The transjugular intrahepatic portosystemic shunt (TIPS) has been shown to be effective in the control of refractory or recidivant ascites. However, the effect of TIPS on survival as compared with that of large-volume paracentesis plus albumin is uncertain. A multicenter, prospective, clinical trial was performed in 66 patients with cirrhosis and refractory or recidivant ascites (16 Child-Turcotte-Pugh class B and 50 Child-Turcotte-Pugh class C) randomly assigned to treatment with TIPS (n = 33) or with large-volume paracentesis plus human albumin (n = 33). The primary endpoint was survival without liver transplantation. Secondary endpoints were treatment failure, rehospitalization, and occurrence of complications. Thirteen patients treated with TIPS and 20 patients treated with paracentesis died during the study period, 4 patients in each group underwent liver transplantation. The probability of survival without transplantation was 77% at 1 year and 59% at 2 years in the TIPS group as compared with 52% and 29% in the paracentesis group (P = .021). In a multivariate analysis, treatment with paracentesis and higher MELD score showed to independently predict death. Treatment failure was more frequent in patients assigned to paracentesis, whereas severe episodes of hepatic encephalopathy occurred more frequently in patients assigned to TIPS. The number and duration of rehospitalizations were similar in the two groups. In conclusion, compared to large-volume paracentesis plus albumin, TIPS improves survival without liver transplantation in patients with refractory or recidivant ascites. (HEPATOLOGY 2004;40:629–635.)
Refractory and recidivant ascites are severe complications of patients with advanced cirrhosis. The first treatment option for both is repeated large-volume paracentesis plus albumin followed by diuretic drugs. Such treatment, however, may be associated with adverse effects and does not improve the outcome of these patients. Elevated portal pressure is the main determinant of ascites, and resolving portal hypertension by means of a portosystemic shunt was shown to improve ascites.1 Transjugular intrahepatic portosystemic shunt (TIPS) was reported to be effective in the treatment of several patients with severe ascites.2–5 However, the few randomized controlled trials which compared TIPS with large-volume paracentesis gave controversial results with regard to patients' survival.6–9 Lebrec et al.6 warned against an increased rate of death in patients with Child C cirrhosis treated with TIPS. Conversely, Rössle et al.7 reported that treatment with TIPS was independently correlated with a longer survival. In two more recent trials, however, the survival of patients with cirrhosis treated with TIPS was similar to that observed in patients treated with large-volume paracentesis.8, 9
We conducted a multicenter, prospective, randomized trial in patients with cirrhosis and severe ascites comparing TIPS with repeated large-volume paracentesis plus albumin.
TIPS, transjugular intrahepatic portosystemic shunt; PRA, plasma rennin activity; PPG, portosystemic pressure gradient; MELD, model of endstage liver disease.
Patients and Methods
A multicenter, unblind, prospective, randomized, clinical trial was performed in three tertiary care settings and approved by the local ethics committees of the Universities of Milan and Rome. Written informed consent was obtained from each patient. The primary endpoint of the study was survival without liver transplantation defined as the time elapsing from randomization to death or liver transplantation.
Secondary endpoints were the failure of treatment, the occurrence of complications and the need for rehospitalization. A treatment was considered to have failed when the patient received at least 4 large-volume paracentesis within the last month for episodes of recurrent tense ascites. These were defined as the development of ascites large enough to cause marked discomfort and impairment of breathing and/or mobility.
Shunt malfunction was defined as the angiographic evidence of stent stenosis associated with a rise of portosystemic pressure gradient (PPG) over 12 mmHg. The lack of success in establishing TIPS or in rehabilitating an occluded stent was not considered a failure per se.
Hepatic encephalopathy was graded on a scale from 1 to 4 according to Conn and Liberthal.10 Episodes reaching grade 3 or 4 were considered severe episodes.
All endpoints of the study were prespecified. Each event (failure or complication) was adjudicated by the main investigators of each center and confirmed by a central control of the criteria employed.
The sample size was calculated on the basis of a 1-year survival rate of 30% for patients with tense ascites treated with repeated paracentesis,11, 12 and 65% for those treated with TIPS.2 To detect such a difference, a minimum of 64 patients was required (type I error, 0.05; type II error, 0.20). Randomization (sealed opaque envelopes), in blocks of six, was centralized in the department of Internal Medicine of Milan, and was independent for each hospital. The correct order of randomization was verified before the analysis of the results.
All patients with cirrhosis and refractory or recidivant ascites, consecutively admitted to our departments from February 1996 to April 2002, were considered. Refractory ascites were defined by the lack of response to a low sodium diet and spironolactone 400 mg/day and furosemide 160 mg/day, or occurrence of serious diuretic-induced complications. Recidivant ascites were defined by the recurrence of at least 3 episodes of tense ascites within a 12-month period despite prescription of low sodium diet and adequate diuretic doses. These definitions were in agreement with those established by a consensus meeting of the International Ascites Club.13 Exclusion criteria were age above 72 years, a history of recurrent episodes of hepatic encephalopathy of grade 2 or more, a serum bilirubin level greater than 6 mg/dL, a serum creatinine level greater than 3 mg/dL, a Child-Turcotte-Pugh score higher than 11, complete portal vein thrombosis, hepatocellular cancer, recent gastrointestinal bleeding (<15 days), serious cardiac or pulmonary dysfunctions, ongoing bacterial infection, and a serum-ascites albumin gradient lower than 11 g/L. In regard to consuming alcohol at presentation, patients were educated to stop their drinking habits, and patients' relatives were informed of the importance of abstinence. Abstinence during follow-up was monitored by means of laboratory tests, relatives' queries and, when abuse was suspected, blood alcohol determinations.
Patients were randomly allocated to TIPS procedure (n = 33) or to treatment with total paracentesis plus albumin (n = 33). The follow-up ended in November 2002.
Care Before and After Randomization
The diagnostic work-up before and after randomization included the determination of renal and hepatic laboratory tests, plasma renin activity (PRA), analysis of urine and ascitic fluid, electrocardiogram, chest radiography, abdominal color Doppler and ultrasound, and the clinical assessment of hepatic encephalopathy. This work-up was performed during the 2 days preceding the first procedure (TIPS or paracentesis). Diuretic drugs, both before and after treatment, were not withdrawn but their doses adjusted according to the clinical needs and tolerability of each patient. A low sodium diet (80 mEq/day) was prescribed allover the study period. After discharge from the hospital, patients were seen after 1, 3, 6 months and then every 3–6 months or when clinically necessary. At the same intervals, in patients with TIPS, the stent patency was controlled by means of color Doppler. When stenosis or occlusion was suspected, a new angiography with measurement of PPG was performed. When PPG was increased over the threshold of 12 mmHg, the stent patency was restored by angioplasty.
Paracentesis was performed as previously described,14 and was combined with the intravenous infusion of human albumin (8 grams per liter of fluid removed). TIPS procedure was performed as previously described.15 A puncture needle was advanced transjugularly into a hepatic vein. After successful puncture of the portal vein, an expandable stent (Wallstent, Schneider, Minneapolis, MN; or Memotherm stent, Angiomed, Karlsruhe, Germany) was placed and dilated to obtain a PPG below 12 mmHg. The interval between randomization and initiation of treatment was always comprised between 2 and 5 days.
In both groups of patients episodes of recurrent tense ascites were treated with a large-volume tap and albumin infusion, and in subjects with TIPS angiography was repeated.
When a failure of treatment was recorded, the patient was allowed to cross to the other treatment, or was recommended for a LeVeen shunt, according to her/his preferences and doctor's advice.
Data are presented as means ± standard error of the means for quantitative variables and as absolute numbers for qualitative variables. All comparisons between the two treatment groups were based on an intention-to-treat principle, and were performed with Mann-Whitney for quantitative data and χ2 or Fisher exact test for qualitative data. The comparisons between baseline and repeated data were performed with Wilcoxon test.
The probabilities of survival without liver transplantation, as well as the other cumulative probabilities, were estimated with the Kaplan-Meier method and compared with the log-rank test. The ability to predict mortality of baseline variables, as well as of treatment assigned, was estimated by univariate and multivariate analyses, using the Cox regression method. For such analyses continuous variables were not dichotomized. Multivariate analysis included only variables which were shown to be significant at the univariate analysis. The results are presented as estimated relative risks with 95% confidence intervals.
All tests of significance were two-sided, and P values <.05 were considered statistically significant. Data processing and analyses were performed with SPSS 11.5 for Windows (SPSS, Chicago, IL).
Of the 137 patients admitted with a diagnosis of refractory or recidivant ascites during the period of study, 63 patients were excluded for the presence of one or more exclusion criteria, and 8 refused to participate (Fig. 1). Of the 66 remaining patients, 33 were allocated to TIPS treatment and 33 to large-volume paracentesis combined with albumin. The number of patients included by each center were, respectively, 32 in Rome (15 patients allocated to TIPS), 24 in Milan (12 patients allocated to TIPS) and 10 in Monza (6 patients allocated to TIPS). The mean follow-up was 18.2±2.3 months in the whole population, 21±3.2 in patients treated with TIPS and 15.1± 3.3 in patients treated with paracentesis (P = .048).
There were no differences in the baseline characteristics between the two groups (Table 1). In particular, the distribution of patients with recidivant ascites was similar in the two groups (9/33 in TIPS group vs. 12/33 in paracentesis group; P = .43). One patient died 24 hours after randomization, before TIPS procedure could be performed. The insertion of TIPS was successful in 29 patients, but failed for technical reasons in 3 patients. Two of these subjects were treated with peritoneovenous shunt and one with paracentesis. The PPG decreased from 22.5±1.1 to 8.7±0.6 mmHg after TIPS. Only one patient experienced a severe complication related to TIPS insertion. He had an acute stent thrombosis with transient ischemic attack caused by cerebrovascular embolism. A persistence of interatrial communication was documented.
Table 1. Baseline Characteristics of the 66 Patients With Cirrhosis Separated According to Assigned Treatment
TIPS (n = 33)
Paracentesis (n = 33)
58 ± 1.3
60 ± 1.3
Refractory ascites (n)
Recidivant ascites (n)
Gastrointestinal bleeding (n)
Hyponatremia (<130 meq/L) (n)
Renal failure (>1.5 mg/dL) (n)
Serum bilirubin (mg/dL)
1.7 ± 0.15
1.9 ± 0.24
1.26 ± 0.03
1.26 ± 0.04
Serum albumin (g/L)
29 ± 0.7
29 ± 0.8
9.4 ± 0.2
9.4 ± 0.2
Child class (B/C)
11.1 ± 0.8
11.1 ± 0.9
10.6 ± 0.3
10.8 ± 0.3
Serum creatinine (mg/dL)
1.12 ± 0.06
1.15 ± 0.09
Serum sodium (meq/L)
133 ± 1
133 ± 1
Serum potassium (meq/L)
4.2 ± 0.1
4.5 ± 0.1
Urine sodium excretion (meq/day)
38 ± 6
38 ± 6
15.7 ± 3.3
12.5 ± 2
Mean arterial blood pressure (mmHg)
84 ± 1
83 ± 1
Heart rate (bpm)
78 ± 2
76 ± 2
Initial paracentesis was successful in all cases and no severe procedure-related complications were recorded.
Thirteen patients in the group of TIPS and 20 in the group of paracentesis died (Table 2). In most patients, death was due to terminal liver failure or hepatorenal syndrome (10 in the TIPS group and 16 in the paracentesis group). The other causes of death were 3 sepsis with multiorgan failure (2 in the TIPS group and 1 in the paracentesis group), 2 cancers in the paracentesis group and 2 unknown (1 for each group).
Table 2. Frequency, Incidence, Hazard Ratio and 95% Confidence Interval of the Main Outcomes of the Study According to Allocation to TIPS Group or Paracentesis Group
Data are reported as absolute number of patients and (percent).
Incidences are reported as cases/100 patient-years.
Only 2 patients from the TIPS group were lost to follow-up, 1 and 20 months after inclusion, respectively. The probability of survival was significantly better in the group of TIPS (Fig. 2; P = .021). The estimated probability to be alive at 1 and 2 years was 77% and 59%, respectively, for the group of TIPS, and 52% and 29% for the group of paracentesis. The difference in the probability of survival persisted even when a post hoc analysis was performed in the subset of Child C patients (P = .001). The survival resulted better in the group of TIPS even when drop-out patients were considered dead (P = .05) or were excluded by the analysis (P = .028).
Four patients from each group underwent liver transplantation. The interval of time between the inclusion in the trial and transplantation was not significantly different between the two groups of patients (20±4 months in patients treated with TIPS vs. 13±4 months in patients treated with paracentesis, P = .25).
Univariate analysis, including all baseline variables and treatment assignment, identified several factors associated with death (Table 3). The Cox proportional hazard model showed that only the treatment assigned and model of endstage liver disease (MELD) score were independent predictors of mortality (Table 4). Bilirubin and creatinine were excluded from this analysis because they are already comprised in the Child-Turcotte-Pugh score and in the MELD score.
Table 3. Factors Predicting Death by Univariate Analysis
RR (95% CI)
Assignment to paracentesis treatment
Urine sodium excretion
Table 4. Factors Independently Predicting Death by Multivariate Analysis
RR (95% CI)
Assignment to paracentesis treatment
Failure of Treatment.
Failure of treatment was recorded in 7 patients from the group of TIPS and in 19 patients from the group of paracentesis (Table 2). The actuarial probability to fail was significantly greater in the group of paracentesis (Fig. 3; P = .0012). Of the patients who failed in the TIPS group, 2 were treated with LeVeen shunt and 5 were treated with repeated paracentesis. Of the 19 patients who failed in the paracentesis group, one was treated with LeVeen shunt, 11 were switched over to TIPS, and 7 preferred to continue with frequent paracentesis.
During the first year of follow-up, the patients in the TIPS group required 47 paracentesis (1.4±0.4 per patient), whereas the patients assigned to paracentesis treatment required 167 paracentesis (5.1± 0.7 per patient; P < .001 vs. TIPS group). Procedure-related complications were rare and not serious. Twenty patients from the group of TIPS and only one from the group of paracentesis did not require further taps during follow-up (P < .001).
Sixty-one percent of patients treated with TIPS and 39% of those treated with paracentesis developed hepatic encephalopathy during the follow-up (Table 2; P = NS). Also the cumulative probability to develop hepatic encephalopathy after inclusion did not differ between the two groups (1 and 2 years probabilities: 63% and 68%, respectively, in patients assigned to TIPS and 43% and 53%, respectively, in patients assigned to paracentesis; P = .086). However the total number of severe episodes of encephalopathy per patient was significantly greater in patients treated with TIPS than in those treated with paracentesis (0.97±0.2 vs. 0.36±0.1, P = .039). All patients with encephalopathy were successfully treated with lactulose and/or neomicin, except one who required a reduction of the stent size 90 days after TIPS implantation.
During follow-up, 3 patients from the TIPS group and 5 from the paracentesis group experienced an episode of gastrointestinal bleeding (Table 2). Esophageal varices were the source of bleeding in one patient from the TIPS group (in this patient TIPS insertion had failed and he received a LeVeen shunt), and in 3 from the paracentesis group. All these episodes were successfully treated with endoscopic procedures.
During follow-up, 28 patients from the TIPS group and 30 from the paracentesis group were readmitted to the hospital at least once (Table 2). The cumulative probability of hospital readmission was similar in the two groups of patients (1 and 2 years probabilities: 91% and 95% in TIPS patients and 90% and 97% in paracentesis patients; P = .43). The number of total days spent in hospital per patient (47±3.8 in the TIPS group vs. 40± 4.2 in the paracentesis group; P = .08) and the number of readmissions per patient (2.1± 0.3 vs. 1.6±0.2; P = .35) were also not different.
Shunt malfunction occurred in 12 patients. The probability to develop shunt insufficiency was 23% and 66% at 1 and 2 years, respectively. In 10 patients shunt patency was restored with angioplasty and/or insertion of a coaxial stent. Irreversible complete obstruction of the stent occurred in 2 patients.
Laboratory Changes During Follow-Up.
Major clinical and laboratory changes at 3 months of follow-up are reported in Table 5. In the group of patients treated with TIPS, bilirubin levels and INR slightly but significantly worsened over time, while creatinine levels tended to improve. MELD score did not change significantly. Conversely, in the group of patients treated with paracentesis, bilirubin levels and creatinine levels, but not INR, worsened over time. The deterioration of renal function probably explains why MELD score, but not Child-Turcotte-Pugh score, significantly worsened in these patients. Albumin levels increased in the group of patients treated with paracentesis probably because of frequent albumin infusions during repeated taps. Whereas no significant changes were found regarding PRA in both groups of patients, urine sodium excretion rate significantly increased only in the group of patients treated with TIPS.
Table 5. Changes of Clinical and Laboratory Parameters at 3 Months After Inclusion Into the Trial
9.4 ± 0.2
8.8 ± 0.3
9.4 ± 0.3
8.7 ± 0.6
10.3 ± 0.9
11.3 ± 0.9
10.4 ± 0.9
12.4 ± 1.0
29 ± 0.8
30 ± 1.0
30 ± 0.9
34 ± 1.3
1.6 ± 0.1
2.1 ± 0.2
1.9 ± 0.3
2.1 ± 0.3
1.22 ± 0.03
1.28 ± 0.04
1.23 ± 0.04
1.28 ± 0.05
1.08 ± 0.06
1.04 ± 0.08
1.07 ± 0.1
1.24 ± 0.11
15.9 ± 4.6
11.8 ± 5.3
14.8 ± 4.4
11.8 ± 3.5
Urine sodium excretion (meq/day)
50 ± 10
116 ± 12
27 ± 8
51 ± 12
During follow-up only 4 patients (2 treated with TIPS and 2 treated with paracentesis) did not completely abstain from alcohol. In all these patients treatment failed; one of these patients died after 15 months (paracentesis group) and one was lost to follow-up after 20 months (TIPS group).
Whether TIPS improves survival of patients with cirrhosis and refractory ascites is still debated.7–9 Therefore, the most important finding of the present study is the improved survival observed in the group of patients assigned to TIPS treatment, as compared with patients treated with large-volume paracentesis. According to the multivariate analysis, the average relative risk to die was 2.95 times greater in patients assigned to paracentesis treatment. The rate of survival of patients assigned to TIPS treatment was 77% and 59% at 1 and 2 years, respectively, which are figures markedly better than those reported for similar patients in studies of the natural history of cirrhosis.11, 12, 16
Improvement of survival could be attributed to a general reduction of portal hypertension–related risks. Accordingly, a factor concurring in improving the survival of patients with TIPS could have been the marked reduction in the number of taps performed in these patients during follow-up. Actually, these procedures, even if combined with albumin infusion, carry a risk of postparacentesis circulatory dysfunction, a complication which can impair long-term survival without evident clinical symptoms.17, 18 Moreover, patients treated with TIPS can experience an improvement of their nutritional status19 and this is another factor with a potential favorable effect on survival.
Interestingly, the TIPS-related improvement of survival was maintained when the analysis was restricted to the subset of patients with Child C cirrhosis, who represented 75% of our population. As one of our exclusion criteria was a Child-Turcotte-Pugh score >11, it can be speculated that we selected the subset of Child C patients in which the benefit obtained with portal decompression exceeds the detrimental effect of TIPS-related liver hypoperfusion. However, as this deduction was based on a post hoc analysis, it should ideally be confirmed in a randomized study specifically designed to investigate this hypothesis.
Unlike our trial, two other recent trials8, 9 reported no difference in the survival between patients with cirrhosis treated with TIPS and those treated with paracentesis and medical therapy. This discrepancy might be explained by some subtle but critical differences in the characteristics of the studied populations. Two of these differences were the exclusion of Child C patients with a Child-Turcotte-Pugh score >11 and the inclusion of a minor number of patients with recidivant ascites. These last patients have a better hepatic reserve than patients with refractory ascites, and thus might be at lower risk of liver failure after TIPS. All these studies, including our trial, were not performed blind. This aspect is a great limitation, even if the investigators who inserted TIPS did not participate in the periodical controls performed during the follow-up of the patients.
In addition to the treatment assigned, the other factor independently associated with death was a high score of MELD. These findings confirm the value of MELD score as a tool to predict short-term survival in patients with cirrhosis.20
In the present trial, patients treated with paracentesis showed a higher number of failures than patients treated with TIPS. This result is in agreement with all previous trials, notwithstanding the criteria employed to define treatment failure,7–9 and justifies the high percentage of subjects (30%) who were crossed over to TIPS, a figure similar to that reported by Rössle et al.7
As for the issue of complications caused by TIPS, the probability of developing new episodes of hepatic encephalopathy between the two groups of patients was not different. This finding differs from the findings of previous trials where TIPS was compared with endoscopic treatment for prevention of variceal bleeding.15 Such discrepancy could have been determined by a different average severity of liver failure, as most patients participating in the trials on variceal bleeding had a compensated liver disease. However, the negative effect of TIPS on the brain function of our patients emerged when the total number of severe episodes of encephalopathy per patient was taken into account. This number was significantly greater in patients treated with TIPS, a finding also reported by Ginès et al.8
TIPS malfunction due to stent stenosis because of pseudointimal proliferation was another drawback of TIPS treatment. It occurred in many patients, but stent patency was restored by angioplasty in all but two cases. Frequent episodes of severe encephalopathy and frequent need of restoring TIPS patency, on one side, and frequent recurrence of tense ascites, on the other side, justify the elevated number and days of rehospitalization reported in both groups of patients. It is conceivable that the recent availability of covered stent might improve the long-term patency of TIPS and, thus, reduce the number of rehospitalizations.21
During follow-up in patients treated with TIPS, hepatic function, but not renal function, slightly deteriorated, whereas in patients treated with paracentesis, renal function deteriorated although hepatic function was quite stable and albumin levels increased due to repeated albumin infusions. As a result in patients treated with paracentesis, MELD significantly worsened after 3 months from inclusion.
In conclusion, TIPS can improve the survival of the subset of patients with cirrhosis and refractory ascites. The necessity to perform paracentesis is also reduced by this procedure. However, the need of hospitalization and the rate of complications is similar to those of patients treated with repeated paracentesis plus albumin.
The authors thank the investigators who participated in the present study for the selection and care of the patients: University of Milan, Policlinico IRCCS: Larry Burdich, M.D.; Daniele Meregaglia, M.D.; Andrea Lovaria, M.D.; Giovanni Pagnozzi, M.D.; Mauro Borzio, M.D.; University of Monza, Policlinico San Gerardo: Elena Radaelli, M.D.; Laura Ratti, M.D. University of Rome “La Sapienza”, Policlinico Umberto I: Plinio Rossi, M.D.; Adriano De Santis, M.D.; Adolfo Francesco Attili, M.D.; Giorgia Nicolini, M.D.; Stefania Angeloni, M.D.; Cesare Efrati, M.D.; Andrea Masini, M.D.