A Prospective Randomized Trial Comparing Repeated Endoscopic Sclerotherapy and Propranolol in Decompensated (Child Class B and C) Cirrhotic Patients S.DASARATHY, MANISHA DWIVEDI,DINESHK. BHARGAVA, K.R. SUNDARAM AND K. RAMACHANDRAN Departments of Gastroenterology and Biostatistics, All India Institute of Medical Sciences, New Delhi 110029, India

A prospective randomized study was conducted to compare the efficacy of long-term endoscopic sclerotherapy vs. propranolol in Child class B and C patients with variceal bleeds within the 30 days before the stub. Forty-five and 46 patients were randomized to r d v e sclerotherapy and propranolol, respectively, after preentry stratification for Child scores. Sclerotherapy was administered with 1% polidocanol at 1O-day intervals until obliteration of varices was achieved. Propranolol was administered to achieve a reduction in resting pulse rate of 25%. Rebleeding occurred in 19 patients undergoing sclerotherapy and in 31 receiving propranolol (p < 0.05). The number of epieodes of mbleeding was higher (p < 0.06) in the propranolol group (n = 64) than in the sclerotherapy group (n = 35). The mean bleeding risk factor,number of hospitalizations for rebleeding and blood transfusion requirement were also significantly higher in the propranolol-treated patients. The median bleed-free period was more than 36 m o in the sclerotherapy group and 2.5 mo in the propranolol group (p c 0.01). The median survival time was dgni6cantly longer in the sclerotherapy group ( > 36 mo) than in the propranolol group (>24 mo). We conclude that in decompnsated cirrhotic patients, long-term endos c h c sclerotherapy is superior to propranolol in preventing rebleeding and improving survival. (HEPATOLOQY

1992;18:89-94.)

Variceal hemorrhage is a major complication of portal hypertension in patients with cirrhosis. It is also responsible for high morbidity and mortality associated with severe liver disease (1).Two nonsurgical modalities for prevention of recurrent bleeding include long-term propranolol treatment and endoscopic sclerotherapy (2-8). Lebrec et al. (2) suggested oral propranolol to be effective and safe as a medical treatment for wellcompensated cirrhotic patients. Three subsequent Received December 11, 1989; accepted December 16,1991 TMS manumfipt WM presented in part during the Sixth Biennial Meeting of the Asian P d c Association for the Study of the Liver, New Delhi, India, Febmmg 14,1988. Address reprint requests to Dinesh K Bhargava, M D , P h D , Add1 Prsfessor of G a ~ t ! r a e ~ ~ l oAll g y ,India Institute of Medical Smences, Ansan Nagar, New hlhi 110029, I d a 3111131410

89

studies (9-11) in a less selected population failed to confirm these results. Recently a metanalysis of the various studies on propranolol found it of value in the reduction of both primary and secondary variceal bleeding (12). Endoscopic sclerotherapy has also been found effective in reducing the episodes of rebleeding (4-8). However, the results were influenced by the liver function status (13-15).Both these nonsurgical modalities have been compared in the past (16-18)and showed conflicting results. None of the studies has specifically addressed the problem of long-term management of decompensated cirrhotic (Child class B and C) patients. This study was conducted to compare the efficacy of propranolol and endoscopic sclerotherapy in cirrhotic patients with Child class B and C status who had bled from esophageal varices. MATERIALS AND METHODS The study population comprised patients with cirrhosis and at least one episode of esophageal variceal bleeding. Consecutive patients with the following criteria were included after informed consent was obtained: (a) Child class B and C groups classified by the Pugh’s modification of Child’s scoring system (191,(b) endoscopic diagnosis of grade 4 esophagealvarices (20) with signs of high risk of bleeding (21) in the absence of any other potential bleeding site, (c) no contraindication to the use of B-blocking agents and (d) absence of previous treatment with B-blockers endoscopic sclerotherapy or surgery for portal hypertension. Patients were included after control of bleeding, which was defined as at least 24 hr of stable hemdynamic condition without further transfusion, stable or rising hematawit, no hematemesis and clear nasogastric aspirates. Consecutive patients were randomized by the sealed-envelope method to receive endoscopic sclerotherapy or propranolol after preentry stratification for Child status. Cirrhosis was verified by liver biopsy or considered the most likely diagnosis according to clinical features, biochemical findings and imaging techniques. The etiology of cirrhosis was identified by history, stigmata of chronic alcoholism, viral markers and histological appearance, when available. All patients had complete hematologid and liver function profiles on each visit. Propranolol was given when the patient was hemodynamically stable, had a resting pulse rate between 60 and 100 beats/min and a hemoglobin of 8 gm% or more. Incremental oral doses were given in a twice-daily schedule until the resting pulse rate was reduced by 25%.Compliance was evaluated by

90

DASARATHY ET AL. TABLE1. Demographic and clinical characteristics Characteristic

No. Age err) Sex (ME) Etiology of cirrhosis Posthepatitic Alcoholic Cryptogenic Child score 7-9 10-15 Time (days) after control of active bleeding and before entry

Sclerotherapy

Propranolol

45 43.8 -t 12.9" 3619

46 46.6 2 11.6" 4016

10 13 22

13 12 21

30 15 4.6 f 2.2"

30 16 3.9 2 1.9"

1-34 9

1-21 11

Range Median aData expressed as mean 2 S.D.

HEPATOLOGY

the episodes of rebleeding, number of hospital visits for bleeding, blood transfusion requirement to maintain hemodynamic stability and bleeding-free period. The mean bleeding risk factor was calculated as the average number of bleeding episodes per month of follow-up per patient. The mortality rate, causes of death, survival time and incidence of complications were compared between the two treatment groups. All patients were analyzed in the group to which they had been randomized (intention-to-treat analysis) when alternative treatment modalities had been offered (e.g., shunt surgery). StatieticuZ h l y s i s . In the comparison of treatment groups, qualitative variables were analyzed with the x2 test (221, and Wilcoxon's rank-sum test for independent samples was used for quantitative variables. The method-of-life-table analysis as described by Kaplan-Meier (23) was applied to the patients' survival time and to the time until the patients experienced rebleeding. The curves were drawn until the number of patients available for analysis was more than 10 (23). The resulting curves were compared with the log-rank test (24).

RESULTS history, resting and exercise pulse rate, compliance with scheduledappointments and the pill-counttechnique. Patients Between June 1, 1986, and June 30, 1990, 104 were subjected to upper gastrointestinal endoscopy at monthly consecutive patients who satisfied the inclusion criteria intervals during follow-up for reassessment of variceal grade. were randomized to receive propranolol or to undergo Sclerotherapy was carried out with an Olympus Q or X QlO endoscopic sclerotherapy. Four patients could not tolforward viewing panendoscope (Olympus Ltd., Tokyo, Japan) and an NM-1K injector. One percent polidocanol solution was erate propranolol; two experienced severe bronchoused as a sclerosing agent. All variceal columns were injected spasm and two had congestive heart failure within 48 hr intravariceally in a circumferential pattern, initially just of the start of therapy. They were not included in the proximal to the gastroesophagealjunction and then at a site 3 analysis because of inadequate treatment period. Three or 4 cm proximal to the gastroesophagealjunction. One or two patients assigned to the sclerotherapy group refused milliliters of sclerosing solution was injected at each site, with further injections. Six patients refused to enter a study a total of 20 dsession. The injections were repeated at 10-day protocol and were excluded. Forty-five and 46 patients intervals until eradication of varices was achieved. Eradication were included in the final analyses in the sclerotherapy of varices was defined as the disappearance of all the variceal and propranolol groups, respectively. The two randomly columns and a smooth esophageal mucous membrane. constructed groups exhibited no significant differences Follow-up of these patients after obliteration of varices was done with endoscopic examination at 3-mo intervals. The new in age, sex, Child score, etiology of liver disease, previous varices observed at endoscopy were subjected to injection at the episodes of gastrointestinal bleeding at inclusion or the duration of follow-up after entry into the trial. The same interval as above until eradication was achieved. Rebleeding episodes were defined as the demonstration of clinical and demographic characteristics of the patients active bleeding from esophagealvarices and the absence of any are shown in Table 1. other associatedbleeding lesions. Bleeding outside the hospital Among the patients in the propranolol group, one was classified as gastrointestinal bleeding of unknown origin patient experienced extreme lethargy requiring disconwhen a convincing history of hematemesis or melena was tinuation of therapy, and four others had frequent obtained. Such bleeding episodes were not classified as being of massive variceal bleeding. These five patients were then variceal etiology unless documented by endoscopy. Each managed with long-term maintenance sclerotherapy. bleeding episode in either group was treated with nasogastric aspiration with lavage and blood transfusion. Emergency However, they were analyzed in the propranolol group endoscopy was performed to identify the site of bleeding. until the time of propranolol treatment (intention-toContinued variceal bleeding was treated with balloon tam- treat basis). During the study period, six patients ponade. Failure to control bleeding or recurrence of bleeding (13.3%)in the sclerotherapy group and seven (15.2%)in after deflation of the balloon was managed with endoscopic the propranolol group were lost to follow-up. The mean sclerotherapy. Patients who continued to bleed despite this dose of propranolol used was 160 k 120 mg/day management underwent surgery. (range = 120 to 360 mdday). All the patients receiving Death was considered to be due to a given factor if it could propranolol had reductions in pulse rate of 25% after not be controlled and the patient's death occurred because of therapy. The time required for this reduction in pulse failure to manage that factor. It was defined as being due to rate was 9.6 k 4.8 days. bleeding if other factors such as progressive liver failure and Rebkeding. Rebleeding occurred in 19 (42.2%)of 45 infection were excluded. A contributory factor was any factor patients treated with sclerotherapy and 31 (67.4%)of 46 present but not of as much clinical significance as the direct managed with propranolol. This difference was statistifactor. Analysis was performed for rebleeding, survival and com- cally significant (p 0.05). There were significantly plications of therapy. The rebleeding parameters assessed were fewer patients (p < 0.05) with frequent rebleeding

-=

Vol. 16, No. 1, 1992

91

S C L E R O T H E W Y AND PROPRANOLOL IN CIRRHOSIS

TABLE 2. Rebleedine characteristics in the two treatment groum~as stratified by Child scores Propranolol

Sclerotherapy Characteristics

Episodes of rebleeding Variceal Unknown origin Bleeding risk" factor Median bleeding-free periodd Hospital admissions for bleeding' Mean duration of hospital stay (for bleedingP No. of units in blood transfusion**

Child class B

Child class C

Child class B

Child class C

19 2 0.08 2 0.01' > 36 mo 12 (11)f 3.4 2 1.3'

12

39

2

5

25 (11)

0.13 ? 0.02" 9 mo

0.28 2 0.09'

7 (6)

24 (18) 3.1 2 1.9'

19 1 0.40 ? 0.12" 2.5 mo 19 (13) 4.6 ? 2.4'

76 (18)

61 (13)

4.3

?

1.8'

28 (6)

6 mo

"Average of the number of bleeding episodes per patient per month of follow-up *Sclerotherapyvs. propranolol:p < 0.05. CDataexpressed as mean ? S.E.M. dSclerotherapyvs. propranolol:p < 0.01. 'sclerotherapy vs. propranolol: p < 0.005. ,Figures in parentheses indicate number of patients. aSclerotherapy vs. propranolol:p > 0.1 (not significant).

episodes (more than two rebleeds) among those treated with sclerotherapy (four patients) than propranolol (fourteen patients). The other rebleeding characteristics in the two treatment groups are shown in Table 2. The number of episodes of rebleeding was significantly higher (p < 0.05) in patients treated with propranolol than in those with sclerotherapy. Patients bled on more than one occassion, which resulted in a number of episodes of bleeding greater than the number of patients who rebled. The mean bleeding risk factor was significantly lower (p < 0.05) in patients managed with sclerotherapy than with propranolol. The distribution of cumulative bleeds per patient per month in the sclerotherapy group showed that the highest risk of rebleeding was in the initial 3 mo. However, in the propranolol group the number of bleeding episodes were evenly distributed throughout the follow-up period (Fig. 1). Two patients receiving propranolol discontinued the drug for 2 and 3 wk each. However, the episodes of rebleeding did not correlate with the period of noncompliance. The median bleeding-free period was more than 36 mo in the sclerotherapy group and 2.5 mo in the propranolol group (Fig.2). This was significantly longer in the sclerotherapy group than in the propranolol group (p c 0.01).The transfusion requirements were significantly greater in patients treated with propranolol than with sclerotherapy (p < 0.01). Similarly, the number of hospital admissions for bleeding was greater in the propranolol-treated patients (p < 0.005). However, the mean duration of stay in each admission was similar in the two treatment groups (p > 0.1). Bleeding was controlled by balloon tamponade in four patients of the sclerotherapy group and in 11 patients in the propranolol group. Three patients in each group underwent emergency sclerotherapy for control of bleeding. Two of the three patients treated with emergency sclerotherapy continued to experience rebleeds and were started on a

long-term maintenance sclerotherapy program. They formed part of the five-patient group treated with maintenance sclerotherapy. Two patients in the sclerotherapy group and four patients in the propranolol group were treated with intravenous vasopressin. Among the patients who rebled, two in each group were subjected to shunt surgery for uncontrolled bleeding. These two patients subjected to shunt surgery were analyzed in the group to which they had been assigned until surgery (intention-to-treat analysis). The remaining episodes of rebleeding were controlled by conservative treatment with blood transfusions alone. VariceaZ Eradication. Reduction by more than two grades or eradication of varices was achieved in 38 patients (84.4%) treated with sclerotherapy though eradication was achieved in 30 (66.7%).No significant difference was seen in the success of eradication of varices (70%vs. 60%),number of sessions required for eradication (7.4 2 2.4 and 7.8 2 2.9) or time for eradiction of varices (12.8 * 4.6 wk vs. 13.2 & 5.9 wk) in Child class B and C patients. No reduction in the size of the varices was seen during the entire period of follow-up in the patients treated with propranolol. There was no significant difference in the dose of propranolol required to achieve reductions in pulse rate between Child class B and C patients. Survival. Ten deaths (22.2%)occurred in the sclerotherapy group, and 19 patients (41.3%) died in the propranolol group (p = 0.08). Kaplan-Meier analysis (Fig. 3) showed a significantlylonger survival (p c 0.05) in patients treated with sclerotherapy (median survival > 36 mo) than with propranolol (median survival > 24 mo). Uncontrolled bleeding was the cause of death in five patients undergoing sclerotherapy and in 14 patients taking propranolol. Liver failure resulted in the death of four patients on sclerotherapy and in three receiving propranolol. Septicemia and bacterial peritonitis were the causes of death in one patient undergoing

92

DASARATHY ET AL

HEPATOLOGY

Cumulative episodes of bleeding

Propranolol I.

60

N

t l I I II

ow

Sclerotherapy

I'

2

O E 1

3

4

5

6

7

8

9

10

12

11

Time (months) FIG. 1. Distribution of cumulative number of episodes of rebleeding over time in the sclerotherapy- and propranolol-treated patients (p c 0.01). Arrows with E and S labels = points at which interventions were performed. S = surgical portacaval shunt; E = emergency endoseopicsclerotherapy. Other arrows = treatment with vasopressin (n = 2 in the sclerotherapy group and n = 4 in the propranolol group) and balloon tamponade (n = 4 in the sclerotherapy group and n = 11 in the propranolol group).

Proprandol

15

I

0

,

,

4

,

, 8

,

, , , , , . , . , . 12 16 20 24 28 BLEEDING FREE PERIOD IN MONTHS

,

.

32

.

36

FIG.2. &plan-Meier curve of bleeding-freeperiod for all bleeds in patients treated with endoscopicsclerotherapy or propranolol. The median bl&g-free pedod w a significantlylonger for patients treated with endoscopic sclerotherapy than for patients given propranolol (p c 0.01). Numbers on the figure refer to the patients at risk in each group.

sclerotherapy and in two patients treated with propranolol. Additional complications in these patients included development of encephalopathy in three patients undergoing sclerotherapy and in two patients taking propranolol who continued to bleed. Encephalopathy also developed in two patients undergoing sclerotherapy and in one taking propranolol who had septicemia and bacterial peritonitis. One of the patients taking propranolol who had encephalopathy also experienced fatal bleeding. Spontaneous bacterial peritonitis was present in all patients with septicemia and in one patient in each

treatment group with encephalopathy. No significant difference was found in the causes of mortality in the various groups of patients. Two patients in each group underwent shunt surgery for uncontrolled bleeding; all of them died within 1wk of surgery of progressive liver failure. Complicatione. Significant complications in the sclerotherapy group occurred in nine patients and in five patients in the propranolol group. Four patients treated with propranolol were excluded from analysis because of early-onset side effects, and the drug was discontinued in one more patient because of intolerable lethargy. Five

93

SCLEROTHERAPY AND PROPRANOLOL IN CmRHOSIS

Vol. 16, No. 1, 1992

Propmndd > I-

-d

27

0-4m

0 (I

CL

0.2-

o

l

.

l

'

r

,

a

.

other patients receiving propranolol complained of extreme lethargy. None of the patients treated with sclerotherapy required discontinuation of therapy because of complications. Most complications of sclerotherapy were minor, with self-limited retrosternal chest pain the most common (22%). Other complications included esophageal ulceration and dysphagia (12%). Sclerotherapy-induced esophageal ulcers subsided without specific therapy on postponement of injection therapy for 1or 2 wk. Esophageal strictures developed in two patients, necessitating dilatation. No statistical difference was seen in the incidence or the number of patients with complications in the two treatment groups (p > 0.1). DISCUSSION

This trial suggests that repeated sclerotherapy is more effective than oral propranolol in the prevention of upper gastrointestinal bleeding in a selected population of cirrhotic patients with portal hypertension. The parameters of rebleeding that favored sclerotherapy included the number of patients with rebleeding, episodes of bleeding and bleeding risk factor after initiation of treatment. These were significantly lower in the sclerotherapy group. Similarly, blood-transfusion requirement and the number of hospital admissions for bleeding as indicators of the severity of bleeding were significantly lower in patients treated with sclerotherapy than propranolol. Cumulative bleeds per patient by month showed that patients treated with sclerotherapy bled in the initial 3 mo, whereas those treated with propranolol continued to bleed throughout follow-up. This was related to the obliteration of esophagealvarices in the sclerotherapy group; no change in varieal grade occurred in the propranolol group. Median bleeding-free period also favored sclerotherapy. A standard protocol for the therapy of acute variceal

,

.

,

.

,

.

,

.

,

.

,

bleeding in all the patients ensured that no confounding factors influenced the rebleeding rates. It has been suggested that endoscopic sclerotherapy performed for acute bleeding influences the rebleeding rates later (25). Avoiding emergency sclerotherapy ensured equal distribution of variables that might influence the rebleeding episodes. Only three patients in each group were treated with emergency sclerotherapy, thus reducing the effects of intervention when they were compared. Results similar to those of this study have been reported by Alexandrino, Alves and Pinto Correiro (18). In their trial, variceal bleeding was significantly lower in the sclerotherapy group than in the propranolol group (29% vs. 62%)during a follow-up of about 4 yr. Other authors have also found that the addition of propranolol to therapy for patients undergoing long-term sclerotherapy confers no benefit ( 17). The rate of recurrent episodes of bleeding distributed throughout the follow-up period while patients received propranolol in this study was similar to those reported by other authors (9-11). The results of these studies differ from those reported by Lebrec et al. (2)in the high incidence of rebleeding, but this may be due to the differences in selection of patients. The poor results with propranolol in the studies by Burroughs et al. (9) and Villeneuve et al. (10) were probably due to the selection of patients with advanced cirrhosis. Most had cirrhosis of nonalcoholic etiology. Our patients were similar and had predominantly nonalcoholic advanced cirrhosis. The high rebleeding and mortality rates in this study were probably due to the poor liver function and high risk factors of rebleeding. The other finding in our study that was at variance with previous reports was the absence of any patient documented to have bled from gastric varices or congestive gastropathy. In the study by Alexandrino, Alves and Pinto Correia (18), gastric erosions were

94

DASARATHY ET AL.

HEPNTOLOGY

responsible for rebleeding in 10.8% of patients. The Yamada S. A prospective, randomized controlled trial of chronic esophageal variceal sclerotherapy. HEPATOLOGY 1985;5:584-589. reason for this discrepancy needs further evaluation. 7. The Copenhagen Esophageal Varices Sclerotherapy Project. ScleThe cumulative survival of both groups of patients rotherapy after first variceal hemorrhage in cirrhosis: a ranwas calculated according to the group into which they domized multicenter trial. N Engl J Med 1984;311:1594-1600. were randomized, even if the patients who rebled were 8. Bhargava DK, Dwivedi M, Acharya SK, Sundaram KR. Effect of low dosage of polidocanol in treatment of esophageal varices in moved to another treatment mode (intention-to-treat cirrhotic patients. Ind J Med Res 1988;88:515-521. basis). The median survival time was longer in patients Burroughs AK, Jenkins WJ, Slierlock S, Dunk A, Walt RP, treated with sclerotherapy than with propranolol. The 9. Ousafor TO, Mackie S. Controlled trial of propranolol for the results were not different if the patient withdrawn from prevention of recurrent variceal haemorrhage in patients with the propranolol group was analyzed with the sclerocirrhosis. N Engl J Med 1983;309:1539-1542. therapy group. Two other controlled trials comparing 10. Villeneuve JP, Pomier-Layrarses G, Infante-Rivard C, Willems B, Huet PM, Marleau D, Vallet A. Propranolol for the prevention sclerotherapy and propranolol showed no significant of recurrent variceal hemorrhage: a controlled trial. HEPATOLOGY differences in survival (16, 18). However, a recent 1986;61239-1243. metanalysis of the various trials with long-term endo- 11. Quienet AM, Czernichow P, Lerebours E, Ducrotte P, Tranvouez JL, Colin R. Etude controllee du propranolol dans la prevention scopic sclerotherapy suggested that survival is improved des recidives hemorrhagiques chez les patients cirrhotiques. with long-term sclerotherapy (26). Clin Biol 1987;11:41-47. Minor complications related to sclerotherapy were 12. Gastroenterol Hayes PC, Davis JM, Lewis JA, Bouchier IAD.Meta analysis of seen in 37%of patients and in 26% of those treated with value of propranolol in prevention of variceal hemorrhage. Lancet propranolol. Propranolol was well tolerated except in 1990;336:153-156. four patients who had severe side effects and in one 13. Bhargava DK, Dwivedi M, Dasarathy S, Sundaram KR. Sclerotherapy after variceal haemorrhage in non-cirrhotic portal fipatient who could not continue therapy beyond 3 mo. brosis: results and long term follow-up. Am J Gastroenterol Deterioration in liver or kidney function did not occur. 1989;84:1235-1238. Nor did difficulty in resuscitation from bleeding episodes 14. Bhargava DK, Dwivedi M, Dasarathy S, Arora A. Endoscopic sclerotherapy for portal hypertension due to extrahepatic oboccur in our patients taking propranolol, unlike previstruction: results and long term follow-up. Gastrointest Endosc ously reports (9). 1989;35:309-311. Thus trials comparing sclerotherapy with placebo 15. Bhargava DK, Dasarathy S, Sundaram KR, Ahuja RK. Efficacy of favor sclerotherapy (5-7), those comparing propranolol endoscopic sclerotherapy on long term management of esophageal with placebo show no significant benefit of propranolol varices: a comparative study of results in patients of cirrhosis of the liver noncirrhotic portal fibrosis and extrahepatic portal (9-11) and the study by Alexandrino, Alves, and Pinto obstruction. J Gastroenterol Hepatol 1991;6:471-475. Correiro (18)comparing sclerotherapy with propranolol 16. venous Fleig WE, Stange EF, Hunecke R, Schonborn W, Hurler V, Rainer favored sclerotherapy. Another trial (16) comparing K, Gaus W, et al. Prevention of recurrent bleeding in cirrhotics sclerotherapy and propranolol found no significant with recent variceal haemorrhage: a prospective, randomised comparison of propranolol and sclerotherapy. HEPATOLOGY 1987; difference between the two. The results of our study 7:355-361. suggest that sclerotherapy is superior to propranolol in D, Melia W, Hegarty J , Gimson AES, Stallon AJ,Williams the long-term treatment of decompensated cirrhotic 17. Westaby R. Use of propranolol to reduce the rebleeding rate during patients with past variceal bleeds. injection sclerotherapy prior to variceal obliteration. HEPATOLOGY 1986;6:673-675. 18. Alexandrino PT, Alves MM, Pinto Correia J. Propranolol or endoscopic sclerotherapy in the prevention of recurrence of variceal bleeding. J Hepatol 1988;7:175-195. 19. Pugh RNH, Murray Lyon IM, Dawson JL, Pietroni MC, Williams R. Transection of the esophagus for bleeding esophageal varices. Br J Surg 1973;60:646-649. REFERENCES 20. Conn HO. Ammonia tolerance in the diagnosis of esophageal 1. Graham DY, Smith JL. The course of patients after variceal varices: a comparison of endoscopic, radiologic and biochemical hemorrhage. Gastroenterology 1981;80:800-809. techniques. J Lab Clin Med 1976;70:442. 2. Lebrec D, Poynard T, Bernuau J, Bercoff E, Nouel 0, Capon JP, 21. Beppu K, Inokuchi K, Koyanogi N, Nakayamo S, Sakata H, Kitano Poupon R, Bouvry M, Bernard R, Behamou JP: A randomized S, Kobayashi M. Prediction of variceal hemorrhage by esophageal controlled study of propranolol for prevention of recurrent endoscopy. Gastrointest Endosc 1981;27:213-218. gastrointestinal bleeding in patients with cirrhosis: a finalreport. 22. Snedecor GW and Cochran WG, eds. Statistical methods. 7th ed. 1984;4:355-358. Ames, IA: Iowa State Press, 1980:64-82. HEPATOLOGY 3. Poynard T, Lebrec D, Hillon P. Sayegh R, Bernuau J, Naveau S, 23. Kaplan EL, Meier P. Non parametric estimation from incomplete observations. J Am stat Assoc 1958;53:457-481. Chaput JC, et al. Propranolol for prevention of recurrent gastrointestinal bleeding in patients with cirrhosis: a prospective 24. Mantel N. Evaluation of survival data and two new rank order study of factors associated with bleeding. HEPATOLOGY statistics arising in its consideration. Cancer Chemother Rep 1987;7: 1966;50: 163-170. 447-451. 4. Terblanche J , Bornman PC, Kahn D, Jonker MA, Campbell JA, 25. Burroughs AK. Long term endoscopic sclerotherapy versus no Wright J, Kirsch R. Failure of repeated injection sclerotherapy to elective treatment for the prevention of variceal rebleeding. In: improve long-term survival after oesophageal variceal bleeding. Burroughs AK, ed. Methodology and review of clinical trials in portal hypertension. Amsterdam: Elsevier Science Publishers BV, Lancet 1983;2:1328-1332. 1987:203-216. 5. Westaby D, MacDougall BRD, Williams R. Improved survival after injection sclerotherapy for esophageal varices: final analysis of a 26. Infant Rivard C, Esnaola S, Villeneuve JP. Role of endoscopic variceal sclerotherapy in the long term management of variceal 1985;5:827-830. controlled trial. HEPATOLOGY bleeding: a meta analysis. Gastroenterology 1989;96:1087-1092. 6. Korula J , Balart LA, Radvan G, Zweiban BE, Larson AW, Kao HW,

Acknowledgments: We thank Prof. B.N. Tandon and Professor S . Nundy for their critical comments during the study and Mr. R.K. Ahuja and Mr. Sudesh for the statistical analysis and preparation of this manuscript.