Uncovered Transjugular Intrahepatic Portosystemic Shunt for Refractory Ascites: A Meta-Analysis

Article Outline

• Abstract
• Materials and Methods
• Results
  • Technical Results
  • Overall Estimates
  • Heterogeneity, Metaregression, and Sensitivity Analyses
  • Pooled Estimates by Excluding the Potentially Outlier Trial
  • Adverse Events and Other Relevant Outcomes
  • Publication Bias
• Discussion
• Acknowledgment
• References
• Copyright

Background & Aims: Several trials showed that uncovered transjugular intrahepatic portosystemic shunt (TIPS) is superior to paracentesis for the control of refractory ascites. However, the results for encephalopathy and mortality were not consistent across trials. We performed a systematic review of randomized controlled trials of TIPS for refractory ascites to assess the overall treatment effects and to explore potential reasons of heterogeneity. Methods: Pertinent studies were retrieved trough MEDLINE (1968–2004), EMBASE (1986–2004), the Cochrane Library (2004;4), and reference lists of key articles. Outcome measures were recurrence of ascites, encephalopathy, and mortality. Metaregression analysis was used to explore heterogeneity. Results: Five trials were identified including 330 patients. Successful TIPS placement ranged from 77% to 100% and portosystemic pressure gradient reduction ranged from 6.0 to 14.0 mm Hg. Metaregression analysis showed that bilirubin levels and successful TIPS placement rates were associated significantly with log–odds ratio for death after TIPS, explained heterogeneity of trials for mortality, and suggested an outlier trial. After exclusion of the outlier trial, pooled odds ratios for recurrence of ascites with TIPS was .14 (confidence interval, .07–.27), for encephalopathy was 2.26 (confidence interval, 1.35–3.76), and for mortality was .74 (confidence interval, .40–1.37), without any significant heterogeneity. Conclusions: Uncovered TIPS is significantly better than paracentesis for control of refractory ascites. Although it increases encephalopathy, it also is associated with a trend toward improvement of survival. Future TIPS trials should select patients on the basis of bilirubin levels and predictors of the risk for post-TIPS encephalopathy, and assess costs and quality of life.

Abbreviations used in this paper:  CI, confidence interval , OR, odds ratio , POR, pooled odds ratio , RCT, randomized controlled trial , TIPS, transjugular intrahepatic portosystemic shunt

Refractory ascites is a serious complication of portal hypertension in patients with advanced cirrhosis.¹ It is associated with an increased risk of hepatorenal syndrome and spontaneous bacterial peritonitis.², ³, ⁴ One-year mortality rates range from 20% to 50%⁵, ⁶ and liver transplantation is the only definitive treatment for most patients. In these patients, repeated large-volume or total paracentesis associated with intravenous albumin infusion is considered the first-line treatment. Although paracentesis is effective and safe in the elimination of ascites, it does not correct portal hypertension and neither prevents recurrence of ascites nor modifies the natural course of the disease. Surgical portal-caval shunts reducing portal pressure are effective in the treatment of refractory ascites but they have been abandoned because of their high morbidity and mortality rates. Transjugular intrahepatic portosystemic shunt (TIPS), which physiologically is equivalent to a side-to-side portal caval shunt, has been proposed for the treatment of refractory ascites⁷, ⁸, ⁹ and several uncontrolled studies have shown encouraging results.⁶ However, the available randomized controlled trials (RCTs) comparing TIPS with repeat paracentesis show conflicting results that hinder conclusions for clinical practice.¹⁰, ¹¹, ¹², ¹³, ¹⁴ The first RCT¹⁰ included a very small number of patients and showed a higher mortality rate in the TIPS than in the control group in Child–Pugh¹⁵ class C patients. In a larger study including 60 patients,¹¹ a multivariate analysis showed that TIPS was associated independently with better survival by adjusting the treatment effect for bilirubin levels, age, sex, and serum sodium levels. Two other subsequent RCTs¹², ¹³ failed to show any difference in survival and the most recent RCT¹⁴ showed improved survival with TIPS. In fact, a recent Cochrane meta-analysis,¹⁶ including the first 4 of these RCTs,¹⁰, ¹¹, ¹², ¹³ found a statistically significant and yet unexplained heterogeneity for the TIPS effect on mortality.

The aim of this meta-analysis was to assess the efficacy of TIPS compared with large-volume or total paracentesis in cirrhotic patients with refractory ascites including all 5 RCTs now available and, where appropriate, to explore the causes of heterogeneity.

Back to Article Outline

Materials and Methods 

Eligible RCTs were those comparing TIPS with large-volume or total paracentesis (± intravenous albumin infusion) for refractory ascites in patients with cirrhosis regardless of the cause of liver disease, irrespective of publication status or language. Quasirandomized trials and observational studies were excluded.

Measures of treatment efficacy were as follows: (1) recurrence of ascites requiring paracentesis: number of patients in whom a new paracentesis was needed after the trial treatment; (2) portosystemic encephalopathy: number of patients with encephalopathy after the trial treatment independent of portosystemic encephalopathy prevalence at randomization and the number of patients developing grades III/IV encephalopathy according to Conn and Liebertal¹⁷ or an equivalent classification; (3) mortality: the number of patients who died during the study period, the number of patients who died of gastrointestinal bleeding, and the number of patients who died of causes other than bleeding. Adverse events associated with treatment, the number of patients crossed-over between treatment groups, and the number of patients undergoing liver transplantation also were assessed.

Retrieval of RCTs was based on the Cochrane Controlled Trials Register, The Cochrane Library, MEDLINE, and ENBASE (until December 2004) using the terms TIPS, paracentesis, and refractory ascites, and limiting the search to randomized clinical trials and human studies. A manual search also was performed using the reference lists from articles, reviews, editorials, and the proceedings of international congresses. When the results of a single study were reported in more than 1 publication, only the most recent and complete data were included in the meta-analysis.

Decisions on which trials to include were taken unblindly by 2 reviewers (A.L. and G.D.). Disagreements were resolved by discussion. Excluded trials were identified with the reason for exclusion. Five RCTs fulfilled the inclusion criteria, all published as full reports.¹⁰, ¹¹, ¹², ¹³, ¹⁴

The methodologic quality of the trials was assessed by 4 major criteria previously validated¹⁸, ¹⁹, ²⁰, ²¹, ²², ²³, ²⁴: adequate generation of the randomization sequence, adequate treatment allocation concealment, blinded outcome assessment, and intention-to-treat analysis. Each quality component was rated as yes, unclear, or no. The quality of trials was reported according to each separate component.²⁰

Data concerning trials, patient characteristics, and treatment outcome (Table 1, Table 2, Table 3, Table 4, Table 5, Table 6, Table 7) were abstracted by 3 independent reviewers (G.D., A.L., and R.M.) and discrepancies were resolved by discussion.

Table 1. Study Characteristics of Each RCT Included in the Meta-Analysis

PA, paracentesis; NR, not reported; PSE, portosystemic encephalopathy; PVT, portal vein thrombosis; HCC, hepatocellular carcinoma; UGIB, upper gastrointestinal bleeding; SBP, spontaneous bacterial peritonitis; HRS, hepatorenal syndrome.

Table 2. Characteristics of Patients Included in Each RCT Using TIPS/Paracentesis Included in the Meta-Analysis

NOTE. Data refer to TIPS/paracentesis.

Table 3. Outcome Definitions Used in the Studies Included in this Meta-Analysis

NR, not reported; PPG, portosystemic pressure gradient.

Table 4. Technical Results of the Treatment in the 4 RCTs Published as Full Reports

Study	Successful stent placement, n/randomized (%)	PPG change from before to after TIPS, mm Hg (mean ± SD)	PPG reduction after TIPS, mm Hg (mean)	TIPS dysfunction n/N (%)	TIPS-assisted patency n/randomized (%)
Lebrec et al10	10/13(77)	From 20 ± 1 to 14 ± 1a	6	3/10(30)	6/13(46)
Rossle et al11	29/29(100)	From 24 ± 6 to 10 ± 4b	14	13/29(45)	27/29(93)
Gines et al12	34/35(97)	From 19.1 ± 0.8 to 8.7 ± 0.4a	10.4	13/34(38)	32/35(91)
Sanyal et al13	49/52(94)	From 19.8 ± 4.8 to 8.3 ± 3.6c	11.5	36/49(73)	>90%d
Salerno et al14	29/33(89)e	From 22.5 ± 1.1 to 8.7 ± 0.6a	13.8	12/29(41)	27/33(82)

PPG, portal pressure gradient; n/N, number with the outcome/patients at risk.

Table 5. Results of Each Study and PORs for Recurrence of Ascites, Overall Encephalopathy, and Severe Encephalopathy by Including and Excluding 1 Outlier Trial

RCT	Recurrence of ascites			Encephalopathy			Severe encephalopathy
	TIPS	PARA	OR (CI)	TIPS	PARA	OR (CI)	TIPS	PARA	OR (CI)
	n/N			n/N			n/N
Lebrec et al10	10/13	11/12	.30(.03–3.41)	3/13	0/12	8.33(.39–1.80)	2/13	0/12	5.43(.24–125)
Rossle et al11	14/29	26/31	.18(.05–.60)	6/29	3/31	2.43(.55–10.8)	—	—	—
Gines et al12	17/35	29/35	.20(.06–.59)	27/35	23/35	1.76(.61–5.05)	21/35	12/35	2.88(1.09–7.60)
Sanyal et al13	22/52	48/57	.14(.06–.34)	22/52	13/57	2.48(1.08–5.68)	15/52	10/57	1.91(.77–4.73)
Salerno et al14	13/33	32/33	.02(.00–.17)	20/33	13/23	2.37(.88–6.35)	14/33	10/33	1.69(.61–4.67)
Total	76/162	146/168	.14(.08–.26)	75/162	51/168	2.34(1.41–3.87)	52/133	32/137	2.17(1.25–3.74)
Heterogeneity χ2 (P value)			4.23(.37)			.96(.92)			.96(.81)
Excluding Lebrec et al10	66/149	135/156	.14(.07–.27)	72/149	51/156	2.26(1.35–3.76)	50/120	32/125	2.10(1.21–3.67)
Heterogeneity χ2 (P value)			3.86(.27)			.28(.96)			.62(.73)

PARA, paracentesis; n/N, number of patients with the outcome/total number of patients; OR, odds ratio.

Table 6. Results of Each Study and Pooled Estimates for Mortality by Including and Excluding 1 Outlier Trial

RCT	Mortality			Mortality for bleeding			Mortality for other causes
	TIPS	PARA	POR (CI)	TIPS	PARA	POR (CI)	TIPS	PARA	POR (CI)
	n/N			n/N			n/N
Lebrec et al10	9/13	4/12	4.50(.84–24.2)	2/13	3/12	.55(.07–4.01)	7/13	1/12	12.8(1.26–1.30)
Rossle et al11	15/29	23/31	.37(.13–1.10)	1/29	4/31	.24(.03–2.30)	14/29	19/31	.59(.21–1.64)
Gines et al12	20/35	18/35	1.26(.49–3.23)	—	—	—	—	—	—
Sanyal et al13	21/52	21/57	1.16(.54–2.51)	0/52	3/57	.15(.01–2.94)	21/52	18/57	1.47(.67–3.22)
Salerno et al14	13/33	20/33	.42(.16–1.13)	0/33	0/33	Not estimable	13/33	20/33	.42(.16–1.13)
Total	78/162	86/168	.90(.44–1.81)	3/127	10/133	.38(.08–1.20)	55/127	58/133	1.05(.39–2.85)
Heterogeneity χ2 (P value)			9.20(.056)			.60(.74)			9.48(.024)
Excluding Lebrec et al10	69/149	82/156	.74(.40–1.37)	1/114	7/121	.20(.03–1.22)	48/114	57/121	.75(.35–1.64)
Heterogeneity χ2 (P value)			5.27(.15)			.07(.80)			4.21(.12)

PARA, paracentesis; n/N, number of patients with the outcome/total number of patients.

Table 7. Metaregression Analysis for the Treatment Effect on Mortality

NOTE. The dependent variable is the observed log–odds ratio for mortality from each trial. Weights have been assigned according to the estimated variance of the log–odds ratios. τ² is a measure of residual between trial heterogeneity, estimated by a restricted maximum likelihood method using an iterative procedure.

SE, standard error of the coefficient; PSE, portosystemic encephalopathy.

To improve the power of the meta-analysis, we combined the available trials independently on whether total or large-volume paracentesis was used as a control treatment, assuming that they were equivalent interventions. This assumption also was supported by the observation that incidence of the outcomes of interest in the control groups of the included RCTs was independent of the type (total or large-volume) of paracentesis used.

Treatment outcomes were expressed as odds ratios (ORs) and pooled odds ratios (PORs), with 95% confidence intervals (CIs) estimated by a random-effects model according to Der Simonian and Laird.²⁵ For statistically significant treatment effects (ie, when the 95% CI of the POR excludes 1), the number needed to treat for benefit or for harm and the corresponding 95% CIs were calculated.²⁶, ²⁷

A χ² test for heterogeneity was performed per each assessed outcome and potential sources of heterogeneity were discussed where appropriate. A P value of less than .10 was considered indicative of statistically significant heterogeneity.²⁸ To explore the potential effect of several patients or trial characteristics on the pooled estimates, a meta-regression analysis was performed for the outcomes showing a statistically significant heterogeneity.²⁹ The dependent variable was the observed log-ORs from each trial for the outcome of interest. Weights were assigned according to the estimated variance of log-ORs.³⁰ The residual between trials heterogeneity was expressed by τ² estimated by a restricted maximum likelihood method using an iterative procedure.³⁰ The following covariates were included in the meta-regression analysis, including covariates previously shown to be associated with survival in patients with refractory ascites treated by TIPS¹¹: age, sex, bilirubin level,¹¹, ¹⁴ albumin level, portosystemic encephalopathy at randomization, Child–Pugh score, rate of successful TIPS placement, rate of TIPS-assisted patency, and the year of publication as a proxy for technical improvement with time. For multivariate analysis the number of covariates was reduced according to the number of retrieved studies to variable clusters (expressing clinical or technical aspects of studies) and to clinical judgment. If the meta-regression showed a significant association between covariates and the intervention effect, stratified meta-analyses were performed according to the covariates of interest when appropriate.

A sensitivity analysis using a fixed-effects model also was performed to assess the consistency of results. To assess the influence of the quality of RCTs on the treatment effect, sensitivity analyses were planned according to each of the predefined quality items.

Publication bias was assessed by the Begg and Mazumadar³¹ adjusted rank-correlation test for publication bias and by the Egger et al³² regression asymmetry test for publication bias.

All the statistical analyses were performed by STATA 7 (Stata Corporation, College Station, TX).

Back to Article Outline

Results 

A total of 64 articles or abstracts were retrieved. We excluded 35 references because they were duplicates or were determined to be irrelevant by reading the abstracts. The other 14 references were excluded because they were not RCTs or TIPS was not compared with medical therapy. The remaining 5 articles were RCTs,¹⁰, ¹¹, ¹², ¹³, ¹⁴ fulfilling the inclusion criteria and were included in the meta-analysis. The 5 RCTs included a total of 330 patients: 162 treated with TIPS and 168 treated with paracentesis. Clinical and methodologic characteristics of the 5 RCTs are shown in Table 1 and patient characteristics are shown in Table 2. The control treatment was large-volume paracentesis in 3 studies¹⁰, ¹¹, ¹⁴ and total paracentesis in the other 2 studies.¹², ¹³ In all 5 studies, intravenous albumin infusion was associated with paracentesis. The dose of infused albumin was 8 g/L of paracentesis in 3 studies¹¹, ¹², ¹⁴ and 6–8 g/L in another study,¹³ whereas it was not reported in 1 study.¹⁰ In 3 RCTs only patients with refractory ascites were included, whereas in the other 2 studies a proportion of patients had recurrent ascites (Table 2). Major exclusion criteria across the 5 trials were encephalopathy greater than grade I, renal function impairment, high serum bilirubin concentration, and hepatocellular carcinoma. However, the cut-off values for creatinine level as an exclusion criterion ranged from 1.7 to 3 mg/dL and for bilirubin level ranged from 5 to 10 mg/dL (Table 1), suggesting relatively heterogeneous trial populations. In fact, the proportion of Child–Pugh class C patients in the randomized groups ranged from 23% to 79% and the prevalence of encephalopathy ranged from 15% to 46% (Table 2). The mean daily urine sodium excretion was less than 10 mmol in 2 studies.¹⁰, ¹² The proportion of patients with previous episodes of gastrointestinal bleeding was reported in 3 RCTs and ranged from 18% to 34%.¹², ¹³, ¹⁴ The mean follow-up period ranged from 7.5 to 44 months.

The outcome definitions and major end points used in the 5 studies are shown in Table 3.

The quality of the studies was not assessable properly mainly because details on the relevant items were not available: the method for generating the randomization list was not reported in any of the studies; treatment assignment concealment was reported in 3 studies¹⁰, ¹², ¹⁴; in none of the studies was the treatment outcome assessed blindly; however, all 5 studies reported the intention-to-treat analysis (Table 1).

Technical Results 

TIPS implantation was successful in 77%–100% of randomized patients (median, 94%) (Table 4). The mean reduction of the portosystemic pressure gradient ranged from 6.0 to 14 mm Hg (median, 11.5 mm Hg). TIPS dysfunction requiring stent revision occurred in 30%–73% of patients and TIPS-assisted patency was achieved in 46%–93% of randomized patients (Table 4).

Overall Estimates 

The results of each study and PORs for recurrence of ascites, encephalopathy (total and severe), and mortality (total, from bleeding and from other causes) are reported in detail in Table 5, Table 6. Combined results of the 5 studies showed that TIPS significantly reduced the recurrence of ascites (number needed to treat for benefit, 2.6; 95% CI, 1.9–3.7) and increased the incidence of both total encephalopathy (number needed to treat for harm, 6; 95% CI, 4–14) and severe encephalopathy (number needed to treat for harm, 6.6; 95% CI, 4–20). Total mortality, mortality from bleeding, and from other causes were not significantly different between the 2 treatments.

Heterogeneity, Metaregression, and Sensitivity Analyses 

Because a significant heterogeneity was found for overall mortality and for mortality from causes other than bleeding, we performed first a metaregression for overall mortality. At univariate analysis significant variables were lower age and higher bilirubin levels (Table 7). The significant association of bilirubin with death risk after TIPS was expected¹¹, ¹⁴; however, the inverse association of age with mortality in TIPS patients was difficult to explain. For this reason we performed a multivariate metaregression analysis including age, bilirubin level, and the successful TIPS placement rate as the best surrogate marker of physician technical skill. In this analysis, age was not significant and by removing it from the analysis the statistical significance of bilirubin level was confirmed and the successful TIPS rate also was significant in the final model (Table 7, Figure 1).

• 
  • View Large Image
  • Download to PowerPoint

• Figure 1. 

  Log-OR for mortality in the 5 trials included in the meta-analysis, according to the 2 variables significant at multivariate metaregression analysis, together with the summary random-effects metaregression line. Each trial is represented by an open circle and may be identified by the corresponding value of each variable as reported in Table 1, Table 3, Table 4. The area of each circle is proportional to the variance of each log-OR estimate.

Altogether, the metaregression analysis (Figure 1) and the distribution across the studies of several covariates according to the ORs for mortality with TIPS (Figure 2) suggested that the results of the metaregression were explained at least in part by the Lebrec et al¹⁰ study, which seemingly behaved as an outlier. In fact, this study was the only study showing a significantly increased OR for mortality with TIPS, and achieved the lowest successful TIPS placement rate, the lowest portal pressure gradient reduction, and the lowest TIPS-assisted patency rate, indicating a lower expertise of physicians performing TIPS compared with the subsequent trials published 4–8 years later (Figure 1, Figure 2).¹¹, ¹², ¹³, ¹⁴ This is also in accordance with the nonsignificant trend toward an association of the year of publication with the log-OR for mortality with TIPS (Table 7). Therefore, we performed separate analyses by excluding this trial.¹⁰

• 
  • View Large Image
  • Download to PowerPoint

• Figure 2. 

  OR for mortality in the 5 trials included in the meta-analysis, according to 6 relevant variables. Each trial is represented by an open circle and may be identified by the corresponding value of each variable as reported in Table 1, Table 3, Table 4. The area of each circle is proportional to the variance of each OR estimate. The horizontal dashed line is the equivalence line, for which the OR is equal to 1. The trial with the highest OR always had the extreme value for each variable.

Pooled Estimates by Excluding the Potentially Outlier Trial 

The 4 RCTs remaining in the analysis¹¹, ¹², ¹³, ¹⁴ included a total of 305 patients, 149 treated by TIPS and 156 treated by paracentesis. TIPS significantly reduced the recurrence of ascites in the 4 remaining RCTs.¹¹, ¹², ¹³, ¹⁴ The overall recurrence rate was .44 with TIPS and .87 with paracentesis. POR was .14 (95% CI, .07–.27) (Table 5; Figure 3); no statistically significant heterogeneity was found. The number needed to treat for benefit was 2.3 (95% CI, 1.8–3.0), meaning that for every 2 to 3 patients treated by TIPS, 1 episode of recurrent ascites was saved as compared with the control treatment.

• 
  • View Large Image
  • Download to PowerPoint

• Figure 3. 

  OR (random-effects model), log scale for recurrence of ascites (top), encephalopathy (middle), and mortality (bottom) by including and excluding 1 outlier trial.¹⁰ Trials are identified by the first author and the year of publication. Open squares indicate the OR per each trial and the size of the squares is proportional to the weight of trials. The horizontal bars denote the 95% CIs of the ORs. The vertical solid line is the equivalence line, for which the OR is equal to 1. ORs on the left of the equivalence line denote benefit with TIPS whereas those on the right denote harm with TIPS. The vertical dashed line indicates the POR by excluding the outlier trial.

TIPS increased the incidence of encephalopathy in all 4 RCTs, significantly in 1 RCT (Table 5; Figure 3).¹³ The overall encephalopathy rate was .48 with TIPS and .33 with paracentesis. POR was 2.26 (95% CI, 1.35–3.76), without significant heterogeneity. The number needed to treat for harm was 6 (95% CI, 4–17). Three RCTs (245 patients) reported the incidence of severe encephalopathy (Table 5)¹², ¹³, ¹⁴: it was increased by TIPS in all, and significantly in 1 RCT (Table 5).¹² The overall rate of severe encephalopathy was .42 with TIPS and .26 with paracentesis. POR was 2.10 (95% CI, 1.21–3.67) (Table 5). No heterogeneity was found. The number needed to treat for harm was 6.6 (95% CI, 4–20).

The OR for mortality was decreased almost significantly by TIPS in 2 RCTs,¹¹, ¹⁴ and was nearly equivalent with both treatments in the other 2 RCTs (Table 6; Figure 3).¹², ¹³ The overall death rate was .46 with TIPS and .52 with paracentesis. The POR was .74 (95% CI, .40–1.37) without heterogeneity (χ² = 5.27, P = .15). Three RCTs involving 235 patients reported on mortality rates from bleeding (Table 6).¹¹, ¹³, ¹⁴ Only 1 patient died of bleeding in the TIPS groups and 7 patients died in the paracentesis groups. The overall rate of death from bleeding was .009 with TIPS and .06 with paracentesis. In 1 trial¹⁴ no patient died of bleeding in either group and the POR, estimated from 2 studies,¹¹, ¹³ was .20 (95% CI, .03–1.22). There was no significant heterogeneity (Table 6). Mortality from causes other than bleeding was reduced almost significantly in 1 study¹⁴ and was not different in 2 studies (Table 6).¹¹, ¹³ The overall rate of death from causes other than bleeding was .42 with TIPS and .47 with paracentesis. The POR was .75 (95% CI, .35–1.64) without statistically significant heterogeneity (Table 6).

The sensitivity analyses by the fixed-effects model yielded very similar results to the random model (data not shown). The available data were insufficient to allow sensitivity analyses according to the methodologic quality of the included studies.

Adverse Events and Other Relevant Outcomes 

Adverse events were reported incompletely in most trials. TIPS-related complications were reported in only 2 studies: liver failure in 3 patients, cardiac failure in 4, and severe hemolytic anemia in 3¹²; acute stent thrombosis with transient ischemic attack caused by cerebral embolism was reported in 1 patient in whom an interatrial communication subsequently was shown.¹⁴ Other adverse events were reported only in 2 studies¹³, ¹⁴: sepsis in 6 patients after TIPS and in 10 patients after paracentesis. Hepatorenal syndrome occurred in 7 patients treated with TIPS and in 18 patients treated with paracentesis.

Among patients randomized to TIPS, 14 of 162 (8.6%) underwent paracentesis mainly because of failure to place the stent; among patients randomized to paracentesis, 26 of 168 (15.4%) underwent TIPS, mainly because of uncontrolled ascites or bleeding.

Liver transplantation was performed in 28 of 162 (17.3%) patients in the TIPS groups and 31 of 168 (18.4%) patients in the paracentesis groups. The difference was not statistically significant and the indication for liver transplantation was not reported in any study.

Publication Bias 

Egger and Begg tests for publication bias showed that the risk for having missed trials was acceptably low: for recurrence of ascites, encephalopathy, and mortality the P values ranged from .11 to .90 by the Egger’s test and from .43 to .81 by the Begg’s test.³¹, ³² The Egger’s publication bias plot for the risk difference for mortality is shown in Figure 4.

• 
  • View Large Image
  • Download to PowerPoint

• Figure 4. 

  Egger’s publication bias plot. The standardized risk difference (TIPS − paracentesis) for mortality is plotted against the precision of the risk difference along with the regression line and the confidence interval (arrows) about the intercept. Failure of this confidence interval to include 0 would indicate asymmetry in the funnel plot and might give evidence of publication bias. Standardized risk difference, risk difference/standard error; precision of risk difference, 1/standard error.

Back to Article Outline

Discussion 

This systematic review showed that in patients with refractory or recurrent ascites, TIPS is superior to large-volume paracentesis for controlling ascites and, although it significantly increases the risk for encephalopathy, it is associated with a trend toward a reduction of mortality.

The effect on survival was significantly heterogeneous across the 5 available studies with 1 trial showing a nearly significant increase of death risk,¹⁰ 2 trials showing the opposite,¹¹, ¹⁴ and 2 trials showing nearly equivalent survival with the 2 treatments.¹², ¹³ The metaregression analysis showed that the mean serum bilirubin concentration and the proportion of successful TIPS procedures were associated independently with a reduction of the log-OR for death with TIPS and almost completely explained the residual between trial heterogeneity (τ² = .01). The association of bilirubin with post-TIPS survival previously was reported consistently¹¹, ³³, ³⁴, ³⁵, ³⁶ and the metaregression we performed provides evidence of this association from a meta-analysis of RCTs. Although we could not assess the best cut-off value of bilirubin to predict mortality after TIPS (individual patient data would be needed for this analysis), it has been suggested that mortality after TIPS is significantly higher when the pre-TIPS bilirubin level is greater than 3 mg/dL.³⁴, ³⁵

It also is important to note that inclusion of patients with a low creatinine level but a high bilirubin level might have favored the paracentesis groups³⁷ by reducing postparacentesis complications without accounting for the increasing risk with increasing bilirubin level for TIPS-treated patients. Although the creatinine level was not significant in the metaregression analysis, the different bilirubin and creatinine thresholds for patient selection may have contributed to the variability of results as well as to the different patient inclusion rate across the available studies (Table 1).³⁷

The proportion of successful TIPS procedures clearly is related to the operators’ skill and its clinical relevance is obvious. Although this variable was not significant at univariate analysis, we decided to include it in the final multivariate analysis because we believed it was the most reliable measure of the operator skill whereas the TIPS-assisted patency rate would be redundant in this regard. In this regard, it is worth noting that we did not include the portal pressure reduction achieved after TIPS in the metaregression analysis because this variable may not be related to the operator expertise. The trend toward an association of the year of publication with the log-OR for mortality at univariate analysis likely was related to the increasing skill in performing TIPS over time, although it was not statistically significant. By contrast, the inverse association of age with mortality in TIPS patients at univariate analysis (Table 7) was difficult to explain and the multivariate analysis showed that this association was not independent and that it most likely depended on the inclusion of the youngest patients in the trial with the lowest successful TIPS rate (Figure 2).¹⁰ Because the covariates significantly associated with the log-OR for mortality at univariate metaregression analysis and several other variables had extreme values in the only trial showing a significant increase in mortality with TIPS, we considered this an outlier trial. In fact, by excluding this study from the meta-analysis, heterogeneity disappeared and a trend toward a reduction in mortality with TIPS was found.

Technical results of the 5 studies clearly indicated that the first of them¹⁰ was performed possibly in an early stage of the learning curve as shown by the lower proportion of successfully placed TIPS (77% vs ≥89% in the subsequent studies), the portal pressure gradient reduction after TIPS (6 vs ≥10.4 mm Hg), and the TIPS-assisted patency (46% vs ≥82%). In this respect some heterogeneity of clinical results of the 5 studies might have been anticipated, as probably reflected by the significant heterogeneity found for mortality, and the result of the metaregression analysis supports this interpretation.

Overall, the number of patients included in the available studies was relatively small, suggesting that the interpretation of the cumulative estimates should be cautious. The quality of the trials was not assessable satisfactorily, mainly because of insufficient information on the generation of the randomization list and on the concealment of treatment assignment. However, all the studies reported the intention-to-treat analysis. Blinding is demanding in this type of trial and in fact all the available studies were unblinded. It should be noted, however, that a blind assessment of encephalopathy would be of special importance in TIPS trials.

The effect of TIPS on ascites was predictable because the recurrence of ascites was reduced in all 5 included studies and the reduction failed to reach statistical significance only in the outlier trial.¹⁰ However, this result was homogeneous across the 5 studies and the homogeneity also was confirmed statistically in this meta-analysis. Therefore, what the meta-analysis adds to our previous knowledge, in this respect, is only the pooled estimate of the OR for recurrence of ascites with TIPS compared with paracentesis (.14; 95% CI, .07–.27 by excluding the outlier trial and .14; 95% CI, .08–.26 by including it). It is well accepted that this effect is plausible pathophysiologically³⁸ and it may be assumed that because it has been confirmed in 5 trials and in their meta-analysis, this effect should be generalizable to the population of patients with cirrhosis and refractory ascites.

However, TIPS increased the risk for encephalopathy by approximately twice compared with large-volume paracentesis. Also, this effect was anticipated by each of the included trials, although the increase in risk was statistically significant only in 1 study.¹³ In fact, for this effect we also did not find any statistically significant heterogeneity. Similar considerations apply to severe encephalopathy, which was increased in TIPS arms in all 4 trials reporting this outcome.¹⁰, ¹², ¹³, ¹⁴ It should be emphasized that the increase of severe encephalopathy was statistically significant in 1 study despite the fact that a gradual dilatation was used both in the insertion and in the surveillance of TIPS to achieve a portal pressure gradient of just less than 12 mm Hg.¹² However, it also should be noted that post-TIPS encephalopathy usually was treated successfully by medical treatment.¹⁴ The increase in encephalopathy was similar to that found after TIPS for the prevention of recurrent variceal bleeding.³⁹ However, it has been noted that in the setting of variceal bleeding the increase of encephalopathy is more predictable because more patients surviving variceal bleeding still have otherwise compensated disease as compared with the decompensated patients with refractory ascites.¹⁴ Moreover, it is noteworthy that, similar to the trials of TIPS for the prevention of rebleeding,³⁹ encephalopathy was not assessed blindly in any of the trials for refractory ascites, although a great effort has been made to define and assess it clearly (Table 3). On the other hand, 17%–40% of patients in the included trials had encephalopathy at randomization, although it is known that pre-TIPS encephalopathy is an important predictor of post-TIPS encephalopathy.⁴⁰ Moreover, several other predictors of post-TIPS encephalopathy such as age greater than 60 years, sex, non–alcohol-induced cause of cirrhosis, low albumin levels, Child–Pugh class, and portal perfusion before TIPS have been suggested.⁴¹, ⁴², ⁴³, ⁴⁴, ⁴⁵, ⁴⁶ Therefore, it is conceivable that a better selection of patients might reduce the risk for post-TIPS encephalopathy in patients with refractory ascites as well, and the future use of a covered stent⁴⁷ may contribute further to improving the overall efficiency of TIPS in this setting.

Other adverse events related to the trial treatments were reported systematically in only 1 trial; therefore no firm conclusions may be drawn on this point. It may be noted, however, that although sepsis and hepatorenal syndrome were reported more frequently in association with paracentesis treatment,¹², ¹³ liver and heart failure and hemolytic anemia were reported more frequently with TIPS.

This meta-analysis showed that TIPS only achieved a trend toward a reduction of mortality, which failed to reach statistical significance, even though it achieved an almost complete control of ascites and a substantial reduction of portal pressure. A reduction of mortality therefore was expected through the reduction of the overall risk for portal hypertension–related unfavorable outcomes. Although information on portal hypertension–related outcomes was not reported systematically in all the studies, TIPS showed a trend toward a reduced mortality that did not satisfy the expectations, possibly because of its negative effect on liver function, which also was expressed by the significant increase of encephalopathy.

It is worth noting that although overall very few patients died of bleeding, the OR for this cause of mortality was reduced almost significantly with TIPS (.20; 95% CI, .03–1.22), suggesting a type II error of this estimate.

The assessment of quality of life and costs of TIPS and paracentesis in patients with refractory ascites is of obvious relevance for clinical practice. However, no conclusions may be drawn from the available studies on these two aspects because each of them was assessed in only 1 study.¹², ¹³

In conclusion, this systematic review showed that TIPS significantly improves the control of refractory ascites although it increases the risk for encephalopathy. However, the important trend toward a reduction of mortality clearly calls for further studies in this field. In fact, post-TIPS encephalopathy might be reduced by excluding patients with pre-TIPS encephalopathy and with other predictors of the risk for post-TIPS encephalopathy. Moreover, the trend toward improved survival shown by excluding from the analysis the outlier trial might become a significant benefit when more studies are added. Costs and quality of life also should be assessed in future TIPS trials.

Back to Article Outline

Drs. G. D’Amico and Luca equally contributed to the study design and protocol, data abstraction, interpretation of results, and text. Dr. G. D’Amico and Professor Morabito performed the statistical analyses. Drs. Miraglia and M. D’Amico contributed to the search of pertinent articles, data recording forms, database, and interpretation of results.

Back to Article Outline

References 

1. Arroyo V , Colmenero J . Ascites and hepatorenal syndrome in cirrhosis (pathophysiological basis of therapy and current management) . J Hepatol . 2003;38:S69–S89
   • View In Article
2. Bosch J , Garcia-Pagan JC . Complications of cirrhosis. I. Portal hypertension . J Hepatol . 2000;32:141–146
   • View In Article
   • Abstract
   • Full-Text PDF (2246 KB)
   • CrossRef
3. Gines P , Guevara M , Arroyo V , Rodes J . Hepatorenal syndrome . Lancet . 2003;362:1819–1827
   • View In Article
   • Abstract
   • Full Text
   • Full-Text PDF (122 KB)
   • CrossRef
4. Moller S , Henriksen H . Review article (pathogenesis and pathophysiology of hepatorenal syndrome—is there a scope for prevention?) . Aliment Pharmacol Ther . 2004;20(Suppl 3):31–41
   • View In Article
   • CrossRef
5. Runyon BA . Refractory ascites . Semin Liver Dis . 1993;13:343–351
   • View In Article
   • MEDLINE
6. Russo MW , Sood A , Jacobson IM , Brown SR . Transjugular intrahepatic portosystemic shunt for refractory ascites (an analysis of the literature on efficacy, morbidity and mortality) . Am J Gastroenterol . 2003;98:2521–2527
   • View In Article
   • MEDLINE
   • CrossRef
7. Ochs A, Rossle M, Haag K, Hauenstein KH, Deibert P, Siegerstetter V, et al. The transjugular intrahepatic portosystemic stent-shunt procedure for refractory ascites . N Engl J Med . 1995;332:1192–1197
   • View In Article
   • MEDLINE
   • CrossRef
8. Quiroga J, Sangro B, Nuñez M, Bilbao I, Longo J, Garcia-Villarreal L, et al  Transjugular intrahepatic portosystemic shunt in the treatment of refractory ascites (effect on clinical, renal, humoral, and hemodynamic parameters) . Hepatology . 1995;21:986–994
   • View In Article
   • MEDLINE
   • CrossRef
9. Forrest EH , Stanley AJ , Redhead DN , McGilchrist AJ , Hayes PC . Clinical response after transjugular intrahepatic portosystemic stent shunt insertion for refractory ascites in cirrhosis . Aliment Pharmacol Ther . 1996;10:801–806
   • View In Article
   • MEDLINE
10. Lebrec D, Giuily N, Hadengue A, Vilgrain V, Moreau R, Poynard T, et al  Transjugular intrahepatic portosystemic shunts: comparison with paracentesis in patients with cirrhosis and refractory ascites: a randomized trial . J Hepatol . 1996;25:135–144
    • View In Article
    • Abstract
    • Full-Text PDF (1184 KB)
    • CrossRef
11. Rossle M, Ochs A, Gulberg V, Siegerstetter V, Holl J, Deibert P, et al  A comparison of paracentesis and transjugular intrahepatic portosystemic shunting in patients with ascites . N Engl J Med . 2000;342:1701–1707
    • View In Article
    • MEDLINE
    • CrossRef
12. Gines P, Uriz J, Calahorra B, Garcia-Tsao G, Kamath PS, Del Arbol LR, et al  Transjugular intrahepatic portosystemic shunting versus paracentesis plus albumin for refractory ascites in cirrhosis . Gastroenterology . 2002;123:1839–1847
    • View In Article
    • Abstract
    • Full Text
    • Full-Text PDF (106 KB)
    • CrossRef
13. Sanyal AJ, Genning C, Reddy KR, Wong F, Kowdley KV, Benner K, et al  The North American Study for the Treatment of Refractory Ascites . Gastroenterology . 2003;124:634–641
    • View In Article
    • Abstract
    • Full Text
    • Full-Text PDF (128 KB)
    • CrossRef
14. Salerno F, Merli M, Riggio O, Cazzaniga M, Valeriano V, Pozzi M, et al. Randomized controlled study of TIPS versus paracentesis plus albumin in cirrhosis with severe ascites . Hepatology . 2004;40:629–635
    • View In Article
    • MEDLINE
    • CrossRef
15. Pugh RN , Murray-Lyon IM , Dawson JL , Pietrni MC , Williams R . Transection of the esophagus for bleeding oesophageal varices . Br J Surg . 1973;60:646–649
    • View In Article
    • MEDLINE
    • CrossRef
16. Saab S , Nieto JM , Ly D , Runyon BA . TIPS versus paracentesis for cirrhotic patients with refractory ascites . Cochrane Database Syst Rev . 2004;3:CD004889
    • View In Article
17. Conn HO , Liebertal MM . The hepatic coma syndromes and lactulose . Baltimore: Williams & Wilkins; 1979; 46–84
    • View In Article
18. Jadad AR, Moore RA, Carroll D, Jenkinson C, Reynolds JM, Gavaghan DJ, et al. Assessing the quality of reports of randomized clinical trials (is blinding necessary?) . Control Clin Trials . 1996;17:1–12
    • View In Article
    • Abstract
    • Full-Text PDF (891 KB)
    • CrossRef
19. Moher D , Pham B , Jones A , Cook DJ , Jadad AR , Moher M , et al.   Does quality of reports of randomised trials affect estimates of intervention efficacy reported in meta-analyses? . Lancet . 1998;352:609–613
    • View In Article
    • Abstract
    • Full Text
    • Full-Text PDF (45 KB)
    • CrossRef
20. Juni P , Altman DG , Egger M . Assessing the quality of controlled clinical trials . BMJ . 2001;323:42–46
    • View In Article
    • MEDLINE
    • CrossRef
21. Kjaergard LL , Villumsen J , Gluud C . Reported methodologic quality and discrepancies between large and small randomized trials in meta-analyses . Ann Intern Med . 2001;135:982–989
    • View In Article
    • MEDLINE
22. Schulz KF , Grimes DA . Generation of allocation sequences in randomised trials (chance not choice) . Lancet . 2002;359:515–519
    • View In Article
    • Abstract
    • Full Text
    • Full-Text PDF (84 KB)
    • CrossRef
23. Schulz KF , Grimes DA . Allocation concealment in randomised trials (defending against deciphering) . Lancet . 2002;359:614–618
    • View In Article
    • Abstract
    • Full Text
    • Full-Text PDF (304 KB)
    • CrossRef
24. Schultz KF , Grimes AD . Blinding in randomised trials . Lancet . 2002;359:696–700
    • View In Article
    • Abstract
    • Full Text
    • Full-Text PDF (250 KB)
    • CrossRef
25. Der Simonian R , Laird N . Meta-analysis in clinical trials . Control Clin Trials . 1986;7:177–188
    • View In Article
    • MEDLINE
    • CrossRef
26. Laupacis A , Sackett DL , Roberts AS . An assessment of clinically useful measures of the consequences of treatment . N Engl J Med . 1988;318:1728–1733
    • View In Article
    • MEDLINE
    • CrossRef
27. Altman DG . Confidence intervals for the number needed to treat . BMJ . 1998;317:1309–1312
    • View In Article
    • MEDLINE
    • CrossRef
28. Petitti D . Approaches to heterogeneity in meta-analysis . Stat Med . 2001;20:3625–3633
    • View In Article
    • MEDLINE
    • CrossRef
29. Thompson SG , Higgins JPT . How should meta-regression analyses be undertaken and interpreted . Stat Med . 2002;21:1559–1573
    • View In Article
    • MEDLINE
    • CrossRef
30. Thompson SG , Sharp SG . Explaining heterogeneity in meta-analysis (a comparison of methods) . Stat Med . 1999;18:2693–2708
    • View In Article
    • MEDLINE
31. Begg CB , Mazumadar M . Operating characteristics of a rank correlation test for publication bias . Biometrics . 1994;50:1088–1098
    • View In Article
    • MEDLINE
    • CrossRef
32. Egger M , Smith DG , Schneider M , Minder C . Bias in meta-analysis detected by a simple, graphical test . BMJ . 1997;315:629–634
    • View In Article
    • MEDLINE
    • CrossRef
33. Malinchoc M , Kamath P , Gordon FG , Peine CJ , Rank J , Ter Borg CGP . A model to predict poor survival in patients undergoing transjugular intrahepatic portosystemic shunts . Hepatology . 2000;31:864–871
    • View In Article
    • MEDLINE
    • CrossRef
34. Gerbes AL , Gulberg V . Benefit of TIPS for patients with refractory or recidivant ascites (serum bilirubin may make the difference) . Hepatology . 2005;41:217
    • View In Article
    • MEDLINE
    • CrossRef
35. Rajan KD , Haskal ZJ , Clark TWI . Serum bilirubin and early mortality after transjugular intrahepatic portosystemic shunts (results of a multivariate analysis) . J Vasc Interv Radiol . 2002;13:156–161
    • View In Article
36. Brensing AK, Raab P, Textor J, Gorich J, Schiedermaier P, Strunk H, et al  Prospective evaluation of a clinical score for 60-day mortality after transjugular intrahepatic portosystemic stent-shunt (Bonn TIPS early mortality analysis) . Eur J Gastroenterol Hepatol . 2002;14:723–731
    • View In Article
    • MEDLINE
    • CrossRef
37. Rossle M . Discussion on the North American study for the treatment of refractory ascites . Gastroenterology . 2004;126:1214–1215
    • View In Article
    • Full Text
    • Full-Text PDF (59 KB)
    • CrossRef
38. Garcia-Tsao G . Treatment of refractory ascites (TIPS or TAPS) . Hepatology . 2001;33:477–479
    • View In Article
    • MEDLINE
    • CrossRef
39. Burroughs AK , Vangeli M . Transjugular intrahepatic portosystemic shunt versus endoscopic therapy. Randomized trials for secondary prophylaxis of variceal bleeding: an updated meta-analysis . Scand J Gastroenterol . 2002;37:249–252
    • View In Article
    • MEDLINE
    • CrossRef
40. Sanyal AJ , Freedman AM , Shiffman ML , Purdum PP , Luketic VA , Cheatham AK . Portosystemic encephalopathy after transjugular intrahepatic portosystemic shunt (results of a prospective controlled study) . Hepatology . 1994;20:46–55
    • View In Article
    • MEDLINE
    • CrossRef
41. Somberg KA , Riegler JL , Doherty M , et al.   Hepatic encephalopathy after transjugular intrahepatic portosystemic shunts (incidence and risk factors) . Am J Gastroenterol . 1995;4:549–555
    • View In Article
42. Cabrera J , Maynar M , Granados R , et al.   Tansjugular intrahepatic portosystemic shunt versus sclerotherapy in the elective treatment of variceal hemorrhage . Gastroenterology . 1996;110:832–839
    • View In Article
    • Abstract
    • Full-Text PDF (123 KB)
    • CrossRef
43. Rössle M , Deibert P , Haag K , et al.   Randomised trial of transjugular-intrahepatic-portosystemic shunt versus endoscopy plus propranolol for prevention of variceal rebleeding . Lancet . 1997;249:1043–1049
    • View In Article
44. Sauer P , Theilmann L , Stremmel W , Benz C , Richter GM , Stiehl A . Transjugular intrahepatic portosystemic stent shunt versus sclerotherapy plus propranolol for rebleeding . Gastroenterology . 1997;113:1623–1631
    • View In Article
    • Abstract
    • Full-Text PDF (369 KB)
    • CrossRef
45. Nolte W, Wiltfang J, Schindler C, Munke H, Unterberg K, Zumhasch U, et al  Portosystemic hepatic encephalopathy after transjugular intrahepatic portosystemic shunt in patients with cirrhosis (clinical, laboratory, psychometric and electroencephalographic investigations) . Hepatology . 1998;28:1215–1225
    • View In Article
    • MEDLINE
    • CrossRef
46. Hassoun Z, Deschenes M, Lafortune M, Dufresne MP, Perrault P, Lepanto L, et al  Relationship between pre-TIPS liver perfusion by the portal vein and the incidence of post-TIPS chronic hepatic encephalopathy . Am J Gastroenterol . 2001;96:1205–1209
    • View In Article
    • MEDLINE
    • CrossRef
47. Bureau C, Garcia-Pagan JC, Otal P, Pomier-Layrargues G, Chabbert V, Cortez C, et al  Improved clinical outcome using polytetrafluoroethylene-coated stents for TIPS (results of a randomized study) . Gastroenterology . 2004;126:469–475
    • View In Article
    • Abstract
    • Full Text
    • Full-Text PDF (179 KB)
    • CrossRef
48. Ginès P, Arroyo V, Vargas B, Planas R, Casafont F, Panés J, et al  Paracentesis with intravenous infusion of albumin as compared with peritoneovenous shunting in cirrhosis with refractory ascites . N Engl J Med . 1991;325:829–835
    • View In Article
    • MEDLINE
    • CrossRef
49. Arroyo V, Ginès P, Gerbes AL, Dudley JF, Gentilini P, Laffi G, et al  Definitions and diagnostic criteria of refractory ascites and hepatorenal syndrome in cirrhosis . Hepatology . 1996;23:164–176
    • View In Article
    • MEDLINE
    • CrossRef
50. Conn HO, Leevy CM, Vlahcevic ZR, Rodgers JB, Maddrey WC, Seef L, et al. Comparison of lactulose and neomycin in the treatment of chronic portal-systemic encephalopathy . Gastroenterology . 1977;72:573–583
    • View In Article
    • Abstract