Potential conflict of interest: Dr. Navickis and Dr. Wilkes received grants from CSL Behring.
This investigation was supported through an unrestricted grant from CSL Behring, King of Prussia, PA.
Albumin infusion reduces the incidence of postparacentesis circulatory dysfunction among patients with cirrhosis and tense ascites, as compared with no treatment. Treatment alternatives to albumin, such as artificial colloids and vasoconstrictors, have been widely investigated. The aim of this meta-analysis was to determine whether morbidity and mortality differ between patients receiving albumin versus alternative treatments. The meta-analysis included randomized trials evaluating albumin infusion in patients with tense ascites. Primary endpoints were postparacentesis circulatory dysfunction, hyponatremia, and mortality. Eligible trials were sought by multiple methods, including computer searches of bibliographic and abstract databases and the Cochrane Library. Results were quantitatively combined under a fixed-effects model. Seventeen trials with 1,225 total patients were included. There was no evidence of heterogeneity or publication bias. Compared with alternative treatments, albumin reduced the incidence of postparacentesis circulatory dysfunction (odds ratio [OR], 0.39; 95% confidence interval [CI], 0.27-0.55). Significant reductions in that complication by albumin were also shown in subgroup analyses versus each of the other volume expanders tested (e.g., dextran, gelatin, hydroxyethyl starch, and hypertonic saline). The occurrence of hyponatremia was also decreased by albumin, compared with alternative treatments (OR, 0.58; 95% CI, 0.39-0.87). In addition, mortality was lower in patients receiving albumin than alternative treatments (OR, 0.64; 95% CI, 0.41-0.98). Conclusions: This meta-analysis provides evidence that albumin reduces morbidity and mortality among patients with tense ascites undergoing large-volume paracentesis, as compared with alternative treatments investigated thus far. (HEPATOLOGY 2012)
Ascites, the most common major complication of cirrhosis, is associated with a poor prognosis.1 As ascites progresses, abdominal distension becomes more marked, and in the advanced grades, the patient can experience major discomfort and impaired breathing, often necessitating hospitalization.2 Spontaneous bacterial peritonitis develops in 25%-30% of patients with cirrhosis and ascites.3 Therapeutic paracentesis has become the first line of treatment for patients with tense (i.e., grade 3) and refractory ascites.4-6 Large-volume paracentesis (LVP) is faster and less likely to exert unwanted side effects than diuretics.7, 8
One drawback of LVP without adjunctive treatment is the associated risk of postparacentesis circulatory dysfunction (PCD). Even before paracentesis, patients with ascites are subject to marked circulatory dysfunction, most notably splanchnic arteriolar vasodilation, hyperdynamic circulation, and decreased effective arterial blood volume. Accompanying activation of the renin-angiotensin-aldosterone system (RAAS) and sympathetic nervous system (SNS) and increase in antidiuretic hormone levels result in sodium and water retention and renal vasoconstriction.
Paracentesis substantially influences systemic hemodynamics.9 In the majority of patients not receiving adjunctive treatment, removal of large ascitic fluid volumes, by reducing intra-abdominal pressure, boosts venous return to the heart. As a result, right atrial pressure decreases and cardiac output and stroke volume increase. However, because of an excessive drop in peripheral vascular resistance, effective circulating volume further declines, leading to a reduction in arterial pressure.10-12 In the days after paracentesis, a pronounced reactivation of RAAS and SNS ensues that can persist for months. These events represent the pathophysiological background of PCD, which is usually defined as an increase in plasma renin activity of 50% or greater. This complication is associated with a high rate of ascites recurrence, development of hepatorenal syndrome, dilutional hyponatremia, and decrease in survival.13, 14
It was first demonstrated in the 1980s that adjunctive albumin infusion favorably influences circulatory function after LVP and prevents the subsequent reactivation of vasoconstrictor systems and occurrence of PCD.2 The value of adjunctive albumin treatment in this setting has been recognized in both European and American clinical practice guidelines, which recommend the administration of albumin when the volume of ascites removed during paracentesis exceeds 5 L.4-6
Since the early 1990s, less-costly alternatives to albumin have been sought, such as artificial colloid volume expanders and vasoconstrictors. Despite numerous randomized trials, it remains uncertain whether the effectiveness of such alternative treatments is comparable to that of albumin. This uncertainty, in part, reflects the limited size and statistical power of the randomized trials reported so far. Quantitatively combining results of all relevant trials by meta-analysis could help resolve this uncertainty. The aim of this meta-analysis was to determine the comparative effectiveness of albumin and alternative treatments in minimizing PCD, hyponatremia, and mortality among ascites patients undergoing LVP. The specific null hypothesis tested was that the effects of albumin infusion in cirrhotic patients with tense ascites undergoing LVP are equivalent to those of alternative treatments.
The primary endpoints of the meta-analysis were PCD, hyponatremia, and mortality. Secondary endpoints consisted of ascites recurrence, renal impairment, hepatic encephalopathy, portal hypertensive bleeding, and hospital readmission. A subsidiary aim was to quantify, as precisely as possible, the magnitude of albumin effects versus no treatment.
Randomized, controlled trials (RCTs) were eligible for inclusion if they specifically evaluated albumin in the acute treatment of cirrhotic patients with tense ascites undergoing LVP and furnished data for one or more primary endpoints of the meta-analysis. Eligible trials could compare LVP plus albumin with LVP plus an alternative treatment, such as another volume expander or a vasoconstrictor. Trials assessing the combined effect of LVP plus albumin versus a control drug, such as diuretic, or nonparacentesis procedure, such as shunting, were not eligible. Trials with ascites fluid withdrawal, concentration by ultrafiltration ex vivo, and reinfusion of the concentrate as the control treatment were excluded. To quantify the effect size of albumin, compared with no treatment, trials were also sought that evaluated LVP plus albumin versus LVP without adjunctive treatment. Both published and unpublished trials could be included, and no limitations were placed on trial size, time period, or language of reporting.
Eligible trials were identified by multiple methods, including computer searches of MEDLINE, EMBASE, the Cochrane Library, the ClinicalTrials.gov website, and the abstract databases from major meetings in hepatology. Full-text searches with the Google search engine were also performed. Search terms included the following: ascites; cirrhosis; paracentesis; albumin; colloid; dextran; gelatin; hydroxyethyl starch; HES; saline; crystalloid; vasoconstrictor; terlipressin; midodrine; norepinephrine; postparacentesis circulatory dysfunction; plasma renin activity; hyponatremia; mortality; survival; and RCTs. Roots and variants of the search terms were also sought. Reference lists in primary study publications and review articles were examined and online contents of specialty journals in hepatology were consulted.
Determinations of trial eligibility and extraction of data were performed independently by at least two investigators. In the case of differences in interpretation, consensus was reached through discussion. The investigators, time periods, patients, and methods were closely compared between candidate trial reports to avoid inclusion of redundant data and ensure the most complete possible data set. Data extracted from the trial reports included, as available, numbers of patients allocated to each study group, year of reporting, methods of randomization, allocation concealment, and blinding, duration of follow-up, age, gender, baseline serum albumin level, indication for treatment, refractoriness of ascites to treatment, Child-Pugh score, etiology of cirrhosis, baseline renal dysfunction, albumin and control treatment regimens, concomitant diuretics, type of paracentesis and volume of ascites fluid removed, study definitions of PCD, hyponatremia, and renal impairment, day of PCD assessment, numbers of patients developing PCD, hyponatremia, renal impairment, hepatic encephalopathy, or portal hypertensive bleeding, experiencing ascites recurrence, or requiring hospital readmission, mortality, and numbers of patients at risk for each outcome. Individual patient and summary data reported in the form of graphical displays were captured by computer digitization.
Heterogeneity was assessed by the Cochran Q test and the I2 statistic. Publication bias was evaluated by linear regression of standardized effect versus precision.15 In the analysis of publication bias, precision was defined as the inverse of the standard error. Upon confirmation that significant heterogeneity was absent, trials were combined under a fixed-effects model, and the pooled odds ratio (OR) and its 95% confidence interval (CI) were computed. The significance of effect size differences between subgroups was determined by meta-regression. An α level of 0.05 was adopted as the criterion for statistical significance. Subgroup analyses were performed on the effects of trial quality, as judged by the randomization method, allocation concealment, and blinding.16 All analyses were performed using R version 2.13.0 (The R Foundation for Statistical Computing, Vienna, Austria) statistical software.
The process of RCT selection for the meta-analysis is outlined in Fig. 1. After the screening of 126 candidate clinical study reports, 41 reports were examined in detail. The most common reasons for the exclusion of reports at that stage were ineligible control treatment (e.g., concentrated ascitic fluid) or a study design that did not specifically test the effect of albumin on outcome (e.g., paracentesis + albumin versus diuretics or shunting). Seventeen candidate RCTs reported from 1988 to 2010 with 1,225 total patients met all selection criteria and were included in the meta-analysis.2, 17-32 None was unpublished. The median number of patients per trial was 54, with an interquartile range of 40-88. In three trials, more than 100 patients each were enrolled (Table 1).
The mean age of the study population ranged from 46.9 to 61.4 years. The pooled mean proportion of male patients enrolled was 73.6% (range, 60.0%-90.0%). In 11 of the 17 RCTs, mean baseline serum albumin ranged from 26 to 29 g·L−1 (Table 1). The respective minium and maximum values for the remaining RCTs were 20.9 and 31.9 g·L−1. In two trials, ascites was characterized as refractory.18, 32
Renal dysfunction and gastrointestinal bleeding were study-exclusion criteria in all 17 trials. Nonetheless, in six trials, at least some degree of renal dysfunction was present at baseline in 9.7%-35.1% of patients. Other common patient-exclusion criteria were infection, including spontaneous bacterial peritonitis and sepsis, in 16 trials, liver cancer in 14, hepatic encephalopathy in 13, low prothrombin activity percentage (<25%-40%) or prolonged prothrombin time in 11, thrombocytopenia (<30,000-50,000 mm−3) in 10, marked hyperbilirubinemia (>5-10 mg·dL−1) in eight, and severe hyponatremia (<120-125 mEq·L−1) in four.
Mean Child-Pugh score was reported for 11 trials, and values ranged from 8.8 to 11.0. Alcoholism was the etiology of cirrhosis for a pooled mean 71.3% of patients enrolled in the included trials (range, 38.9%-94.1%).
Random numbers from a table or computer-generated sequence were used for randomization in 10 trials. Method of randomization was not indicated for the remaining seven trials. Concealment of allocation to randomized groups was adequate for five trials and was unspecified for 12. Both the physician and patient were blinded in two trials and the laboratory in one. Blinding procedures, if any, were not described for any of the remaining trials.
Diuretics were discontinued before study treatment in 16 trials, whereas in the remaining trial, the prestudy diuretic regimen could be continued without change. Total paracentesis was performed in 13 trials and repeated LVP in three. Type of paracentesis was unspecified for one trial. The mean volume of ascites fluid removed during paracentesis ranged from 5.5 to 15.9 L (Table 1).
In 12 trials, albumin was administered at a dose of 8 g per L ascites fluid removed. The dose in two trials was 6 g·L−1, and in one trial each 5 g·L−1, 10 g·L−1 and, depending on ascites volume removed, 7-10 g·L−1. The concentration of albumin infused was 20% in 12 trials, 25% in one trial, and unspecified in four trials.
The control group received no treatment in three trials, an artificial colloid or hypertonic saline in nine trials, or a vasoconstrictor in five. The nine trials comparing albumin with other volume expanders formed the predominant category, accounting for 74% (903 of 1,225) of all patients in the meta-analysis, compared with 13% each for trials with no treatment or vascoconstrictor as the control regimen. Artificial colloids were the control volume expanders in eight of nine trials, namely, dextran in three trials, gelatin in two trials, either of those two artificial colloids in one trial, and hydroxyethyl starch in two trials. In only one trial did hypertonic (3.5%) saline serve as the control volume expander. Vasoconstrictors evaluated in the control group were terlipressin and midodrine in two trials each and norepinephrine in one trial.
Length of follow-up was reported for eight trials. Median follow-up for those eight trials was 76 days (range, 6-426).
Data for the PCD endpoint were available from all the included trials except one.27 Across all 16 included trials with PCD data, there was no significant heterogeneity (P = 0.31; I2, 12.8%) or detectable publication bias (P = 0.87).
Among the three trials comparing albumin with no treatment,2, 20, 21 31 of the 44 untreated control patients (72.7%) developed PCD, compared with 7 of 41 albumin recipients (17.1%). In these trials, albumin reduced the odds of PCD by 93% (pooled OR, 0.07; CI, 0.02-0.22). The OR for each of the three individual trials were remarkably consistent: 0.07 (CI, 0.02-0.28),2 0.07 (CI, 0.00-1.50),20 and 0.08 (CI, 0.01-0.75).21
Thirteen trials comparing albumin with alternative treatments provided PCD data (Fig. 2). Across those trials, 148 of 471 patients receiving other treatments (31.4%) developed PCD, compared with 58 of 386 albumin-treated patients (15.0%). Albumin reduced the odds of PCD by 61% versus alternative treatments. In subgroup analyses, the reductions in odds of PCD by albumin were comparable across strata defined by age, refractory ascites, baseline serum albumin level, type of control volume expander, day of PCD assessment, paracentesis regimen, ascites volume removed, trial quality, control group mortality, and time period (Table 2).
Table 2. Subgroup Analyses of Albumin Effects Versus Alternative Treatments
The effects of albumin versus other volume expanders were highly consistent, with OR for all eight trials in this subgroup falling within the comparatively tight range of 0.17-0.80 (Fig. 2). Albumin significantly reduced the odds of PCD, in comparison with each of the four other volume expanders evaluated, and the magnitudes of the OR were similar for all four comparisons (Table 2).
All five trials in the subgroup comparing albumin with vasoconstrictor were relatively small, with the total enrolled population in no cases exceeding 40 patients. Results were more variable in this subgroup, as evidenced by the broad range of individual trial OR from 0.30 to 5.54. Although the pooled OR for the subgroup (0.79) was higher than that for the subgroup comparing albumin with other volume expanders (0.34), the difference in effect sizes between the two subgroups was not statistically significant (P = 0.10).
Data on hyponatremia were available for all 17 included trials. However, in one trial comparing albumin with vasoconstrictor,31 there were no cases of hyponatremia in either study group, and so the OR for that trial could not be computed or included in the calculation of pooled OR. Across the other 16 included trials there was no evidence of significant heterogeneity (P = 0.97; I2, 0%) or publication bias (P = 0.24) with respect to the hyponatremia endpoint.
In the three trials comparing albumin with no treatment, 13 of 79 untreated control patients (16.5%) developed hyponatremia versus 3 of 77 patients receiving albumin (3.9%). Albumin reduced the odds of hyponatremia by 80%, compared with no treatment (pooled OR, 0.20; CI, 0.05-0.74).
Over all 14 trials evaluating alternative treatments, 86 of 577 control patients (14.9%) and 39 of 478 patients allocated to albumin treatment (8.2%) experienced hyponatremia. Among 13 of those trials with at least one case of hyponatremia, albumin decreased the odds of hyponatremia by 42% (Fig. 3). The effects of albumin on the odds of hyponatremia were similar across strata of age, refractory ascites, baseline serum albumin level, type of control volume expander, day of PCD assessment, paracentesis regimen, ascites volume removed, trial quality, control group mortality, and time period (Table 2).
Data on mortality were unavailable for four trials, and in one trial,20 there were zero deaths. Across the remaining 12 trials, the mortality endpoint did not exhibit significant heterogeneity (P = 0.91; I2, 0%) or publication bias (P = 0.24). Only 1 of the 12 trials compared albumin with no treatment,2 and there was no significant effect of albumin on mortality in that trial (P = 0.37).
In 11 trials with mortality data comparing albumin with alternative treatments, 74 of 513 control patients (14.4%) died versus 50 of 414 assigned to albumin infusion (12.1%). The odds of death were lower by 36% in the albumin group (P = 0.038; Fig. 4). Differences in the effect of albumin on mortality between subgroup strata of age, refractory ascites, baseline serum albumin level, type of control volume expander, day of PCD assessment, paracentesis regimen, ascites volume removed, trial quality, control group mortality, and time period were minor (Table 2).
Of the trials comparing albumin with alternative treatments, the number with available data for a particular secondary endpoint and at least one event ranged from 6 to 11 (Table 3). Albumin administration was associated with 15%-19% reductions in the odds of acites recurrence, renal impairment, and hospital readmission. Smaller reductions were observed for hepatic encephalopathy and portal hypertensive bleeding. However, none of these effects with respect to secondary endpoints were statistically significant.
Table 3. Other Effects of Albumin Versus Alternative Treatments
Odds Ratio (CI)
Portal hypertensive bleeding
Though PCD is clinically silent, it is the most powerful independent predictor of mortality in patients with tense ascites treated by LVP.22 In three trials included in the meta-analysis comparing albumin with no treatment, 73% of the untreated patients developed PCD, and albumin reduced the odds of this complication by 93%, confirming the value of adjunctive treatment in patients undergoing LVP. Albumin also reduced the odds of hyponatremia 80% versus no treatment.
The clinical question addressed in 14 of the 17 included trials was whether there may be alternative treatments of comparable effectiveness to albumin. However, this meta-analysis indicates that adjunctive albumin infusion is significantly more effective than alternative treatments evaluated to date in preventing the development of PCD and hyponatremia and reducing mortality after LVP. Effect sizes were substantial, with albumin reducing the odds of PCD by 66%, hyponatremia by 42%, and death by 36%. In subgroup analyses, albumin was significantly superior to each of the four other volume expanders tested in averting PCD. Oncotic forces exerted by 20%-25% albumin, and hence effectiveness in draining water and sodium from the extravascular space, may be greater than those of the alternative volume expanders.33 It is also possible that nononcotic properties contribute to the observed effects of albumin. For instance, evidence in patients with spontaneous bacterial peritonitis suggests that albumin may modulate endothelial cell function.34
No particular volume of ascites removed was required for eligibility in this meta-analysis. Nevertheless, in all eligible trials identified and included, average volume removed exceeded 5 L, the customary cutoff below which albumin infusion is not recommended. Within the range of volumes removed among the included studies, no major differences in the effects of albumin were apparent between 5.5 and 8.0 L of ascites fluid removed and >8 L.
In the majority of included studies, albumin was administered at a dose of 8 g per L of ascites fluid removed, although doses as low as 5 or 6 g·L−1 were also evaluated in three trials. Lower albumin doses, if effective, might aid in reducing treatment costs. A pilot study directly comparing 4- and 8-g·L−1 doses has recently been reported.35
The pathogenesis of PCD is not completely understood, but it is thought to be secondary to accentuation of already established arteriolar vasodilation. This concept has prompted the evaluation of the vasoconstrictors vasopressin, midodrine, and norepinephrine, as alternatives to albumin. The systematic search process undertaken as part of this meta-analysis revealed that randomized trial evidence on the clinical utility of vasopressors in ascites patients undergoing LVP is limited to small trials with 40 or fewer total patients each, and the results were variable. On the other hand, the combination of albumin and vasoconstrictors is the first-line treatment for type-1 hepatorenal syndrome and is being tested in patients awaiting liver transplantation. The possibility that such combination treatment might benefit ascites patients needing LVP has not been investigated thus far.
Although some data have suggested that hyponatremia may occur less frequently in patients receiving albumin as an adjunct to LVP, such an effect has not previously been established. The present meta-analysis provides clear-cut evidence that albumin effectively prevents hyponatremia, compared with alternative treatments. This finding is of potential clinical importance, because hyponatremia is a risk factor for hepatic encephalopathy and death.36-38 Patients with hyponatremia also display greater susceptibility to the development of refractory ascites, lower responsiveness to diuretics in terms of change in body weight, higher requirement for LVP to manage their ascites, and a shorter interval between needed paracentesis procedures.39 Hyponatremia develops because of an impairment in the renal capacity to eliminate solute-free water, causing disproportionate retention of water relative to sodium and thereby reducing serum sodium concentration. Arterial underfilling unloads high-pressure baroreceptors, stimulating a nonosmotic hypersecretion of vasopressin and inducing solute-free water retention and dilutional hyponatremia.
Albumin infusion has yielded favorable results as a treatment for hyponatremia in patients with cirrhosis.40, 41 In one RCT, daily albumin infusion for severe hyponatremia in patients with refractory ascites increased serum sodium levels and reduced the incidence of infection, hepatic encephalopathy, and in-hospital mortality.41
The clinical value of albumin administration as an adjunct to LVP has been questioned because a survival benefit could not previously be shown.5, 42 Individual trials have not been powered to detect a survival benefit. By “borrowing strength” from all available RCTs, this meta-analysis has provided evidence that albumin can indeed bestow a survival benefit, compared with alternative treatments in ascites patients undergoing LVP.
The mechanisms underlying the therapeutic effects exerted by albumin in liver failure are not fully understood, but multiple mechanisms appear to be at work. Albumin, the most abundant circulating protein in the plasma, is endowed with an array of nononcotic properties beyond its role as the predominant plasma colloid and its capacity to expand intravascular volume.43 These include ligand binding and antioxidant and anti-inflammatory activity. Patients with ascites are usually hypoalbuminemic as a consequence of reduced hepatic synthesis and secretion, dilution by increased intravascular and interstitial volume, and loss to the ascites fluid. Moreover, the ability of albumin to transport protein-bound substances and drugs and act as a detoxification agent is severely compromised in patients with cirrhosis.44 The combination of hypoalbuminemia and impaired albumin function leads to a marked disturbance in the transport, metabolism, and excretion of many endogenous and exogenous substances. One consequence is alteration in the pharmacokinetics and pharmacodynamics of many drugs, thus affecting their efficacy and side-effect profile.45 Infusion of exogenous albumin in ascites patients may serve the dual purposes of repleting the levels of circulating albumin and the functional activity of the albumin pool. These are effects specific to albumin not shared by alternative treatments and may contribute to the superior effectiveness of albumin documented in this meta-analysis.
The trials included in the meta-analysis concern acute treatment with albumin at the time of LVP. But, albumin has also shown promise for the management of ascites as a weekly or biweekly long-term treatment in combination with diuretics.46-48 The data are much more limited for this indication, however. In an RCT of 100 patients, long-term albumin administration in combination with diuretics to cirrhotic patients after their first ascitic episode significantly improved transplant-free survival and decreased the risk of ascites recurrence, compared with diuretics alone.48 A large, multicenter RCT in Italy is evaluating the effects of the addition of weekly albumin infusion to diuretic treatment in patients with uncomplicated ascites (ClinicalTrials.gov identifier NCT01288794). Primary outcome measures are survival and incidence of refractory ascites.
This meta-analysis is the first to focus on the comparative effectiveness of albumin as an adjunct to therapeutic paracentesis. A 2008 systematic review on diagnostic paracentesis also summarized data on hyponatremia, survival, and certain other endpoints, though not PCD, in patients receiving albumin after therapeutic paracentesis.49 Results of albumin infusion were not quantitatively combined across different control treatments in that review, however. By combining results of all available RCTs comparing albumin with alternative treatments, it has been possible, in the present meta-analysis, to show significant reductions in PCD, hyponatremia, and mortality in patients receiving albumin. Thus, the meta-analysis provides evidence-based support for the well-accepted clinical practice of infusing albumin as the first choice in adjunctive treatment for patients requiring LVP.