Although the outcome of variceal hemorrhage has improved over the past two decades, variceal hemorrhage is still the most serious complication of portal hypertension and chronic liver disease.1 Although bleeding occurs less often from gastric varices (GV) than from esophageal varices (EV) (esophageal variceal hemorrhage [EVH] accounts for 70% to 80% of variceal hemorrhage), it has a poorer prognosis and is associated with more severe blood loss, a higher rebleeding rate, and a higher mortality rate.1–3 However, limited data exist on the best treatment for GV hemorrhage (GVH). Endoscopic treatment is an alternative in the management of GVH and includes endoscopic injection of sclerosants or thrombin,4, 5 endoscopic band ligation,6, 7 and others. The success rate in controlling GVH by endoscopic injection of N-butyl-2-cyanoacrylate (GVO) appears higher than for other sclerosants according to previous non-randomized trials.8, 9 Though endoscopic variceal ligation is regarded as the optimal endoscopic treatment for EVH,10 its safety and efficacy for the treatment of GVH is uncertain.7, 11 The rebleeding rate of GVO is variable and ranges from 10% to 42%.12 At the time we started this trial, there were no randomized control trials, and even now, there is only one relatively small randomized clinical trial,13 which was conducted to compare these two potential treatment modalities (endoscopic band ligation [GVL] vs. GVO) on the GVH; it showed results in favor of GVO for the treatment of GVH. We assumed that the rebleeding rate was higher in GVO than that in GVL based on the major studies before this trial.12 Therefore, we designed a prospective randomized trial to test the hypothesis by comparing the efficacy of cyanoacrylate injection (GVO) and band ligation (GVL) in the treatment of acute GVH in liver patients with cirrhosis with or without concomitant hepatocellular carcinoma (HCC).
Progression of gastric variceal hemorrhage (GVH) is poorer than esophageal variceal bleeding. However, data on its optimal treatment are limited. We designed a prospective study to compare the efficacy of endoscopic band ligation (GVL) and endoscopic N-butyl-2-cyanoacrylate injection (GVO). Liver patients with cirrhosis with or without concomitant hepatocellular carcinoma (HCC) and patients presenting with acute GVH were randomized into two treatment groups. Forty-eight patients received GVL, and another 49 patients received GVO. Both treatments were equally successful in controlling active bleeding (14/15 vs. 14/15, P = 1.000). More of the patients who underwent GVL had GV rebleeding (GVL vs. GVO, 21/48 vs. 11/49; P = .044). The 2-year and 3-year cumulative rate of GV rebleeding were 63.1% and 72.3% for GVL, and 26.8% for both periods with GVO; P = .0143, log-rank test. The rebleeding risk of GVL was sustained throughout the entire follow-up period. Multivariate Cox regression indicated that concomitance with HCC (relative hazard: 2.453, 95% CI: 1.036-5.806, P = .041) and the treatment method (GVL vs. GVO, relative hazard: 2.660, 95% CI: 1.167-6.061, P = .020) were independent factors predictive of GV rebleeding. There was no difference in survival between the two groups. Severe complications attributable to these two treatments were rare. In conclusion, the efficacy of GVL to control active GVH appears not different to GVO, but GVO is associated with a lower GV rebleeding rate. (HEPATOLOGY 2006;43:690–697.)
Patients and Methods
From July 1996 to June 2002, all patients with cirrhosis with or without concomitant HCC who presented to our hospital with acute gastrointestinal bleeding, or who were already hospitalized and who developed acute gastrointestinal bleeding, received emergency endoscopy unless prevented by severe encephalopathy, severe hemodynamic instability, or the patient's refusal. Only patients who were aged between 18 and 80 years and had endoscopy-proven acute GVH were included. GVH was diagnosed using the following: (1) clinical signs of hematemesis, coffee ground vomitus, hematochezia, or melena; (2) endoscopic signs of an active spurting or oozing from the GV; (3) adherent blood clots, white nipple signs, or erosions on the GV; or (4) in the presence of distinct large GV with a red-color sign and absence of EV and other bleeding sources.8, 12, 14, 15 Cases with concomitant large GV and large EV, but without stigmata of recent bleeding, were not enrolled in this study, because the exact bleeder could not be differentiated. Informed consent from the patients or their families was obtained. Patients were excluded from the study if they fulfilled the following criteria: (1) had previous endoscopic, surgical treatment or transjugular intrahepatic portal systemic shunt for GVH; (2) had a terminal illness of any major organ system, such as heart failure, uremia, chronic obstructive pulmonary disease, or nonhepatic malignancy. The diagnosis of liver cirrhosis was based on needle liver biopsy findings or, if unavailable, the combination of clinical, biochemical, and radiological findings of hepatic failure and portal hypertension, as well as identification of a known cause of cirrhosis. The diagnosis of HCC was based on cytohistological criteria or liver biopsy or, if unavailable, two coincident imaging studies as well as one imaging study associated with α-fetoprotein of more than 400 ng/mL.16 This study was approved by the Clinical Research Committee of the Veterans General Hospital in Taipei.
Vasoactive drugs (terlipressin or somatostatin) were used before diagnostic endoscopy. Patients who fulfilled the inclusion criteria were immediately randomized into the two treatment groups using consecutively numbered envelopes that contained the treatment assignments, which were generated by a system using computer-allocated random digit numbers. Most of the procedures were performed with the endoscope retroflexed to inject or ligate the varices. N-butyl-2-cyanoacrylate (Histoacryl blue, Braun, Melsungen, Germany) intravariceal injection was performed using an Olympus XQ-20 fiber endoscope or XQ-230 video endoscope (Olympus Optical Co. Ltd., Tokyo, Japan) and a 23-gauge disposable injection needle (EIS 01943, Top Co., Tokyo, Japan). Each shot contained 0.5 mL cyanoacrylate and 0.5 mL Lipiodol (Guerbet Laboratory, Aulnay-Sous-Bris, France). No more than six shots were performed in each session.
Ligation was performed using an Olympus XQ-20 fiber endoscope or XQ-230 video endoscope and Sumitomo pneumoactive ligator (Sumitomo Bakelite. Co., Ltd., Tokyo, Japan). No more than 10 rubber bands were applied in each session. The device is capable of releasing the rubber bands easily, even with the endoscope retroflexed. The bleeding site was ligated or injected first, and then the surrounding prominent varices were also ligated or injected as possible. Concomitant EV was ligated subsequently after eradication of GV in both groups. Immediately after the endoscopic treatment, a proton pump inhibitor (omeprazole 40 mg every 12 hours) was infused for 2 days, followed by an oral proton pump inhibitor (omeprazole 20 mg twice daily) for 12 days.
Clinical Assessment and Follow-Up.
Information regarding presentation of variceal hemorrhage was carefully gathered from the patients and their families. Vital signs, the amount of blood transfusion, and infection status before and after endoscopic treatment were recorded. After successful control of active bleeding, GVL and GVO were performed at regular intervals of 3 to 4 weeks until the varices were eradicated. Follow-up endoscopy was subsequently performed every 3 months and, if unremarkable twice, was moved to every 6 months. If rebleeding occurred, vasoactive agents, including terlipressin or somatostatin, or balloon tamponade, were allowed, and emergency endoscopy was performed to identify the bleeding site. GV rebleeding was treated again using the same endoscopic method. If treatment failure was encountered, conservative treatment, switching to endoscopic treatment, transjugular intrahepatic portal systemic shunt, or surgery was offered based on preference of the patients or their families. The outcomes assessed in this study were control of active bleeding, rebleeding, and mortality. Well-trained nurses and physicians who were blinded to group assignment conducted the assessments. Patients were followed-up until death or 6 months after the last patient was included and the desired sample size was reached.
GV were graded using the system suggested by Sarin et al3: type 1 (GOV1), varices continuous with EV and extending along the lesser curve for approximately 2 to 5 cm below the gastroesophageal junction; type 2 (GOV2), varices extending from the esophagus below the gastroesophageal junction toward the fundus; type 3 (IGV1, type 1 isolated GV), varices located in the fundus; type 4 (IGV2), ectopic varices in the antrum, corpus, and around the pylorus. The size of GV was classified according to the system suggested by Hashizume et al17: Form 1 (F1), tortuous winding varices; F2, nodular-shaped varices; F3, tumorous huge varices. The severity of cirrhosis was classified according to Pugh's modification of Child's classification.18 Acute GVH was defined as GVH less than 24 hours between its clinical presentation and endoscopic treatment. Active bleeding was defined when the endoscopic findings showed active spurting or oozing of blood from GV. Rebleeding was defined as new onset of hematemesis, coffee-ground vomitus, hematochezia, or melena, with an increasing pulse rate over 100 beats/min and decreasing blood pressure below 90 mmHg after a 24-hour period of stable vital signs and hemoglobin after endoscopic treatment. Treatment failure was defined as failure to control acute bleeding after two attempts with the same endoscopic methods, more than one GV rebleeding episode, bleeding death or change of modality (i.e., if the endoscopist judged that hemorrhage could not be controlled, the modality might be changed to another endoscopic method or operation). Rebleeding index for each patient was calculated by dividing the months of follow-up by the number of rebleeding episodes plus 1. Eradication was defined as non-visualization of patent GV. If mucosal prominence could not be differentiated from scarring tissue or patent GV, endoscopic ultrasound was performed to assess the obliteration.
Because the study was performed on an emergency basis, enrollment error was inevitable. Therefore, the results were based on modified intention-to-treat (ITT) analysis.19, 20 Definition of modified ITT population comprised all randomized patients who met the pre-specified criteria and who received at least one-time treatment. Patients who switched from GVL to GVO (or vice versa) were still counted in their originally assigned treatment group. The determination of non-hepatic malignancy and major systemic disease required further clinical studies after randomization. Therefore, all patients with non-hepatic malignancy or terminal major systemic disease included after enrollment error were excluded from the modified ITT population, even if they were randomized and underwent endoscopic treatment. Each continuous variable between the two treatment groups was expressed as mean ± SD and analyzed with two-sample Student t tests or as the median (minimum, maximum) and analyzed with the Mann-Whitney nonparametric test. Categorical data were examined using the chi-square test with Yates' correction. Ninety-five percent confidence intervals (CI) for differences in proportions were computed assuming a normal distribution. Kaplan-Meier analysis was used to examine the time of first recurrent bleeding and the time to death, and the log-rank test was used to compare differences between the groups. Univariate analysis and stepwise multivariate analyses were performed to assess the potential risk factors of recurrent bleeding and survival using Cox proportional hazards regression with SPSS 11.0 for Windows (SPSS, Chicago, IL). The rebleeding rate of GV was 0% to 18.5% for GVL and 10% to 42% for GVO.12 Because of highly prevalent HCC in this area,21–25 a larger number of patients with HCC, a risk factor for variceal rebleeding,26–29 were included in this study. Therefore, estimates of sample size were based on a rebleeding rate assumed to be 15% for the GVL group and 40% for the GVO group. The type I (α) error and type II (β) error were set to 0.05 and 0.2, respectively. The proposed sample size was 47 per group calculated by MedCalc (MedCalc for Windows, version 4.20.011, Mariakerke, Belgium). The significance level was P < .05.
From July 1996 to June 2002, 522 patients with cirrhosis with esophagogastric variceal bleeding were encountered at our hospital. Among them, 115 patients with acute GVH were recruited. Eight patients (two patients had previous endoscopic and surgical treatment for GVH, one had uremia, three had heart failure, and two had non-hepatic malignancy) in the GVL group, and 10 patients (two had previous endoscopic and surgical treatment for GVH, one had uremia, two had chronic obstructive pulmonary disease, three had heart failure, and two had non-hepatic malignancy) in the GVO group were excluded because of enrollment error (Fig. 1). Therefore, 48 patients in the GVL group and 49 patients in the GVO group (84.3%) were included for the final modified ITT analysis. Both groups had otherwise similar demographic data, association of HCC, hepatic functional reserve, severity of bleeding, form and extent of GV, infection status, and period of follow-up (Table 1).
|∴||Band Ligation (n = 48)||Cyanoacrylate Injection (n = 49)|
|Age (year)||61.77 ± 12.35||61.35 ± 14.63|
|Child-Pugh's score||8.10 ± 2.22||7.96 ± 1.86|
|Albumin (g/dL)||2.89 ± 0.54||2.90 ± 0.47|
|Bilirubin (mg/dL)||1.93 ± 1.36||2.42 ± 3.37|
|Prolonged prothrombin time (sec)||2.63 ± 2.26||2.44 ± 1.55|
|Hemoglobulin (g/dL)||8.98 ± 2.29||9.18 ± 1.87|
|Creatinine (mg/dL)||1.19 ± 0.68||1.37 ± 1.12|
|Platelet (K/mm3)||90.56 ± 55.20||105.90 ± 67.87|
|Hematemesis or hematochezia||30||34|
|Active spurting or oozing||15||15|
|Blood transfusion (unit)||5.15 ± 3.18||5.91 ± 3.50|
|Sarin classification (type 1/2/3)||26/16/6||27/9/13|
|Size of gastric varices (F1/F2/F3)||10/24/14||10/22/17|
|Secondary gastric varices||10*||10†|
|Portal vein thrombosis||10||11‡|
|Time, presentation to endoscopic treatment (hr)||9.21±6.51||8.57±7.41|
|Follow-up period (day)||610.58 ± 603.04||680.67 ± 710.54|
The mean number of bands used per session in GVL was 4.5 ± 2.6 (range, 1-10). The mean number of shots used per session in GVO was 2.7 ± 1.3 (range, 1-6). Success in arresting active bleeding was not different in the GVL and GVO groups (14/15 vs. 14/15, P = 1.000) (Table 2). The rebleeding of GV was higher in the GVL group (GVL vs. GVO, 21/48 vs. 11/49, P = .044). The rebleeding rate was still higher in IGV1 patients undergoing GVL than those undergoing GVO (P = .003) in exploratory subgroups analysis. The overall difference in the cumulative rebleeding rate of GV, computed by Kaplan-Meier analysis and compared by a log-rank test, was also higher in patients in the GVL group even after exclusion of patients with concomitant HCC (Fig. 2A-B). The cumulative rebleeding rate of all portal hypertension–related sources (EV, portal hypertensive gastropathy, and other undetermined) was also higher in patients in the GVL group (P = .0451). The cumulative rebleeding rate of GV was 33.5% (95% confidence interval [CI], 17.6%-49.4%), 63.1% (95% CI, 44.5%-81.7%), 72.3% (95% CI, 51.3%-93.3%), 81.6% (95% CI, 61.4%-101.8%) at 1, 2, 3, and 4 years for GVL and 22.8% (95% CI, 10.1%-35.5%), 26.8% (95% CI, 12.5%-41.1%), 26.8% (95% CI, 12.5%-41.1%), and 26.8% (95% CI, 12.5%-41.1%) at 1, 2, 3, and 4 years for GVO; P = .0143, log-rank test (Fig. 2A). The first episode of rebleeding occurred throughout the entire follow-up period in patients undergoing GVL; however, that did not occur beyond 1½ years of follow-up in patients undergoing GVO. If the cases of error enrollment in each group were included for actual ITT analysis, either the crude rebleeding rate [21/56 (37.5%) vs. 11/59 (18.6%) (P = .041) for GVL vs. GVO] or actuarial proportion rebleeding rate (comparisons by log-rank test, P = .014) would also be higher in the GVL group. Univariate analysis showed that the risk of GV rebleeding was significantly linked to the presence of HCC and treatment method (Table 3). In multivariate analysis of these two risk factors, HCC (presence vs. absence of HCC, relative hazard: 2.453, 95% CI: 1.036-5.806, P = .041) and treatment method (GVL vs. GVO, relative hazard: 2.660, 95% CI: 1.167-6.061, P = .020) remained as two independent determinants of GV rebleeding. Eradication rates in GVL vs. GVO group was 32 of 48 (66.7%) versus 31 of 49 (63.3%) (P = .89). Seven patients in each group needed endoscopic ultrasound to clarify the obliteration. The recurrence rate of GV in GVL vs. GVO was 59.4% (19/32) versus 22.6% (7/31) (P = .007) (Table 2). Regarding treatment failure (Table 2), in the GVO group, seven patients had two episodes of GV rebleeding, five patients bled to death (including one who died of index bleeding, three who died of the first GV rebleeding, and one who died of the second GV rebleeding). In the GVL group, 10 patients had two episodes of GV rebleeding, four patients bled to death (including one who died of index bleeding and three who died of 2nd GV rebleeding), and four patients switched from GVL to GVO. These four patients undergoing GVL were switched to Histocryl injection because rubber bands could not be deployed on the GV when rebleeding occurred.
|Band Ligation (n = 48)||Cyanoacrylate Injection (n = 49)||P-value|
|Control of active bleeding||14/15||14/15||NS|
|No. of GV rebleeding||21/48||11/49||.044|
|GV rebleeding index* (months/episode + 1)||7.61 (0.03-50.60)||9.12 (0.17-66.07)||NS|
|No. of sessions||1.8 ± 1.4 (range: 1-6)||1.5 ± 0.7 (range: 1-3)||NS|
|Time to eradication (day)||59.3 ± 31.3||48.3 ± 22.5||NS|
|Relative Hazard||95% CI||P||Relative Hazard||95% CI||P|
|Unit of blood transfusion||1.016||0.507–2.038||.965||1.538||0.904–2.616||.113|
There was no difference in complications between the GVL and GVO groups (Table 4). Most of the complications were infections. One patient who underwent GVO developed portal vein thrombosis.
|Band Ligation (n = 48)||Cyanoacrylate Injection (n = 49)|
|Infection after treatment|
|Spontaneous bacterial peritonitis||1||1|
|Urinary tract infection||1||1|
|Fever and leukocytosis||7||7|
|Portal vein thrombosis||0||1|
|Cause of death|
|Cerebral vascular accident||1||0|
Mortality and Survival.
Thirty-three patients in the GVL group and 27 patients in the GVO group died (Table 4). Hepatic failure was the most common cause of death. Thirty-day mortality and the cause of mortality were not different between these two groups. The overall rate of survival was similar between these two groups even after excluding patients with concomitant HCC (Fig. 3A-B). The cumulative mortality rate at 1, 2, 3, and 4 years for GVL versus GVO were 44.5% (95% CI, 30.3%-58.7%) versus 42.4% (95% CI, 28.2%-56.6%); 55.3% (95% CI, 41.1%-69.5%) versus 54.1% (95% CI, 39.6%-68.6%); 62.8% (95% CI, 48.7%-76.9%) versus 56.9% (95% CI, 42.2%-71.6%); 79.3% (95% CI, 64.9%-93.7%) versus 60.8% (95% CI, 45.6%-76.0%); P = .311, log-rank test. Univariate analysis showed that the survival was significantly linked to the presence of HCC, Child-Pugh's score, and serum creatinine (Table 3). In multivariate analysis, the association of HCC (presence vs. absence of HCC, relative hazard: 4.473, 95% CI: 2.547-7.855, P = 0) and serum creatinine (relative hazard: 2.015, 95% CI: 1.177-3.448, P = .011) were two independent risk factors determining survival.
GV rupture has the characteristics of more severe blood loss and higher mortality and represents a tougher problem than EVH.1–3 The contributing factors are anatomical and technical. Anatomically, GV lie deeper within the submucosa than EV, on average are larger than EV, drain directly into large veins (such as the gastrorenal shunt) without intervening smaller veins (which are used to drain the collaterals from EV), and are exposed to acid and pepsin.30, 31 Technically, diagnosis of GVH is more difficult because the gastric mucosal folds, blood pooling in the fundus, and high posterior wall (the usual site of GVH) are confusing.
Whereas injection sclerotherapy has been applied to treat active bleeding from GV, its use in GVH is associated with a high rebleeding rate and a frequent need to resort to surgical intervention and thus is regarded as only a temporary hemostatic measure.2, 32 The success rate to control GVH by endoscopic injection of N-butyl-2-cyanoacrylate (GVO) appeared higher than other sclerosants according to previous non-randomized trials.8, 9 The advantage of endoscopic variceal ligation for EVH has been documented and has been suggested as the treatment of choice for EVH.10 The hematostatic effect of endoscopic variceal ligation (GVL) for GVH appeared promising, but evidence is still limited.6, 7 However, a relatively small randomized trial13 was conducted to compare the efficacy of GVL versus GVO on GVH, and its conclusions were against the use of GVL for the treatment of GVH. More randomized trials are needed to clarify the best treatment methods.
In comparison with previous studies,6, 7, 33, 34 this study has three important distinguishing features. First, it is the second randomized trial13 and the largest study population to compare the two promising methods: GVL versus GVO. Second, the study focuses on patients treated acutely rather than on an elective basis, as evidenced by the fact that all of our patients were treated within 24 hours of bleeding. Third, a longer follow-up course was provided.13
The success rate in controlling active bleeding was 93.3% for GVO, which was similar to previous reports9, 13, 35 and to that of GVL. However, equal efficacy arresting active bleeding cannot be concluded in both groups because of the small case number. The power to conclude similar efficacy (93.3%) is 80 cases in each group. Reaching such a large population of active bleeding is very difficult. To set a conclusion of equal efficacy in controlling active bleeding, a large case number trial based on active bleeding is needed. The success rate of GVL was also comparable to that of previous studies6, 7, 36 but higher than the rate in Lo et al.'s study.13 The difference in the success rates of bleeding control between studies may be attributable to different technical applications. In our study, more rubber bands were used to ligate the prominent varices where possible, not just on the bleeding site. The difference was reflected by 4.5 ± 2.6 rubber bands per session used as compared with one to four bands per session by Lo et al. Application of more rubber bands may have led to complete interruption of surrounding communicating vessel or feeding vessels to bleeders, arresting the blood flow more effectively.
The rebleeding rate of GV is still high even after improvements in pharmacological and endoscopic treatment.1–3 The cumulative rate of GV rebleeding was higher in patients who underwent GVL when compared to GVO. The rebleeding rate (11/49) of GVO was comparable to those of previous studies (18.5%-31%).9, 13, 35 The rebleeding rate (21/48) of GVL was higher than previous studies6, 7, 36 but similar to Lo et al.'s study.13 The discrepancy is possibly attributable to heterogenicity between these studies such as acute treatment basis versus elective basis, concomitance of HCC, different technical application, and length of follow-up. Indeed, in comparison with the first 18 months after GVO, the rebleeding risk after GVL sustained throughout the entire follow-up period in this study. From the pathophysiological view, the higher rebleeding and recurrence rate in GVL patients may be attributable to the limitation of GVL's effect on only the superficial collaterals in the mucosal and submucosal layers. In contrast, GVO may obliterate collaterals over a wider area and in deeper layers. Actually, other than the treatment method, HCC was the only independent factor influencing GV rebleeding with a hazard ratio of approximately 2.5, which is consistent with previous reports that have shown that the presence of HCC is an extremely good predictor of rebleeding in patients with EVH.26–28
Except for infections, complications were unusual in patients who underwent GVL or GVO. Actually, bacterial infection often coexists with acute variceal hemorrhage and has been documented in 35% to 60% of patients with cirrhosis who have variceal hemorrhage.37 Because the infectious complications were similar in both GVL and GVO groups, the high infection rate in both groups may have been due to massive bleeding per se and not to endoscopic treatment. This study was designed long before the publication of Baveno III consensus statements that suggested antibiotics prophylaxis. These findings further emphasize the need for antibiotic prophylaxis to decrease infections in patients with acute EV or GV hemorrhage.37, 38
The rate of treatment failures was not different between these two groups. Although the GV rebleeding rate was higher in patients with GVL, the rebleeding events were usually controlled with the same method, and most patients survived the bleeding episode. Therefore, the decreased rebleeding rate did not reflect in decreasing mortality. Actually, most patients died of hepatic failure regardless of whether they were treated with GVL or GVO. HCC and serum creatinine are the main factors determining the survival of patients with GVH. Not surprisingly, HCC is the most critical determinant of survival.38, 39 Consistent with previous reports, serum creatinine is also a determinant of survival in patients with cirrhosis and variceal bleeding.40–42
In summary, the efficacy of GVL to control active GVH appears not different to GVO. However, the GV rebleeding rate was lower in those treated with GVO than in GVL. Therefore, GVO is the superior choice for treating GVH. If GVO is unavailable because of a lack of Histoacryl or technical expertise, GVL might be used as a temporary means to arrest active GVH.
The authors thank Pui-Ching Lee for preparing the manuscript.