Hepatitis C Information Central

Original Articles

Transjugular Intrahepatic Portosystemic Shunt Versus Endoscopic Sclerotherapy for the Prevention of Variceal Bleeding in Cirrhosis: A Randomized Multicenter Trial

Manuela Merli1, Francesco Salerno⁵, Oliviero Riggio¹, Roberto de Franchis⁵, Franco Fiaccadori³, Patrizia Meddi¹, Massimo Primignani^{5/SUP>, Giovanni Pedretti}3, Alessandra Maggi⁵, Livio Capocaccia¹, Andrea Lovaria⁶, Ugo Ugolotti⁴, Filippo Salvatori², Mario Bezzi², Plinio Rossi², and Gruppo Italiano Studio TIPS (G.I.S.T.)

From the¹ Institutes of II Gastroenterology,² II Radiology, La Sapienza University, Rome; ³ Institutes of Infectious Diseases and ⁴ Radiological Sciences, University of Parma; ⁵ Institutes of Internal Medicine, University of Milan, ⁶ Radiology Department, IR CCS Ospedale Maggiore, Milan, Italy

ABSTRACT

Transjugular intrahepatic portosystemic shunt (TIPS), a new technique for the treatment of portal hypertension, has been successful in preliminary studies to treat acute variceal hemorrhage and to prevent variceal rebleeding. The purpose of this multicenter, randomized controlled trial is to compare the efficacy of TIPS with that of endoscopic sclerotherapy in the prevention of variceal rebleeding in cirrhosis. Eighty-one cirrhotic patients, with endoscopically proven variceal bleeding, were randomized to either TIPS (38 patients) or endoscopic sclerotherapy (43 patients). Randomization was stratified according to the following: if bleeding occurred < 1 week (stratum I); if bleeding occurred 1 to 6 weeks (stratum II); and if bleeding occurred 6 weeks to 6 months (stratum III) before enrollment. Follow-up included clinical, biochemical, Doppler Ultrasound, and endoscopic examinations every 6 months. During a mean follow-up of 17.7 months, 51% of the patients treated with sclerotherapy and 24% of those treated with TIPS rebled (P = .011). Mortality was 19% in sclerotherapy patients and 24% in TIPS patients (P = .50). Hepatic encephalopathy (HE) developed in 26% and 55%, respectively (P = .006). A separate analysis of the three strata showed that TIPS was significantly more effective than sclerotherapy (P = .026) in preventing rebleeding only in stratum I patients. TIPS is significantly better than sclerotherapy in preventing rebleeding only when it is performed shortly after a variceal bleed; however, TIPS does not improve survival and is associated with a significantly higher incidence of HE. The overall performance of TIPS does not seem to justify the adoption of this technique as a first-choice treatment to prevent rebleeding from esophageal varices in cirrhotic patients. (HEPATOLOGY 1998;27:40-45.)

INTRODUCTION

Cirrhotic patients who survive an episode of bleeding from esophageal varices have an extremely high risk of rebleeding.¹ For this reason, several treatment modalities aimed at preventing variceal rebleeding have been tested by means of randomized controlled trials.² So far, pharmacological therapy with beta-blockers and endoscopic injection sclerotherapy are the most widely used treatments.³ Nevertheless, both treatments are not fully satisfactory, as the average rebleeding rate with each therapy is about 48%.²

Recently, a new angiographic technique, i.e., the transjugular intrahepatic portosystemic shunt (TIPS) has been proposed to treat portal hypertension.⁴ This procedure creates a communication between the hepatic and the portal vein within the liver, thus decompressing portal hypertension.⁵ Patency of the shunt is maintained by an expandable metal stent. Currently, TIPS has been successfully used in the following: in acute variceal hemorrhage uncontrolled by medical and endoscopic treatment ^6,7 ; in preventing rebleeding in patients in whom sclerotherapy failed5; in refractory ascites⁸; in the Budd-Chiari Syndrome⁹; and in patients who bleed while awaiting liver transplantation.¹⁰ However, the question of whether TIPS is better than other accepted treatments in the prevention of variceal rebleeding remains unanswered. The rate of rebleeding after TIPS, in preliminary studies, ^5,6,11 seems to be lower than that achieved by sclerotherapy.² Thus, more controlled trials comparing these two treatments have been recommended.¹²

The purpose of this multicenter, randomized controlled trial is to compare the efficacy of TIPS with that of endoscopic sclerotherapy in the prevention of recurrent variceal bleeding in cirrhotic patients. Encephalopathy, survival and side effects are also evaluated in the two treatment groups.

PATIENTS AND METHODS

Patients

Selection. As tertiary referral centers, our units observe patients belonging to three broad categories: 1) acutely bleeding patients; 2) patients referred to us for specialized treatment from other hospitals shortly after bleeding that was not treated with sclerotherapy or other definitive measure; and 3) patients referred to our clinics for further treatment and advice at various intervals after a bleed, and not under prophylactic treatment for rebleeding. Because we wanted to design the study to match our everyday situation as closely as possible, we decided to include also patients in category 3 in the study.

Accordingly, from November 1992 to June 1995 all cirrhotic patients consecutively admitted to our departments with active or recent (< 24 weeks) upper gastrointestinal bleeding were considered eligible for the trial if the bleeding was proven or presumed to be from esophageal varices according to the following criteria: 1) endoscopy performed during the bleeding episode identified the esophageal varices as the source of bleeding; or 2) endoscopy carried out after cessation of bleeding, but within 48 hours, showed large esophageal varices in the absence of other potential sources of bleeding.

Patients were excluded if they had the following: complete portal vein thrombosis; previous episode(s) of chronic recurrent hepatic encephalopathy; advanced hepatocellular carcinoma; previous multiple sessions of endoscopic sclerotherapy; ongoing pharmacological prophylaxis of rebleeding (one emergency session during the acute bleeding phase was permissible); severe cardiovascular contraindications; or concomitant morbid condition(s) with a life expectancy of less than 1 year.

Patients admitted for active bleeding were included only after bleeding had stopped for at least 24 hours and when they were hemodynamically stable as defined later.

Randomization.

Randomization was conducted through sealed envelopes in blocks of 4. Because the risk of bleeding decreases with time,¹³ patients were randomized into 3 separate strata according to the interval between bleeding and the time of randomization, as follows: stratum I: 1 to 7 days; stratum II: 1 to 6 weeks; and stratum III: 7 weeks to 6 months. Each center had a separate randomization list.

The study was approved by the ethics committees of each department, and informed consent to participate in the study was sought from each patient.

Baseline Evaluation.

At baseline, the following investigations were performed on all patients: 1) medical history and physical examination; 2) laboratory assessment of liver status and renal function; 3) Child-Pugh score; 4) evaluation of HE; 5) abdominal ultrasound scan; and 6) upper GI endoscopy.

Methods

Endoscopic Sclerotherapy. Sclerotherapy was performed by standard technique using 1% to 2% polidocanol as the sclerosing agent. Sclerotherapy sessions were carried out at 7 to 10 day intervals until varices were eradicated.

TIPS

All patients were treated under sedation or general anesthesia according to the technique described by Rössle et al.⁵ The stents used were a 10 × 52 mm Wallstent (Schneider Europe AG, Zurich, Switzerland) or a 10 × 70/80 mm Nitinol Strecker stent (Ultraflex Biliary Stent System, Meditech, Boston Scientific Co., Natick, MA). The portal vein pressure gradient was measured before and after constructing the shunt. Unless severe coagulopathy was observed, subcutaneous heparin treatment was started immediately after the procedure and continued for the first month. Antibiotics were administered immediately before the procedure and for 48 hours afterwards.

Follow-Up After Discharge.

The supportive care and monitoring provided during follow-up were similar in both groups. After discharge, the patients underwent upper gastrointestinal endoscopy and Doppler ultrasound at 6-month intervals or whenever clinically necessary.

In the endoscopic sclerotherapy group, recurrent varices were treated with additional sclerotherapy sessions. In patients treated with TIPS, repeat angiography was performed at 6 months or whenever stent malfunction was suspected. Episodes of stent stenosis were treated with balloon dilatation alone or by placing a new stent within the old one.

End Points.

Rebleeding (defined later) from any source was the primary end point of the study. An analysis including all rebleeding events, together with additional separate analyses for individual sources, was performed. Rebleeding episodes from unknown sources were analyzed together with portal hypertensive bleeds. Secondary end points were defined as death from any cause, HE, and complications. Any death within six weeks from the rebleeding episodes was considered a bleeding-related death.3

Drop-Outs.

All patients lost to follow-up were included in all analyses until the time of the last visit. Patients who refused the assigned treatment were included in all analyses.

Change of Therapy.

Patients could be crossed over to the alternative treatment or be managed by other means in the case of treatment failure or by physician recommendation.

Definitions of Events

Hemodynamic Stabilization. Hemodynamic stabilization was defined as the stabilization of hematocrit, arterial pressure, and heart rate for at least 24 hours in the absence of blood transfusions and/or vasoactive drug treatment.

Rebleeding.

Rebleeding was defined as hematemesis or melena occurring after 24 hours of hemodynamic stabilization.

Sites of Hemorrhage.

Variceal hemorrhage was defined as either: 1) certain when endoscopy showed blood spurting or oozing from a varix, or a `fibrin plug' on a varix with no other possible source of bleeding; and 2) presumed if endoscopy carried out within 24 hours showed varices in the absence of any other potential source of bleeding. Nonvariceal hemorrhage was defined as when bleeding from a source other than varices is seen. Unknown hemorrhage is defined in any other case.

TIPS Occlusion.

TIPS occlusion was defined as the failure of the angiographic contrast medium to flow through the stent.

TIPS Stenosis.

TIPS stenosis was defined as more than 50% reduction of stent lumen on angiography.

TIPS Malfunction.

An increase of portosystemic gradient to more than 15 mm hg.

Portosystemic Encephalopathy.

Portosystemic encephalopathy was defined as and graded according to Parsons-Smith criteria,¹⁴ as moderate (grade I-II) and severe (grade III-IV).

Treatment Failure.

Any uncontrolled rebleeding that occurred at any time was considered a failure for both treatments. In addition, sclerotherapy was considered to have failed when a second episode of rebleeding from esophageal varices occurred within one month after a first one managed by sclerotherapy or when rebleeding from gastric varices occurred at any time after randomization. TIPS was considered to have failed when it was impossible to place the stent, or when a rebleeding episode could not be managed by dilatation or replacement of the stent with a new one, or when a second rebleeding episode occurred after a first one managed by stent dilatation or replacement.

Statistical Analysis

Results are expressed as mean ± SEM, unless otherwise indicated.

Sample Size Calculation.

The average risk of rebleeding in patients undergoing endoscopic sclerotherapy is about 50%.² We assumed that TIPS could reduce this risk by 75%. To show that reduction with an error of 0.05 and a power of 0.90, we calculated that the study should include 35 patients per group. Allowing for a 10% dropout rate, we planned to enroll 40 patients per group. The study was closed on December 31, 1995, when the last patient included had a follow-up of 6 months. Rebleeding and survival were analyzed by means of life-table analysis (Kaplan-Meier method), and differences between treatment groups were analyzed by the log-rank, ², or the paired or unpaired t test. We carried out both `intention-to-treat' and `per protocol' analyses. Separate analyses were carried out for each endpoint. For the sake of these analyses, the patients were censored at the time of the chosen event.

A rebleeding index (N° of months of follow-up divided by the n° of bleedings plus 1) was calculated in each patient to elevate the interval without rebleeding in each group.

RESULTS

A total of 120 potential candidates were considered for the study. Thirty-eight patients were not randomized, as follows: 4 patients for total portal vein thrombosis; 3 patients for concomitant diseases with short life expectancy (2 human immunodeficiency virus-positive drug addicts and 1 lymphoma); 18 patients for previous treatments with multiple endoscopic sclerotherapy sessions; 9 patients for advanced hepatocellular carcinoma; 3 patients for cardiac failure; and 1 patient who refused randomization.

Eighty-two patients were included in the study: 34 patients entered stratum I; 36 patients entered stratum II; and 12 patients entered stratum III (table 1). Thirty-nine patients were randomly assigned to TIPS and 43 to sclerotherapy. The patients enrolled at the 3 study sites were a total of 45, 21, and 16, respectively. One patient in the TIPS group was erroneously randomized (bleeding had not stopped before randomization) and was excluded from further analysis.

Thirty-three patients (87%) in the TIPS group and 39 (91%) in the sclerotherapy group received the allocated treatment. One patient in each group refused the assigned treatment. Two patients in the sclerotherapy group and 1 in the TIPS group rebled and died after randomization but before treatment could be started. One patient in the sclerotherapy group developed a severe esophageal stricture as a consequence of the emergency sclerotherapy; this complication became evident only after randomization and made it impossible to perform further sclerotherapy sessions. In three patients the TIPS procedure was unsuccesful because of the inability to place the stent (2 cases) and stent migration (1 case).

The two groups were similar in their baseline characteristics (table 1). Although there were more patients with alcoholic cirrhosis in the sclerotherapy arm, the difference was not statistically significant (P= .198). The interval between index bleeding and randomization in the patients was as follows: stratum I was 3.5 ± 0.5 days in the TIPS group and 3.1 ± 0.5 days in the sclerotherapy group (P = .58); stratum II was 21 ± 3.4 for TIPS, and 19 ± 2.8 days for sclerotherapy (P = .66); in the third, 63 ± 16 and 65 ± 8 days, respectively (P= .9). In the 33 patients in whom TIPS was successfully implanted, the hepatic vein pressure gradient decreased from 24 ± 0.9 to 10 ± 0.7 mm hg (P = 2.16 × 10-16).

The 39 patients assigned to sclerotherapy and who did receive the treatment underwent 3.6 ± 0.27 sclerotherapy sessions over 6.2 ± 0.94 weeks. Esophageal varices were eradicated in 20 patients (51.3%).

The mean duration of follow-up was 77.7 ± 7.12 weeks in the sclerotherapy group and 73.9 ± 7.3 weeks in the TIPS group. One patient in each group was lost to follow-up six months after randomization. Three patients were transplanted, two in the sclerotherapy and one in the TIPS group.

Rebleeding.

Twenty-two patients in the sclerotherapy group (51%) and 9 in the TIPS group (24%) rebled during follow-up, with an absolute risk reduction of 27%. The number of patients to treat with TIPS to prevent one rebleed that would have occurred if all patients had been treated by sclerotherapy was 3.7. The odds ratio for rebleeding was 3.38 (95% confidence intervals: 1.30-8.79). Two patients in each group rebled before the assigned treatment could be started; two further rebleeds in the TIPS group occurred in patients in whom the TIPS insertion failed. The cumulative probability of not rebleeding at one and two years was 79% ± 7% and 69% ± 11% for TIPS and 48% ± 8% and 40% ± 9% for sclerotherapy (log-rank test; P = .011; Fig 1.). This difference was maintained if a per protocol analysis was performed (log-rank test; P = .005; data not shown).

View Larger Version

Fig. 1. Cumulative proportion (Kaplan-Meier plot) of patients free of rebleeding.

In the sclerotherapy group, 17 (40%) patients rebled from esophageal varices while in the TIPS group, the corresponding figure was 7 (18%) (log-rank test: intention-to-treat. P = .054, per protocol P = .015).

Overall, 41 episodes of rebleeding occurred in the sclerotherapy group and 12 in the TIPS group. Esophageal varices were the source of bleeding in 68% of all rebleeds in the sclerotherapy group and in 58% in the TIPS group (table 2). The rebleeding index was 12.9 ± 1.6 in the sclerotherapy group and 15.6 ± 1.6 in the TIPS group (P = .24).

Separate analyses of rebleeding in stratum I and II + III were performed: in stratum I, the cumulative probability of remaining free of rebleeding was significantly higher in the TIPS than in the sclerotherapy group (odds ratio for rebleeding 5.50; 95% confidence intervals: 1.28-23.69; log-rank test: P = .026, Fig. 2). In strata II + III the difference was not significant (odds ratio 2.36; 95% confidence intervals: 0.65-8.51; log-rank test: P = .14) (data not shown).

View Larger Version

Fig. 2. Cumulative proportion (Kaplan-Meier plot) of patients free of rebleeding in stratum I.

Treatment Failures.

Endoscopic sclerotherapy failed in 12 patients and TIPS failed in 7. Six sclerotherapy failures were crossed over to TIPS: one was treated with portacaval shunt, three were treated conservatively and the remaining 2 died of rebleeding. Two TIPS failures were crossed over to sclerotherapy, one was treated with surgical portacaval shunt, two were treated conservatively and the remaining two died of rebleeding.

Mortality.

Overall mortality was 21% (17 patients: 8 in the sclerotherapy and 9 in the TIPS group). The causes of death were similar in the two groups (table 3). The actuarial probability of survival (Fig. 3) was 86% ± 5% at one year and 79% ± 8% at two years in patients receiving sclerotherapy. The corresponding figures in the TIPS group were 84% ± 6% and 73% ± 8%. The odds ratio for death was 0.74 (95% confidence intervals: 0.25-2.15). The difference between the two groups was not significant (log-rank test: P = .50). Separate analyses of mortality in stratum I and in II + III showed no differences in mortality between the two treatments.

View Larger Version

Fig. 3. Cumulative proportion (Kaplan-Meier plot) of patients surviving.

HE. Ten patients (26 %) in the sclerotherapy group developed a total of 12 episodes of HE as compared with 21 patients (55%) and 39 episodes in the TIPS group (P = .006). (table 4 and Fig. 4). The odds ratio for HE was 0.25 (95% confidence intervals: 0.09 to 0.64). Three episodes of hepatic encephalopathy in the sclerotherapy group and 8 in the TIPS group were grade III-IV.

View Larger Version

Fig. 4. Cumulative proportion (Kaplan-Meier plot) of patients free of hepatic encephalopathy.

Complications. The incidence of complications did not differ significantly in the two groups (table 5).

Stent Malfunction.

Of 33 patients in whom TIPS was successfully implanted, 21 (66%) developed stent malfunction. Twelve of these patients suffered more than one episode of either stenosis or occlusion of the shunt (range, 2-4). Overall 40 episodes of stent malfunction were diagnosed during follow-up (occlusion in 4 and stenosis in 36 cases). Stent revision by means of angioplasty was successfully performed in 24 of these episodes while an additional stent was needed in 13. Three patients, 2 with complete stent thrombosis and 1 with stent stenosis caused by surrounding neoplastic tissue, were moved to other therapies. The cumulative rate of stent malfunction at one year was 61%.

DISCUSSION

Preliminary data suggest that TIPS controls variceal bleeding in a high number of patients with cirrhosis⁷ and is also effective in preventing rebleeding. ^6,11 However, more data are necessary before the real place of this technique in the therapeutic armamentarium is defined. The present study compared TIPS with endoscopic sclerotherapy in a group of cirrhotic patients who survived a recent episode of bleeding from esophageal varices. All patients had stopped bleeding for at least 24 hours and were considered hemodynamically stable . In patients treated with sclerotherapy, rebleeding occurred in 51% of cases during a mean follow-up of 18 months. This figure is close to the mean rebleeding rate reported after sclerotherapy in controlled studies.²

TIPS was better than sclerotherapy in preventing overall rebleeding, since only 24% of patients rebled during a follow-up of 17 months. The proportion of patients who remained free of rebleeding one year after TIPS (79%) is very close to the figures observed in three previous studies. ^5,6,11 Both the number of first rebleeding episodes and the number of total rebleeds were markedly lower after TIPS than after sclerotherapy. This is probably caused by the ability of TIPS to reduce the hepatic vein pressure gradient to values lower than 12 mm hg, a level considered a threshold for variceal rupture and bleeding.^15-16 Moreover, the reduction in portal pressure occurs immediately after TIPS placement, whereas multiple sclerotherapy sessions are necessary to eradicate varices. In addition, sclerotherapy-induced mucosal ulcerations contribute to increases in rebleeding rates in sclerotherapy patients during the first months. This explains why a greater number of rebleeds were prevented by TIPS in the first months of the study.

The odds ratio for rebleeding was 3.38, with large 95% confidence intervals (1.30-8.79). Thus, the overall power of the study to detect a difference in the rebleeding rate in the order of 25% was 0.73 with an error of 0.05. There is no doubt that the power of the study was somewhat weakened by the inclusion of patients in stratum II, especially in stratum III, since the contribution of these strata to the overall power of the study is small, owing to the lower rebleeding risk of the patients belonging to these two strata. However, the relatively weak power of the study is more the result of a higher-than expected rebleeding rate in the TIPS arm than of a lower rebleeding rate in the sclerotherapy arm. Separate analyses of the three strata showed that TIPS was significantly more effective than sclerotherapy (P = .026) in preventing rebleeding only in patients of stratum I. This evidence seems solid, when taking into consideration the relatively small number of patients included in this stratum and suggests that the higher the risk of rebleeding, the greater is the efficiency of TIPS. It should also be considered that the lack of difference in the rebleeding rates between TIPS and sclerotherapy in stratum II and III may be caused by the lower risk of rebleeding in patients treated later than one week after the bleeding episode and to a possible type-II error (odds ratio for rebleeding, strata II and III = 2.36, 95% confidence interval 0.65-8.51; power 0.25; error 0.05).

A reported drawback of TIPS is the high incidence of HE. In our study, 39 episodes of HE occurred in 21 patients. Usually lactulose or lactitol allowed a rapid resolution of these episodes, and an increase of the daily doses of these disaccharides prevented new episodes of severe HE in all but one patient. Only 8 episodes were severe, and in 10 cases hospitalization was recommended. Moreover, most episodes occurred early after the procedure, and after six months the incidence of HE in the two groups was no longer different. The high incidence of HE in patients treated with TIPS underscores the fact that, to be effective, TIPS must achieve a delicate balance. Too little shunting might not sufficiently decrease portal pressure to prevent rebleeding, while too much shunting might cause HE. ^17,18

Although the risk of rebleeding was considerably reduced in patients treated by TIPS, the survival of these patients did not improve in comparison with that of patients treated by sclerotherapy. One might argue that since 6 sclerotherapy failures were crossed over to TIPS, this might have obscured a potential benefit of TIPS for survival. However, it should be also noted that 2 TIPS failures were crossed over to sclerotherapy. Even when considering the worst case hypothesis, i.e., that all patients crossed over to TIPS would have died if they had not been crossed over and that all TIPS failures crossed over to sclerotherapy would have survived, the difference between groups would have remained far from significant.

Another major drawback of TIPS that emerges in the present as in previous studies¹⁹ is the frequent occurrence of stent malfunction. Stent stenosis or occlusion occurred in 21 patients, and in half of those patients the problem recurred again after a first revision of the stent. The high rate of stenosis of the stent makes a strict follow-up of patients with Doppler ultrasound or angiography mandatory. The cost of patient care increases consequently.

In conclusion, TIPS was significantly superior to sclerotherapy in the prevention of rebleeding only in patients treated within one week of the bleeding episode; it did not improve survival and it led to a higher incidence of HE. In addition, stent malfunction requiring dilatation or re-stenting occurred in a sizable proportion of patients. Therefore, the overall performance of TIPS with currently available materials does not seem to justify the adoption of this technique in place of endoscopic treatments as a first-line treatment for preventing rebleeding from esophageal varices.

Footnotes

Acknowledgement: We are indebted to Luca Carpinelli, M.D., for enlightening discussion and valuable assistance in the statistical analyses.

Abbreviations: TIPS, transjugular intrahepatic portosystemic shunt; HVPG, hepatic vein pressure gradient; HE, hepatic encephalopathy.

Received December 30, 1996; accepted September 16, 1997.

Address reprint requests to: Oliviero Riggio, M.D., II Gastroenterologia, Università La Sapienza, Viale dell'Università 37, 00185 Rome, Italy. Fax: 39-6-444-0806.

REFERENCES

table Of Contents

Copyright © 1998 by the American Association for the Study of Liver Diseases.