Randomised trial of transjugular-intrahepatic-portosystemic shunt versus endoscopy plus propranolol for prevention of variceal rebleeding

Randomised trial of transjugular-intrahepatic-portosystemic shunt versus endoscopy plus propranolol for prevention of variceal rebleeding

THE LANCET Articles Randomised trial of transjugular-intrahepatic-portosystemic shunt versus endoscopy plus propranolol for prevention of variceal r...

59KB Sizes 0 Downloads 81 Views

THE LANCET

Articles

Randomised trial of transjugular-intrahepatic-portosystemic shunt versus endoscopy plus propranolol for prevention of variceal rebleeding M Rössle, P Deiber t, K Haag, A Ochs, M Olschewski, V Siegerstetter, K-H Hauenstein, R Geiger, C Stiepak, W Keller, H E Blum

Summary

Introduction

Background The transjugular-intrahepatic-por tosystemic shunt is a new inter ventional treatment for por tal hypertension. The aim of our study was to compare the transjugular shunt with endoscopic treatment for the prophylaxis of recurrent variceal bleeding. Methods Between March, 1993, and March, 1996, 126 patients with variceal bleeding were randomly assigned either transjugular shunt (n=61) or endoscopic treatment (n=65). Patients were followed up for a median of 14 (IQR 8–25) months and 13 (8–25) months, respectively. In 31 (51%) of the shunted patients, simultaneous transjugularvariceal embolisation was done at the time of shunt placement. Endoscopic treatment consisted of sclerotherapy and/or banding ligation and was combined with propranolol medication. Findings Technical success was achieved in all patients assigned to the shunt group. During follow -up, the cumulative 1-year variceal rebleeding rates in the shunted and endoscopically treated patients were 15% and 41% and the 2-year rates were 21% and 52% (p=0·001), respectively. In nine (12%) patients from the endoscopic group treatment failed and the patients received the transjugular-shunt treatment. A total of 19 bleeding episodes from any source occurred in 15 patients in the shunt group compared with 100 episodes in 33 patients in the endoscopic group. There was no difference in sur vival with estimated 1-year sur vival rates for shunted and endoscopically treated patients of 90% and 89%, and 2-year sur vival rates of 79% and 82%, respectively. The incidence of clinically significant hepatic encephalopathy after 1 year was higher in the shunt group (36% vs 18%, p=0·011). Interpretation These results suggest, that the transjugular shunt is more effective than endoscopic treatment in prevention of variceal rebleeding but has a considerable risk of hepatic encephalopathy. Sur vival is similar in the two groups.

Lancet 1997; 349: 1043–49 Department of Gastroenterology and Hepatology (Prof M Rössle P Deiber t MD, K Haag MD, A Ochs MD, V Siegerstetter MD, Prof H E Blum MD); Institute of Medical Biometry (M Olschewski MSc ), and Department of Radiology (Prof K-H Hauenstein MD), University Hospital, University of Freiburg, Freiburg, Germany; Department of Internal Medicine, General Hospital of Offenburg, Offenburg (R Geiger MD); and Department of Internal Medicine, General Hospital of Rastatt, Rastatt (C Stiepak MD, W Keller MD) Correspondence to: Prof Mar tin Rössle, Depar tment of Gastroenterology and Hepatology, University Hospital, Hugstetterstrasse 55, D-79106 Freiburg, Germany

Vol 349 • April 12, 1997

MD,

The transjugular-intrahepatic-portosystemic shunt is an interventional procedure consisting of an anastomosis between a hepatic vein and an intrahepatic branch of the portal vein. This type of shunt was introduced into clinical practice in 1988 and its use has become widely accepted despite the lack of comparative studies with conventional treatments.1,2 The transjugular shunt functions in a similar way to surgical shunts. There are, however, many differences between results when surgical shunts are compared with transjugular shunts. First, the procedural and hospital mortality of the transjugular shunt is about 1% compared with 3% to 15% for surgical shunts.3,4 Second, the invasiveness of the transjugular shunt is low and justifies its use in older patients (>70 years), in emergency cases, and in patients of Child-Pugh class C who are excluded from most surgical-shunt studies.3–7 Third, the transjugular shunt provides the option of calibration of the shunt—ie, enlargement or reduction—at any time after its establishment. However, the transjugular shunt has a major disadvantage in that it has a high rate (40% to 50%) of dysfunction that usually needs correction.3,4,8 This randomised study was designed to compare the transjugular shunt with endoscopic treatment plus propranolol in the prevention of recurrent variceal bleeding. The number of bleedings from any source, survival, procedure-related complications, hepatic encephalopathy, Child-Pugh class, and liver function, as well as duration of hospital stay were also assessed. The addition of propranolol and the optional use of both polidocanol injection and banding ligation were allowed to compare the transjugular shunt with the most effective non-shunt strategy.

Methods Between March, 1993, and March, 1996, 126 patients (table 1) with variceal bleeding were randomly assigned either transjugular shunt or endoscopic treatment. None of the patients’ clinical characteristics were significantly different between the two groups including the haemodynamic findings, portal vein flow-velocity, and congestion index9 obtained by duplex-sonography. The study protocol was approved by the local ethics committee. All patients were informed about the procedure in detail and written consent was obtained. The primary endpoint of the study was rebleeding from varices. Secondary endpoints were death, bleeding from non-variceal sources, procedure-related complications, and hepatic encephalopathy. Based on the results of comparable trials10 we assumed an incidence of rebleeding of 50% within the first year of endoscopic treatment. The shunt was expected to reduce this rate by half. To detect a 50% reduction with a type I error rate of 5% and a type II error of 20% a sample size of 54 patients per group was calculated. Assuming a drop-out rate of about 10% the minimum sample size needed was calculated to be at least 60 patients per group.

1043

THE LANCET

The inclusion criteria were: liver cirrhosis, variceal bleeding within 2 weeks before randomisation, and age over 18 years. Cirrhosis was diagnosed by clinical criteria and sonography or histology (table 1). Variceal bleeding was confirmed endoscopically in all patients. Exclusion criteria were hepatic encephalopathy grade 3 and 4, liver insufficiency with total bilirubin of more than 5 mg/dL (except patients with primary biliary cirrhosis), cavernomatous portal-vein thrombosis, advanced malignancy, contraindications for propranolol (severe heart insufficiency, obstructive lung disease, severe hypotension), and bleeding emergency. Treatment was randomly assigned to patients within 24 h and treatment was started within 48 h after admission. Randomisation was stratified according to Child-Pugh class11 and age (<60 years or 肁60 years). Patients with acute bleeding at admission were treated by injection of polidocanol (59 patients), a mixture of n-butyl-2cyanoacrylate (bucrylate; Histoacryl) and lipiodol (1/1, vol/vol; five patients), or fibrin glue (Tissucol; four patients) or polidocanol plus bucrylate (eight patients) to stop the bleeding. In six patients in whom endoscopic treatment failed a Sengkstaken tube was put in place. No haemostatic medication was given. After discharge patients were seen after 4 weeks, 3 months, 6 months, 9 months, 12 months, and then every 6 months or when needed for clinical reasons. The follow-up included a clinical examination, blood chemistry and assessment of hepatic encephalopathy by clinical examination, the Reitan tests A and B,12 and a structured questionnaire. Patients who had a transjugular shunt were routinely investigated with duplexsonography at each visit.13 Radiological revision of the shunt was not done routinely; it was done only in patients with lack of shunt flow, or increase in the portal vein flow velocity by less than 50% compared with the preshunt value, or reduction of initial postshunt portal vein flow velocity by more than 50%, and/or varices with a risk of bleeding (grade >2, red spots) at endoscopy. In patients randomly assigned the endoscopic treatment, endoscopy was done at each visit or when needed for clinical reasons and treatment was given if necessary (varices grade >1).

Statistical analysis For descriptive purposes data are presented as means with standard deviations or medians with quartiles for quantitative variables, and as absolute and relative frequencies for qualitative variables. Comparison of baseline data was done with Wilcoxon rank tests for quantitative and Fisher’s exact test for qualitative variables. Probabilities of being free of rebleeding, of the occurrence of hepatic encephalopathy, and of survival were analysed on an intention-to-treat basis and estimated by the Kaplan-Meier method. The prognostic relevance of treatment and selected variables with respect to these endpoints were analysed univariately by log-rank tests and multivariately by Cox’s proportional hazards model. Results of the multivariate analysis are presented as estimated relative risks with corresponding 95% CIs and p values from Wald’s test.14 All significance tests were two-sided. Patients were assigned to study groups by computer-generated random numbers that were read by a person not involved in the clinical setting. The randomisation sequence could not be previewed and the patients had an equal chance of being assigned each treatment.

Definitions Throughout the study the following definitions of the endpoints were used.

Variceal index bleeding —endoscopically verified active spurt or ooze bleeding; or recent bleeding with a visible thrombus on varix or blood in stomach without detection of other sources of bleeding; or recent haematemesis or melaena together with a decrease in the haemoglobin concentration by more than 1·2 mmol/L (>2 g/dL) and exclusion of other sources of bleeding.

1044

Transjugular shunt (n=61)

Endoscopic treatment plus propranolol (n=65)

Age (years) Mean (SD) Range

54·3 (11·9) 32–79

56·6 (12·4) 23–86

Sex (male/female)

40/21

44/21

Aetiology of liver disease Alcoholic Viral Alcoholic plus viral Miscellaneous

42 11 3 5

42 10 4 10

Histology

24

28

Child-Pugh class A/B/C

17/33/11

22/31/12

Previous variceal bleedings 0 1 2 >2

23 17 12 9

26 18 13 8

Previous sclerotherapies (patients) 0 1 2 3 >3

25 13 8 6 9

36 14 7 2 6

Index bleeding (patients) Active bleeding at randomisation Oesophageal varicose

39 53

37 59

Transfusions before randomisation (patients) 0 units 7 1–2 16 3–5 17 >5 21

4 19 29 13

Mean (SD) time between index bleed and randomisation (days) Relevant diseases Diabetes mellitus (insulin no/yes) Coronar y heart disease Heart insufficiency (NYHA 1,2) Severe organic renal insufficiency Miscellaneous Maximum flow velocity in portal vein (cm/s) Congestion index (cm ⫻ s)*

6·3 (5·5)

4·4 (5·0)

8 (7/1) 6 2 1 5

14 (10/4) 5 4 3 5

17·2 (7·6)

17·2 (8·1)

0·17

0·18

*The congestion index (normal: 0·07 [0·03]) was calculated as the ratio of the cross-sectional area to mean velocity of the blood flow of the portal vein.10

Table 1: Clinical characteristics of the patients

Clinically significant bleeding —bleeding requiring two or more transfusions within 24 h of onset and/or with a haemoglobin concentration of less than 8 g/dL, together with a systolic blood pressure of less than 100 mm Hg. Rebleeding —bleeding occurring after randomisation and after 24 h from the point of stable vital signs after index bleeding. Bleeding emergency/failure to control acute bleeding —ongoing or recurrent bleeding within 24 h leading to shock and/or progressive organ failure (eg, coma, severe jaundice, renal failure, or respiratory insufficiency). Failure of endoscopic treatment —three or more clinically significant rebleedings within 1 year of follow-up, or a bleeding emergency. Failure of the transjugular shunt —technical failure of stent placement or rebleeding due to shunt insufficiency in patients in whom shunt revision failed, or the need of more than three revisions within 1 year. Shunt insufficiency —angiographic finding of stenosis/occlusion of the shunt together with a raised portosystemic pressure gradient of more than 12 mm Hg.

Vol 349 • April 12, 1997

THE LANCET

190 eligible patients

126 randomised 64 not randomised 42 refused 22 excluded

Varices Mallor y-Weiss Sclerosing ulcer Duodenal or gastric ulcer Portal hypertensive gastropathy Miscellaneous/unknown All sources

65 endoscopic therapy plus propranolol

61 transjugular-intrahepaticportosystemic shunt

Received treatment as allocated

61 received treatment as allocated

56 endoscopy

44 propranolol

9 failure

17 not compliant 4 withdrawn

Transjugular shunt

Endoscopic treatment

Total (n=15)

Clinically significant (n=14)

Total (n=33)

Clinically significant (n=25)

11 2 1 2 0 3* 19

9 2 0 2 0 2 15

56 8 8 3 7 3/15† 100

47 7 5 2 6 5 72

*Arterial ulcer bleeding, reflux oesophagitis, and unknown source (one patient each). †Miscellaneous: reflux oesophagitis (three patients), unknown (five patients without endoscopically detectable bleeding source and ten patients who did not have endoscopy).

Table 2: Source and number of upper gastrointestinal bleedings after randomisation by intention-to-treat

Results

62 followed up Time: 14 months median

60 followed up Time: 13 months median

62 completed trial

60 completed trial

Figure 1: Trial profile

Hepatic encephalopathy —grade 1: minor limitations of concentration, fatigue; grade 2: somnolence, asterixis; grade 3 deep somnolence or stupor; grade 4: coma. Clinically significant hepatic encephalopathy —more than one episode of overt encephalopathy per 3 months (grade 肁2, >6 h duration), or chronic persistent encephalopathy of at least grade 1. Refractor y hepatic encephalopathy —clinically significant encephalopathy not responding to diet, lactulose, or oral branched-chain aminoacid supplementation.

Techniques The technical procedure of the transjugular-intrahepaticportosystemic shunt is described in detail elsewhere.3 Briefly, a catheter is introduced transjugularly into a hepatic vein, a puncture needle is then pushed through the catheter, and the portal vein is punctured under fluoroscopic and sonographic guidance. After predilatation of the tissue tract, a stent is introduced and expanded with balloon catheters. The following stents were used: Palmaz stent (39 patients, 92 stents), Memotherm stent (16 patients, 19 stents), Wallstent (six patients, six stents). The final diameter of the stent was adjusted to achieve the desired portal venous pressure gradient (portal pressure minus inferior vena cava pressure). In 31 (51%) patients with huge varices or in whom variceal perfusion persisted after creation of the shunt, embolisation with bucrylate/lipiodol was done. Anticoagulation to prevent early thrombosis of the shunt was given as previously described.3 For patients assigned endoscopic treatment, acute bleeding was controlled and then they were treated either with injections of polidocanol (16 [SD 8] mL per session) or banding ligation (3·2 rubber bands) in intervals of 2 to 5 days until eradication of the varices was achieved or at least six treatment sessions were applied. Gastric varices were treated by intravariceal injection of bucrylate/lipiodol. 33 patients were treated with sclerotherapy only, 31 had a combination of sclerotherapy and banding ligation, and one patient had banding ligation only. Propranolol was given in dose of 63 (33) mg/day to decrease the heart rate by 25%.

Vol 349 • April 12, 1997

Source of bleeding

Of the 190 eligible patients 64 could not be included in the study (figure 1). The number of patients who refused to take part was comparable to the number in each study group with respect to age, sex, Child-Pugh class, bleeding site, and number of previous bleeds and scleropathies. The median observation time of the shunted patients was 14 (IQR 8–23) months and of the endoscopically treated patients 13 (8–25) months. One patient in the shunt group and three in the endoscopic group were lost to follow-up. Their data were censored at the time of their last examination. Propranolol was taken by 44 (68%) of the 65 patients who had endoscopic treatment. 17 patients were not compliant and in four patients medication was withdrawn because of severe side effects. In nine of the endoscopic group treatment failed and a transjugularintrahepatic-portosystemic shunt was implanted.

Technical results and complications The shunt implantation was successful in all patients with a mean intervention time of 64 (13) min. The mean shunt diameter, measured by sonography, was 9·1 (1·0) mm (range 6–11 mm), resulting in a reduction of the portosystemic pressure gradient by 64% from a mean of 22 (5) mm Hg to 8 (4) mm Hg. During follow-up, 18 (30%) patients needed interventional revision of the shunt after a mean of 7·6 (8) months (2–882 days) because of shunt insufficiency associated with recurrence of varices or rebleeding. A second revision was required in two (3%) patients. Technical complications were seen in 13 patients and consisted of bleeding into the biliary system (three patients), peritoneal cavity (two patients), or liver parenchyma (one patient); bradycardia requiring atropine (four patients); septicaemia (one patient); and stent migration or dislocation (two patients). In addition, 11 patients developed uncomplicated fever and six patients had transient abdominal pain which was probably due to bucrylate-induced phlebitis of the gastric collaterals. Long-term morbidity was seen in one patient whose stent migrated into the right ventricle. This patient developed bradyarrhythmia requiring pace-maker implantation. No patient died from a procedure-related complication. In the endoscopic group, 581 endoscopies were done (9·7 [4·8] per patient), including 331 interventions (5·3 [3·4] per patient). In only 11 (17%) of the patients was eradication of the varices achieved—ie, no varices visible without introducing air through the endoscope. In 41 (63%) of the patients the size and number of the varices were reduced without achieving eradication. A minority of 1045

100

100 Hepatic encephalopathy (%)

Proportion of rebleeding (%)

THE LANCET

80 60 40

Shunt

20

Variceal rebleeding Rebleeding from any source Endoscopy Variceal rebleeding Rebleeding from any source

0

0

6

12

Shunt Endoscopy

80 60 40 20 0

18

24

0

6

Time (months) Patients at risk Shunt Variceal 61 Total 65 Endoscopy Variceal 61 Total 65

37 51

23 37

12 23

8 15

37 51

23 37

12 23

8 15

The difference between the treatment groups is statistically significant for variceal rebleedings (p=0·001) and bleedings from any source (p<0·001).

13 (20%) of the patients did not respond sufficiently to the endoscopic treatment despite at least six interventions during follow-up. Endoscopy was associated with the following complications: dysphagia (five patients), mediastinitis (one patient), and hypopyon with subsequent eye enucleation most probably caused by sclerotherapyinduced bacteraemia (one patient). No procedure-related death was observed.

Rebleeding Nine (15%) of the shunted patients had 11 variceal rebleeds, and 29 (45%) of the endosopically treated patients had 56 variceal rebleeds. Overall, 15 of the shunted patients and 33 endoscopically treated patients had 19 and 100 upper gastrointestinal rebleedings, respectively. Most of the bleeds were of clinical significance and originated from oesophageal or gastric varices (table 2). Bleeds from non-variceal sources were 100

Survival (%)

80 60 40 20 Shunt Endoscopy 0

6

12

18

24

21 25

14 19

Time (months) Patients at risk Shunt 61 Endoscopy 65

46 57

37 36

Figure 3: Estimated probability of survival according to intention-to-treat

1046

18

24

12 23

8 15

Time (months)

Figure 2: Estimated probability of being free of rebleeding according to intention-to-treat

0

12

Patients at risk Shunt 61 Endoscopy 65

37 51

23 37

Figure 4: Estimated probability of the development of hepatic encephalopathy according to intention-to-treat The difference between the groups is statistically significant (p=0·011).

seen predominantly in the endoscopically treated patients. The analysis of the time of first rebleeding from varices or from any source showed a significant difference between the groups (p=0·001 for variceal and p<0·001 for bleeding from any source; figure 2). The 1-year rates of variceal rebleedings were 15% and 41% and of bleedings from any source 23% and 55% in the shunted and endoscopic group, respectively. In nine patients treatment failed and the patients received the transjugular shunt. Six of these patients are still alive without further bleeds and three patients died, two from liver failure and one from hepatocellular carcinoma. Three of the nine patients presented to hospital as emergency cases because of bleeding and the transjugular shunt operation was done as an emergency treatment. In the shunted group, no treatment failure was observed. During follow-up, 25 (38%) patients in the endoscopic group required a total of 272 units of packed erythrocytes compared with 12 (20%) patients in the shunt-group requiring 96 units of packed erythrocytes (Fisher’s exact test for number of patients: p=0·031).

Hospital stay and sur vival After randomisation, the endoscopically treated patients spent a mean of 34 (28) days in the hospital and had 661 outpatient visits. By comparison, the shunted patients spent a mean of 27 (17) days in the hospital and had 386 outpatient visits. This trend in favour of the shunt group was not statistically significant. Overall, eight patients died in each group. Survival was comparable with estimated 1-year rates of 90% and 89% and 2-year rates of 79% and 82% in the shunted and endoscopically treated patients, respectively (figure 3).The causes of death in the shunted patients were septicaemia (four patients), ulcer bleeding (two patients, one due to non-steroidal anti-inflammatory medication), liver transplantation following liver failure (one patient), and status epilepticus (one patient). Causes of death for those in the endoscopic group were: liver failure (three patients), variceal bleeding (two patients), septicaemia, suicide, and hepatocellular carcinoma. Two of the patients with liver failure had liver transplants 3 and 6 months after receiving the transjugular shunt.

Vol 349 • April 12, 1997

THE LANCET

Child-Pugh-score

Bilirubin (mg/L)

Albumin (g/L)

Ammonia (µg/dL)

Prothrombin time (%)

All patients before treatment

8·0 (2·1)

2·2 (1·5)

3·3 (0·5)

70·4 (43·7)

63·4 (16·8)

Patients with 1 year follow-up Shunt (n=35) Before treatment 1 month after randomisation 6 months 12 months

8·1 (2·1) 7·5 (1·7) 7·5 (1·6) 7·2 (1·7)

2·3 (1·5) 2·3 (1·3) 2·3 (1·5) 2·3 (1·4)

3·3 (0·5) 3·6 (0·4) 4·0 (0·5) 4·0 (0·6)

64·1 (27·4) 66·4 (24·0) 74·1 (33·1) 65·2 (19·4)

63·3 (13·5) 65·9 (13·1) 66·2 (13·6) 65·9 (14·6)

Patients with 1 year follow-up Endoscopy (n=37) Before treatment 1 month after randomisation 6 months 12 months

7·6 (2·0) 6·7 (1·9) 6·4 (1·5) 6·6 (1·8)

2·0 (1·6) 1·7 (1·1) 1·7 (1·3) 2·9 (6·6)

3·5 (0·5) 3·8 (0·4) 3·8 (0·6) 3·9 (0·5)

66·3 (32·2) 59·3 (45·8) 61·4 (36·2) 53·3 (18·3)

65·6 (18·0) 75·8 (15·5) 73·1 (17·1) 74·7 (20·2)

To convert values for bilirubin to µmol/L, multiply by 17·1; to convert values for ammonia to µmol/L, multiply by 0·5872. Changes during follow-up were not significant.

Table 3: Comparison of laboratory data (mean [SD]) of both treatment groups during a follow-up of 12 months

Hepatic encephalopathy The analysis of the time to first occurrence of hepatic encephalopathy is shown in figure 4. The difference between the curves of the two treatment groups is significant (p=0·011). Hepatic encephalopathy was seen in 22 (36%) patients of the shunt group and 12 (18%) patients of the endoscopic group. Clinically significant or refractory hepatic encephalopathy were seen in 16 (26%) and two (3%) patients of the shunt group and in seven (11%) and two (3%) patients in the endoscopic group, respectively. Three of the seven patients with clinically significant and the two patients with refractory encephalopathy in the endoscopic group were treatment failures and therefore had the transjugular shunt treatment. In all four patients with refractory encephalopathy a reducing stent was implanted that resulted in an improvement of the encephalopathy.15 Liver function, Child-Pugh class The Child-Pugh scores and the values for bilirubin, albumin, prothrombin time, and ammonia before treatment and during follow-up are shown in table 3. The pretreatment values of these subgroups with a follow-up of more than 1 year were comparable with those of the entire treatment groups. None of the variables showed a significant change during follow-up. In the shunt-group, the Child-Pugh score, albumin concentration, and prothrombin time improved, while the bilirubin and ammonia concentrations remained unchanged or increased. In the endoscopic group, all variables except bilirubin tended to improve. The increase in bilirubin at 12 months was due to the inclusion of two patients who did not respond to the endoscopic treatment and who developed liver failure after having received a transjugular shunt.

Multivariate analysis The results of a multivariate analysis of treatment and selected baseline variables with respect to the endpoints variceal rebleeding, death, and hepatic encephalopathy are shown in table 4. Compared with endoscopic treatment the transjugular shunt independently reduced the risk of variceal rebleeding and increased the risk of hepatic encephalopathy, but had a comparable rate of survival. In addition, patients with one or more previous variceal bleeding episodes and with alcohol-related liver disease had a reduced risk of developing hepatic encephalopathy. No other baseline variable showed a significant effect on the endpoints assessed. In a second step we investigated the possibility of differential treatment effects in groups of patients subdivided by clinical variables. For this purpose we recalculated a Cox regression model to prove the interaction effects between treatments and selected covariables. In this analysis no statistical significant effects of the treatments were observed with respect to rebleeding and hepatic encephalopathy. However, with respect to mortality we observed a statistically significant interactive effect of treatment with the presence or absence of ascites (p=0·031), as well as with the ammonia concentration (p=0·033) at the time of randomisation. The transjugular shunt looked more favourable in patients that presented with ascites or an ammonia concentration of less than 35 µmol/L (<60 µg/dL). By contrast, patients without ascites or an ammonia concentration greater than 35 µmol/L (>60 µg/100 mL) seemed to have an advantage from endoscopic treatment. However, an additional test to prove the qualitative nature of these interactions did not reach significance.14 This finding has to be treated with caution due to the post-hoc stratification.

Criterion

Variceal rebleeding

Death

Encephalopathy

Shunt treatment Male sex Age >60 years Diabetes mellitus Alcoholic aetiology Bilirubin >3·0 Ammonia >60 Ascites Previous bleeding Active index bleeding Need of more than three blood units Child-Pugh A vs Child-Pugh C Child-Pugh B vs Child-Pugh C

0·27 (0·12–0·62) 1·39 (0·64–3·00) 0·67 (0·26–1·73) 1·42 (0·53–3·83) 2·55 (0·79–8·28) 0·47 (0·15–1·51) 1·16 (0·56–2·42) 1·35 (0·58–3·12) 1·58 (0·72–3·47) 1·28 (0·61–2·68) 1·49 (0·63–3·54) 0·32 (0·06–1·64) 0·84 (0·24–2·97)

1·28 (0·39–4·18) p=0·69 2·24 (0·70–7·21) p=0·18 3·06 (0·82–11·41) p=0·10 1·40 (0·37–5·31) p=0·62 0·90 (0·24–3·36) p=0·87 0·97 (0·11–8·38) p=0·98 1·11 (0·37–3·27) p=0·86 0·70 (0·17–2·91) p=0·63 0·44 (0·15–1·25) p=0·12 1·57 (0·49–5·02) p=0·45 0·55 (0·19–1·59) p=0·27 0·31 (0·02–5·01) p=0·41 0·37 (0·37–3·70) p=0·40

4·83 (1·84–12·67) p=0·001 1·02 (0·44–2·36) p=0·96 1·36 (0·49–3·77) p=0·55 1·06 (0·34–3·30) p=0·92 0·37 (0·14–1·00) p=0·05 4·38 (0·50–38·80) p=0·18 0·49 (0·20–1·20) p=0·12 2·06 (0·83–5·11) p=0·12 0·27 (0·11–0·66) p=0·004 1·51 (0·58–3·96) p=0·40 0·82 (0·34–1·98) p=0·66 5·53 (0·41–74·16) p=0·20 4·94 (0·53–46·04) p=0·16

p=0·002 p=0·40 p=0·41 p=0·49 p=0·12 p=0·21 p=0·70 p=0·48 p=0·26 p=0·52 p=0·36 p=0·17 p=0·79

*Values presented are estimated relative risks with corresponding 95% CIs and p values from two-sided Wald’s test of Cox model estimates.

Table 4: Multivariate analysis of the prognostic value of selected clinical variables on the first occurrence of variceal rebleeding, death, and clinically significant or refractory encephalopathy*

Vol 349 • April 12, 1997

1047

THE LANCET

Discussion Our study compares the transjugular-intrahepaticportosystemic shunt with endoscopic treatment plus propranolol with the optional use of transjugular variceal embolisation and banding ligation, respectively. These additional measures were allowed to optimise the respective baseline treatments. So far, the combination of a transjugular shunt and variceal embolisation has not been studied. The transjugular embolisation before the stent implanation is quick and is, in our experience with more than 200 patients, safe and effective. This is confirmed by a randomised study by Smith-Laing and colleagues16 in which they found a reduction or delay of rebleedings after percutaneous transhepatic obliteration of varices. Similarly, the addition of propranolol and the optional use of polidocanol or banding ligation may have improved the results of endoscopic treatment. This assumption is based on the results of a meta analysis of nine studies17 that showed that either treatment, sclerotherapy and propranolol or banding ligation, were better than sclerotherapy alone.18–20 In our study, sclerotherapy was used for the treatment of the acute haemorrhage and of minor varices in patients who had received multiple previous sclerotherapies. By contrast, banding ligation was mostly used in elective situations and in patients who had had only few sclerotherapies or who had poor response to sclerotherapy. Patients admitted with acute bleeding who could not be stabilised and who proceeded to a bleeding emergency were excluded. The cumulative 1-year variceal rebleeding rate in the shunted patients was 15% and is comparable with our previous findings,3 but lower than that reported by LaBerge and colleagues.4 This may, at least in part, be due to the higher proportion of Child-Pugh class C patients in the latter study and the associated higher risk of rebleeding. Our results are comparable with the 1-year rebleeding rates after surgical shunts,21–24 which showed a high efficacy of the transjugular shunt despite its relatively frequent primary insufficiency. Considering the endoscopic group, our 1-year variceal rebleeding rate of 41% is within the lower range or rebleeding rates (30% to 60%) reported previously after sclerotherapy23–29 or banding ligation.18–20 Survival was similar between the groups with an estimated cumulative 1-year survival of 90% and a 2-year survival of about 80%. A combined analysis of patients with bleeding fatalities who were rescued by the shunt treatment and deaths is in favour of the shunt-treatment although the difference is not significant. Liver transplantation was carried out in only three patients, one of whom died from multiorgan failure. The other 13 patients who died were not eligible for transplantation because of their age or because of sudden death. Hepatic encephalopathy is a problem intrinsic to portosystemic shunts. Thus, it is not surprising that, compared with the endoscopic treatment, the shunted patients had a significantly higher incidence of hepatic encephalopathy that was comparable to surgical shunts.21–24 Although most of the patients responded to medical treatment or shunt reduction it is advisable to better select the patients according to their risk of shunt-induced hepatic encephalopathy. Multivariate analysis of our data shows an independent overall effect of the shunt treatment compared with the endoscopic treatment on rebleeding and hepatic 1048

encephalopathy. Considering the clinical variables, only alcoholic liver disease or previous variceal bleeding episodes correlated with a reduced risk of developing hepatic encephalopathy.The finding of possible interaction effects of ammonia concentration and ascites on the treatments should be interpreted with caution due the post-hoc character of the interaction hypotheses which were not inlcuded as part of the protocol, and to the increased number of significance tests done. It may, therefore, rather be of exploratory and hypothesisgenerating nature at this time. Longer follow-up and additional trials are needed to further confirm or refute these important observations. The quality of life in this patient population may, to a large extent, be assessed by factors such as rebleeding, hepatic encephalopathy, long-term procedural morbidity, effect of treatment on Child-Pugh class, and length of hospitalisation.These factors seem to be balanced between the treatment groups. However, considering the interventions during follow-up, 24 shunt revisions are comparable with 331 endoscopic interventions. Assuming that these interventions are comparable with respect to invasiveness, complications, and discomfort, the shunt may overall be the less compromising procedure. Our findings confirm to a large extent a recently published randomised trial that compared the transjugular shunt and sclerotherapy.30 However, this study had a considerably higher 1-year variceal rebleeding rate (shunt 27%, sclerotherapy 52%). This may be due to our study design allowing additional measures to optimise treatment outcome. Numerous randomised studies have been published comparing surgical shunts and scleroptherapy. Their meta-analysis did not show a difference in survival.17 This may in part be attributed to a selection bias favouring patients with low operative risk and excluding Child-Pugh class C patients in four of the six studies. Thus, only 336 from a total of 1131 patients were randomised. It can be assumed that the inclusion of a higher proportion of patients would have affected survival in a negative way for the surgical shunts. By contrast, the low invasiveness of the transjugular shunt procedure allowed inclusion of most patients who agreed to randomisation. Only 22 (12%) of 190 eligible patients were not randomised because they met exclusion criteria. Thus, by contrast with surgical shunts, the transjugular shunt may be a true alternative to endoscopic treatment not only in properly selected cases but in most patients with variceal bleeding. References 1

2

3

4

5

6

7

Rössle M, Richter GM, Nöldge G, et al. Performance of an intrahepatic portacaval shunt using a catheter technique: a case report. Hepatology 1988; 8: 1348 (abstr). Richter GM, Nöldge G, Palmaz JC, et al. Transjugular intrahepatic portacaval stent shunt: preliminary clinical results. Radiology 1990; 174: 1027–30. Rössle M, Haag K, Ochs A, et al. The transjugular intrahepatic portosystemic stent-shunt procedure for variceal bleeding. N Engl J Med 1994; 330: 165–71. LaBerge JM, Somberg KA, Lake JR, et al. Two-year outcome following transjugular intrahepatic portosystemic shunt for variceal bleeding: results in 90 patients. Gastroenterology 1995; 108: 1143–51. Haag K, Rössle M, Hauenstein KH, et al. The transjugular intrahepatic portosystemic stent-shunt (TIPS) for emergency treatment of variceal bleeding. Intensivmed 1993; 30: 479–83. Simpson KJ, Chalmers N, Redhead DN, Finlayson NDC, Bouchier IAD, Hayes PC. Transjugular intrahepatic portosystemic stent shunting for control of acute and recurrent upper gastrointestinal haemorrhage related to portal hypertension. Gut 1993; 34: 968–73. Perarnau J-M. Anastomose porto-cave intra-hepatique par voie transjugulaire. Gastroenterol Clin Biol 1993; 17: 22–27.

Vol 349 • April 12, 1997

THE LANCET 8

9 10

11

12

13

14 15

16

17 18

19

20

Lind CD, Malisch TW, Chong WK, et al. Incidence of shunt occlusion or stenosis following transjugular intrahepatic portosystemic shunt placement. Gastroenterology 1994; 106: 1277–83. Moriyasu F, Nishida O, Ban N, et al. “Congestion index” of the portal vein. Am J Roentgenol 1986; 146: 735–39. Pagliaro L, Burroughs AK, Sorensen TIA, et al. Therapeutic controversies and randomised controlled trials (RTCs): prevention of bleeding and rebleeding in cirrhosis. Gastroenterol Int 1989; 2: 71–84. Pugh RNH, Murray-Lyon MM, Dawson IL, Pietroni MC, Williams R. Transsection of the oesophagus for bleeding oesophageal varices. Br J Surg 1973; 60: 646–49. Conn HO. Trailmaking and number connection test in the assessment of mental state in portal systemic encephalopathy. Dig Dis Sci 1977; 22: 541–50. Haag K. Duplex-sonographic evaluation before and after TIPS; monitoring of shunt function. In: Conn HO, Palmaz JC, Rösch J, Rössle M, eds. Transjugular intrahepatic portosystemic shunts. New York, Tokyo: Igaku-Shoin, 1996: 319–30. Gail M, Simon R. Testing for qualitative interactions between treatment effects and patient subsets. Biometrics 1985; 41: 361–72. Hauenstein KH, Haag K, Ochs A, Langer M, Rössle M. The reducing stent: treatment for transjugular intrahepatic portosystemic shuntinduced refractory hepatic encephalopathy and liver failure. Radiology 1995; 194: 175–79. Smith-Laing G, Scott J, Long RG, Dick R, Sherlock S. Role of percutaneous transhepatic obliteration of varices in the management of hemorrhage from gastroesophageal varices. Gastroenterology 1981; 80: 1031–36. D’Amico G, Pagliaro L, Bosch J. The treatment of portal hypertension: a meta-analytic review. Hepatology 1995; 22: 332–54. Van Stiegmann GV, Goff JS, Michaletz-Onody PA, et al. Endoscopic sclerotherapy as compared with endoscopic ligation for bleeding esophageal varices. N Engl J Med 1992; 326: 1527–32. Laine L, El-Newihi HM, Migikowski B, Sloane R, Garcia F. Endoscopic ligation compared with sclerotherapy for the treatment of bleeding esophageal varices. Ann Intern Med 1993; 119: 1–7. Gimson AES, Ramage JK, Panoz MZ, et al. Randomised trial of

Vol 349 • April 12, 1997

21

22

23

24

25

26

27

28

29

30

variceal banding ligation versus injection sclerotherapy for bleeding oesophageal varices. Lancet 1993; 342: 391–49. Rikkers LF, Burnetti DA, Volentine GD, Buchi KN, Cormier RA. Shunt surgery versus endoscopic sclerotherapy for long-term treatment of variceal bleeding. Early results of a randomized trial. Ann Surg 1987; 206: 261–71. Henderson JM, Kuttner MH, Millikan WJ, et al. Endoscopic variceal sclerosis compared with distal splenorenal shunt to prevent variceal bleeding in cirrhosis. A prospective randomized trial. Ann Intern Med 1990; 112: 262–69. Teres J, Bordas JM, Bravo D, et al. Sclerotherapy versus distal splenorenal shunt in the elective treatment of variceal hemorrhage: a randomized controlled trial. Hepatology 1987; 7: 430–63. Warren WD, Henderson JM, Millikan WJ, et al. Distal splenorenal shunt versus endosopic sclerotherapy for long-term management of variceal bleeding. Ann Surg 1986; 203: 454–62. Fleig WE, Stange EF, Hunecke R, et al. Prevention of recurrent bleeding in cirrhotics with recent variceal hemorrhage: prospective, randomized comparison of propranolol and sclerotherapy. Hepatology 1987; 7: 355–61. Ink O, Martin T, Poynard T, et al. Does elective sclerotherapy improve efficacy of long-term propranolol for prevention of recurrent bleeding in patients with severe cirrhosis? A prospective multicenter, randomized trial. Hepatology 1992; 16: 912–19. Westaby D, Polson RJ, Gimson AES, Hayes PC, Hayllar K, Williams R. A controlled trial of oral propranolol compared with injection sclerotherapy for the long-term management of variceal bleeding. Hepatalogy 1990; 11: 353–59. Terblanche J, Kahn D, Bornman P. Long-term injection sclerotherapy treatment for esophageal varices. A 10-year prospective evaluation. Ann Surg 1989; 210: 725–31. O’Connor KW, Lehman G,Yune H, et al. Comparison of three nonsurgical treatments for bleeding esophageal varices. Gastroenterology 1989; 96: 899–906. Cabrera J, Maynar M, Granados R, et al. Transjugular intrahepatic portosystemic shunt versus sclerotherapy in the elective treatment of variceal hemorrhage. Gastroenterology 1996; 110: 832–39.

1049