SYSTEMATIC REVIEWS AND META-ANALYSES

Transcription

1 CLINICAL GASTROENTEROLOGY AND HEPATOLOGY 2013;11: SYSTEMATIC REVIEWS AND META-ANALYSES Fasiha Kanwal, Section Editor Albumin Infusion Improves Outcomes of Patients With Spontaneous Bacterial Peritonitis: A Meta-analysis of Randomized Trials FRANCESCO SALERNO,* ROBERTA J. NAVICKIS, and MAHLON M. WILKES *Dipartimento di Medicina Interna, Università degli Studi di Milano, Policlinico IRCCS San Donato, Milano, Italy; and Hygeia Associates, Grass Valley, California BACKGROUND & AIMS: Renal impairment increases mortality among patients with spontaneous bacterial peritonitis (SBP), despite administration of non-nephrotoxic antibiotics. Albumin infusion has been reported to reduce renal impairment and mortality in patients with SBP. We performed a meta-analysis of randomized controlled trials (RCTs) to quantify the effect of albumin infusion on renal impairment and mortality in patients with SBP. METHODS: We searched MEDLINE, EMBASE, the Cochrane Library, and ClinicalTrials. gov for RCTs that evaluated albumin treatment for patients with SBP; we also performed searches by additional methods. Four trials of 288 total patients were included in our analysis. Data were quantitatively combined under a fixed-effects model. RESULTS: We found no evidence of statistically significant heterogeneity or publication bias among the studies analyzed. Albumin was compared with no albumin in 3 trials and with artificial colloid in 1 trial. All patients received antibiotics. The incidence of renal impairment in control groups was 44 of 144 (30.6%), compared with 12 of 144 (8.3%) in groups given albumin. The pooled odds ratio for a reduction in renal impairment after albumin infusion was 0.21 (95% confidence interval, ). Odds ratios for renal impairment after albumin therapy ranged from among the individual studies. Mortality among controls was 51 of 144 (35.4%), compared with 23 of 144 (16.0%) among patients who received albumin. The pooled odds ratio for decreased mortality after infusion of albumin was 0.34 (95% confidence interval, ). Odds ratios for mortality in individual RCTs ranged from CONCLU- SIONS: In a meta-analysis of 4 RCTs (288 patients), albumin infusion prevented renal impairment and reduced mortality among patients with SBP. Keywords: Cirrhosis; Ascites; Infection; Kidney Failure. Spontaneous bacterial peritonitis (SBP), one of the most common bacterial infections in cirrhotic patients, is an often lethal complication of ascites. 1 Mortality can reach 30% despite prompt treatment with non-nephrotoxic antibiotics. 2 Bacterial translocation is the most frequent cause of SBP. Intestinal bacterial overgrowth and slowed intestinal motility in patients with ascites result in bacterial translocation from the intestinal lumen to the mesenteric lymph nodes and eventual colonization of the ascitic fluid via the systemic circulation. Infection causes the release of multiple inflammatory mediators, but the immune system of cirrhotic patients is compromised, and their ability to clear cytokines, bacteria, endotoxins, and other vasoactive mediators is decreased. 3 Consequently, the development of SBP is accompanied by accumulation of cytokines and nitric oxide (NO) in plasma and ascitic fluid and an amplified proinflammatory response. 4,5 The splanchnic vasodilation characteristic of patients with cirrhosis and ascites results in reduced effective arterial blood volume, which leads to compensatory activation of the reninangiotensin-aldosterone system and sympathetic nervous system and vasoconstriction in nonsplanchnic vascular beds, including the kidney. 6 The increased cytokines and NO in SBP further enhance this circulatory dysfunction and consequent kidney hypoperfusion. Renal failure is prevalent and particularly severe in patients with SBP. 7 Renal impairment develops in 30% 40% of patients with SBP, even after the infection is controlled. 2,6 SBP is the most common precipitating factor of type 1 hepatorenal syndrome. 8 Renal dysfunction in SBP is the most powerful independent predictor of mortality, which is 67% in the presence of renal dysfunction vs only 11% in its absence. 9 Intravenous infusion of concentrated albumin is used to restore the reduced effective arterial blood volume associated with ascites and is widely used in conjunction with largevolume paracentesis (LVP) for tense ascites, with vasoconstrictors for type 1 hepatorenal syndrome and with antibiotics for SBP. Albumin, the most abundant plasma protein, also serves ligand-binding, antioxidant, and anti-inflammatory functions that may contribute to its beneficial action in these indications. 10 In a seminal randomized controlled trial (RCT) of 126 patients with SBP published in 1999, albumin infusion was shown to be effective in markedly reducing renal impairment and mortality. 11 On the strength of those data, recommendations that patients with SBP receive albumin have been incorporated in clinical guidelines. 12,13 However, under current guidelines of the American Association for the Study of Liver Diseases (AASLD), the recommendation to infuse albumin in patients with SBP is accorded only Level of Evidence B, since it is based on a single RCT. 12 Level A is reserved by the AASLD for recom- Abbreviations used in this paper: AASLD, American Association for the Study of Liver Diseases; BUN, blood urea nitrogen; CI, confidence interval; HES, hydroxyethyl starch; LVP, large-volume paracentesis; NO, nitric oxide; RCT, randomized controlled trial; SBP, spontaneous bacterial peritonitis by the AGA Institute /$

2 124 SALERNO ET AL CLINICAL GASTROENTEROLOGY AND HEPATOLOGY Vol. 11, No. 2 mendations supported by meta-analysis or multiple RCTs. Current guidelines of the European Association for the Study of the Liver explicitly recognize the need for further studies on the efficacy of albumin in the management of SBP. 13 A previous systematic review sought to identify predictors of mortality in cirrhotic patients with SBP; however, albumin therapy was not tested as a prognostic indicator because albumin was not used in the majority of included studies. 9 The present meta-analysis has quantitatively combined the results of all currently available RCTs evaluating the effects of albumin infusion on renal impairment and mortality in patients with SBP. Methods Study Selection The meta-analysis was limited exclusively to RCTs evaluating albumin treatment in patients with SBP. Trials designed to assess the joint effect of albumin plus paracentesis or to compare different albumin doses were not eligible. Both published and unpublished trials were sought. No restrictions were placed on language of reporting or on time period. Search Strategy Computer searches for eligible trials were performed in MEDLINE, EMBASE, the Cochrane Library, the ClinicalTrials. gov Web site, the abstract databases from major meetings in hepatology and gastroenterology, and other information resources indexed by Google. Search terms included spontaneous bacterial peritonitis, albumin, renal impairment, acute renal failure, acute kidney injury, mortality, survival, and randomized controlled trials. Roots and variants of those terms were also incorporated in the searches. A detailed MEDLINE search strategy is presented in Supplementary Table 1. Supplementary search methods consisted of examining reference lists of published articles related to cirrhosis and the online tables of contents for hepatology and gastroenterology journals. Data Extraction All 3 investigators independently determined trial eligibility, and 2 investigators extracted data from the trial reports. Differences in interpretation were resolved through discussion. Data were extracted for numbers of patients and study centers; year of reporting; methods of randomization, allocation concealment, and blinding; age; gender; ethnicity; etiology of cirrhosis; baseline serum bilirubin and albumin levels; Child Pugh score; albumin and control treatment regimens; duration of follow-up; definition and incidence of renal impairment; and mortality. Methodological details and supplementary unpublished data were requested from the trial investigators as needed. Statistical Analysis The end points of the meta-analysis were renal impairment, as defined in the included trials, and mortality. In the absence of heterogeneity, as judged by Cochran Q test and the I 2 statistic, trial results were combined under a fixed-effects model. Publication bias was assessed by linear regression of standardized effect vs the inverse of the standard error. 14 The significance of subgroup differences was determined by test of interaction. 15 The level for significance was Trial quality Figure 1. RCT selection process. was evaluated by randomization method, allocation concealment, and blinding. 16 R version (The R Foundation for Statistical Computing, Vienna, Austria) statistical software was used for all analyses. Results Included Trials Figure 1 summarizes the RCT selection process for the meta-analysis. After identification and screening of 64 candidate clinical study reports, 50 reports were judged not to be eligible at the screening stage, most often because they were literature reviews or described studies that did not evaluate albumin infusion. The remaining 14 reports were examined in detail, and 10 were found not eligible at that stage, most frequently because they did not concern albumin infusion. Four RCTs published from with 288 total patients fulfilled all eligibility criteria and were included in the metaanalysis. 11,17 19 One was a multicenter trial. 11 More than 100 patients were enrolled in each of 2 included trials, whereas 20 patients participated in a third trial and 30 patients in the fourth (Table 1). Patient characteristics are summarized in Table 1. Ranges of baseline means for serum bilirubin, serum albumin, and Child Pugh score were mg/dl 1, g/l 1, and , respectively. SBP was diagnosed on the basis of ascitic fluid polymorphonuclear cell counts 250 mm 3 in 3 trials or 250 mm 3 in 1 trial, with no evidence suggesting secondary peritonitis. In all 4 trials, enrollment was limited to adults, and the following exclusion criteria were applied: antibiotic treatment in the preceding week, cardiac or renal disease, human immunodeficiency virus or unspecified other infection, grade 3 4 hepatic encephalopathy, gastrointestinal bleeding, and advanced age. Additional exclusion criteria in 2 trials each were shock, septic shock, pulmonary disease, -blocker treatment or therapeutic paracentesis in the preceding week, advanced hepatocellular carcinoma, and ileus.

3 February 2013 ALBUMIN IN SPONTANEOUS BACTERIAL PERITONITIS 125 Table 1. Included Trials Trial N Age a (SD), y Male, % Ethnicity Etiology Treatment regimen Sort et al, (7.9) 64.3 Caucasian 29.4% alcohol, 70.6% other Cefotaxime intravenously dosed according to creatinine without vs with 1.5g/kg 1 20% albumin within first 6 h plus 1.0 g/kg 1 on day 3 Xue et al, b Chinese Ceftriaxone intravenously dosed according to creatinine without vs with g/kg 1 20% albumin within first 6 h and every third day for 21 days Fernández et al, Chen et al, (9.5) 55.0 Caucasian 60.0% HCV, 40.0% other Ceftriaxone intravenously 2gatdiagnosis and then 1 g/d plus 1.5 g/kg 1 at baseline and 1.0g/kg 1 on day 3 of 20% albumin vs 6% HES 200/ (11.5) 60.0 Chinese 60.0% HBV, 23.3% HCV, 16.7% alcohol HBV, hepatitis B virus; HCV, hepatitis C virus; SD, standard deviation. a Mean, except as otherwise indicated. b Age range eligible for this trial; mean age not reported. Cephalosporins without vs with 50 ml 20% albumin (0.14 g/kg 1 for a 70-kg patient) on days 1 3 Treatment assignment was determined by computer-generated sequence of random numbers in 3 trials 11,18,19 and coin toss in 1 trial. 17 In 3 trials, concealment of allocation to randomized groups was accomplished by using sealed envelopes. 11,18,19 Allocation concealment method was unspecified for the fourth trial. 17 Thus, method of randomization was adequate for all 4 included trials and of allocation concealment for 3 trials. On the other hand, only 1 of the 4 trials was blinded. 11 In all trials, the patients of both groups received antibiotic (Table 1). Albumin was infused as a 20% solution in all trials. The treatment period with albumin was 3 days in 3 trials and 3 weeks in the remaining trial. An albumin dose range of g/kg 1 was administered in 3 trials, whereas in the remaining trial, 3 daily fixed dosages of 10 g (0.14 g/kg 1 for a 70-kg patient) were given on the basis of National Health Insurance coverage. Control treatment in 1 trial consisted of hydroxyethyl starch (HES) at the same dose as that of albumin. In the other 3 trials, albumin was compared with no albumin. Length of follow-up was 90 days in 1 trial and unspecified in the others. Renal Impairment In 3 trials, renal impairment was defined according to the presence or absence of renal insufficiency at baseline. For patients without preexisting renal insufficiency, renal impairment was defined as an increase from baseline in serum creatinine to 1.5 mg/dl 1 or, alternatively in 2 of the 3 trials, in blood urea nitrogen (BUN) to 30 mg/dl 1. In addition, in 2 of the 3 trials the relative magnitude of serum creatinine or BUN increase must have been 50%. Only the 50% relative increase criterion was applied in cases of preexisting renal insufficiency. In the fourth trial, renal impairment was defined as a serum creatinine level 1.5 mg/dl 1 or BUN 25 mg/dl 1. Across all 4 trials, 12 of 144 albumin recipients (8.3%) developed renal impairment, compared with 44 of 144 control group patients (30.6%). The odds ratios for renal impairment in the 4 trials were highly consistent, ranging narrowly from (Figure 2). The pooled odds ratio was There was no evidence of significant heterogeneity (P.99; I 2, 0%) or publication bias (P.11) with respect to the renal impairment end point. The trials by Sort et al 11 and Xue et al 17 accounted for the preponderance of meta-analytic weight. The pooled odds ratio for renal impairment was 0.21 with 95% confidence interval (CI) of after excluding the trial by Sort et al and 0.23 (CI, ) after excluding the trial by Xue et al. These results are the same as or closely similar to the pooled odds ratio shown in Figure 2 with no exclusion (0.21; CI, ). For one included trial, 11 outcomes were reported with stratification according to significant baseline predictors. Nevertheless, in the report of that trial no formal analysis was presented Figure 2. Renal impairment. CI is shown by error bars. Data points for individual trials scaled according to meta-analytic weight.

4 126 SALERNO ET AL CLINICAL GASTROENTEROLOGY AND HEPATOLOGY Vol. 11, No. 2 Table 2. Comparative Effectiveness of Albumin in Subgroups a Outcome Albumin Control Subgroup Event Total Event Total Odds ratio (CI) P value b Renal impairment Bilirubin 4 mg/dl ( ) Bilirubin 4 mg/dl ( ) Ratio of odds ratios (CI) 0.48 ( ).47 Bilirubin 4 mg/dl 1 or creatinine 1 mg/dl ( ) Bilirubin 4 mg/dl 1 and creatinine 1 mg/dl ( ) Ratio of odds ratios (CI) 0.76 ( ).88 Hospital mortality Bilirubin 4 mg/dl ( ) Bilirubin 4 mg/dl ( ) Ratio of odds ratios (CI) 0.36 ( ).34 BUN 30 mg/dl ( ) BUN 30 mg/dl ( ) Ratio of odds ratios (CI) 5.97 ( ).27 Bilirubin 4 mg/dl 1 or BUN 30 mg/dl ( ) Bilirubin 4 mg/dl 1 and BUN 30 mg/dl c Bilirubin 4 mg/dl 1 and BUN 30 mg/dl ( ) Bilirubin 4 mg/dl 1 or BUN 30 mg/dl ( ) Ratio of odds ratios (CI) 2.00 ( ).54 a Data from included trial of Sort et al. 11 Subgroups defined in terms of baseline levels for serum bilirubin, serum creatinine, or BUN. In multivariate analyses reported by Sort et al, independent predictors of renal impairment were baseline levels of serum bilirubin (P.001) and serum creatinine (P.01) and antibiotic treatment without albumin (P.02). Independent predictors of hospital mortality included baseline levels of serum bilirubin (P.01) and BUN (P.001) and antibiotic treatment without albumin (P.05). b Difference in effectiveness of albumin between the subgroups by test of interaction. c Odds ratio inestimable for 0 events in both groups. on the comparative effectiveness of albumin in the subgroups defined by the baseline predictors. Such a subgroup analysis is shown in Table 2. The control group incidence of renal impairment was 14 of 29 (48%) when baseline serum bilirubin equaled or exceeded 4 mg/dl 1, as contrasted with 7 of 34 (21%) when serum bilirubin was below 4 mg/dl 1 at baseline. The odds ratio for the prevention of renal impairment as the result of albumin treatment was 0.15 in patients with baseline serum bilirubin 4 mg/dl 1, compared with 0.32 in those with lower baseline serum bilirubin; however, this subgroup difference in the effectiveness of albumin for averting renal impairment was not statistically significant (P.47). With regard to renal impairment, furthermore, the effectiveness of albumin was similar in the subgroup defined by either serum bilirubin 4 mg/dl 1 or serum creatinine 1 mg/dl 1 at baseline vs the subgroup with lower values of those 2 parameters (Table 2). Mortality Total mortality was 23 of 144 patients allocated to albumin treatment (16.0%) and 51 of 144 to the control regimen (35.4%). Odds ratios for mortality in individual RCTs were similar, ranging from (Figure 3). The pooled odds ratio was There was no significant heterogeneity (P.72; I 2, 0%) or publication bias (P.69) with respect to mortality. The pooled odds ratio for mortality with exclusion of either the trial by Sort et al 11 (0.28; CI, ) or by Xue et al 17 (0.41; CI, ) was comparable to that with no exclusion shown in Figure 3 (0.34; CI, ). Figure 3. Mortality. CI is shown by error bars. Data points for individual trials scaled according to meta-analytic weight.

5 February 2013 ALBUMIN IN SPONTANEOUS BACTERIAL PERITONITIS 127 In a subgroup analysis that was based on data from 1 included trial (Table 2), control group mortality was 13 of 29 (45%) in patients with baseline serum bilirubin 4 mg/dl 1, compared with 5 of 34 (15%) in the subgroup with lower baseline serum bilirubin. The effectiveness of albumin in reducing mortality appeared to be more pronounced in the subgroup with higher baseline serum bilirubin (odds ratio, 0.18 vs 0.48), although this difference was not significant (P.34). Control group mortality rates were 13 of 28 (46%) and 5 of 35 (14%) in the respective subgroups, with baseline BUN at least 30 mg/dl 1 vs lower levels. The odds ratio for reducing mortality by albumin treatment was 6-fold as great in the subgroup with lower than higher BUN at baseline; nevertheless, this difference was not significant (P.27). There was also no significant difference between the subgroup with higher baseline values of both serum bilirubin and BUN as compared with the other subgroup (Table 2). Discussion In this meta-analysis, albumin infusion in patients with SBP decreased renal impairment and mortality. These salutary effects of albumin infusion were remarkably consistent from trial to trial. The value of a meta-analysis is considered to be especially clear when the included studies show clinically important effects of similar magnitude. 20 A major strength of the meta-analysis was exclusive reliance on RCTs. However, only 4 eligible RCTs were identified, 2 of which were relatively small. Possibly, some investigators may have refrained from conducting confirmatory trials because of perceived lack of equipoise about the benefits of albumin after the RCT of Sort et al. 11 In any case, the number of trials in this meta-analysis is far from unusual. For instance, 25% of Cochrane reviews on therapeutic interventions include no more than 4 studies. 21 Furthermore, the precision for quantifying the pooled odds ratio in a fixed-effects meta-analysis is determined entirely by the numbers of patients and events, not the number of studies per se. 22 The precision afforded by the 4 studies in this meta-analysis was excellent, as judged by the narrow CI for the pooled odds ratios (Figures 2 and 3). Although the trial of Sort et al 11 was a major component of this meta-analysis, the other 3 included RCTs contributed 56% of the total patients and approximately half the total metaanalytic weight of evidence for both the renal impairment (Figure 2) and mortality (Figure 3) end points. They also added substantially to precision, as reflected by 37% shrinkage of the pooled 95% CI for renal impairment relative to that in the RCT of Sort et al alone and by 41% shrinkage for mortality. The 4 included RCTs were similar in design and patient characteristics. Criteria applied in establishing the diagnosis of SBP, selecting patients, and defining renal impairment coincided closely. Age ranges, gender apportionments, and baseline serum bilirubin and serum albumin levels and Child Pugh scores were comparable. In all RCTs, patients promptly received antibiotic and intravenous 20% albumin, and albumin therapy was continued for a minimum of 3 days. These similarities may at least partly explain the fact that there was no evidence of significant statistical heterogeneity. Nevertheless, studies included in a meta-analysis are rarely, if ever, identical clones, and some degree of clinical heterogeneity is expected. 20 For example, a relatively low albumin dose was administered in one trial, 19 albumin therapy was more prolonged in another, 17 and the control group received artificial colloid in another. 18 Despite these elements of clinical heterogeneity, consistent effects on renal impairment and mortality were observed. Although potentially limiting the certainty with which the effects of one particular treatment regimen can be determined, clinical heterogeneity can conversely help indicate the robustness and broader applicability of meta-analytic findings. The trial of Sort et al 11 was the only included RCT providing results stratified according to subgroups defined by risk factors such as serum bilirubin, serum creatinine, and BUN. Subgroup analysis can potentially aid in patient selection, and indeed, the subgroup data of Sort et al have given impetus to proposals that albumin infusion be reserved for higher-risk patients with SBP However, subgroup differences need to be interpreted with caution, because they can occur by chance and may prove impossible to replicate in subsequent studies It is well established that subgroup differences need to be assessed by test of interaction rather than by consideration of significant withinsubgroup effects. 15,27 The interaction test results in Table 2 did not reveal any significant subgroup differences. Table 2 provides a rigorous statistical analysis of RCT subgroup data and indicates persisting uncertainty as to whether the benefits of albumin may be restricted to identifiable subgroups of patients with SBP. In subgroups with low event rates, albumin infusion might not be needed. The control group rate of renal impairment in the subgroup with serum bilirubin 4 mg/dl 1 and serum creatinine 1 mg/dl 1 was 6.7% (CI, 0.2% 31.9%), on the basis of a single event (Table 2). In addition, there were no hospital deaths in the subgroup of control patients with serum bilirubin 4 mg/dl 1 and BUN 30 mg/dl 1, and corresponding mortality was 0.0% (CI, 0.0% 16.8%). The large upper 95% confidence bounds for control group renal impairment and mortality in lower-risk patients indicate low precision in estimating the true population event rates. In a retrospective study, no albumin was administered to 15 SBP patients at low risk by the criterion of serum bilirubin 4 mg/dl 1 and serum creatinine 1 mg/dl 1 at baseline. 24 None of these patients developed renal impairment or died during the hospital stay. Accordingly, AASLD clinical guidelines 12 advocate that albumin be reserved for high-risk patients. Two larger subsequent retrospective studies have, however, furnished data that the incidence of poor outcomes in low-risk patients not treated with albumin may not be negligible. 29,30 Table 3 summarizes outcomes among untreated low-risk patients in the RCT of Sort et al 11 and the 3 retrospective studies. The pooled incidence rates of renal impairment and hospital mortality were 2.8% and 5.8%, respectively. From those rates and the pooled odds ratios in Figures 2 and 3, it can be estimated that the number of low-risk patients needed to treat with albumin to avoid 1 case of renal impairment is 45 and to prevent 1 hospital death is 27. These estimates, which are based on the evidence currently available, suggest the prudence of treating all SBP patients with albumin. The dose-response relationship between infused albumin and outcomes of SBP needs to be further delineated. Albumin dosages of g/kg 1 for at least 3 days were administered in 3 of the included trials. A lower fixed dose of 10 g daily for 3 days yielded favorable results in the fourth trial. 19 However, that trial was relatively small, and more data will be needed to establish optimal dosing. In a recently described pilot RCT, 46 cirrhotic patients with SBP were allocated to a standard-dose regimen of 1.5 g/kg 1 albumin at diagnosis and 1.0 g/kg 1 48 hours thereafter or to a

6 128 SALERNO ET AL CLINICAL GASTROENTEROLOGY AND HEPATOLOGY Vol. 11, No. 2 Table 3. Renal Impairment and Hospital Mortality in Low-Risk Patients Not Receiving Albumin a Renal impairment Hospital mortality Study Design Event Total Incidence (CI), % Event Total Incidence (CI), % Sort et al, RCT ( ) ( ) Sigal et al, Retrospective ( ) ( ) Terg et al, Retrospective 1 38 b 2.6 ( ) ( ) Poca et al, Retrospective 1 c 39 d 2.6 ( ) 4 39 d 10.3 ( ) Total ( ) ( ) NNT to prevent 1 event (CI) e 45 (40 62) 27 (21 44) NNT, number needed to treat. a The criterion for low risk was serum bilirubin 4 mg/dl 1 and serum creatinine 1 mg/dl 1 at baseline. The only exception was that, with respect to hospital mortality end point only, Sort et al 11 defined low risk as serum bilirubin 4 mg/dl 1 and BUN 30 mg/dl 1 at baseline. For individual studies, exact binomial CI calculated. Incidence pooled across studies by mixed-effects logistic regression to accommodate studies with 0 events. 43 b Patients at risk computed from reported 2.6% incidence rate. c In the publication, only cases of renal impairment occurring before SBP resolution were reported. In response to a query, the investigators indicated that 1 low-risk patient developed renal impairment after SBP resolution and before discharge. d The number of unique low-risk patients, as indicated by the investigators in response to a query. In the publication, totals were reported only on a per SBP episode rather than a per patient basis. e NNT with albumin computed from pooled incidence rates in this table and pooled odds ratios in Figures 2 and 3. No differences in the effectiveness of albumin between low- and high-risk patients were assumed on the basis of the subgroup analyses in Table 2. reduced-dose regimen at corresponding dosages of 1.0 g/kg 1 albumin and 0.5 g/kg 1. Between-group differences in renal impairment and mortality could not be detected. However, mean plasma renin activity declined by 50% in response to the standarddose regimen but only 11% with the reduced dosages. 31 Another RCT is comparing a standard regimen of 1.5 g/kg 1 albumin on day 1 and 1.0 g/kg 1 on day 3 with an alternative regimen in which the second dose, to be administered on day 2, is reserved for patients whose serum creatinine at that time exceeds 1.0 mg/dl 1 or whose serum creatinine or BUN is elevated above baseline (ClinicalTrials.gov Identifier: NCT ). Circulatory dysfunction similar to that experienced by patients with SBP can also contribute to poor outcomes in cirrhotic patients without SBP who undergo LVP to relieve tense ascites. Albumin infusion is also recommended in that setting. 12,13 In a recent meta-analysis of 17 RCTs, albumin infusion as an adjunct to LVP was found to reduce postparacentesis circulatory dysfunction, hyponatremia, and mortality as compared with alternative treatments, namely other volume expanders and vasoconstrictors. 32 The benefits of albumin infusion in SBP remain to be explained, although multiple direct and indirect mechanisms may be operating. As a colloid, albumin effectively expands intravascular volume, thereby increasing cardiac preload, enhancing left ventricular function, and correcting the deficit in effective arterial blood volume that provokes the activation of vasoconstrictor systems and contributes to renal hypoperfusion. 33 Albumin may both directly and indirectly modify the effects of endotoxin. Each albumin molecule can bind up to 10 molecules of endotoxin. 34 The endotoxin inactivation rate of plasma in cirrhotic patients was found to be significantly positively correlated with the binding capacity of albumin to endotoxin. 35 Albumin has been shown to block lipopolysaccharide-stimulated oxidative burst activity of neutrophils. 36 Albumin selectively suppressed neutrophil oxidative burst activity stimulated by tumor necrosis factor- and prevented tumor necrosis factor- induced neutrophil spreading by binding to and preventing the shedding of CD43 from the neutrophil surface. 37 NO and albumin can combine to form covalent S-nitrosoalbumin adducts, and albumin may serve as a reservoir, traffic protein, and modulator of NO activity. 38 Benefits bestowed by volume expansion per se might also be attainable through infusion of artificial colloids. However, other putative mechanisms described above depend with a high degree of biological specificity on the structure and function of the albumin molecule. Benefits mediated by those mechanisms would not be expected from artificial colloids. In one included RCT, more favorable effects were achieved with albumin than HES. 18 Also, the safety record of albumin, the natural endogenous colloid, is superior to that of artificial colloids. 39 For instance, meta-analyses of RCTs have shown HES to be nephrotoxic. 40,41 Evidence has also been reported of hepatic impairment associated with lysosomal storage of HES in Kupffer cells and hepatocytes. 42 This meta-analysis provides the basis for a Level A recommendation in the AASLD guidelines that patients with SBP be treated with albumin. In conjunction with data from retrospective studies, the meta-analysis also supports the current European Association for the Study of the Liver recommendation 13 that all patients with SBP receive albumin. Nevertheless, evidence available at present is limited on outcomes in low-risk SBP patients not receiving albumin as well as on the responsiveness of low-risk patients to albumin infusion. Hence, further studies are warranted to inform the selection of patients for albumin therapy. Supplementary Material Note: To access the supplementary material accompanying this article, visit the online version of Clinical Gastroenterology and Hepatology at and at dx.doi.org/ /j.cgh References 1. Fernández J, Navasa M, Gómez J, et al. Bacterial infections in cirrhosis: epidemiological changes with invasive procedures and norfloxacin prophylaxis. Hepatology 2002;35:

7 February 2013 ALBUMIN IN SPONTANEOUS BACTERIAL PERITONITIS Follo A, Llovet JM, Navasa M, et al. Renal impairment after spontaneous bacterial peritonitis in cirrhosis: incidence, clinical course, predictive factors and prognosis. Hepatology 1994;20: Bonnel AR, Bunchorntavakul C, Reddy KR. Immune dysfunction and infections in patients with cirrhosis. Clin Gastroenterol Hepatol 2011;9: Bories PN, Campillo B, Azaou L, et al. Long-lasting NO overproduction in cirrhotic patients with spontaneous bacterial peritonitis. Hepatology 1997;25: Navasa M, Follo A, Filella X, et al. Tumor necrosis factor and interleukin-6 in spontaneous bacterial peritonitis in cirrhosis: relationship with the development of renal impairment and mortality. Hepatology 1998;27: Ruiz-del-Arbol L, Urman J, Fernández J, et al. Systemic, renal, and hepatic hemodynamic derangement in cirrhotic patients with spontaneous bacterial peritonitis. Hepatology 2003;38: Ginès P, Schrier RW. Renal failure in cirrhosis. N Engl J Med 2009;361: Arroyo V, Fernandez J, Ginès P. Pathogenesis and treatment of hepatorenal syndrome. Semin Liver Dis 2008;28: Tandon P, Garcia-Tsao G. Renal dysfunction is the most important independent predictor of mortality in cirrhotic patients with spontaneous bacterial peritonitis. Clin Gastroenterol Hepatol 2011;9: Quinlan GJ, Martin GS, Evans TW. Albumin: biochemical properties and therapeutic potential. Hepatology 2005;41: Sort P, Navasa M, Arroyo V, et al. Effect of intravenous albumin on renal impairment and mortality in patients with cirrhosis and spontaneous bacterial peritonitis. N Engl J Med 1999;341: Runyon BA. Management of adult patients with ascites due to cirrhosis: an update. Hepatology 2009;49: Ginès P, Angeli P, Lenz K, et al. EASL clinical practice guidelines on the management of ascites, spontaneous bacterial peritonitis, and hepatorenal syndrome in cirrhosis. J Hepatol 2010;53: Egger M, Davey Smith G, Schneider M, et al. Bias in metaanalysis detected by a simple, graphical test. BMJ 1997;315: Altman DG, Bland JM. Interaction revisited: the difference between two estimates. BMJ 2003;326: Jüni P, Altman DG, Egger M. Systematic reviews in health care: assessing the quality of controlled clinical trials. BMJ 2001;323: Xue HP, Lin B, Zhong J, et al. Effect of albumin infusion on preventing the deterioration of renal function in patients with spontaneous bacterial peritonitis. Chin J Dig Dis 2002;3: Fernández J, Monteagudo J, Bargallo X, et al. A randomized unblinded pilot study comparing albumin versus hydroxyethyl starch in spontaneous bacterial peritonitis. Hepatology 2005;42: Chen TA, Tsao YC, Chen A, et al. Effect of intravenous albumin on endotoxin removal, cytokines, and nitric oxide production in patients with cirrhosis and spontaneous bacterial peritonitis. Scand J Gastroenterol 2009;44: Higgins JP, Thompson SG, Deeks JJ, et al. Measuring inconsistency in meta-analyses. BMJ 2003;327: Moher D, Tetzlaff J, Tricco AC, et al. Epidemiology and reporting characteristics of systematic reviews. PLoS Med 2007;4:e Fleiss JL. The statistical basis of meta-analysis. Stat Methods Med Res 1993;2: Perdomo Coral G, Alves de Mattos A. Renal impairment after spontaneous bacterial peritonitis: incidence and prognosis. Can J Gastroenterol 2003;17: Sigal SH, Stanca CM, Fernandez J, et al. Restricted use of albumin for spontaneous bacterial peritonitis. Gut 2007;56: Narula N, Tsoi K, Marshall JK. Should albumin be used in all patients with spontaneous bacterial peritonitis? Can J Gastroenterol 2011;25: Oxman AD, Guyatt GH. A consumer s guide to subgroup analyses. Ann Intern Med 1992;116: Wang R, Lagakos SW, Ware JH, et al. Statistics in medicine: reporting of subgroup analyses in clinical trials. N Engl J Med 2007;357: Sun X, Briel M, Walter SD, et al. Is a subgroup effect believable? updating criteria to evaluate the credibility of subgroup analyses. BMJ 2010;340:c Terg R, Gadano A, Cartier M, et al. Serum creatinine and bilirubin predict renal failure and mortality in patients with spontaneous bacterial peritonitis: a retrospective study. Liver Int 2009;29: Poca M, Concepción M, Casas M, et al. Role of albumin treatment in patients with spontaneous bacterial peritonitis. Clin Gastroenterol Hepatol 2012;10: de Araujo A, de Barros Lopes A, Rossi G, et al. Low-dose albumin in the treatment of spontaneous bacterial peritonitis: should we change the standard treatment? Gut 2012;61: Bernardi M, Caraceni P, Navickis RJ, et al. Albumin infusion in patients undergoing large-volume paracentesis: a meta-analysis of randomized trials. Hepatology 2012;55: Arroyo V, Fernandez J. Pathophysiological basis of albumin use in cirrhosis. Ann Hepatol 2011;10(Suppl 1):S6 S Jürgens G, Müller M, Garidel P, et al. Investigation into the interaction of recombinant human serum albumin with Re-lipopolysaccharide and lipid A. J Endotoxin Res 2002;8: Fukui H, Tsujita S, Matsumoto M, et al. Endotoxin inactivating action of plasma in patients with liver cirrhosis. Liver 1995;15: Stadlbauer V, Mookerjee RP, Wright GA, et al. Role of Toll-like receptors 2, 4, and 9 in mediating neutrophil dysfunction in alcoholic hepatitis. Am J Physiol Gastrointest Liver Physiol 2009; 296:G Nathan C, Xie QW, Halbwachs-Mecarelli L, et al. Albumin inhibits neutrophil spreading and hydrogen peroxide release by blocking the shedding of CD43 (sialophorin, leukosialin). J Cell Biol 1993; 122: Marley R, Patel RP, Orie N, et al. Formation of nanomolar concentrations of S-nitroso-albumin in human plasma by nitric oxide. Free Radic Biol Med 2001;31: Groeneveld AB, Navickis RJ, Wilkes MM. Update on the comparative safety of colloids: a systematic review of clinical studies. Ann Surg 2011;253: Zarychanski R, Turgeon AF, Fergusson DA, et al. Renal outcomes and mortality following hydroxyethyl starch resuscitation of critically ill patients: systematic review and meta-analysis of randomized trials. Open Med 2009;3:E196 E Dart AB, Mutter TC, Ruth CA, et al. Hydroxyethyl starch (HES) versus other fluid therapies: effects on kidney function. Cochrane Database Syst Rev 2010;1:CD Christidis C, Mal F, Ramos J, et al. Worsening of hepatic dysfunction as a consequence of repeated hydroxyethylstarch infusions. J Hepatol 2001;35: Bradburn MJ, Deeks JJ, Berlin JA, et al. Much ado about nothing: a comparison of the performance of meta-analytical methods with rare events. Stat Med 2007;26: Reprint requests Address requests for reprints to: Francesco Salerno, MD, Dipartimento di Medicina Interna, Università degli Studi di Milano, Policlinico

8 130 SALERNO ET AL CLINICAL GASTROENTEROLOGY AND HEPATOLOGY Vol. 11, No. 2 IRCCS San Donato, Via Morandi 30, Milano, Italy. francesco.salerno@unimi.it; fax: Acknowledgments The authors acknowledge the cooperation of Prof Vicente Arroyo, Dr Angela Chen, Dr Javier Fernández, Dr Bing Lin, Dr Marie Poca, Prof German Soriano, and Prof Hui Ping Xue in providing supplementary data. Conflicts of interest These authors disclose the following: Drs Navickis and Wilkes have received past unrestricted grant support from CSL Behring. The remaining author discloses no conflicts. Funding Supported under an unrestricted grant from CSL Behring, King of Prussia, Pennsylvania.

9 130.e1 SALERNO ET AL CLINICAL GASTROENTEROLOGY AND HEPATOLOGY Vol. 11, No. 2 Supplementary Table 1. MEDLINE Search Strategy Set Query 1 spontaneous bacterial peritonitis 2 albumin 3 acute renal failure OR acute kidney injury 4 renal OR kidney* OR nephr* 5 failure OR injur* OR impair* OR dysfunc* 6 #4 AND #5 7 dialysis OR hemodiafiltr* OR renal replacement therapy 8 #3OR#6OR#7 9 mortality OR surviv* OR death* 10 #8 OR #9 11 random* [tiab] OR random allocation [mh] OR randomized controlled trial [pt] 12 #1 AND #2 AND #10 AND #11