgery trial 36 the no-albumin control group received a cumulative mean volume of 24 l non-colloid fluids over the course of the 5-day study. In two trials of high-risk neonates, the no albumin control regimen consisted of routine fluid 56 or 5 ml/kg infant maintenance fluid 42. No albumin control group volume expansion, if any, in three other trials was unspecified 355153. If control patients in any of those three trials did not receive volume expanders, then observed differences between the hyperoncotic albumin and control arms might have reflected volume expansion rather than a specific effect of hyperoncotic albumin per se. In that scenario, it is possible that an alternative fluid could exert effects similar to those observed with hyperoncotic albumin.

These limitations highlight the need for further trials. There is consistent evidence that hyperoncotic albumin has the capacity to reduce edema; however, the impact of edema reduction on outcomes requires further investigation. Short follow-up has been a deficiency of most hyperoncotic albumin trials to date, and future trials with longer follow-up are essential to assess outcomes. Especially in the areas of surgery, trauma and sepsis existing data need to be supplemented by additional trials. Lastly, standardized and well-specified control regimens will be important in future trials so that clearer inferences can be drawn concerning the specific effects of hyperoncotic albumin vs those of particular alternative fluids or generic volume expansion.

A small volume can be administered more rapidly, and so the period during which the patient remains at increased risk of poor outcome can be commensurately reduced. In the pre-hospital setting, hyperoncotic albumin could accelerate hemodynamic stabilization and transport to hospital. Another advantage in the pre-hospital setting is the portability of small hyperoncotic albumin volumes.

In surgical patients, small-volume hyperoncotic albumin resuscitation may simplify fluid management by maintaining a stable hemodynamic state. In a retrospective study of 28 patients undergoing pelvic exenterations for gynecological malignancies, recipients of 25% albumin required fewer boluses of fluid (p < 0.01), electrolyte (p < 0.01) and diuretic (p < 0.01) and could be started sooner on central hyperalimentation (p < 0.05) than the group receiving crystalloid 61.

Speed of resuscitation may also be of the essence in rapidly progressing conditions such as spontaneous bacterial peritonitis 53 and in the urgent care of high-risk neonates 20. In pre-term infants with respiratory distress syndrome, intravenous administration of 1 g/kg 25% albumin over a 10 min period increased blood volume (p < 0.0005) and mean arterial blood pressure (p < 0.05) within 10 min after infusion was completed 20. Albumin also significantly augmented glomerular filtration in that study, as indicated by increased creatinine clearance (p < 0.005).

Formation of edema serves as a harbinger of organ failure 62 and may, depending upon the tissues affected, lead to impairments in gas exchange, myocardial compliance, neurocognitive function, gut barrier competence, nutrient absorption and wound healing. Pulmonary edema, for instance, is associated with prolonged ventilator dependence and intensive care unit stay 63. Furthermore, positive fluid balance appears to be predictive of poorer survival in sepsis 64. Encephalopathy may occur at least as frequently as other forms of organ damage during sepsis 65. Contributors to septic encephalopathy are reduced cerebral blood flow and oxygen extraction by the brain and cerebral edema 65. While edema may be adequately tolerated by many patients, such as young serious trauma victims, more grave consequences may be faced by elderly or frail patients in whom edema may retard oxygen delivery to the lungs, myocardium and brain.

The impact of cerebral edema on outcome after subarachnoid hemorrhage (SAH) was investigated in 374 patients 66. In a multivariate analysis assessing 16 significant univariate predictors, cerebral edema demonstrated by computed tomography was an independent risk factor both for death (odds ratio 2.5; CI 1.1 to 5.6) and for either death or severe disability (odds ratio 2.5; CI 1.2 to 5.4). Severe disability was defined as a modified Rankin scale score of > 3. These observations support a causal role for cerebral edema in producing poor outcomes and prompted the investigators to conclude: 'Critical care management strategies that minimize edema formation after SAH may improve outcome' 66. Similarly, in a prospective study of 113 non-traumatic SAH patients evaluated by comprehensive neuropsychological testing at 3 months, cerebral edema was an independent risk factor for cognitive impairment 67.

Hyperoncotic albumin may be of particular clinical value in edematous states such as those encountered in liver disease, high-risk neonates, brain injury and nephrotic syndrome (Table 3). While edema may pose problems in sepsis as well, this review suggested advantages of HES over albumin in sepsis, e.g., higher cardiac index and oxygenation. However, follow-up in all the included sepsis trials was only 5 days. With follow-up of 30 days or longer HES has been shown in randomized trials to increase the incidence of acute renal failure among patients with severe sepsis or septic shock 6068. Hyperoncotic albumin displays protective effects on the kidney 4153 and hence would likely offer a safer fluid management alternative than HES in sepsis.

The randomized trial evidence showing the capacity of hyperoncotic albumin to reduce edema is supplemented by non-randomized clinical studies and animal models in the areas of high-risk neonates, brain injury and nephrotic syndrome. Among 10 normotensive premature infants with idiopathic respiratory distress syndrome, 20% albumin reduced body weight (p < 0.01), indicating lessened edema, and improved urine output (p < 0.05) within 6 h 69.

Decreased cerebral edema after hyperoncotic albumin administration was shown in one included randomized trial 56. Reduction in cerebral edema and other sequelae of brain injury by hyperoncotic albumin has also been repeatedly demonstrated in other studies. In a non-randomized trial of 22 patients with putaminal hemorrhage, 50 to 100 ml/day of 25% albumin with concomitant furosemide significantly reduced brain edema as assessed by computed tomography 70. No albumin recipient died or remained in a vegetative state compared with 27% of crystalloid recipients. In a cohort study of 10 patients with elevated intracranial pressure and brain edema, infusion of 2 g/kg 25% albumin over 60 min produced a significant and long-lasting (≥ 9 h) decline in intracerebral pressure 71. Among hemorrhagic stroke patients, rapid infusion of 50 ml 20% albumin promptly and significantly reversed electroencephalographic abnormalities 72. The randomized, double-blinded, placebo-controlled multi-center Albumin in Acute Stroke (ALIAS) clinical trial funded by the US National Institutes with a target enrollment of 1,800 patients is evaluating the neuroprotective effects of 25% albumin as compared with saline 73.

According to experimental cerebral ischemia studies, hyperoncotic albumin can reduce infarct volume and edema, augment cortical perfusion and improve functional outcome 7475. In head trauma models, hyperoncotic albumin increased neurological score and reduced brain tissue edema and damage 767778.

All three included randomized trials in nephrotic syndrome 385057 indicated the ability of hyperoncotic to promote diuresis, and one of these trials 38 also provided direct evidence of edema reduction. In a non-randomized study of 14 severely edematous children with minimal change nephrotic syndrome, 20% albumin infused at an albumin dose of 0.5 g/kg over 1 h followed by furosemide reduced pretibial edema and body weight within 1 h (p < 0.05) 79. These effects persisted for at least 24 h (p < 0.005).

Starling forces may be sufficient to explain the effectiveness of hyperoncotic albumin for small-volume resuscitation and edema reduction. However, additional mechanisms such as anti-inflammatory and antioxidant activity may contribute to the clinical benefits of hyperoncotic albumin. Furthermore, at least some of these mechanisms are specific to hyperoncotic rather than 4 to 5% albumin. Thus, in an in vitro study of blood specimens from 10 healthy adult human volunteers, dilution with HES 450/0.7 produced a dose-dependent increase in neutrophil activation to 19-fold the baseline level 80. The increase after dilution with various crystalloids was to 13- to 19-fold the baseline level and with 5% albumin twofold. In contrast, there was no evidence of neutrophil activation by 25% albumin.

The reduced thiol moiety on cysteine 34 of albumin can play a direct antioxidant role. In a randomized clinical trial of 20 patients with acute lung injury due to trauma, pneumonia, sepsis and other insults, repeated administration of 25% albumin elevated plasma thiol concentration (p = 0.0001) and total antioxidant capacity (p = 0.033) compared with saline placebo 81.

The capacity of hyperoncotic albumin to shift fluid from the interstitium to the intravascular space may be viewed as a two-edged sword, depending upon the hydration state of the patient. In edematous states, such fluid shifts may be desirable. Conversely, hyperoncotic albumin should probably be avoided in patients with severe dehydration 3.

High doses of hyperoncotic albumin could potentially increase pulmonary edema and capillary leak. In one widely cited randomized trial of Lucas et al. 82, patients receiving very high hyperoncotic albumin doses (1,142 g) and total fluid volumes (44.9 l) displayed impairment in pulmonary function. Commentators have ascribed that result to fluid overload rather than hyperoncotic albumin per se, however 8384. The trial of Lucas et al. was not included in the present review because albumin was administered with the aim of elevating serum albumin concentration rather than expanding intravascular volume, and indeed the attained central venous pressure (mean 18.6 cm H₂O) was supranormal.

In a dose-escalation study of 25% albumin treatment in 82 patients with acute ischemic stroke, mild or moderate pulmonary edema was observed in 13% of the patients 85. This adverse effect was readily managed with diuretics, and 25% albumin doses up to 2 g/kg could be tolerated without major dose-limiting complications.

None of the trials included in the present review or in a prior systematic review of both hyperoncotic and 4 to 5% albumin 28 has indicated harmful effects attributable to albumin. A Cochrane meta-analysis encompassing both 4 to 5% and 20 to 25% solutions suggested increased mortality among albumin recipients (RR 1.7) 26. That finding could not be replicated either in the SAFE trial 25 or a more comprehensive subsequent meta-analysis 27. In the present survival meta-analysis, which was adequately powered to detect a relative mortality risk increase only half as great as that reported by the Cochrane investigators, there was no evidence of poorer survival among hyperoncotic albumin recipients.