A Systematic Review of the Comparative Safety of Colloids

Hypothesis  Safety differences exist among colloids widely used for fluid management in acutely ill patients, as judged according to the comparative incidence of adverse events.

Data Sources  Colloid safety data for human subjects were sought, without language or time period restrictions, by means of computer searches of bibliographic and clinical trial databases, hand searches of medical journals and Index Medicus, inquiries with investigators and colloid suppliers, and examination of reference lists. Search terms included "colloids", "morbidity", and "mortality".

Study Selection  Controlled trials, cohort studies, pharmacovigilance studies, and prior meta-analyses were independently selected by 2 unblinded investigators. Of 189 candidate studies, 113 were included, with safety data encompassing 1.54 × 10⁶ patients and 1.09 × 10⁸ colloid infusions.

Data Extraction  Two unblinded investigators independently extracted data. Study limitations and confounding factors were tabulated.

Data Synthesis  With albumin as the reference colloid, the incidence rate ratio for anaphylactoid reactions was 4.51 (95% confidence interval, 2.06-9.89) after hydroxyethyl starch administration, 2.32 (95% confidence interval, 1.21-4.45) after dextran, and 12.4 (95% confidence interval, 6.40-24.0) after gelatin. Pruritus occurrence was significantly increased by hydroxyethyl starch exposure (odds ratio, 1.78; 95% confidence interval, 1.23-2.58). Artificial colloid administration was consistently associated with coagulopathy and clinical bleeding, most frequently in cardiac surgery patients receiving hydroxyethyl starch. On the basis of large-scale pharmacovigilance study results, albumin infusion resulted in a low rate of both total adverse events (3.1 to 8.6 per 10⁵ infusions) and serious adverse events (1.29 per 10⁶ infusions).

Conclusions  Significant safety differences exist among colloids. Therefore, conclusions regarding the clinical usefulness of colloids as a fluid class should be formed with caution.

Colloids promote retention of fluid in the intravascular space, with concomitant reduction of the potential for edema that might compromise the function of organs such as the lungs, myocardium, and gastrointestinal tract.¹ The chief colloids currently in routine clinical use worldwide are albumin, hydroxyethyl starch (HES), dextran, and gelatin. Clinically available albumin is a 69-kDa protein purified from human plasma. Hydroxyethyl starch is synthesized by partial hydrolysis of amylopectin plant starch and hydroxyethylation at the C2, C3, and C6 positions of the constituent glucose molecules. Dextran is composed of naturally occurring glucose polymers synthesized by Leuconostoc mesenteroides bacteria growing in sucrose-containing media. Gelatin for clinical use is derived from hydrolysis of bovine collagen followed by being either succinylated or linked to urea.

All 3 artificial colloids are polydisperse molecules in a range of sizes. Hydroxyethyl starch is clinically available in an array of forms differing on the basis both of average molecular weight and extent of molar substitution, although in the United States only the HES of high molecular weight (450 kDa) of 0.7 molar substitution ratio has been used in routine fluid management. In the United States, dextran is less extensively used for fluid management than is HES, and gelatin is unavailable for clinical use.

Clinically available colloids have generally exhibited similar effectiveness in maintaining colloid oncotic pressure. Thus, colloids have often been viewed as a class of essentially interchangeable inert fluids, and selection of colloid has commonly been based on cost and convenience. Nevertheless, differences in safety profiles among colloids are well recognized.^2,3 Such differences underlie, for example, the recommended 1500 mL (20 mL per kilogram of body weight) dose limitation for HES.⁴

The clinical importance of differences in colloid safety has been debated. Firm conclusions have been difficult to draw, in part because comparative colloid safety has not been systematically reviewed. We here present the results of such a review.

We systematically sought all studies of acutely ill patients with data on the safety of the natural colloid albumin and the artificial colloids HES, dextran, and gelatin. Randomized controlled trials (RCTs), nonrandomized controlled trials (NCTs), cohort studies, pharmacovigilance studies (PVSs), and meta-analyses (MAs) were eligible for inclusion.

Clinical studies fulfilling the selection criteria were identified, without language or time period restrictions, by computer searches of the MEDLINE and EMBASE bibliographic databases, the Cochrane Controlled Trials Register, and the Cochrane Medical Editors Trial Amnesty of unpublished trials. Search terms included "colloids", "morbidity", and "mortality". Hand searches were conducted of general medical journals and Index Medicus. We contacted the authors of published clinical studies related to colloids and the medical directors of colloid suppliers and examined the reference citations from completed reviews and protocols in the Cochrane Database of Systematic Reviews, other MAs, review articles, and controlled and uncontrolled studies involving colloids.

Two unblinded investigators (M.M.W. and R.J.N.) independently selected studies for inclusion and extracted data about study design, numbers of patients enrolled and/or infusions administered, clinical setting, fluid regimen, and major study findings, as well as study limitations and confounding factors. Differences in interpretation were resolved through discussion.

Study results were generally assessed qualitatively. However, quantitative MAs were performed of data for 2 end points: anaphylactoid reactions and HES-associated pruritus. The meta-analytic methodology was generalized mixed modeling with study-level random effects. Such models are intended to accommodate expected between-study heterogeneity.

The incidence rate ratio of anaphylactoid reactions with individual colloids, as compared with albumin as reference standard, was calculated together with the corresponding confidence interval (CI) by means of random-effects Poisson regression. Pruritus associated with HES was modeled by means of random-effects logistic regression. Results were expressed as the odds ratio for occurrence of HES-associated pruritus. For both incidence rate ratio and odds ratio, the absence of the number 1 from the CI signifies a statistically significant effect.

Of 189 candidate studies initially identified, 113 studies published from 1944 through 2002 were included.^1-5556-113 One RCT was excluded because of fluid overload in the albumin group.¹¹⁴ Numbers of study patients, which were reported in 107 of the included studies, totaled 1.54 × 10⁶ patients. The median number of patients per study was 60, with an interquartile range of 29 to 200. In the remaining 6 studies, the numbers of infusions were reported, and these totaled 1.09 × 10⁸ infusions. The median number of infusions per study was 8.5 × 10⁵ (interquartile range, 1.20-74.0 × 10⁵). Safety data about albumin, HES, dextran, and gelatin were available from 60, 75, 17, and 25 included studies, respectively.

Twenty-one of the included studies involved acute illness generally (Table 1), 35 cardiac surgery (Table 2), 19 noncardiac surgery (Table 3), 5 ascites, 4 sepsis, 13 brain injury, 3 dialysis, 6 plasma exchange, and 7 acute hearing loss. (Tables summarizing studies of ascites, sepsis, brain injury, dialysis, plasma exchange, and acute hearing loss are available from the authors.) Of the cardiac surgery studies, 14 were evaluations of extracorporeal circuit pump priming; 19, volume expansion; and 2, both.

The most frequently represented study design was the RCT, which accounted for 54 studies. Twenty-eight studies were NCTs, 22 were cohort studies, 6 were PVSs, and 3 were MAs.

In large-scale PVSs, the reported incidence for adverse events of any severity in albumin recipients was 6.1 to 6.8 per 10⁵ infusions of 5% albumin and 3.1 to 8.6 per 10⁵ infusions of 20% to 25% albumin.^7,28 For serious adverse events, an incidence of 1.29 per 10⁶ infusions was reported.¹¹¹ The PVSs are generally based on spontaneous adverse event reporting and are subject to underreporting. One of these PVSs also included data for gelatin, and the reported incidence of adverse events was similar to that for albumin.⁷ In a cohort study of 379 patients, the incidence of all HES-associated adverse effects was 4.5%.⁴²

One MA of RCTs indicated poorer survival in critically ill patients receiving albumin vs crystalloid or no albumin.⁸⁵ However, authors of a subsequent MA¹¹⁰ considered RCT evidence approximately 3-fold more extensive than that of the first MA, and there was no evidence of increased albumin-associated mortality. Results of higher quality trials suggested a potential survival benefit of albumin.¹¹⁰ Thus, in a multivariate analysis of blinded larger RCTs, mortality was significantly reduced by albumin (odds ratio, 0.78; 95% CI, 0.76-0.81). A large-scale PVS provided evidence that deaths after albumin administration are rare (5.24 per 10⁸ infusions).¹¹¹ Hemodilution with HES was investigated in 1 RCT of patients with acute ischemic stroke.³² The trial was stopped prematurely because of a significant increase in mortality related to cerebral edema among HES recipients.

In 9 studies, data were reported on anaphylactoid reactions after 3.63 × 10⁶ total colloid infusions.^{2,5,6,8,16,26,57,75,84} The pooled incidence of anaphylactoid reactions after albumin administration was 9.44 per 10⁵ infusions (95% CI, 5.04-17.7 per 10⁵ infusions). Infusions of all 3 artificial colloids, as compared with albumin, were associated with significantly increased anaphylactoid reactions ( Table 4).

In 1 study, there was evidence of dextran-associated pruritus in some patients.⁴⁵ Otherwise, however, reports of this adverse effect were restricted to HES exclusively. Pruritus associated with HES was reported in 14 studies involving a total of 2598 patients, of whom 2173 (83.6%) received HES and 425 (16.4%) did not.^{45,52,53,55,58,59,76,84,86,95,96,99,105,106} The odds of pruritus were significantly increased by HES exposure ( Table 5). The effect of HES on pruritus occurrence depended on dose. Neither HES molecular weight nor HES molar substitution exerted a statistically significant effect on pruritus.

In 1 study of patients receiving intensive care, 44% of the patients developing pruritus experienced a severe, persistent, and refractory form of the condition.⁹⁵ Pruritus associated with HES was typically delayed in onset and manifested as pruritic crises,^52,99,105 prompting patients to seek medical attention and seriously detracting from their quality of life.⁹⁵ Pruritus associated with HES is generally unresponsive to currently available forms of therapy.⁵²

Results of numerous studies indicate that HES administration can lead to reduction in circulating factor VIII and von Willebrand factor levels, impairment of platelet function, prolongation of partial thromboplastin time and activated partial thromboplastin time, and increase in bleeding complications.^{27,29,34,40,61,62,69,78,80,82,94,103,104} Coagulopathy and hemorrhage associated with HES are often encountered in cardiac surgery, a setting in which susceptibility to such complications is heightened by transient acquired platelet dysfunction resulting from the procedure. Thus, in cardiac surgery studies with albumin as the control, HES has resulted in platelet depletion and dysfunction, prothrombin time and activated partial thromboplastin time prolongation, and increased postoperative bleeding.^{3,11,13,15,17,46,48,50,65,92} One NCT of 444 patients revealed significant increases in blood product use, as well as postoperative blood loss, in patients receiving HES for volume expansion.⁹⁸ The effects of HES on clinical bleeding in cardiac surgery patients depend on dose; however, excessive postoperative bleeding has been reported with HES doses less than the recommended maximum.⁸⁷

Results of 1 RCT suggested that postoperative bleeding might be greater in patients receiving HES of high rather than medium molecular weight.⁵⁰ This result was not supported, however, by results of 1 NCT of 200 patients showing postoperative blood loss significantly greater after pump priming with HES of medium molecular weight than with albumin.⁹² Furthermore, in 1 MA of RCTs, bleeding was increased by HES, as compared with albumin, and the effects of HES of high and medium molecular weight were similar.¹⁰⁹

Dextran, as compared with albumin, has been shown to reduce platelets and increase postoperative bleeding in cardiac surgery patients.⁵⁶ Postoperative blood loss was linearly correlated with the volume of gelatin used to prime the extracorporeal circuit.⁶⁰

All 3 artificial colloids have been associated with renal impairment, and HES has been demonstrated to increase sensitive markers of renal tubule damage in surgical patients.^74,93 In 1 RCT of sepsis patients, HES exposure was recently shown to be an independent risk factor for acute renal failure.¹⁰⁸

In the renal transplantation setting, HES reduced urinary output, increased creatinine levels and dopamine requirement, and increased the need for hemodialysis or hemodiafiltration.^64,66 By contrast, in 1 NCT, no significant difference was evident in delayed graft function after renal transplantation with administration of HES to the donor.⁸⁹

In a cohort study of patients with acute ischemic stroke, 4.7% experienced acute renal failure associated with dextran infusion.⁷² Gelatin, as compared with albumin as pump prime in cardiac surgery, elevated creatinine levels.⁴¹

The occurrence of circulatory dysfunction marked by increased plasma renin activity and aldosterone has been investigated in patients who have ascites and are undergoing large-volume paracentesis. The incidence of circulatory dysfunction was significantly higher after infusion of dextran than of albumin in 2 RCTs^38,67 but not in a third.⁴⁷ Circulatory dysfunction was also more frequent in patients receiving gelatin than in those receiving albumin.⁶⁷

Repeated infusion of HES in conjunction with dialysis resulted in the development of ascites that necessitated Denver shunt implantation in 1 case.¹⁸ In 1 recent study, HES deposition in hepatic Kupffer cells was associated with worsening of hepatic dysfunction after HES infusion.¹⁰¹ Pump priming with HES, as compared with albumin, during cardiac surgery has been found to increase levels of liver enzymes during and after cardiopulmonary bypass surgery.¹⁵

Hydroxyethyl starch is deposited in a variety of tissues, including skin, liver, muscle, spleen, intestine, trophoblast, and placental stroma.^{18,40,55,96,101} Such deposition often has been described in association with pruritus.^55,96 Tissue deposits persist as long as 54 months after HES administration.⁹⁶ Organ deposition of dextran has been reported in 1 NCT of patients undergoing long-term hemodialysis.³⁵ Tissue deposition of administered albumin has not been detected at necroscopic examination.⁵

The major conclusion emerging from this systematic review is that there are clinically important differences in safety among colloids. Many of the differences have been demonstrated between albumin and HES, possibly because these 2 colloids have been more extensively investigated than have dextran and gelatin.

The incidence of adverse events in albumin recipients was low. Albumin administration was not consistently associated with any characteristic types of adverse events. This observation is perhaps unsurprising, because albumin infusion serves to replenish the normal endogenous colloid. Although albumin is isolated from human plasma, we could identify no evidence of viral disease transmission attributable to albumin.

Bleeding associated with artificial colloid administration has been widely reported.^{3,21,27,40,49,50,56,60,62,73,87,92,98,104,109} Such complications have been particularly frequent in the cardiac surgery setting, necessitating increased blood product use and increased costs of care.^87,98 Bleeding complications can be particularly troublesome because of the long half-life of HES and because discontinuation of HES infusion cannot immediately resolve coagulopathy.

Although potentially life threatening, anaphylactoid reactions were relatively infrequent for all colloids. Hydroxyethyl starch, as compared with albumin, more than quadrupled the incidence of anaphylactoid reactions, whereas dextran more than doubled them. The incidence of these reactions in recipients of gelatin was greater by more than an order of magnitude than that after albumin infusion. Because artificial colloids are derived from nonhuman source materials, they may be recognized as foreign and hence are more likely to provoke an immune-mediated response. The foreign nature of artificial colloids may also hinder metabolic clearance and promote tissue deposition.

Although HES has been widely used for several decades, HES-related pruritus has been widely described only since the early 1990s.⁴⁵ Its late recognition as a clinical entity appears to be at least partly because of the lengthy delay in onset of symptoms, in many cases occurring after discharge. This adverse effect was initially characterized in otologic patients receiving relatively high HES doses to improve microcirculation. Besides otologic indications, we identified evidence of HES-associated pruritus in studies of general⁹⁶ and cardiac⁹⁹ surgery, patients receiving intensive care,^95,106 and subarachnoid hemorrhage.¹⁰⁵ In our MA, the pooled odds of pruritus were significantly increased by HES exposure. Although the effect was dose related, many patients receiving less than the recommended 20 mL/kg HES maximum were affected.^95,99 In a cardiac surgery study, more than half of the patients developing pruritus had received less than the recommended HES maximum.⁹⁹

We compiled substantial evidence linking HES exposure to increased risk of renal failure. This effect has been characterized in the settings of surgery,^74,93 sepsis,¹⁰⁸ and kidney transplantation.^64,66 Adverse effects of HES in kidney transplantation have, however, been controversial.⁸⁹

Outside the United States, HES products of varying molecular weight and/or molar substitution have been introduced. It has often been argued that adverse effects of HES are primarily attributable to preparations of higher molecular weight and greater molar substitution.^61,70 The basis for such putative differences is the more rapid clearance of HES of lower molecular weight and less highly substituted forms. In our MA, neither HES molecular weight nor molar substitution significantly affected the odds of pruritus. More broadly, our systematic review failed to reveal consistent differences among HES preparations with respect to safety.

On the basis of extensive evidence, albumin appears to be in general the safest colloid of the 4 we reviewed. In some settings, other factors such as the desirability of anticoagulant activity might militate in favor of artificial colloids. In any case, results of our review suggest the need to consider the contrasting safety profiles of colloids in clinical decision making.

Accepted for publication August 13, 2003.

Corresponding author and reprints: Mahlon M. Wilkes, PhD, Hygeia Associates, 17988 Brewer Rd, Grass Valley, CA 95949, (e-mail: mwilkes@hygeiaassociates.com).