A Controlled Trial of Oral Propranolol Compared with Injection Sclerotherapy for the Long-term Management of Variceal Bleeding DAVIDWESTABY, REXJ. POLSON, ALEXANDER E. S. GIMSON,PETER C. HAYES,KARENHAYLLAR AND ROGERWILLIAMS Liver Unit, King’s College Hospital and School of Medicine and Dentistry, Denmark Hill, London SE5 9RS, United Kingdom
This trial was carried out to assess the value of propranolol for the prevention of recurrent variceal bleeding in patients with well-compensatedcirrhosis. We also compared propranolol therapy to long-term injection sclerotherapy. One hundred and eight patients, in whom the original variceal hemorrhage stopped spontaneously (before diagnostic endoscopy) and without sclerotherapy or surgical intervention, were included. All were Pugh grade A or B; 55% had alcoholic cirrhosis. Patients were chosen randomly to receive oral propranolol (in a dosage to reduce resting pulse rate by 25%) or to undergo long-term injection sclerotherapy. In both groups, episodes of repeat bleeding that did not stop spontaneously were managed with selerotherapy. Patients considered to have failed propranolol therapy were treated with long-term sclerotherapy. Follow-up ranged from 12 to 64 mo. In the propranolol group, 28 (54%) of the 52 patients had repeat bleeding from varices with a total of 57 episodes; 14 received long-term sclerotherapy. In the sclerotherapy group, 25 (45%) of the 56 patients had repeat bleeding, with a total of 40 episodes (p < 0.20). On an intention-to-treat basis, the risk of bleeding expressed per patient-month of follow-up was similar for the two groups, at 0.05 and 0.037, respectively. Survival as assessed by cumulative life analysis was also similar, with 55% and 66% alive at 3 yr (p < 0.40). Stepwise regression analysis of possible factors predicting further bleeding in patients taking propranolol selected only two variables-the pretreatment pulse rate and the extent of pulse-rate reduction in response to propranolol. These data support propranolol as an alternative first-line measure to long-term injection for the management of variceal bleeding. The pretreatment pulse rate and subsequent response to propranolol may provide means of selecting those most likely to benefit. (HEPATOLOGY 1990;11:353-359.)
Received March 7, 1989; accepted September 3, 1989. Address reprint requests to: Dr. David Westaby, Liver Unit, King’s College Hospital, Denmark Hill, London SE59RS, United Kingdom. 31/1/18425
Between 1981 and 1985 Lebrec and colleagues reported two interim and one final analysis of a placebobased controlled trial showing that oral propranolol, in a dose sufficient to reduce resting pulse rate by 25%, significantly reduced the frequency of repeat bleeding from gastroesophageal varices and gastric erosions in patients with well-compensated alcoholic cirrhosis (1-3). Subsequently, three placebo- or no-treatmentbased controlled trials of @-receptorblockade to prevent repeat bleeding failed to confirm significant benefit (4-6).More recently, two placebo-based controlled trials of nadolol (7) and propranolol (8) showed significant benefits from the use of @-receptor blockade. There is also evidence to suggest that such therapy prevents the first episode of variceal bleeding (when used as a prophylactic agent) (9-10). In this study we compared oral propranolol to longterm injection sclerotherapy, a treatment regimen we previously showed to significantly reduce repeat bleeding and improve survival rates in patients with cirrhosis and previous variceal hemorrhage (11).In a recent interim report of a similar study, Fleig and colleagues (12) found the two measures to be of equal efficacy, although sclerotherapy was found to be superior in a much smaller trial reported by Alexandrino, Alves and Pinto Correia (13). In this analysis of the current study, which was begun in 1982, the 108 patients were followed for 12 to 64 mo. PATIENTS AND METHODS The trial was carried out in patients who were seen in the Liver Unit between April 1982 and September 1986 with recent variceal bleeding (requiring 2 or more units of blood transfusion or showing a fall in hemoglobin > 2 gm)that stopped spontaneously by the time of diagnostic endoscopy, which was done within 4 hr of admission. In each case, gastroesophageal varices were the only potential source of bleeding. Underlying cirrhosis was histologically confirmed in all cases. At the time of trial design, there was considerable concern about possible adverse effects of propranolol in patients with decompensated liver disease (14).Therefore pa-
353
354
HEPATOLOGY
WESTABY ET AL.
TABLE1. Clinical data for two groups of patients at entry to
T R I A L PROTOCOL
the trial
PATIENTS W I T H CIRRHOSIS A N 0 V A R I C E A L BLEEDING No active bleeding at diagnostic endoscopy (within 4hrs of admission)
0 Pugh Grade A or B 0 No contraindication to 0-adrenoreceptor blockade RANDOMISATION
4 PROPRANOLOL* Oral dose t o reduce resting pulse rate by >25%
1 4 SCLEROTHERAPY Weekly intervals to obliteration
1 MANAGEMENTOFREBLEEDS
5+
Diagnostic endoscopy No active bleeding
+
Continue propranolol
1
1
Sclerotherapy
Active bleeding
t
Mean age f S.E. Female/male Origin of cirrhosis AlcohoIic Primary biliary Cryptogenic Chronic active hepatitis Other Pugh grade A B
Propranolol (52 patients)
Sclerotherapy (56 patients)
54 f 1.5 21:31
52 f 1.1 22:34
32 (9)=
7 6 2 5
26 (8)" 11 7 6 6
22 30
22 34
"Patients who continued to drink heavily (>40 gm daily).
Sclerotherapy
FIG.1. Flow diagram of trial protocol. *Those patients in propranolol group with life-threatening(>10-unit transfusion requirement) or multiple (three of more) episodes of repeat bleeding were treated with long-term sclerotherapy.
tients classed as Pugh grade C (15) were excluded, as were those with specific contraindications to P-adrenoreceptor blockade. Patients were chosen using a system of random numbers to receive oral propranolol or to undergo long-term injection sclerotherapy. Figure 1 summarizes the trial protocol. Propranolol was given at a dosage of 40 mg twice daily and adjusted over 5 to 7 days with the aim of reducing the resting pulse rate by 25%. For those selected to take propranolol, an estimate of portal pressure was made before therapy by the wedged hepatic venous technique (16). This was done only after 12 h r of hemodynamic stability. To prevent delays in the start of propranolol therapy, the measurement, of portal pressure was omitted if for logistical reasons it could not be made within 24 hr of the initial 12-hr period of stability. A second estimate of pressure was made after approximately 4 wk of propranolol therapy. Compliance was assessed using the patient's history, supplemented by evidence from close relatives and supported by the level of resting pulse rate and random propranolol levels. Injection sclerotherapy was begun at the time of diagnostic endoscopy and was repeated at 1-wk intervals for the first three sessions. A period of 3 to 4 wk was then allowed before the need for further sclerotherapy was assessed by endoscopy. An intravariceal technique of injection sclerotherapy was used with treatment restricted to approximately 1 to 3 cm above the squamocolumnarjunction of the lower esophagus. An outer esophageal sheath (17) was used at the operator's discretion for the first treatment sessions of patients with large varices or for the management of active bleeding. Obliteration of varices was confirmed if there were no visible or patent variceal cords within approximately 3 to 5 cm of the squamocolumnar junction. After obliteration of varices, the patients underwent surveillance endoscopy at 3- to 4-mo intervals for the first year, and yearly thereafter. Similar surveillance endoscopy was carried out in patients taking propranolol, at which time variceal size was noted (17). On each hospital visit alcohol
consumption histories were obtained from the patient and close relatives. This has been adequately shown in our population to reflect findings on random urine-alcohol levels (9). Episodes of repeat bleeding (defined as for the index hemorrhage) were investigated by endoscopy. Patients in the sclerotherapy group were treated with further injection therapy at the time of diagnostic endoscopy. Management of repeat bleeding in patients treated with propranolol was determined by the presence or absence of bleeding at the time of endoscopy (within 4 h r of admission); those found to be bleeding were treated with a single session of sclerotherapy. Those who had spontaneously stopped bleeding continued taking propranolol. Patients in the propranolol group who experienced major hemorrhages (>10-unit transfusion requirement) or multiple (three or more) episodes were treated with long-term injection sclerotherapy. With respect to trial analysis, such patients were assessed on an intention-to-treat basis. Differences between the two groups in clinical data at presentation, repeat bleeding and overall number of deaths were compared using the x2 test. The cumulative percentage of patients surviving and free of repeat bleeding was analyzed using the Kaplan-Meier method (18).The number of patients recruited into the trial was calculated to allow a difference of 25% in the frequency of repeat bleeding to be detected with a power of 80% at the 5% level of significance. In patients receiving propranolol, a logistical regression analysis was carried out to assess possible risk factors predicting repeat bleeding (Table 1).The distribution of all noncategorical variables was examined and when appropriate a log transformation was used to normalize the data. Data were analyzed for the whole group and also for those with alcoholic cirrhosis. In the first analysis the variables of portal pressure and percentage of change in portal pressure (in response to propranolol) were included and failed to show a correlation. Because the data for both variables were incomplete for the whole group, they were omitted to enable analysis of the larger data set. Because the data had a large number of covariate patterns, Brown's goodness-of-fit test was used as a measure of variation found in each model (19). The trial was carried out under King's College Ethical Committee guidelines and informed consent was obtained from all patients.
Vol. 11, No. 3, 1990
355
PROPRANOLOL FOR PREVENTION OF RECURRENT VARICEAL BLEEDING
TABLE2. The wedged hepatic venous pressure, free hepatic venous pressure and portal pressure gradient before and approximately 1 mo after taking oral propranolol Baseline (mm Hg) Case
I
This trial was carried out to assess the value of propranolol for the prevention of recurrent variceal bleeding in patients with well-compensatedcirrhosis. We also compared propranolol therapy to long-term injection sclerotherapy. One hundred and eight patients, in whom the original variceal hemorrhage stopped spontaneously (before diagnostic endoscopy) and without sclerotherapy or surgical intervention, were included. All were Pugh grade A or B; 55% had alcoholic cirrhosis. Patients were chosen randomly to receive oral propranolol (in a dosage to reduce resting pulse rate by 25%) or to undergo long-term injection sclerotherapy. In both groups, episodes of repeat bleeding that did not stop spontaneously were managed with selerotherapy. Patients considered to have failed propranolol therapy were treated with long-term sclerotherapy. Follow-up ranged from 12 to 64 mo. In the propranolol group, 28 (54%) of the 52 patients had repeat bleeding from varices with a total of 57 episodes; 14 received long-term sclerotherapy. In the sclerotherapy group, 25 (45%) of the 56 patients had repeat bleeding, with a total of 40 episodes (p < 0.20). On an intention-to-treat basis, the risk of bleeding expressed per patient-month of follow-up was similar for the two groups, at 0.05 and 0.037, respectively. Survival as assessed by cumulative life analysis was also similar, with 55% and 66% alive at 3 yr (p < 0.40). Stepwise regression analysis of possible factors predicting further bleeding in patients taking propranolol selected only two variables-the pretreatment pulse rate and the extent of pulse-rate reduction in response to propranolol. These data support propranolol as an alternative first-line measure to long-term injection for the management of variceal bleeding. The pretreatment pulse rate and subsequent response to propranolol may provide means of selecting those most likely to benefit. (HEPATOLOGY 1990;11:353-359.)
Received March 7, 1989; accepted September 3, 1989. Address reprint requests to: Dr. David Westaby, Liver Unit, King’s College Hospital, Denmark Hill, London SE59RS, United Kingdom. 31/1/18425
Between 1981 and 1985 Lebrec and colleagues reported two interim and one final analysis of a placebobased controlled trial showing that oral propranolol, in a dose sufficient to reduce resting pulse rate by 25%, significantly reduced the frequency of repeat bleeding from gastroesophageal varices and gastric erosions in patients with well-compensated alcoholic cirrhosis (1-3). Subsequently, three placebo- or no-treatmentbased controlled trials of @-receptorblockade to prevent repeat bleeding failed to confirm significant benefit (4-6).More recently, two placebo-based controlled trials of nadolol (7) and propranolol (8) showed significant benefits from the use of @-receptor blockade. There is also evidence to suggest that such therapy prevents the first episode of variceal bleeding (when used as a prophylactic agent) (9-10). In this study we compared oral propranolol to longterm injection sclerotherapy, a treatment regimen we previously showed to significantly reduce repeat bleeding and improve survival rates in patients with cirrhosis and previous variceal hemorrhage (11).In a recent interim report of a similar study, Fleig and colleagues (12) found the two measures to be of equal efficacy, although sclerotherapy was found to be superior in a much smaller trial reported by Alexandrino, Alves and Pinto Correia (13). In this analysis of the current study, which was begun in 1982, the 108 patients were followed for 12 to 64 mo. PATIENTS AND METHODS The trial was carried out in patients who were seen in the Liver Unit between April 1982 and September 1986 with recent variceal bleeding (requiring 2 or more units of blood transfusion or showing a fall in hemoglobin > 2 gm)that stopped spontaneously by the time of diagnostic endoscopy, which was done within 4 hr of admission. In each case, gastroesophageal varices were the only potential source of bleeding. Underlying cirrhosis was histologically confirmed in all cases. At the time of trial design, there was considerable concern about possible adverse effects of propranolol in patients with decompensated liver disease (14).Therefore pa-
353
354
HEPATOLOGY
WESTABY ET AL.
TABLE1. Clinical data for two groups of patients at entry to
T R I A L PROTOCOL
the trial
PATIENTS W I T H CIRRHOSIS A N 0 V A R I C E A L BLEEDING No active bleeding at diagnostic endoscopy (within 4hrs of admission)
0 Pugh Grade A or B 0 No contraindication to 0-adrenoreceptor blockade RANDOMISATION
4 PROPRANOLOL* Oral dose t o reduce resting pulse rate by >25%
1 4 SCLEROTHERAPY Weekly intervals to obliteration
1 MANAGEMENTOFREBLEEDS
5+
Diagnostic endoscopy No active bleeding
+
Continue propranolol
1
1
Sclerotherapy
Active bleeding
t
Mean age f S.E. Female/male Origin of cirrhosis AlcohoIic Primary biliary Cryptogenic Chronic active hepatitis Other Pugh grade A B
Propranolol (52 patients)
Sclerotherapy (56 patients)
54 f 1.5 21:31
52 f 1.1 22:34
32 (9)=
7 6 2 5
26 (8)" 11 7 6 6
22 30
22 34
"Patients who continued to drink heavily (>40 gm daily).
Sclerotherapy
FIG.1. Flow diagram of trial protocol. *Those patients in propranolol group with life-threatening(>10-unit transfusion requirement) or multiple (three of more) episodes of repeat bleeding were treated with long-term sclerotherapy.
tients classed as Pugh grade C (15) were excluded, as were those with specific contraindications to P-adrenoreceptor blockade. Patients were chosen using a system of random numbers to receive oral propranolol or to undergo long-term injection sclerotherapy. Figure 1 summarizes the trial protocol. Propranolol was given at a dosage of 40 mg twice daily and adjusted over 5 to 7 days with the aim of reducing the resting pulse rate by 25%. For those selected to take propranolol, an estimate of portal pressure was made before therapy by the wedged hepatic venous technique (16). This was done only after 12 h r of hemodynamic stability. To prevent delays in the start of propranolol therapy, the measurement, of portal pressure was omitted if for logistical reasons it could not be made within 24 hr of the initial 12-hr period of stability. A second estimate of pressure was made after approximately 4 wk of propranolol therapy. Compliance was assessed using the patient's history, supplemented by evidence from close relatives and supported by the level of resting pulse rate and random propranolol levels. Injection sclerotherapy was begun at the time of diagnostic endoscopy and was repeated at 1-wk intervals for the first three sessions. A period of 3 to 4 wk was then allowed before the need for further sclerotherapy was assessed by endoscopy. An intravariceal technique of injection sclerotherapy was used with treatment restricted to approximately 1 to 3 cm above the squamocolumnarjunction of the lower esophagus. An outer esophageal sheath (17) was used at the operator's discretion for the first treatment sessions of patients with large varices or for the management of active bleeding. Obliteration of varices was confirmed if there were no visible or patent variceal cords within approximately 3 to 5 cm of the squamocolumnar junction. After obliteration of varices, the patients underwent surveillance endoscopy at 3- to 4-mo intervals for the first year, and yearly thereafter. Similar surveillance endoscopy was carried out in patients taking propranolol, at which time variceal size was noted (17). On each hospital visit alcohol
consumption histories were obtained from the patient and close relatives. This has been adequately shown in our population to reflect findings on random urine-alcohol levels (9). Episodes of repeat bleeding (defined as for the index hemorrhage) were investigated by endoscopy. Patients in the sclerotherapy group were treated with further injection therapy at the time of diagnostic endoscopy. Management of repeat bleeding in patients treated with propranolol was determined by the presence or absence of bleeding at the time of endoscopy (within 4 h r of admission); those found to be bleeding were treated with a single session of sclerotherapy. Those who had spontaneously stopped bleeding continued taking propranolol. Patients in the propranolol group who experienced major hemorrhages (>10-unit transfusion requirement) or multiple (three or more) episodes were treated with long-term injection sclerotherapy. With respect to trial analysis, such patients were assessed on an intention-to-treat basis. Differences between the two groups in clinical data at presentation, repeat bleeding and overall number of deaths were compared using the x2 test. The cumulative percentage of patients surviving and free of repeat bleeding was analyzed using the Kaplan-Meier method (18).The number of patients recruited into the trial was calculated to allow a difference of 25% in the frequency of repeat bleeding to be detected with a power of 80% at the 5% level of significance. In patients receiving propranolol, a logistical regression analysis was carried out to assess possible risk factors predicting repeat bleeding (Table 1).The distribution of all noncategorical variables was examined and when appropriate a log transformation was used to normalize the data. Data were analyzed for the whole group and also for those with alcoholic cirrhosis. In the first analysis the variables of portal pressure and percentage of change in portal pressure (in response to propranolol) were included and failed to show a correlation. Because the data for both variables were incomplete for the whole group, they were omitted to enable analysis of the larger data set. Because the data had a large number of covariate patterns, Brown's goodness-of-fit test was used as a measure of variation found in each model (19). The trial was carried out under King's College Ethical Committee guidelines and informed consent was obtained from all patients.
Vol. 11, No. 3, 1990
355
PROPRANOLOL FOR PREVENTION OF RECURRENT VARICEAL BLEEDING
TABLE2. The wedged hepatic venous pressure, free hepatic venous pressure and portal pressure gradient before and approximately 1 mo after taking oral propranolol Baseline (mm Hg) Case
I
