This was a retrospective case-control study conducted in 341 patients who were diagnosed with bleeding GV and had undergone GVO using NBC at Chonnam National University Hospital (Gwangju, Korea) from January 2004 to July 2013. Patients who were diagnosed with bleeding GV and treated at other institutions before being referred to our center were excluded from the study. We also excluded isolated GV bleeding. Patients were followed up until death or until July 2013. The patients were divided into two groups: the PPI and non-PPI groups. Patients in the PPI group (n = 219) had received 40 mg pantoprazole daily for at least 1 month from the day of hospital admission. Patients in the non-PPI group (n = 122) had not received any anti-secretory treatment. All patients had received intravenous third-generation cephalosporins and vasoactive drugs, such as terlipressin or somatostatin, immediately before endoscopy or during endoscopy, and these medications were continued for up to 5 days afterwards, in accordance with published guidelines [16]. The dose of a non-selective β-blocker was titrated according to the patient’s heart rate (55 to 60 beats per minute) and systolic blood pressure (90 to 100 mmHg) if possible.

All data, including patient demographics, etiology of GVs, Child-Pugh scores, endoscopic findings, the effectiveness of endoscopic treatments, and clinical and endoscopic follow-up data, were recorded and analyzed. After GVO, follow-up endoscopies were performed before the patients were discharged from the hospital, if possible (Fig. 1). All patients provided written informed consent to undergo endoscopy.

Patients who had upper gastrointestinal (UGI) bleeding and features indicative of cirrhosis had undergone upper endoscopy within 12 hours after admission to the hospital. Endoscopy was performed using a forward-viewing endoscope (Olympus GIF Q 260, Olympus, Tokyo, Japan).

Histoacryl was prepared as described below. The injection needle was primed with distilled water, followed by a 1:1 mixture of Histoacryl and Lipiodol (iodized oil, Laboratoire Guerbet, Roissy-Charles-de-Gaulle Cedex, France) injected as a bolus dose of 1 to 4 mL, depending on the size of the gastric varix. Sclerotherapy was performed using a disposable 2.3-mm diameter sclerotherapy needle (MTW, Endoskopie, Wesel, Germany). The catheter was introduced into the working channel of the endoscope and advanced. Once the bleeding varix was located, the needle was exposed and inserted into the base of the varix and the Histoacryl-Lipiodol mixture was injected by an endoscopy assistant. Upon removal of the needle from the varix, the assistant injected normal saline to flush the remaining glue from the catheter. Injections were repeated until GVs appeared to occlude, as judged by probing using a blunt probe. Both the endoscopist and assistant wore safety spectacles to comply with safety guidelines and to reduce the likelihood of accidentally spraying NBC into the eyes, which could cause glue-induced damage.

GVs were grouped into two categories using the classification of Sarin and Kumar [1]. In this classification, type 1 gastroesophageal varices (GOV1) appear as a continuation of esophageal varices (EVs) and extend 2 to 5 cm below the gastroesophageal junction (GEJ), while type 2 GOV (GOV2) extend beyond the GEJ into the fundus of the stomach [1]. GVs were classified into three types, tortuous (F1), nodular (F2), and tumorous (F3), using the system proposed by Hashizume et al. [17]. Active GV bleeding was defined as blood actively spurting or oozing from a varix, and evidence of recent bleeding was defined as the presence of a white nipple or red plugs.

The efficacy of Histoacryl in successful GVO was assessed by probing the varix with the injection catheter. If it remained soft, an additional injection was initiated to achieve complete obliteration, defined as absolute firmness of the injected varix. Initial hemostasis was considered successful when bleeding had been arrested and no recurrence was observed for 2 days [10]. Rebleeding was defined as a new onset of hematemesis and the presence of coffee-ground vomitus, hematochezia, or melena accompanied by an increased pulse rate of > 100 beats per minute and a decrease in blood pressure to < 90 mmHg, following a 24-hour period of stable vital signs and hemoglobin levels after endoscopic treatment [18]. Bleeding-related death was defined as death within 6 weeks of the index bleed [19]. A Histoacryl-induced ulcer was defined as a mucosal break ≥ 3 mm deep at a site where Histoacryl had been injected previously (Fig. 2).

Written informed consent was obtained from all the patients regarding the nature and purpose of the treatment, and this study was approved by the Institutional Review Board of Chonnam National University Hospital (CNUH-2013-191).

Statistical analyses were performed using SPSS version 20.0 (IBM Co., Armonk, NY, USA). Continuous variables are expressed as means ± standard deviation. Student t test and Pearson chi-square test were used to compare baseline characteristics of the patients. Significant factors identified by univariate analysis were entered into a stepwise multivariate logistic regression analysis to distinguish the risk factors that retained statistical significance from those that merely depended on other factors. The rebleeding rate was estimated using Kaplan-Meier curves and compared using the log-rank test. The null hypothesis of no difference was rejected if p < 0.05, or if the 95% confidence intervals (CI) of the odds ratio (OR) estimates excluded one.

The baseline demographic and clinical characteristics of the patients are shown in Tables 1 and 2. There were no significant differences in the baseline clinical characteristics between the PPI and non-PPI groups except the follow-up esophagogastroduodenoscopy (EGD) rate, which was higher in the PPI groups than the non-PPI group (p = 0.027).

A total of 341 patients were enrolled in this study. Of these, 219 (64.2%) in the PPI group received 40 mg pantoprazole daily for at least 1 month from the time of hospital admission, and 122 (35.8%) in the non-PPI group did not receive any anti-secretory treatment. The mean age of the patients was 57.8 years (range, 22 to 96), with 278 males (81.5%) and 63 females (18.5%). Of the patients evaluated, 100 (29.3%) were positive for the hepatitis B surface antigen, 27 (7.9%) were positive for hepatitis C virus antibodies, 155 (45.5%) were chronic alcohol drinkers and 59 (17.3%) had a combined etiology.

Patients were grouped according to the Child-Pugh classification to assess the prognosis of chronic liver disease. There were 76 class A patients (22.3%), 188 class B patients (55.1%), and 77 class C patients (22.6%). Of the patients evaluated, 100 (29.3%) had concomitant hepatocellular carcinomas, and 162 (47.5%) had a previous history of variceal bleeding. Patients were also classified based on the type of GV. There were 213 patients (62.5%) classified as GOV1, 128 (37.5%) as GOV2, 34 (10.0%) had an F1 form of GV, 141 (41.3%) had an F2 form of GV, and 166 (48.7%) had an F3 form of GV. The rate of initial hemostasis was 97.1%. The mean follow-up duration was 554 days (range, 1 to 2,939).

Kaplan-Meier analysis demonstrated that rebleeding occurred in 0.3% of patients within 1 week, 2.2% of patients within 2 weeks, 3.9% of patients within 4 weeks, 18.9% of patients within 6 months, and 27.6% of patients within 12 months (Table 3, Fig. 3A). The probability of rebleeding in the PPI group was lower than that in the non-PPI group (log rank test, p = 0.009) (Fig. 3B). Bleeding-related death occurred in 7% of patients. There were also 123 (36.1%) adverse events during the study, including the follow-up period. These adverse events included GVO-induced ulceration in 74 of 214 patients (34.6%), fever in 26 (7.6%), abdominal pain in 13 (3.8%), diarrhea in four (1.2%), bacteremia in two (0.6%), spontaneous bacterial peritonitis in two (0.6% ), embolism in one (0.3%), and pseudomembranous colitis in one (0.3%) (Table 3).

Univariate analysis showed that use of PPIs (relative risk [RR], 0.496; 95% CI, 0.307 to 0.801; p = 0.004) and a previous history of variceal bleeding (RR, 1.918; 95% CI, 1.195 to 3.078; p = 0.007) were associated with rebleeding. Multivariate analyses of the risk factors for rebleeding are shown in Table 4.

Univariate analysis showed that use of PPIs (RR, 0.436; 95% CI, 0.222 to 0.857; p = 0.016) and a previous history of variceal bleeding (RR, 3.257; 95% CI, 1.701 to 6.25; p < 0.001) were associated with rebleeding in the follow-up EGD group. Multivariate analyses of risk factors for rebleeding were use of PPIs (RR, 0.487; 95% CI, 0.262 to 0.903; p = 0.022) and a previous history of variceal bleeding (RR, 3.067; 95% CI, 1.742 to 5.405; p < 0.001). However, GVO-induced ulcer was not associated with rebleeding (p = 0.799).

Univariate analysis showed that Child-Pugh class C (RR, 16.564; 95% CI, 4.764 to 57.594; p < 0.001), failure of the initial hemostasis (RR, 15.027; 95% CI, 2.318 to 97.408; p = 0.004), use of a β-blocker (RR, 0.153; 95% CI, 0.040 to 0.584; p = 0.006), and presence of concomitant red-colored EV (RR, 5.341; 95% CI, 1.263 to 222.589; p = 0.023) were associated with bleeding-related death, while PPI use (RR, 0.582; 95% CI, 0.196 to 1.730; p = 0.330) was not. Multivariate analyses of the risk factors for rebleeding are shown in Table 5.