Preventive Therapy of First Gastrointestinal Bleeding in Patients with Cirrhosis: Results of a Controlled Trial Comparing Propranolol, Endoscopic Sclerotherapy and Placebo TONIANDREANI.'RENEE E. POUPON,' BEVERLEY J. BALKAU,'JEAN-CLAUDE TRINCHET,' * NICOLASPEIGNEY,' MICHELBEAUGRAND' AND RAOUL POUPON' JEAN-DIDIER GRANGE,' 'Hbpital Saint-Antoine. 75.571 Paris Cedex 12, "INSERM-Unite21, 94807 Villejuif Cedex, 'Hbpital Jean-Verdier, 93140 Bond?/.France
Propranolol and endoscopic sclerosis of esophageal varices are the two approaches currently used in prophylaxis of the first gastrointestinal hemorrhage in the cirrhotic patient. One hundred twenty-six cirrhotic patients with esophageal varices and no histories of bleeding were included in the trial regardless of the gravity of the cirrhosis or the size of the esophageal varices. Patients with hepatocarcinomas or other cancers, clearly impossible follow-up, previous treatment for portal hypertension or contraindication to p-blockers were excluded. After randomization, 43 patients received propranolol twice daily at a dose reducing the heart rate by 25%; 42 patients were treated with intravariceal and extravariceal injections of Polidocanol; 41 control patients received vitamin K orally as placebo. The patients were seen at 3-mointervals for 2 yr. On entry to the trial the three groups were comparable in terms of clinical and biological parameters, including size of esophageal varices (grade I = 51, grade Il = 54, grade I11 = 17). Child-Pugh classification (A = 29, B = 61, C = 32) and the origin of cirrhosis (alcoholic in 79% of cases). Twenty-four patients bled (two bled in the propranolol group, nine bled in the endoscopic sclerosis of esophageal varices group and 13 bled in the placebo group). Actuarial estimates (Kaplan-Meier)of the time of onset of first bleeding showed that the differences were significant between propranolol and placebo (p < 0.004) and between propranolol and sclerotherapy (p < 0.03) but not between sclerotherapy and placebo. The multivariate Cox model indicated that the size of esophageal varices and the Child-Pugh class were prognostic factors for the onset of GI bleeding; the treatment effect remained after adjustingfor these factors. No significant difference in survival was observed among the three treatment groups. The multivariate Cox model indicated that female gender and Received February 22. 1990: accepted J u n e :XI, 1990 'Current address: HBpital Tenon. Paris. France Address reprint requests to' Dr R e n d E Poupon. Unit6 de Recherches Cliniques & Epidhiologiques. INSERM I ' 21, 16. avenue Paul-Vaillant Couturier, 94807 Villejuif Cedex. France 31/1/24861
Child-Pugh class were factors predictive of survival and that treatment played no part in predicting survival. In conclusion, our study suggests that propranolol is effective in preventing the first gastrointestinal bleeding in the cirrhotic patient but does not improve the survival rate. (HEPATOLOGY 1990;12:14131419.)
Gastrointestinal (GI) bleeding is a frequent and grave complication of cirrhosis. It is estimated that 19%to 57% of cirrhotic patients with esophageal varices (EV) suffer at least one GI hemorrhage and that the first episode is fatal in 28% to 66%of cases (1-12). Two treatments are proposed for the prevention of the first GI bleeding: endoscopic sclerosis of esophageal varices (ESEV) and p-blockers, which reduce portal pressure. Four controlled studies published as articles have investigated the efficacy of p-blockers (propranolol or nadolol) for this indication in comparison with a control group (1-4).Treatment with p-blockers significantly reduced the risk of GI bleeding in all four studies. However, this effect was observed in the overall patient group in only two of these reports (1, 3) and in the subgroup of compliant patients (2) or those with mild cirrhosis (4) in the other two. Survival was significantly improved in only one study (3).Only patients with large EV were included in the four studies. It should be noted that patients with one of the following signs of severity were excluded: Child class C (2),Child-Pugh score > 14 (3), intractable ascites (l),intractable ascites or bilirubinemia >30 p,mol/L (4). Seven controlled studies published as articles have investigated the efficacy of ESEV in preventing the onset of the first GI hemorrhage in comparison with a control group (5-11). In four of these reports (5-7, 11) ESEV significantly reduced the risk of bleeding and significantly improved overall survival (6, 7, 11) or survival in the Child class A patient subgroup (5). In two other studies ( 8 , 9), ESEV led to no significant modification of the risk of bleeding or of survival. Finally, in one study (10) ESEV significantly increased
1413
ANDREANI ET AL.
1414
the risk of bleeding without affecting survival. In five of these seven studies (6-10)only patients with large EV were included, whereas i n the remaining t w o (5, 111 size was not a factor of inclusion. Gravity of cirrhosis (Child-Pugh score > 12) w a s only a factor of exclusion in o n e s tu d y (9). A metanalysis of controlled studies concluded that p-blockers probably reduce the risk of first bleeding; n o conclusion was possible concerning t hei r effect o n survival. Similarly, no conclusion could be draw n concerning ESEV (13). We report the first controlled s t udy comparing the efficacy of propranolol with that of ESEV in the prophylaxis of the first GI hemorrhage i n the cirrhotic patient. The concomitant study of a control g roup w as considered necessary given the contradictory results i n previous studies of ESEV and doubts abou t the improvement of survival w i t h p-blockers. T o gain a m ore general view of the results of these t r eat m ent s, it w as decided to include patients regardless of the size of the EV o r the severity of the cirrhosis. PATIENTS AND METHODS Patients. All adult patients seen in two Paris centers were considered eligible for inclusion if they met the following criteria: (a)cirrhosis proven by histological examination or, if this was unavailable, on the basis of clinical and laboratory test results, regardless of origin; (b) presence of EV on endoscopy, regardless of size; and (c) no history of GI bleeding by rupture of EV. Noninclusion criteria were: (a) existence of HCC; tb) contraindication to the use of propranolol (cardiac insufficiency, asthma, disturbance of auriculoventricular conduction); (c) refusal or unfeasibility of one of the three modes of treatment; (d)unfeasibility of regular surveillance (for reasons of distance or apparent indiscipline); (e) serious associated illness reducing life expectancy to < 1 yr; and tD previous treatment with ESEV, propranolol or surgery for portal hypertension. Informed consent was given by all patients and the study protocol was approved by the Ethics Committee (HBpital Saint-Antoine). Study Protocol. After inclusion, the patients were stratified according to severity of cirrhosis using the Child-Pugh score ( 14)as follows: one group with scores 5 to 8 and the other group with scores 9 to 15. According to this stratification, the patients in each center were randomly assigned to receive one of the three treatments. Propranolol doses were titrated to achieve a 25% reduction in resting heart rate. ESEV was carried out as follows: after premedication with intravenous diazepam or midazolam, as required, sclerotherapy was performed using a flexible endoscope (Olympus GIFQ or XQ, Olympus Corp., Lake Success, NY) and an ABS (Advanced Biochemical Supply, Saint-Die, France) or Olympus sclerosing needle. The sclerosant, 1% or 2% polidocanol, was administered by intravariceal and extravariceal injections (15) in the distal 10 cm of the esophagus. The quantity used varied according to the size of the EV, generally between 15 and 40 ml; gastric varices were not sclerosed. Treatment was repeated every week or 2 wk until complete disappearance of the EV was effected. Thereafter the results were verified endoscopically every 3 mo; in case of recurrence, sclerotherapy was repeated according to the same treatment modality. The control patients received one tablet of vitamin K (10 mg) morning and evening as placebo. As in a previous study (41,
HEPXI'OI.OGY
vitamin K was considered suitable given the lack of any known effect on GI bleeding in the cirrhotic patient and because compliance might be increased by the fact that liver indications are mentioned on the packet. For obvious reasons, these treatments were not administered blindly. Other associated treatments were authorized, with the exception of p-blockers. On inclusion in the trial, a questionnaire was filled out containing clinical, endoscopical and laboratory data (differential blood count, creatinine, bilirubin, AST, y-glutamyltransferase, prothrombin time, factor V and albumin). EV were classified in three grades as follows: grade I = nonconfluent EV flattened by insufflation; grade 11 = EV separated by zones of normal esophagus and not flattened by insufflation; and grade 111 = confluent EV not flattened by insufflation. Patients were seen 1 mo after inclusion in the trial and then at 3-mo intervals. Clinical and biological parameters and compliance and tolerance to the treatment were recorded at each examination; patients with alcoholic cirrhosis were asked whether their drinking habits had changed. Abstinence was defined as the total arrest of alcoholic consumption stated by the patients. When necessary, the dosage of propranolol was adjusted to obtain the desired heart rate. In addition, once a year, patients were screened for HCC with ultrasonography and tr-fetoprotein assay and fiberoptic examination was again performed. When patients failed to attend the follow-up examinations, attempts were made to contact them by mail, telephone or through their family doctor; failing this, public records were consulted to determine whether the patient was still living. Treatment and follow-up were planned to last 2 yr. Patients presenting significant GI bleeding (requiring the transfusion of at least 2 U of RBCs) underwent endoscopical examination as soon as this was feasible to determine the mechanism. Patients with EV rupture were withdrawn from the assigned treatment and given standard therapy. The criteria for variceal bleeding was active bleeding from the varices or the presence of clot on a varix and no other detectable cause of hemorrhage. Standard therapy applied in the two centers included esophageal tamponade and if necessary intravenous vasopressin. Emergency sclerosis was not performed. After bleeding from other causes, the assigned treatment was continued as soon as possible. The assigned treatment was also interrupted in the case of severe side effects or if surgery was required because of an aggravation of the cirrhosis (liver transplantation or portacaval shunt for refractory ascites). Patients were nonetheless monitored until the end of the trial after withdrawal from the assigned treatment. Statistical Analysis. Assuming that 60% of patients in the control group would be free of bleeding after 2 yr of follow-up and that treatment would increase this figure to 90%, 38 patientshreatment group were required for a type I error of 5% and a power of 80% if the comparison were made using the two-sided log-rank test (16). Biological and clinical characteristics of the patients at entry to the trial were compared between the three treatment groups using ANOVA for quantitative variables and the x 2 test for qualitative variables. The data were analyzed on an intentionto-treat basis. The percentages of patients free of bleeding and surviving during the 2 yr of follow-up were estimated using the Kaplan-Meier method (17) and compared using the Breslow statistic ( 18). Quantitative prognostic variables were either divided into the three classes as used in the Child-Pugh score (prothrombin time and bilirubin and albumin concentrations) or into two classes as defined by the median. The multivariate Cox model (19) was used to identify factors predictive of GI bleeding and survival and to compare the treatments after
TAHIA1. Characteristics of patients at entry to study according to treatment" Characteristics
Propranolol
Sclerotherapy
Plareho
Significance
adjustment for possible prognostic factors. Variables were Follow-up. One patient with HCC was inadvertently considered eligible for inclusion i n the (:ox models if they were included. Fourteen patients were lost to follow-up after a predictive at the 20% level of significance or more in the period of 4.9 2 1.9 mo (propranolol = six, sclerosis = univariate analysis. Age was always forced into the propor- six, placebo = two). The remaining patients were all tional hazards model, and t h e other predictive variables were treated until one of the end-points defined in the confirmed by exhaustive searching of numerous models. '1'0 symmetrize distributions. t h e logarithms of bilirubin. pro- protocol: EV rupture = 15; discontinuation of treatthrombin, albumin and AS'I were used. Either the Child-Pugh ment because of severe iatrogenic complications = 3 : classification or the variables used to define it were entered refusal to continue treatment = 7; and secondary surgiinto the model. The proportional hazards assumptions of the cal treatment required for aggravation of the cirrhomodels were validated. Statistical analyses were performed sis = 6. With the exception of those lost to follow-up, using BMDP software 1 2 0 )on ;I VAX 8.530 at the INSERM patients were monitored for 2 yr or until death. computing center, Villijuif. In the propranolol group, severe side effects occurred i n five patients (12%) (hypotension in three and severe HESULI'S asthma in two) requiring discontinuation in three Patients. Between November 1985 and February patients. Nine patients ( 2 1 % )did not regularly take 1988, 126 patients were included in the study. During propranolol. Among these patients, one bled during the same period, a prospective study carried out in one follow-up. In the sclerotherapy group, severe side effects of the centers showed that of 165 consecutive patients were observed in three patients (8%) (one esophageal meeting the inclusion criteria only 78 (47%) could stenosis that regressed after a single dilation and two actually be enrolled. The remaining 87 patients were febrile episodes that regressed under antibiotics); these excluded for the following reasons: follow-up deemed not side effects did not require discontinuation of treatment. feasible (33%0), HCC (26%), other cancers (8%), previous Moderate dysphagia was frequent during the first few treatment (propranolol, sclerotherapy or surgery for days after sclerotherapy and was not considered a portal hypertension) (10%), moribund (8%), propranolol complication. Eleven patients (26%1 were not compliant contraindicated (12%) and refusal to participate ( 1%). for sclerotherapy; four refused to undergo sclerotherapy Of the 126 patients included. 43 were treated with despite initially consenting to participate in the trial and propranolol, 42 by ESEV and 41 with placebo. At seven patients refused to complete sclerotherapy after inclusion, no significant difference was found between one to three sessions. In the propranolol group the dosage initially conthe three treatment groups with regard to clinical and sidered effective was 91 ? 10 mg. This required adapbiological data (Table 1) .
1416
HEPATOLOGY
ANDREANI ET AL.
n
e
v)
60.
.--
n c
0
0
12
6
18
i
a?
24
0
Months
i
12
6
18
24
Months
FIG. 1. Cumulative percentages of patients free of bleeding in the propranolol, sclerotherapy and placebo groups (Kaplan-Meier estimates). Percentages were significantly different between propranolol and placebo (p < 0.004) and between propranolol and sclerotherapy groups (p c 0.03).
FIG.2. Cumulative percentages of surviving patients treated with propranolol, sclerotherapy and placebo (Kaplan-Meier estimates). Significant differences in survival were not observed among the three treatment groups.
TABLE 2. Type of gastrointestinal bleeding (variceal or other) according to severity of cirrhosis and treatment Propranolol Classification
Varices
Other
Sclerotherapy Varices
Other
Child-Pugh class A
B C TOTAL
Size of varices Grade I Grade I1 Grade I11 TOTAL ~
~~
~~~~
Placebo Varices
Other
Total
0 8 2 10
2 12 10 24
2 5 3 10
5 13 6 24
~~~
"Erosive gastritis (n = 2); gastric ulcer (n = 1). bErosive gastritis (n = 1); fundic varices (n = 1);gastric ulcer (n
=
tation for six patients after follow-up examinations. In the ESEV group, EV disappeared completely in 24 patients after a mean of three sessions (range = 1 to 7). The quantity of polidocanol injected at each session was 22 2 1.8 ml (range = 5 to 60 ml). Relapse occurred in eight of these patients and further sclerotherapy was performed in seven. Among the patients with alcoholic cirrhosis, no significant difference between the three groups in terms of alcohol withdrawal was observed during the trial. Gastrointestinal Bleeding. Twenty-four patients had GI bleeding during the study period; two were in the propranolol group, nine were in the ESEV group and 13 were in the placebo group. The timing of these events is shown in Figure 1. The 1-yr and 2-yr bleeding rates were, respectively, 30% and 39% in the placebo group, 23%and 31%in the sclerotherapy group and 6% and 6% in the propranolol group. The differences were significant between propranolol and placebo (p < 0.004) and between propranolol and sclerotherapy (p < 0.03). These results were not affected if only those patients having started treatment or those having completed
1).
treatment were considered. Table 2 gives details of GI bleeding in terms of origin, size of the EV and gravity of cirrhosis. The difference between propranolol and placebo remains significant (p < 0.02) if only the 18 patients who had GI bleeding as a result of EV rupture are considered (propranolol = 2, ESEV = 6 and placebo = 10). GI bleeding was the cause of death in seven of the 24 patients (29%).Two of these deaths were due to massive bleeding while the patient was at home where no treatment could be attempted. The prothrombin time and the Child-Pugh class were both factors predictive of GI bleeding after univariate analysis (Table 3). The multivariate Cox model indicated that the size of EV and the Child-Pugh class were prognostic factors for the onset of GI bleeding; the treatment effect remained after adjusting for these factors (Table 4). Suriual. The 1-yr and 2-yr survival rates were, respectively, 66%and 54% in the placebo group, 71%and 52%in the sclerotherapy group and 79%and 66% in the propranolol group. No significant difference in survival was observed between the three treatment groups (Fig.
TABLE 3. Statistical significance (univariate analysis) of factors thought to be prognostic for gastrointestinal bleeding and survival Significance Classes for factof'
Factor
GI bleeding
Sex i male 1 Age lyr) Bilirubin I pmol 1.1 Prothrombin time 1 ' ; 1 Factor V (0) Albumin (gm 1.1 ASI' i multiple of normal GGT ( W L i Creatinine (pinol 1, I Size of EV Ascites Encephalopathy Child-Pugh class Type of cirrhosis Duration of cirrhosis I nio' Clinical center Treatment
survival 0.04
07 0.5
0.3 0.02 0.02 0.1 0.006
0.09 0.00.5 0.6 0.3 0.3 07 02 0. 1
1
0.007 0.5 1 0.4
0.08 0 :3
0.06 0.5
0.02 0.3 0.5 0.6 0.02
0.001
0.2 0.5 0.9 0.7
"Factors were analyzed in the classeh indicated GGT = y - g l u t a m y l t r a n s f e r ~ i ~ ~
TABLE 4. Prognostic factors for GI bleeding and survival identified by the
Cox model after adjustment for age at entry to
the trial Model
Significance
Relative risk (95% confidence interval)
GI bleeding model Size of EV' Treatment
iII 1111 Vh I Placebo vs propranolol ESEV vs propranolol I
Rvh A .A
I) 01
Child-Pugh clash
c Vh
Survival model Child-Pugh class
Sex
x
0 00 1
H V h
0.01
CVAA Male vs female
4.1 (1.4. 12.11 9.3 (1.9. 43.71 7.6 (1.5. 37.01 X 2 (1.2. 18.21 6.1 (1.2. 30.31 8 9 (2.0. 38.81 16.7 (3.7, 74.91 2.5 (1.2. 5.1)
~~
"Variables are shown in t h e order of entry to the model. "p Values indicate the significancc~of'each factor in the model
'rAs1.e 5. Causes of death during the 2 yr of the study according to treatment Cause of death
Treatment complicat ion Variceal bleeding Other GI bleeding Complications of cirrhosiNot associated with cirrho5is Unknown cause roi
\I
Propranolol
0 1
0 S
Sclerotherapy
0 > 0 11
-
1
1
:I
4 18
13
2). However, a slight tendency toward an improvement in survival was seen in the sclerotherapy group and a stronger tendency appeared in the propranolol group. Causes of death are given in Table 5 according to gravity of cirrhosis and the treatment groilp. A number of
Placebo
Total 0 7
1
0 28 4 10 49
f'actors were predictive of survival after the univariate analysis (Table 3). The multivariate Cox model indicated that female gender and Child-Pugh class were factors predictive of survival (Table 4); treatment played no part in the survival of patients.
1418
ANDREANI Err AL.
DISCUSSION
This study confirms the frequency and the gravity of the first GI hemorrhage in the cirrhotic patient: during a 2-yr period GI bleeding occurred in 31% of our untreated patients and was fatal in 30% of these cases. Our results are comparable to those of previous studies in which the occurrence of GI bleeding in patients with EV, regardless of size, was 19%to 57% (5, 11, 12), while the rate of mortality directly related to GI bleeding was 28%to 66%(1-11).It is noteworthy that the incidence of GI bleeding was relatively high in the untreated patients with grade I EV: 10%in our study but less than in two previous reports, 35% and 33% ( 5 , 11). Two factors found to be predictive for the occurrence of the first GI hemorrhage in previous studies (12, 21)-the large size of EV and the Child-Pugh score of cirrhosis- were also found here. It is difficult to explain the finding that men are more susceptible to death than women. The most important result of this study is that it shows that propranolol is superior to ESEV in the prevention of the first GI hemorrhage in the cirrhotic patient. Survival, however, was not significantly different between these two treatment groups. The frequency of side effects was similar, but it should be noted that 12% of the patients could not be included because propranolol was contraindicated. We also demonstrated that, as in the four previous studies (1-41, propranolol significantly reduces the risk of bleeding compared with a control group. This is the first time that this beneficial effect has been observed in a group of unselected cirrhotic patients, regardless of the severity of the cirrhosis or the size of the EV. The survival rate was not, however, significantly improved. In contrast, ESEV did not significantly improve either the risk of bleeding or the survival rate. It should, nonetheless, be noted that the results of ESEV were slightly better than in the control group; however, this negative result might have been due to a lack of power. Indeed, 62 patientslgroup would have been required for these results to reach the threshold of significance. In light of these results and those of previous studies, two questions arise: what is the explanation for the observed inefficacy of ESEV and why does propranolol not improve the survival rate when it reduces the risk of bleeding? If one accepts that EV completely eradicated cannot bleed, hemorrhage in these patients can occur, in theory, by one of two mechanisms-i.e., the persistence of EV (lack of patient compliance or difficulties in persistent eradication) or bleeding resulting from another cause (gastric varices, ulcers or congestive gastropathy). In this study, the main cause of failure was lack of compliance. In the ESEV group, of the patients who bled from EV, 50% had refused to begin or to continue treatment sessions. In previous studies, the distinction between a lack of compliance and a difficulty in eradicating EV was not clearly established, but it seems probable that the two factors are related because patients who require the largest number of treatment sessions stand the highest risk of dropping out. It should be noted that in previous studies, complete eradication
HEPAIWOGY
of EV was impossible in 30% to 51% of patients. In addition, several authors have stressed that the risk of bleeding in this subgroup is particularly high (7,8).The study by Santangello et al. (10) and two other reports published in abstract form (22,231 focused attention on a possible increase in the risk of GI bleeding after ESEV. Such bleeding was attributed to EV ulcerations produced by the sclerosing product. The distinction between this and spontaneous EV rupture is difficult to establish, particularly if one takes into account the fact that asymptomatic ulceration is almost always observed in the first few days after ESEV. This is considered to be part of the process leading to sclerosis (24). Two possible explanations can be proposed to explain the discrepancy between the three studies where ESEV increased the rate of bleeding and the seven studies (including ours) where this rate was lowered compared with the control group. First, it is tempting to incriminate differences in technique: polidocanol was used in the seven “positive” studies, all carried out in Europe, whereas tetradecylsodium was used in two “negative” studies performed in the United States. However, the third “negative” study was European and also involved polidocanol. The second possible explanation for the discrepancy is a difference in the rate of spontaneous bleeding, which was clearly higher in the seven “positive” studies (304 to 66%1 than in the three “negative” ones (8% to 15%). In contrast, the difference in the risk of bleeding in the sclerotherapy-treated group between the “positive” studies (9%’to 29%)and “negative” studies (22%to 35%) was far less marked. It might thus be deduced that when the risk of spontaneous bleeding is high, ESEV would be beneficial despite a certain number of hemorrhagic events caused directly by the treatment. Given the high rate of mortality directly linked to the first GI hemorrhage in the cirrhotic patient, it is surprising that the significant reduction in the risk of bleeding obtained by the use of P-blockers only led to a significant improvement in the survival rate in one (3) of five studies (1, 2, 41, including ours. In all these studies death among the patients with no bleeding was mainly related to other complications of cirrhosis. The failure of P-blockers to improve survival in our study might be related to a lack of power and to too-small sample sizes. The two metanalyses of controlled trials (13, 25) have both concluded that P-blockers probably reduced the risk of first bleeding. However, regarding survival, only one of the two metanalyses (25) has concluded that p-blockers improved survival. In contrast, in the four studies (5,6, 7, 11) in which ESEV significantly reduced the incidence of bleeding, the survival rate was significantly increased. In conclusion, it appears that P-blockers should be prescribed for the prevention of the first GI hemorrhage in the cirrhotic patient, given that their efficacy has been demonstrated in all studies published to date despite uncertainty concerning their effect on survival. This therapeutic efficacy of P-blockers has been so far established in cirrhotic patients in good condition and with large varices. Our study, which has to be confirmed,
Vol 12, N o 6 . 1990
I ' K O I ' I I ~ I , . \ ~ " I ~ THERAPY I~~ OF (;IIS'I'KOIN'I'ES'TINAI,
suggests that patients - whatever the severity of cirrhosis and the size of varicee - might benefit from thc drug. ESEV should not be recommended as first-line treatment but might be envisaged for patients who do not tolerate propranolol or for whom it is contraindicated when the risk of spontaneous bleeding is particularly high.
Acknowledgments: We thank Patricia Grand and Beatrice Pineau for their technical assistance. REFERENCES 1 . Ideo G, Bellati G. Fesce E.Griinoldi L). Nadolol can prevent the first gastrointestinal bleeding i n cirrhotics: a prospective. randomized study. HEPA.I.(II o(;~ 198H:X:6-9. 8 . Lebrec D. Poynard T, Capron J P . Hillon P. Geoffroy P, Roulot D. Chaput J-C. et al. Nadolol for prophylaxis of gastrointestinal bleeding in patients with cirrhosis: a randomized trial. J IIepatol 1988;7:1 18-125. :3. Pascal J P . Cales P, Multicenter Study Group Propranolol in the prevention of first upper gastrointestinal tract hemorrhage in patients with cirrhosis of the liver and esophageal varices N Engl J Med 1987;317:856-861. 4. The Italian Multicenter Project for Propranolol In Prevention of Bleeding. Propranolol prevents first gastrointestinal bleeding in non ascitic cirrhotic patients. J Hepatol 1989;9:75-83. 5. Koch H , Henning H. Griinm H , Soehendra N. Prophylactic sclerosing of esophageal varices: rt-sultsof a prospective controlled study. Endoscopy 1986: 18:40-43 6. Paquet KJ. Prophylactic endoscopic sclerosing treatment of the oesophageal wall in varices a prospective controlled randomized trial. Endoscopy 1982;14:4-5. 7. Piai G , Cipolletta L. Claar M. Marone C;. Uianco MA. Forte (;, Iodice G. Prophylactic sclerotherapy of' high-risk esophageal varices: results of a multicentric prospective controlled trial Hw.vroi.o(;~ 1988;8:1495-1500. 8. Potzi R, Bauer P . Reichel W. tierstan E. Kenner F'. Gang1 .4 Prophylactic endoscopic sclerotherapy of' oesophageal varices in liver cirrhosis: a multicentre prospective controlled randomised trial in Vienna. Gut 1989;:30:87s'3-879. 9. Sauerbruch T , Wotzka H. Kijpcke W. t-Iarlin M. Heldwein W. Bayerdorffer E, Sander R. et al. Prophylactic sclerotherapy before the first episode of variceal hemorrhage in patienh with cirrhosis. N Engl J Med 1988;319:8-15
BLEEDIN(; IN C'IRKHOSIS
1419
(;J. Prophylactic sclerotherapy of large esophageal varices. N Engl J Med 1988: 3 18:8 1 4-8 18. 1 1 . Witzel L. Wolbergs E. Merlu I I . I'rophylactic endoscopic sclerotherapy of oesophageal varices. a prospective controlled study. Lancet 1985;1:773-775. 12 Burroughs AK, d'Heygere F. Mclntyre N . Pitfalls in studies o f ' prophylactic therapy for variceal bleeding in cirrhotics. HEPA N)L,IN;Y 1986;6:1407-1413. 1 : j Pagliaro L. Burroughs AK. Sorensen T1.4,Lehrec D, Morabito A, D'Amico G, Tine F. Therapeutic controversies and randomised controlled trials (RCTs):prevention of bleeding and rebleeding in cirrhosis. Gastroenterol Int 1989;2:71-84. 1-1. Pugh RNH. Murray-Lyon IM. Dawson J L , Pietroni MC, Williams R . Transection of the oesophagus for bleeding oesophageal varices. Br J Surg 1973;60:646-649. 16 Soehendra N, de Heer K. Kempeneers I. Frommelt L. Morphological alterations of the esophagus after endoscopic sclerotherapy of varices. Endoscopy 1983;15:291-296. 16. Freeman LS. Tables of the number of patients required in clinical trials using the logrank test. Stat Med 1982;1:121-129. 17 Kaplan SA, Meier P . Nonparametric estimation from incomplete observations. J Am Stat Assoc 1958;53:457-481. 18. Breslow N. A generalised Kruskal-Wallis test for comparing k samples subject to unequal patterns of censorship. Biometrika 1970;57:579-594. 19 Cox DR. Regression models and life tahles. J R Stat Soc 197234: 187-220. 2 0 . Dixon WJ. BMDP statistical software. Vol. 2. Berkeley. CA. University of California Press, 1988:689-742. 21. The North Italian Endoscopic Club for the Study and Treatment of Esophageal Varices. Prediction of the first variceal hemorrhage in patients with cirrhosis of the liver and esophageal varices: a prospective multicenter study. N Engl J Med 1988;319:983-989. 22. Gregory P, Hartigan P, Amodeo D. Baum R, Camara D, Colcher H. Fye C. et al. Prophylactic sclerotherapy for esophageal varices in alcoholic liver disease: results of a VA cooperative randomized trial [Abstract]. Gastroenterology 1987;92:1414. 2 3 . Planas R, Boix J. Dominguez M. Abad A, Quer J C , de Leon R. Villagrasa M, et al. Prophylactic sclerosis of esophageal varices (EV):prospective trial IAbstract I . J Hepatol 1989;9(suppl 1 ): 144 24. Sarin SK, Nanda R, Vij *JC, Anand BS. Oesophageal ulceration after sclerotherapy: a complication or an accompaniment? Endoscopy 1986;18:44-45. 25. Pignon J P , Poynard T , Naveau S. Chaput J-C. Metaanalyse des essais randomi& des p-bioquants dans la prevention des hemorragies digestives des sujets atteints de cirrhose IAbstract I. Gastroenterol Clin Biol 1988;12:A175. 1 0 Santangelo WC, Dueno MI. Estes BL. KreJs
Propranolol and endoscopic sclerosis of esophageal varices are the two approaches currently used in prophylaxis of the first gastrointestinal hemorrhage in the cirrhotic patient. One hundred twenty-six cirrhotic patients with esophageal varices and no histories of bleeding were included in the trial regardless of the gravity of the cirrhosis or the size of the esophageal varices. Patients with hepatocarcinomas or other cancers, clearly impossible follow-up, previous treatment for portal hypertension or contraindication to p-blockers were excluded. After randomization, 43 patients received propranolol twice daily at a dose reducing the heart rate by 25%; 42 patients were treated with intravariceal and extravariceal injections of Polidocanol; 41 control patients received vitamin K orally as placebo. The patients were seen at 3-mointervals for 2 yr. On entry to the trial the three groups were comparable in terms of clinical and biological parameters, including size of esophageal varices (grade I = 51, grade Il = 54, grade I11 = 17). Child-Pugh classification (A = 29, B = 61, C = 32) and the origin of cirrhosis (alcoholic in 79% of cases). Twenty-four patients bled (two bled in the propranolol group, nine bled in the endoscopic sclerosis of esophageal varices group and 13 bled in the placebo group). Actuarial estimates (Kaplan-Meier)of the time of onset of first bleeding showed that the differences were significant between propranolol and placebo (p < 0.004) and between propranolol and sclerotherapy (p < 0.03) but not between sclerotherapy and placebo. The multivariate Cox model indicated that the size of esophageal varices and the Child-Pugh class were prognostic factors for the onset of GI bleeding; the treatment effect remained after adjustingfor these factors. No significant difference in survival was observed among the three treatment groups. The multivariate Cox model indicated that female gender and Received February 22. 1990: accepted J u n e :XI, 1990 'Current address: HBpital Tenon. Paris. France Address reprint requests to' Dr R e n d E Poupon. Unit6 de Recherches Cliniques & Epidhiologiques. INSERM I ' 21, 16. avenue Paul-Vaillant Couturier, 94807 Villejuif Cedex. France 31/1/24861
Child-Pugh class were factors predictive of survival and that treatment played no part in predicting survival. In conclusion, our study suggests that propranolol is effective in preventing the first gastrointestinal bleeding in the cirrhotic patient but does not improve the survival rate. (HEPATOLOGY 1990;12:14131419.)
Gastrointestinal (GI) bleeding is a frequent and grave complication of cirrhosis. It is estimated that 19%to 57% of cirrhotic patients with esophageal varices (EV) suffer at least one GI hemorrhage and that the first episode is fatal in 28% to 66%of cases (1-12). Two treatments are proposed for the prevention of the first GI bleeding: endoscopic sclerosis of esophageal varices (ESEV) and p-blockers, which reduce portal pressure. Four controlled studies published as articles have investigated the efficacy of p-blockers (propranolol or nadolol) for this indication in comparison with a control group (1-4).Treatment with p-blockers significantly reduced the risk of GI bleeding in all four studies. However, this effect was observed in the overall patient group in only two of these reports (1, 3) and in the subgroup of compliant patients (2) or those with mild cirrhosis (4) in the other two. Survival was significantly improved in only one study (3).Only patients with large EV were included in the four studies. It should be noted that patients with one of the following signs of severity were excluded: Child class C (2),Child-Pugh score > 14 (3), intractable ascites (l),intractable ascites or bilirubinemia >30 p,mol/L (4). Seven controlled studies published as articles have investigated the efficacy of ESEV in preventing the onset of the first GI hemorrhage in comparison with a control group (5-11). In four of these reports (5-7, 11) ESEV significantly reduced the risk of bleeding and significantly improved overall survival (6, 7, 11) or survival in the Child class A patient subgroup (5). In two other studies ( 8 , 9), ESEV led to no significant modification of the risk of bleeding or of survival. Finally, in one study (10) ESEV significantly increased
1413
ANDREANI ET AL.
1414
the risk of bleeding without affecting survival. In five of these seven studies (6-10)only patients with large EV were included, whereas i n the remaining t w o (5, 111 size was not a factor of inclusion. Gravity of cirrhosis (Child-Pugh score > 12) w a s only a factor of exclusion in o n e s tu d y (9). A metanalysis of controlled studies concluded that p-blockers probably reduce the risk of first bleeding; n o conclusion was possible concerning t hei r effect o n survival. Similarly, no conclusion could be draw n concerning ESEV (13). We report the first controlled s t udy comparing the efficacy of propranolol with that of ESEV in the prophylaxis of the first GI hemorrhage i n the cirrhotic patient. The concomitant study of a control g roup w as considered necessary given the contradictory results i n previous studies of ESEV and doubts abou t the improvement of survival w i t h p-blockers. T o gain a m ore general view of the results of these t r eat m ent s, it w as decided to include patients regardless of the size of the EV o r the severity of the cirrhosis. PATIENTS AND METHODS Patients. All adult patients seen in two Paris centers were considered eligible for inclusion if they met the following criteria: (a)cirrhosis proven by histological examination or, if this was unavailable, on the basis of clinical and laboratory test results, regardless of origin; (b) presence of EV on endoscopy, regardless of size; and (c) no history of GI bleeding by rupture of EV. Noninclusion criteria were: (a) existence of HCC; tb) contraindication to the use of propranolol (cardiac insufficiency, asthma, disturbance of auriculoventricular conduction); (c) refusal or unfeasibility of one of the three modes of treatment; (d)unfeasibility of regular surveillance (for reasons of distance or apparent indiscipline); (e) serious associated illness reducing life expectancy to < 1 yr; and tD previous treatment with ESEV, propranolol or surgery for portal hypertension. Informed consent was given by all patients and the study protocol was approved by the Ethics Committee (HBpital Saint-Antoine). Study Protocol. After inclusion, the patients were stratified according to severity of cirrhosis using the Child-Pugh score ( 14)as follows: one group with scores 5 to 8 and the other group with scores 9 to 15. According to this stratification, the patients in each center were randomly assigned to receive one of the three treatments. Propranolol doses were titrated to achieve a 25% reduction in resting heart rate. ESEV was carried out as follows: after premedication with intravenous diazepam or midazolam, as required, sclerotherapy was performed using a flexible endoscope (Olympus GIFQ or XQ, Olympus Corp., Lake Success, NY) and an ABS (Advanced Biochemical Supply, Saint-Die, France) or Olympus sclerosing needle. The sclerosant, 1% or 2% polidocanol, was administered by intravariceal and extravariceal injections (15) in the distal 10 cm of the esophagus. The quantity used varied according to the size of the EV, generally between 15 and 40 ml; gastric varices were not sclerosed. Treatment was repeated every week or 2 wk until complete disappearance of the EV was effected. Thereafter the results were verified endoscopically every 3 mo; in case of recurrence, sclerotherapy was repeated according to the same treatment modality. The control patients received one tablet of vitamin K (10 mg) morning and evening as placebo. As in a previous study (41,
HEPXI'OI.OGY
vitamin K was considered suitable given the lack of any known effect on GI bleeding in the cirrhotic patient and because compliance might be increased by the fact that liver indications are mentioned on the packet. For obvious reasons, these treatments were not administered blindly. Other associated treatments were authorized, with the exception of p-blockers. On inclusion in the trial, a questionnaire was filled out containing clinical, endoscopical and laboratory data (differential blood count, creatinine, bilirubin, AST, y-glutamyltransferase, prothrombin time, factor V and albumin). EV were classified in three grades as follows: grade I = nonconfluent EV flattened by insufflation; grade 11 = EV separated by zones of normal esophagus and not flattened by insufflation; and grade 111 = confluent EV not flattened by insufflation. Patients were seen 1 mo after inclusion in the trial and then at 3-mo intervals. Clinical and biological parameters and compliance and tolerance to the treatment were recorded at each examination; patients with alcoholic cirrhosis were asked whether their drinking habits had changed. Abstinence was defined as the total arrest of alcoholic consumption stated by the patients. When necessary, the dosage of propranolol was adjusted to obtain the desired heart rate. In addition, once a year, patients were screened for HCC with ultrasonography and tr-fetoprotein assay and fiberoptic examination was again performed. When patients failed to attend the follow-up examinations, attempts were made to contact them by mail, telephone or through their family doctor; failing this, public records were consulted to determine whether the patient was still living. Treatment and follow-up were planned to last 2 yr. Patients presenting significant GI bleeding (requiring the transfusion of at least 2 U of RBCs) underwent endoscopical examination as soon as this was feasible to determine the mechanism. Patients with EV rupture were withdrawn from the assigned treatment and given standard therapy. The criteria for variceal bleeding was active bleeding from the varices or the presence of clot on a varix and no other detectable cause of hemorrhage. Standard therapy applied in the two centers included esophageal tamponade and if necessary intravenous vasopressin. Emergency sclerosis was not performed. After bleeding from other causes, the assigned treatment was continued as soon as possible. The assigned treatment was also interrupted in the case of severe side effects or if surgery was required because of an aggravation of the cirrhosis (liver transplantation or portacaval shunt for refractory ascites). Patients were nonetheless monitored until the end of the trial after withdrawal from the assigned treatment. Statistical Analysis. Assuming that 60% of patients in the control group would be free of bleeding after 2 yr of follow-up and that treatment would increase this figure to 90%, 38 patientshreatment group were required for a type I error of 5% and a power of 80% if the comparison were made using the two-sided log-rank test (16). Biological and clinical characteristics of the patients at entry to the trial were compared between the three treatment groups using ANOVA for quantitative variables and the x 2 test for qualitative variables. The data were analyzed on an intentionto-treat basis. The percentages of patients free of bleeding and surviving during the 2 yr of follow-up were estimated using the Kaplan-Meier method (17) and compared using the Breslow statistic ( 18). Quantitative prognostic variables were either divided into the three classes as used in the Child-Pugh score (prothrombin time and bilirubin and albumin concentrations) or into two classes as defined by the median. The multivariate Cox model (19) was used to identify factors predictive of GI bleeding and survival and to compare the treatments after
TAHIA1. Characteristics of patients at entry to study according to treatment" Characteristics
Propranolol
Sclerotherapy
Plareho
Significance
adjustment for possible prognostic factors. Variables were Follow-up. One patient with HCC was inadvertently considered eligible for inclusion i n the (:ox models if they were included. Fourteen patients were lost to follow-up after a predictive at the 20% level of significance or more in the period of 4.9 2 1.9 mo (propranolol = six, sclerosis = univariate analysis. Age was always forced into the propor- six, placebo = two). The remaining patients were all tional hazards model, and t h e other predictive variables were treated until one of the end-points defined in the confirmed by exhaustive searching of numerous models. '1'0 symmetrize distributions. t h e logarithms of bilirubin. pro- protocol: EV rupture = 15; discontinuation of treatthrombin, albumin and AS'I were used. Either the Child-Pugh ment because of severe iatrogenic complications = 3 : classification or the variables used to define it were entered refusal to continue treatment = 7; and secondary surgiinto the model. The proportional hazards assumptions of the cal treatment required for aggravation of the cirrhomodels were validated. Statistical analyses were performed sis = 6. With the exception of those lost to follow-up, using BMDP software 1 2 0 )on ;I VAX 8.530 at the INSERM patients were monitored for 2 yr or until death. computing center, Villijuif. In the propranolol group, severe side effects occurred i n five patients (12%) (hypotension in three and severe HESULI'S asthma in two) requiring discontinuation in three Patients. Between November 1985 and February patients. Nine patients ( 2 1 % )did not regularly take 1988, 126 patients were included in the study. During propranolol. Among these patients, one bled during the same period, a prospective study carried out in one follow-up. In the sclerotherapy group, severe side effects of the centers showed that of 165 consecutive patients were observed in three patients (8%) (one esophageal meeting the inclusion criteria only 78 (47%) could stenosis that regressed after a single dilation and two actually be enrolled. The remaining 87 patients were febrile episodes that regressed under antibiotics); these excluded for the following reasons: follow-up deemed not side effects did not require discontinuation of treatment. feasible (33%0), HCC (26%), other cancers (8%), previous Moderate dysphagia was frequent during the first few treatment (propranolol, sclerotherapy or surgery for days after sclerotherapy and was not considered a portal hypertension) (10%), moribund (8%), propranolol complication. Eleven patients (26%1 were not compliant contraindicated (12%) and refusal to participate ( 1%). for sclerotherapy; four refused to undergo sclerotherapy Of the 126 patients included. 43 were treated with despite initially consenting to participate in the trial and propranolol, 42 by ESEV and 41 with placebo. At seven patients refused to complete sclerotherapy after inclusion, no significant difference was found between one to three sessions. In the propranolol group the dosage initially conthe three treatment groups with regard to clinical and sidered effective was 91 ? 10 mg. This required adapbiological data (Table 1) .
1416
HEPATOLOGY
ANDREANI ET AL.
n
e
v)
60.
.--
n c
0
0
12
6
18
i
a?
24
0
Months
i
12
6
18
24
Months
FIG. 1. Cumulative percentages of patients free of bleeding in the propranolol, sclerotherapy and placebo groups (Kaplan-Meier estimates). Percentages were significantly different between propranolol and placebo (p < 0.004) and between propranolol and sclerotherapy groups (p c 0.03).
FIG.2. Cumulative percentages of surviving patients treated with propranolol, sclerotherapy and placebo (Kaplan-Meier estimates). Significant differences in survival were not observed among the three treatment groups.
TABLE 2. Type of gastrointestinal bleeding (variceal or other) according to severity of cirrhosis and treatment Propranolol Classification
Varices
Other
Sclerotherapy Varices
Other
Child-Pugh class A
B C TOTAL
Size of varices Grade I Grade I1 Grade I11 TOTAL ~
~~
~~~~
Placebo Varices
Other
Total
0 8 2 10
2 12 10 24
2 5 3 10
5 13 6 24
~~~
"Erosive gastritis (n = 2); gastric ulcer (n = 1). bErosive gastritis (n = 1); fundic varices (n = 1);gastric ulcer (n
=
tation for six patients after follow-up examinations. In the ESEV group, EV disappeared completely in 24 patients after a mean of three sessions (range = 1 to 7). The quantity of polidocanol injected at each session was 22 2 1.8 ml (range = 5 to 60 ml). Relapse occurred in eight of these patients and further sclerotherapy was performed in seven. Among the patients with alcoholic cirrhosis, no significant difference between the three groups in terms of alcohol withdrawal was observed during the trial. Gastrointestinal Bleeding. Twenty-four patients had GI bleeding during the study period; two were in the propranolol group, nine were in the ESEV group and 13 were in the placebo group. The timing of these events is shown in Figure 1. The 1-yr and 2-yr bleeding rates were, respectively, 30% and 39% in the placebo group, 23%and 31%in the sclerotherapy group and 6% and 6% in the propranolol group. The differences were significant between propranolol and placebo (p < 0.004) and between propranolol and sclerotherapy (p < 0.03). These results were not affected if only those patients having started treatment or those having completed
1).
treatment were considered. Table 2 gives details of GI bleeding in terms of origin, size of the EV and gravity of cirrhosis. The difference between propranolol and placebo remains significant (p < 0.02) if only the 18 patients who had GI bleeding as a result of EV rupture are considered (propranolol = 2, ESEV = 6 and placebo = 10). GI bleeding was the cause of death in seven of the 24 patients (29%).Two of these deaths were due to massive bleeding while the patient was at home where no treatment could be attempted. The prothrombin time and the Child-Pugh class were both factors predictive of GI bleeding after univariate analysis (Table 3). The multivariate Cox model indicated that the size of EV and the Child-Pugh class were prognostic factors for the onset of GI bleeding; the treatment effect remained after adjusting for these factors (Table 4). Suriual. The 1-yr and 2-yr survival rates were, respectively, 66%and 54% in the placebo group, 71%and 52%in the sclerotherapy group and 79%and 66% in the propranolol group. No significant difference in survival was observed between the three treatment groups (Fig.
TABLE 3. Statistical significance (univariate analysis) of factors thought to be prognostic for gastrointestinal bleeding and survival Significance Classes for factof'
Factor
GI bleeding
Sex i male 1 Age lyr) Bilirubin I pmol 1.1 Prothrombin time 1 ' ; 1 Factor V (0) Albumin (gm 1.1 ASI' i multiple of normal GGT ( W L i Creatinine (pinol 1, I Size of EV Ascites Encephalopathy Child-Pugh class Type of cirrhosis Duration of cirrhosis I nio' Clinical center Treatment
survival 0.04
07 0.5
0.3 0.02 0.02 0.1 0.006
0.09 0.00.5 0.6 0.3 0.3 07 02 0. 1
1
0.007 0.5 1 0.4
0.08 0 :3
0.06 0.5
0.02 0.3 0.5 0.6 0.02
0.001
0.2 0.5 0.9 0.7
"Factors were analyzed in the classeh indicated GGT = y - g l u t a m y l t r a n s f e r ~ i ~ ~
TABLE 4. Prognostic factors for GI bleeding and survival identified by the
Cox model after adjustment for age at entry to
the trial Model
Significance
Relative risk (95% confidence interval)
GI bleeding model Size of EV' Treatment
iII 1111 Vh I Placebo vs propranolol ESEV vs propranolol I
Rvh A .A
I) 01
Child-Pugh clash
c Vh
Survival model Child-Pugh class
Sex
x
0 00 1
H V h
0.01
CVAA Male vs female
4.1 (1.4. 12.11 9.3 (1.9. 43.71 7.6 (1.5. 37.01 X 2 (1.2. 18.21 6.1 (1.2. 30.31 8 9 (2.0. 38.81 16.7 (3.7, 74.91 2.5 (1.2. 5.1)
~~
"Variables are shown in t h e order of entry to the model. "p Values indicate the significancc~of'each factor in the model
'rAs1.e 5. Causes of death during the 2 yr of the study according to treatment Cause of death
Treatment complicat ion Variceal bleeding Other GI bleeding Complications of cirrhosiNot associated with cirrho5is Unknown cause roi
\I
Propranolol
0 1
0 S
Sclerotherapy
0 > 0 11
-
1
1
:I
4 18
13
2). However, a slight tendency toward an improvement in survival was seen in the sclerotherapy group and a stronger tendency appeared in the propranolol group. Causes of death are given in Table 5 according to gravity of cirrhosis and the treatment groilp. A number of
Placebo
Total 0 7
1
0 28 4 10 49
f'actors were predictive of survival after the univariate analysis (Table 3). The multivariate Cox model indicated that female gender and Child-Pugh class were factors predictive of survival (Table 4); treatment played no part in the survival of patients.
1418
ANDREANI Err AL.
DISCUSSION
This study confirms the frequency and the gravity of the first GI hemorrhage in the cirrhotic patient: during a 2-yr period GI bleeding occurred in 31% of our untreated patients and was fatal in 30% of these cases. Our results are comparable to those of previous studies in which the occurrence of GI bleeding in patients with EV, regardless of size, was 19%to 57% (5, 11, 12), while the rate of mortality directly related to GI bleeding was 28%to 66%(1-11).It is noteworthy that the incidence of GI bleeding was relatively high in the untreated patients with grade I EV: 10%in our study but less than in two previous reports, 35% and 33% ( 5 , 11). Two factors found to be predictive for the occurrence of the first GI hemorrhage in previous studies (12, 21)-the large size of EV and the Child-Pugh score of cirrhosis- were also found here. It is difficult to explain the finding that men are more susceptible to death than women. The most important result of this study is that it shows that propranolol is superior to ESEV in the prevention of the first GI hemorrhage in the cirrhotic patient. Survival, however, was not significantly different between these two treatment groups. The frequency of side effects was similar, but it should be noted that 12% of the patients could not be included because propranolol was contraindicated. We also demonstrated that, as in the four previous studies (1-41, propranolol significantly reduces the risk of bleeding compared with a control group. This is the first time that this beneficial effect has been observed in a group of unselected cirrhotic patients, regardless of the severity of the cirrhosis or the size of the EV. The survival rate was not, however, significantly improved. In contrast, ESEV did not significantly improve either the risk of bleeding or the survival rate. It should, nonetheless, be noted that the results of ESEV were slightly better than in the control group; however, this negative result might have been due to a lack of power. Indeed, 62 patientslgroup would have been required for these results to reach the threshold of significance. In light of these results and those of previous studies, two questions arise: what is the explanation for the observed inefficacy of ESEV and why does propranolol not improve the survival rate when it reduces the risk of bleeding? If one accepts that EV completely eradicated cannot bleed, hemorrhage in these patients can occur, in theory, by one of two mechanisms-i.e., the persistence of EV (lack of patient compliance or difficulties in persistent eradication) or bleeding resulting from another cause (gastric varices, ulcers or congestive gastropathy). In this study, the main cause of failure was lack of compliance. In the ESEV group, of the patients who bled from EV, 50% had refused to begin or to continue treatment sessions. In previous studies, the distinction between a lack of compliance and a difficulty in eradicating EV was not clearly established, but it seems probable that the two factors are related because patients who require the largest number of treatment sessions stand the highest risk of dropping out. It should be noted that in previous studies, complete eradication
HEPAIWOGY
of EV was impossible in 30% to 51% of patients. In addition, several authors have stressed that the risk of bleeding in this subgroup is particularly high (7,8).The study by Santangello et al. (10) and two other reports published in abstract form (22,231 focused attention on a possible increase in the risk of GI bleeding after ESEV. Such bleeding was attributed to EV ulcerations produced by the sclerosing product. The distinction between this and spontaneous EV rupture is difficult to establish, particularly if one takes into account the fact that asymptomatic ulceration is almost always observed in the first few days after ESEV. This is considered to be part of the process leading to sclerosis (24). Two possible explanations can be proposed to explain the discrepancy between the three studies where ESEV increased the rate of bleeding and the seven studies (including ours) where this rate was lowered compared with the control group. First, it is tempting to incriminate differences in technique: polidocanol was used in the seven “positive” studies, all carried out in Europe, whereas tetradecylsodium was used in two “negative” studies performed in the United States. However, the third “negative” study was European and also involved polidocanol. The second possible explanation for the discrepancy is a difference in the rate of spontaneous bleeding, which was clearly higher in the seven “positive” studies (304 to 66%1 than in the three “negative” ones (8% to 15%). In contrast, the difference in the risk of bleeding in the sclerotherapy-treated group between the “positive” studies (9%’to 29%)and “negative” studies (22%to 35%) was far less marked. It might thus be deduced that when the risk of spontaneous bleeding is high, ESEV would be beneficial despite a certain number of hemorrhagic events caused directly by the treatment. Given the high rate of mortality directly linked to the first GI hemorrhage in the cirrhotic patient, it is surprising that the significant reduction in the risk of bleeding obtained by the use of P-blockers only led to a significant improvement in the survival rate in one (3) of five studies (1, 2, 41, including ours. In all these studies death among the patients with no bleeding was mainly related to other complications of cirrhosis. The failure of P-blockers to improve survival in our study might be related to a lack of power and to too-small sample sizes. The two metanalyses of controlled trials (13, 25) have both concluded that P-blockers probably reduced the risk of first bleeding. However, regarding survival, only one of the two metanalyses (25) has concluded that p-blockers improved survival. In contrast, in the four studies (5,6, 7, 11) in which ESEV significantly reduced the incidence of bleeding, the survival rate was significantly increased. In conclusion, it appears that P-blockers should be prescribed for the prevention of the first GI hemorrhage in the cirrhotic patient, given that their efficacy has been demonstrated in all studies published to date despite uncertainty concerning their effect on survival. This therapeutic efficacy of P-blockers has been so far established in cirrhotic patients in good condition and with large varices. Our study, which has to be confirmed,
Vol 12, N o 6 . 1990
I ' K O I ' I I ~ I , . \ ~ " I ~ THERAPY I~~ OF (;IIS'I'KOIN'I'ES'TINAI,
suggests that patients - whatever the severity of cirrhosis and the size of varicee - might benefit from thc drug. ESEV should not be recommended as first-line treatment but might be envisaged for patients who do not tolerate propranolol or for whom it is contraindicated when the risk of spontaneous bleeding is particularly high.
Acknowledgments: We thank Patricia Grand and Beatrice Pineau for their technical assistance. REFERENCES 1 . Ideo G, Bellati G. Fesce E.Griinoldi L). Nadolol can prevent the first gastrointestinal bleeding i n cirrhotics: a prospective. randomized study. HEPA.I.(II o(;~ 198H:X:6-9. 8 . Lebrec D. Poynard T, Capron J P . Hillon P. Geoffroy P, Roulot D. Chaput J-C. et al. Nadolol for prophylaxis of gastrointestinal bleeding in patients with cirrhosis: a randomized trial. J IIepatol 1988;7:1 18-125. :3. Pascal J P . Cales P, Multicenter Study Group Propranolol in the prevention of first upper gastrointestinal tract hemorrhage in patients with cirrhosis of the liver and esophageal varices N Engl J Med 1987;317:856-861. 4. The Italian Multicenter Project for Propranolol In Prevention of Bleeding. Propranolol prevents first gastrointestinal bleeding in non ascitic cirrhotic patients. J Hepatol 1989;9:75-83. 5. Koch H , Henning H. Griinm H , Soehendra N. Prophylactic sclerosing of esophageal varices: rt-sultsof a prospective controlled study. Endoscopy 1986: 18:40-43 6. Paquet KJ. Prophylactic endoscopic sclerosing treatment of the oesophageal wall in varices a prospective controlled randomized trial. Endoscopy 1982;14:4-5. 7. Piai G , Cipolletta L. Claar M. Marone C;. Uianco MA. Forte (;, Iodice G. Prophylactic sclerotherapy of' high-risk esophageal varices: results of a multicentric prospective controlled trial Hw.vroi.o(;~ 1988;8:1495-1500. 8. Potzi R, Bauer P . Reichel W. tierstan E. Kenner F'. Gang1 .4 Prophylactic endoscopic sclerotherapy of' oesophageal varices in liver cirrhosis: a multicentre prospective controlled randomised trial in Vienna. Gut 1989;:30:87s'3-879. 9. Sauerbruch T , Wotzka H. Kijpcke W. t-Iarlin M. Heldwein W. Bayerdorffer E, Sander R. et al. Prophylactic sclerotherapy before the first episode of variceal hemorrhage in patienh with cirrhosis. N Engl J Med 1988;319:8-15
BLEEDIN(; IN C'IRKHOSIS
1419
(;J. Prophylactic sclerotherapy of large esophageal varices. N Engl J Med 1988: 3 18:8 1 4-8 18. 1 1 . Witzel L. Wolbergs E. Merlu I I . I'rophylactic endoscopic sclerotherapy of oesophageal varices. a prospective controlled study. Lancet 1985;1:773-775. 12 Burroughs AK, d'Heygere F. Mclntyre N . Pitfalls in studies o f ' prophylactic therapy for variceal bleeding in cirrhotics. HEPA N)L,IN;Y 1986;6:1407-1413. 1 : j Pagliaro L. Burroughs AK. Sorensen T1.4,Lehrec D, Morabito A, D'Amico G, Tine F. Therapeutic controversies and randomised controlled trials (RCTs):prevention of bleeding and rebleeding in cirrhosis. Gastroenterol Int 1989;2:71-84. 1-1. Pugh RNH. Murray-Lyon IM. Dawson J L , Pietroni MC, Williams R . Transection of the oesophagus for bleeding oesophageal varices. Br J Surg 1973;60:646-649. 16 Soehendra N, de Heer K. Kempeneers I. Frommelt L. Morphological alterations of the esophagus after endoscopic sclerotherapy of varices. Endoscopy 1983;15:291-296. 16. Freeman LS. Tables of the number of patients required in clinical trials using the logrank test. Stat Med 1982;1:121-129. 17 Kaplan SA, Meier P . Nonparametric estimation from incomplete observations. J Am Stat Assoc 1958;53:457-481. 18. Breslow N. A generalised Kruskal-Wallis test for comparing k samples subject to unequal patterns of censorship. Biometrika 1970;57:579-594. 19 Cox DR. Regression models and life tahles. J R Stat Soc 197234: 187-220. 2 0 . Dixon WJ. BMDP statistical software. Vol. 2. Berkeley. CA. University of California Press, 1988:689-742. 21. The North Italian Endoscopic Club for the Study and Treatment of Esophageal Varices. Prediction of the first variceal hemorrhage in patients with cirrhosis of the liver and esophageal varices: a prospective multicenter study. N Engl J Med 1988;319:983-989. 22. Gregory P, Hartigan P, Amodeo D. Baum R, Camara D, Colcher H. Fye C. et al. Prophylactic sclerotherapy for esophageal varices in alcoholic liver disease: results of a VA cooperative randomized trial [Abstract]. Gastroenterology 1987;92:1414. 2 3 . Planas R, Boix J. Dominguez M. Abad A, Quer J C , de Leon R. Villagrasa M, et al. Prophylactic sclerosis of esophageal varices (EV):prospective trial IAbstract I . J Hepatol 1989;9(suppl 1 ): 144 24. Sarin SK, Nanda R, Vij *JC, Anand BS. Oesophageal ulceration after sclerotherapy: a complication or an accompaniment? Endoscopy 1986;18:44-45. 25. Pignon J P , Poynard T , Naveau S. Chaput J-C. Metaanalyse des essais randomi& des p-bioquants dans la prevention des hemorragies digestives des sujets atteints de cirrhose IAbstract I. Gastroenterol Clin Biol 1988;12:A175. 1 0 Santangelo WC, Dueno MI. Estes BL. KreJs
