Notice: Wiley Online Library will be unavailable on Saturday 27th February from 09:00-14:00 GMT / 04:00-09:00 EST / 17:00-22:00 SGT for essential maintenance. Apologies for the inconvenience.
If you can't find a tool you're looking for, please click the link at the top of the page to "Go to old article view". Alternatively, view our Knowledge Base articles for additional help. Your feedback is important to us, so please let us know if you have comments or ideas for improvement.
Nonselective β-adrenergic antagonists are widely used in the prevention of first hemorrhage from esophageal varices in patients with cirrhosis and portal hypertension.1 These drugs reduce the risk of bleeding by about 45% and the mortality rate by 20% at 2 yr.2 Unfortunately, many patients have contraindications or severe side effects making it necessary to suspend therapy.3, 4
This is why other treatments have been used to prevent the first episode of variceal bleeding in high risk patients, in particular 2 endoscopic techniques: sclerotherapy and endoscopic variceal ligation (EVL). Sclerotherapy has been abandoned because several trials have reported increased mortality, as well as morbidity due to complications.5, 6 EVL, which results in fewer complications than sclerotherapy in the prevention of rebleeding,7–10 is the currently used endoscopic therapy for primary prophylaxis of esophageal variceal bleeding. Several trials comparing EVL with beta-blockers (BB) in primary prophylaxis have been performed,11–16 but whether BB or endoscopic banding is the best therapy is still unclear and still subject to debate.17, 18
The aim of the present study was to compare, by means of a randomized controlled trial, the safety and the results of EVL vs. BB, in the prevention of the first variceal bleeding in patients who are candidates for liver transplantation (LT).
A prospective randomized trial was conducted in the Department of Surgical and Gastroenterological Sciences of Padova University, where LT is performed. Study subjects were selected from those patients referred for LT evaluation. The patients enrolled met the inclusion and exclusion criteria summarized in Table 1. The study was performed according to the 1975 Declaration of Helsinki and the study protocol was approved by the ethics committee of the University of Padova, Italy.
Table 1. Inclusion and Exclusion Criteria for Patient Enrollment
1. Diagnosis of liver cirrhosis on the basis of clinical, biochemical, or histological analysis
2. Child-Turcotte-Pugh ≥ B7
3. Studied for liver transplantation
4. Age between 18 and 65 yr
5. No previous bleeding from esophageal varices
6. Signed informed consent
7. Esophageal varices F3 or F2 blue with red signs, according to Beppu et al.19
1. Esophageal varices less than F3 or F2 blue with red signs
2. Presence of gastric varices
3. Previous endoscopic, radiological, or surgical treatment of esophageal varices
5. Portal vein thrombosis
6. Severe heart, respiratory, or renal failure
7. Contraindications to beta-blockers (severe chronic obstructive pulmonary disease, severe asthma, severe insulin-dependent diabetes mellitus, bradyarrhythmia)
8. Treatment with nitrates, calcium antagonists, or other antiarrhythmic drugs, including beta-blockers, that can not be suspended
11. An uncooperative attitude or the suspicion that the candidate could or would not return for routine follow-up examinations
Randomization, Treatments, and Follow-Up Schedule
After the patients were enrolled, treatment assignment was made by opening a sealed opaque envelope that designated 1 of the 2 treatments: EVL or BB. Randomization of numbers was assigned by a statistical software package. After starting treatment all patients had an esophagogastroduodenoscopy and a clinical examination every 6 months and were provided with the emergency numbers of investigators so that they could report all new events about treatment complications or bleeding. During the screening and follow-up endoscopies, varices were graded according to Beppu et al.'s19 classification. Their extension in length was also recorded. Moreover, gastric varices and hypertensive portal gastropathy were detected and their severity described.
At the first session a diagnostic esophagogastroduodenoscopy without the ligation device was performed before banding to confirm the presence of high-risk esophageal varices and the absence of gastric varices. Banding placement was then performed using a multiband ligator with 6 or 7 bands (Sixshooter; Wilson-Cook, Winston-Salem, NC; SpeedBand SuperView Super7; Boston Scientific, Natick, MA). The endoscopist entered the esophagus only once and banded the varices starting from the cardias and progressing proximally every 1 cm, in an upward spiral fashion, attempting to avoid placing the bands too close together. Patients were treated during a 1-day hospital stay, followed by a liquid diet for 24-hours and semiliquid diet for 1 week. Subsequent sessions were performed every 2 weeks until the varices were completely eradicated, avoiding placing the bands near to the scars left by the previous treatment. During the period of eradication patients were also prescribed proton-pump inhibitors. Recurrent varices detected during the follow-up esophagogastroduodenoscopy were banded again using the same method as above.
Patients were treated with propranolol, starting with a low dosage of 10 mg twice a day and increasing by 20 mg/day until a 25% reduction of the baseline heart rate was obtained. The amount of 160 mg twice daily was considered the maximal dosage. Patients recorded their systolic and diastolic blood pressure and their pulse rate twice a day. After the first week's time these recordings were sent by FAX to the study investigator, who decided if the propranolol dosage needed to be modified. The process was repeated weekly until the maximum tolerated dose was achieved.
Treatment was interrupted when the systolic blood pressure fell to below 90 mmHg, or the heart rate was under 50 beats/minute, or when patients developed severe disabling side effects. At every 6-month checkup, pulse rate and blood pressure were taken, and all home recordings that the patients brought were reviewed by the study investigators.
Treatment Endpoints and Treatment Failure
These were defined as the following:
1Death from esophageal variceal bleeding.
2Esophageal variceal bleeding confirmed by an esophagogastroduodenoscopy.
3Major complications due to endoscopic treatment.
4Severe adverse effects caused by propranolol making it necessary to discontinue treatment.
All of the patients who developed upper gastrointestinal bleeding were hospitalized and underwent endoscopy for diagnosis. Patients with variceal hemorrhage were treated by sclerosant injection or, when possible, by banding. Treatment was suspended in patients who did not tolerate BB, and EVL was then used. All bleeding events were recorded for the analysis of results on an intention-to-treat basis. The suspension of BB treatment was decided by a collegial discussion of 3 investigators and the consultation of a specialist was obtained (for example, a cardiologist) when deemed appropriate. Patients who received LT were censored at this point for analysis.
All the treatment costs were calculated using Italian Health Ministry cost assignments current at the time, for both patient groups. They include treatment medications, endoscopic treatment, follow-up endoscopies, and visits, hospitalization due to treatment-related complications, bleed-related hospitalization, and readmission for rebleeding. In particular, for the costs of propranolol therapy, we considered the dosage and time of assumption of this drug, while costs of ligation sessions, endoscopies, visits, and admission for bleeding were calculated according to the Diagnosis-Related Groups.
Patients' data were collected prospectively by a study investigator, entered into data files and checked for completeness and accuracy by a second data manager.
Sample-Size Determination and Statistical Analysis
This study was carried out to verify the superiority of EVL with respect to BB in prevention of first variceal bleeding, which was the primary endpoint.
When this study was set up, Sarin et al.12 had published the first paper comparing BB with EVL in primary prophylaxis of variceal bleeding. On the basis of their results reporting a 28% reduction of bleeding incidence in EVL-treated (15%) vs. BB-treated (43%) patients at 18 months, we hypothesized a smaller 23% difference of bleeding incidence between the 2 treatment groups. With this treatment difference, a sample size of 51 patients per group was calculated to provide 80% power with a 2-sided α of 0.05 by log-rank test. A dropout rate of 15% was assumed, so that 60 patients needed to be included in each treatment group to provide 120 study patients. An interim analysis was planned when 50% of the patients had been enrolled. For continuous variables, comparison of the baseline characteristics of the 2 groups of patients was made using a t-test or the Mann Whitney test if a Gaussian model was not appropriate. When the variables were categorical, we used the chi-squared or the Fisher's exact test. Discrete time-to-event outcomes (time to hemorrhage and death) were compared in the 2 groups by the Kaplan Meier method and significance testing by log-rank test. Statistical difference was set at P < 0.05. Statistical calculations were made by using SAS software for Windows (SAS Institute, Cary, NC).
Between September 2001 and December 2005, 99 patients who met the inclusion criteria were considered for this trial. Of these, 37 were excluded because they presented 1 or more exclusion criteria, while 62 patients were enrolled and randomly assigned to 1 of the 2 treatment groups: 31 EVL and 31 BB. At randomization, the 2 groups were well-matched with respect to baseline characteristics (Table 2). All of the banding group patients, except for 2 who had severe complications, completed the treatment and came to the scheduled follow-up sessions. According to our follow-up, of the patients under BB, at least 95% maintained their maximally-tolerated dose. Their mean compliance in reporting their pulse rate after the first week of treatment was 94%. The compliance rate of attending the 6-monthly esophagogastroduodenoscopy controls for both treatments was 92%.
Table 2. Baseline Characteristics: Comparison of the Two Groups
All patients (n = 62)
EVL (n = 31)
BB (n = 31)
NOTE: Means and standard deviations are given.
Abbreviations: HCV, hepatitis C virus.
Mean age (yr)
52.6 ± 7.1
52.5 ± 6.1
51.7 ± 8.0
Mean length varices (cm)
13.0 ± 3.5
12.8 ± 3.42
13.2 ± 3.63
Mean Child-Turcotte-Pugh score
8.4 ± 1.7
8.51 ± 1.95
8.29 ± 1.44
Mean follow-up (months)
14.6 ± 10.3
16.8 ± 11
12.29 ± 9
Out of 31 patients treated with EVL, 2 (6.5%, 95% confidence interval, 0.0-15.2%) presented with a dramatic hemorrhage due to postbanding ulcer that required emergency treatment with sclerosant injection. In both patients hemorrhage developed a few days after the first banding session (Fig. 1). One of the 2 patients survived and underwent LT 3 months later, the other died despite treatment. Three sessions were sufficient in the other 29 patients for a complete eradication of varices. During banding and after variceal eradication, none bled from varices, while 3 patients presented with melena due to portal hypertension gastropathy, in 2 cases requiring hospital admission. Variceal recurrence was reported by 2 patients at follow-up and these patients underwent another banding session. Two patients died because of liver failure after 1 and 7 months, respectively. A total of 14 patients underwent LT after a mean follow-up of 14.64 months (range 2-29 months). Mean costs per patient and mean standard error (including treatment, follow-up, and bleeding-related costs) were 3,383 ± 225 Euros (4,289 ± 285 U.S. dollars).
Patients were treated with a mean dose of 30 mg per day of propranolol (range 20-80 mg). During treatment, the mean heart rate decreased from 79.7 ± 13.0 to 62.3 ± 6.9 beats/minute (mean reduction percentage 22%). Treatment was suspended in 2 patients for persistent, symptomatic bradycardia (<55 bpm) despite the low dosage (10 mg twice a day), in 2 for symptomatic persistent hypotension (systolic pressure <90 bpm), and in 1 for vertigo. Of these 5 patients, 1 underwent a transplant 1 month later without complications, 3 were banded prophylactically: 1 subsequently had a LT without prior bleeding, another had no bleeding throughout the 16 month follow-up, and the third developed a dramatic bleeding episode 6 months after endoscopic treatment from a recurrent cardiac varix and subsequently died. The fifth patient who did not tolerate BB experienced a bleeding episode 1 month later, before prophylactic treatment banding was initiated. He received in emergency sclerosant injection and subsequently underwent banding. He died 2 months later of liver failure. Of the 26 patients who were able to continue propranolol treatment, 1 had variceal bleeding after 11 months follow-up and was treated with EVL and another was hospitalized for upper gastrointestinal bleeding from portal hypertensive gastropathy. In both, the reduction of the pulse rate was suboptimal (<25%). One other patient died due to liver failure 1 month later. Globally, on an intention-to-treat basis, 3 patients from the propranolol group bled from esophageal varices (9.7%, 95% confidence interval, 0-20.1%) and 2 patients died of esophageal bleeding (6.5%, 95% confidence interval, 0-15.2%). A total of 10 patients underwent LT after a mean follow-up of 7.6 months (range 1-19 months). Mean costs for BB patients and mean standard error (including treatment with propranolol, follow-up, and bleeding-related costs) were 1,124 ± 363 Euros (1,425 ± 460 U.S. dollars).
Study Outcomes According to Treatment Allocation
The planned interim analysis was performed on an intention-to-treat basis. As summarized in Table 3, the difference between the rates of first esophageal variceal hemorrhage, the overall mortality, and the bleeding-related mortality were not significantly different. Also the bleedings from portal hypertensive gastropathy were not significantly different.
Table 3. Study Outcomes: Comparison of the Two Groups
EVL (n = 31 patients)
BB (n = 31 patients)
NOTE: Percentage and 95% confidence interval are given in parentheses.
2 (6.5%, 0-15.2%)
3 (9.7%, 0-20.1%)
Severe adverse event
2 (6.5%, 0-15.2%)
5 (16.1%, 3.2-29%)
2 (6.5%, 0-15.2%)
6 (19.4%, 5.5-33.3%)
14 (45.2%, 27.7-62.7%)
10 (32.3%, 15.8-48.8%)
3 (9.7%, 0-20.1%)
3 (9.7%, 0-20.1%)
1 (3.2%, 0-9.4%)
2 (6.5%, 0-15.2%)
Bleeding from gastropathy
3 (9.7%, 0-20.1%)
1 (3.2%, 0-9.4%)
The cumulative incidence of bleeding at 1 and 2 yr were 6.5% and 6.5% for the banding and 9.6% and 27.7% for the propranolol group, respectively (Fig. 2). The cumulative global mortality at 1 and 2 yr was 10% and 10% for banding and 6% and 25% for the propranolol group, respectively (Fig. 3). The bleeding related mortality at 1 and 2 yr was 3.2% and 3.2% for banding and 3.5% and 22.9% for propranolol, respectively. The rates of LT were also not significantly different (14/31 in the banding and 10/31 in the propranolol group).
Both groups had adverse events requiring interruption of treatment (2 in the banding and 5 in the propranolol group, P = not significant [n.s.]), but only variceal ligation was associated with fatal complications. According to an intention-to-treat analysis, treatment failure (variceal bleeding or treatment interruption) was not significantly different between the 2 treatment groups (2/31 in the banding and 6/31 in the propranolol group, P = n.s.).
Recalculation of the sample size showed that more than 1,000 patients per group would need to be enrolled to achieve a statistical significance for the small difference observed with 62 patients. For this reason and because of 2 iatrogenic bleedings that occurred after banding, 1 of which was fatal, we decided, after a collegial meeting, to bring the trial to an end.
Costs were significantly higher for EVL than for BB treatment (P < 0.001) (Table 4).
Table 4. Comparison of the Costs for the Two Treatments
EVL (n = 31 patients)
BB (n = 31 patients)
NOTE: Values are expressed in U.S. dollars as mean ± standard error.
Medicaments for BB and ligation sessions for EVL.
Including esophagogastroduodenoscopy and visits.
Including complication or bleeding-related hospitalizations.
From the results of this study, both propranolol and endoscopic banding reduced the expected incidence of bleeding in the presence of high-risk varices of more than 30% after 1 yr20 to a lower percentage (6.5-9.6% at 1 yr).
The 2 treatments were not significantly different at the interim analysis, and when the sample size was recalculated it would have been impractical to continue randomization.
Other authors likewise interrupted their studies for similar reasons: Thuluvath et al.,13 reporting only 3 bleeding episodes (1 BB and 2 EVL) in 31 patients (15 BB and 16 EVL), calculated that 424 patients would have been required in each group to show a statistically significant difference. Schepke et al.,14 who treated 152 cirrhotic patients (75 EVL and 77 BB), observed no significant difference between the 2 groups in terms of bleeding or mortality. This interim analysis of the planned randomization of 400 patients showed the impossibility of demonstrating a significant difference with completion of the study. Importantly, Schepke et al.14 also reported 2 fatal bleedings due to ligation but no life-threatening complications in the propranolol group. Two other studies11, 15 found no significant difference in terms of first esophageal bleeding between EVL and BB.
The meta-analysis by Khuroo et al.21 included 596 patients and showed that prophylactic EVL significantly reduces bleeding episodes and adverse severe events in comparison to BB, but without any effect on mortality. However, although adverse severe events were found more frequently in the BB group, fatal complications were described only after EVL. The authors' conclusion was that endoscopic banding should be used only in those patients who cannot be treated with BB. Jutabha et al.16 subsequently published a comparison of the 2 treatments on 62 patients who were transplant candidates with high-risk varices. This study was also halted following the interim analysis because a significant increase in both variceal bleeding and cumulative mortality was found in the propranolol group, although there was no difference in bleeding-related deaths. This result could have been influenced by the 2 non-bleeding-related deaths in the propranolol group vs. no deaths in the banding group and thus this trial appears to be an outlier.18 According to Jutabha et al.16 the 2 treatments had the same costs. In contrast, in a similar group of patients, we found significantly higher costs for banding compared to propranolol.
No bleeding event due to banding was reported in Jutabha et al.'s16 trial, which may have been fortuitous as bleeding from dropped bands and esophageal ulcers are well-described complications after variceal banding.22 During prophylactic treatment, at least 11 episodes of bleeding complications due to endoscopic banding have been described by some authors, 3 of which were fatal.14, 22 We also observed 2 cases of esophageal bleeding caused by endoscopic treatment that required emergency treatment, 1 of which was lethal. In the propranolol group, we suspended therapy for intolerance or side effects in 16% of the patients, none of whom died. To reduce the frequency of side effects, we started with the lowest dosage of 20 mg/day, increasing it gradually to obtain a reduction of 25% in the heart rate. Although the mean dosage of propranolol utilized in our study was low, about half of that utilized by other authors,11, 12, 14, 16 we nonetheless obtained a mean reduction of 22% in the pulse rate. Moreover, the low incidence of bleeding would seem to confirm that the drug was effective, particularly considering that we enrolled only patients with Child-Turcotte-Pugh B or C, while in other studies patients with Child-Turcotte-Pugh A cirrhosis were also enrolled.11, 13–16 We believe the lower dose of propranolol tolerated by our cohort of patients reflects their severity of liver disease.
All our patients were liver transplant candidates, similar to the cohort randomized by Jutabha et al.,16 but not by other trials. This meant that at the time 24 patients underwent a LT (after a mean of 1 yr) they were censored for follow-up. Although this made the overall mean follow-up time shorter (15 months), this was similar to that of Jutabha et al.16 and longer than that of Sarin et al.12 We believe that the low incidence of bleeding we observed with both therapies, despite having a more severely ill population in comparison to other studies, was related to the effectiveness of treatment and not to the shorter follow-up.
Bleeding after suspension of BB is considered by some authors as an indirect complication of the pharmacologic treatment.14, 16 Thus, for ethical reasons, endoscopic ligation was offered to all the patients intolerant to BB, as banding compared to no therapy has been shown to be effective.23 However, a recent randomized study in patients intolerant to BB reported iatrogenic bleeding complications with banding.24 The authors concluded that ligation may be harmful as a primary form of prophylaxis, just as sclerotherapy is.18, 24
In our study, the finding of a low incidence of variceal bleeding following ligation (6.45% at 15 months) in patients in whom the expected incidence if untreated was superior to 30%, supports the use of banding when BB are contraindicated in high-risk patients. However, physicians and patients should be aware that there is risk of iatrogenic death during endoscopic treatment. This must be considered in the risk-benefit balance, considering on one hand the variceal risk of bleeding and on the other hand the fact that in patients intolerant to propranolol fatal bleeding may still occur despite banding, as happened in 1 of 3 patients in our study. Thus, BB should remain the first choice of prophylactic therapy in candidates for LT.
In conclusion, both propranolol and endoscopic banding are similarly effective in reducing the incidence of variceal bleeding in cirrhotic patients with high-risk varices, candidates for LT, but ligation can be complicated by severe and fatal bleeding and is significantly more expensive. Our results suggest that banding should not be utilized as the primary prophylaxis in candidates for LT who can be treated with BB.
We thank Dr. Alberto Brolese (Department of Surgical and Gastroenterological Sciences, 1st Surgical Clinic, University of Padova, Italy) for his help with patient recruitment and Dr. Piergirolamo Polese (Medical Hospital Director, Villa Salus Classified Hospital, Mestre-Venezia, Italy) for his assistance in calculating the treatments' costs.