Materials and methods

Between February 2005 and January 2007, all patients with cirrhosis in our hospital presenting with either haematemesis or melaena were assessed for inclusion in this trial. The inclusion criteria were: (1) age range between 18 and 75 years old; and (2) patients presenting with acute inactive oesophageal variceal bleeding (defined below) proven by emergency endoscopy. Acute inactive oesophageal variceal bleeding was defined as when patients presented with red colour signs on their oesophageal varices with blood in the oesophagus or stomach and no other potential site of bleeding identified. Patients were excluded if they presented with one or more of the following: (1) no cirrhotic portal hypertension; (2) evident active oesophageal variceal bleeding during endoscopy; (3) association with hepatocellular carcinoma or other malignancy; (4) association with cerebral vascular accident, uraemia, sepsis, bedridden or with other debilitating disease; (5) a history of gastric variceal bleeding; (6) received a β-blocker within 1 month prior to entry; (7) a history of contraindication to the use of terlipressin, such as myocardial ischaemia, peripheral vascular disease or hypersensitivity; (8) a history of prior shunt operation, TIPS (transjugular intrahepatic portosystemic stent shunt), sclerotherapy or EVL; (9) deep jaundice (serum bilirubin >10 mg/dl); (10) encephalopathy greater than stage II; (11) were moribund; (12) pregnancy; or (13) refusal to participate in the trial.

Treatment of patients

All the patients considered suitable for the trial received resuscitation and nasogastric irrigation initially, and then terlipressin infusion was instituted for patients suspected of acute variceal bleeding. If patients became haemodynamically stable, emergency endoscopy was performed within 12 h of admission. Terlipressin (Glypressin, Ferring, Kiel, Germany) was started with a 2 mg intravenous bolus injection. After endoscopic confirmation of inactive acute oesophageal variceal bleeding, eligible subjects were randomised to two groups. The Terlipressin group received terlipressin infusion only, whereas the Combined group received terlipressin infusion plus EVL treatment. Randomisation was performed by a study nurse, who was blind to the severity of liver disease of each study subject. The method of randomisation used opaque sealed envelopes numbered according to a table of random numbers.

Among the Terlipressin group, terlipressin infusion 1 mg intravenously every 6 h was continued for 5 days after enrolment. Among the Combined group, patients received immediate EVL after endoscopic examination soon after randomisation. In addition, intravenous terlipressin infusion 1 mg every 6 h was continued for only 48 h. Our EVL technique has been described previously.3 Briefly, premedication with 20 mg of buscopan intramuscularly was undertaken. The Saeed Six-Shooter or Four-Shooter (Wilson-Cook Medical, Winston-Salem, North Carolina, USA) attached to a video endoscope (Olympus XQ 230, Tokyo, Japan) was utilised. Ligation was initiated at the haematocystic spots or white nipple signs, if present, or at the gastro-oesophageal junction and advanced proximally. After the EVL procedure, sucralfate granules were routinely administered to hasten ligation-induced ulcers.18

Clinical follow-up

Standard treatment, including blood and frozen plasma transfusion, replacement of fluid and electrolytes, and lactulose, was administered to patients in both groups as clinically indicated. Prophylactic antibiotics using cefazolin were given to all the patients for 5 days and subsequently adjusted by sensitivity test in patients proven to have an infection.19 Blood pressure, pulse rate and body temperature were recorded every 6 h for 5 days. Haemoglobin and haematocrit were measured at presentation of acute bleeding, and after endoscopic examination, and then every day for 5 days. Haemoglobin and haematocrit were also closely monitored in suspected cases of rebleeding. Blood transfusion was aimed at maintaining the haematocrit between 27% and 30%. The amounts of blood transfusion prior to endoscopic examination, during 48 h and 48–120 h after enrolment were recorded. The severity of liver disease was based on Child–Pugh classification,20 which was evaluated at admission. Treatment-related complications were recorded. All the recruited patients signed an informed consent form. Our study was approved by our hospital Ethics Committee. After successful control of acute bleeding without very early rebleeding, patients were allocated to receive β-blockers to prevent further rebleeding.

Definition of treatment failure, 5-day haemostasis, failure to control acute variceal bleeding and very early rebleeding

Treatment failure was defined as failure to control acute bleeding episodes or very early rebleeding or death within 5 days. Failure to control acute variceal bleeding was defined as occurrence of any of the following events within 48 h of enrolment, based on the modified criteria of Baveno consensus III 21: (1) fresh haematemesis after start of enrolment; (2) sudden onset of reduction in blood pressure of ⩾20 mm Hg and/or increase in pulse rate of ⩾20 beats/min with a 2 g drop in haemoglobin in those not transfused; (3) transfusion of 4 units of blood required to increase the haematocrit to above 27% or haemoglobin to above 9 g/dl; and (4) death. Very early rebleeding was defined when criteria for failure to control acute variceal bleeding occurred between 48 and 120 h after enrolment in patients achieving control of acute bleeding. Control of acute bleeding was defined as when criteria for failure did not occur within 48 h of enrolment. Five-day haemostasis was defined when criteria for failure to control acute variceal bleeding and very early rebleeding did not occur within 5 days of enrolment. A nasogastric tube was not routinely inserted after initial endoscopy; however, a nasogastric tube was inserted in cases of failure to control acute bleeding or episodes of very early rebleeding. Failure to control acute bleeding and very early rebleeding were assessed by two experienced clinicians.

Failures of control of acute bleeding or very early rebleeding in both groups were treated with EVL (first choice), sclerotherapy, balloon tamponade or somatostatin infusion, at the discretion of the attending doctor.

Primary endpoints were treatment failure and very early rebleeding. Secondary endpoints were failure to control acute bleeding, treatment-related complications, amount of blood transfusion and mortality at 42 days.

Sample size calculation and statistical analysis

A haemostatic rate for terlipressin in the control of acute variceal bleeding ranging from 19% to 96% has been recorded.22 23 24 Our previous trials showed that emergency EVL could achieve a 97% haemostatic rate with low rebleeding rate in patients with active variceal bleeding.3 15 Thus, if we presumed that the failure to control acute bleeding within 48 h is 20% in the Terlipressin group and 5% in the Combined group, and the very early rebleeding rate is 20% in the Terlipressin group and 10% in the Combined group—that is, the presumed treatment failure was 40% in the Terlipressin group and 15% in the Combined group—approximately 46 patients in each group were required with a two-tailed test to achieve a β value of 0.1 and an α error of 5%.

The data were expressed as mean (SD). Statistical analysis was based on an intention-to-treat principle. Quantitative variables were compared according to Student t test, and qualitative variables were compared with the χ² test and Fisher exact test when appropriate. Actuarial curves were calculated by the Kaplan–Meier method and compared by the log rank test. Cox regression analysis was used to detect independent predictive variables for rebleeding and survival. All p values were two-tailed. A p value <0.05 was considered significant. Analyses were performed using SPSS 12.0 software (SPSS, Chicago, Illinois, USA).

Results

A total of 201 patients were screened; 108 patients were not eligible for the following reasons: hepatocellular carcinoma, 36 patients; >75 years old, 23 patients; <18 years old, 1 patient; actively bleeding at endoscopy, 15 patients; other malignancy, 2 patients; chronic renal insufficiency, 6 patients; previous TIPS, 1 patient; previous propranolol, 9 patients; previous EVL, 5 patients; serum bilirubin >10 mg/dl, 5 patients; septic shock, 1 patient; myocardial infarction, 1 patient; refused to participate, 3 patients. Finally 93 patients were enrolled in the trial, 46 patients in the Terlipressin group and 47 patients in the Combined group (fig 1). Both groups were comparable in baseline characteristics (table 1). The proportion of Child–Pugh class C was slightly higher in the Combined group (30%) than in the Terlipressin group (15%). No subjects were lost to follow-up.

• Download figure
• Open in new tab
• Download powerpoint

Figure 1

Flow chart of enrolment. EVL, endoscopic variceal ligation; TIPS, transjugular intrahepatic portosystemic stent shunt.

The time between admission and endoscopic examination was 8.0 (3.5) h in the Terlipressin group and 6.7 (4.5) h in the Combined group (p = 0.57). The time between terlipressin infusion and endoscopic examination was 2.5 (2.2) h in the Terlipressin group and 1.8 (2.0) h in the Combined group (p = 0.63). Terlipressin infusion was continued in all the patients after endoscopic examination. EVL was performed in all patients in the Combined group with a mean of 4.0 (0.6) rubber bands (range 3–6).

Rebleeding

The main outcomes are shown in table 2. Failure to control acute bleeding was encountered in 4 patients (9%) in the Terlipressin group and 1 patient (2%) in the Combined group (p = 0.20). Two patients in the Terlipressin group presented massive haematemesis within 10 h of endoscopy and died of torrential haemorrhage despite institution of balloon tamponade and additional bolus infusion of terlipressin. The other 2 patients were successfully rescued by EVL. The only one failure in the Combined group developed ulcer bleeding from the ligation site on the second day.

Very early rebleeding occurred in 7 patients in the Terlipressin group but in no patients in the Combined group (p = 0.006). Thus, treatment failure at 5 days was 24% in the Terlipressin group and 2% in the Combined group (p = 0.002). All 7 patients with very early rebleeding in the Terlipressin group were successfully treated with EVL. The proportion of patients remaining free of rebleeding at 7 days was significantly higher in patients treated with terlipressin plus EVL than those treated with terlipressin alone (p = 0.003, fig 2). Univariate analysis including age, gender, aetiologies, albumin, bilirubin, ascites, prothrombin time, presence of encephalopathy, variceal size and red colour signs on varices showed that only variceal size and treatment were predictive factors of very early rebleeding. Multivariate analysis revealed that treatment (odds ratio (OR) 0.081; 95% CI 0.010 to 0.627) was the only factor predictive of very early rebleeding (table 3).

• Download figure
• Open in new tab
• Download powerpoint

Figure 2

The actuarial probability of being free from early rebleeding at 7 days in both groups.

The amount of blood transfusion was significantly higher in the Terlipressin group than in the Combined group, with respect to within 48 h or within 48–120 h of enrolment (table 2). There were significantly more patients requiring intensive care in the Terlipressin group. Hospital stay was slightly longer in the Terlipressin group than in the Combined group, but the difference was not significant.

Complications and mortality

The complications are shown in table 4. Abdominal pain was more commonly noted in the Terlipressin group, whereas chest pain was seen more often in the Combined group. Oesophageal ulcer bleeding induced by banding ligation was encountered in 2 patients in the Combined group; both were treated successfully with somatostatin infusion 250 μg/h for 3 days. None of the differences was statistically significant.

Three patients in the Terlipressin group and 1 patient in the Combined group died within 42 days. The survival was similar between the two treatment groups (fig 3). The causes of death in the Terlipressin group were variceal haemorrhage in 2 patients and aspiration pneumonia in 1 patient. The only death in the Combined group was due to sepsis. Multivariate analysis did not reveal any factor predictive of mortality.

• Download figure
• Open in new tab
• Download powerpoint

Figure 3

The Kaplan–Meier survival curve of both groups at 42 days.

Discussion

Following control of acute oesophageal variceal haemorrhage, very early rebleeding constitutes a major determinant of poor outcome, posing a great challenge for clinicians. Consequently, it is imperative for clinicians to control acute bleeding episodes as well as preventing very early rebleeding.19 24 25 26 Vasoconstrictors such as somatostatin and terlipressin have been proven to be as effective as endoscopic treatment in the prevention of very early rebleeding, but with fewer complications.6 8 10 Though a combination of vasoconstrictor and endoscopic treatment is the mainstay for the management of acute oesophageal variceal bleeding,11 27 28 the role of endoscopic treatment in patients with no active variceal bleeding at endoscopy is disputed.7 8 9 10 12 13 14

Our study enrolled patients with endoscopically proven acute oesophageal variceal bleeding. Variceal haemorrhage was shown to have already stopped during endoscopic examination. Among the patients excluded from the current trial, only 15 patients (7.5%) were noted to be actively bleeding at endoscopy, similar to the figure of 10% reported by Sarin et al.29 The lower incidence of active bleeding than in previous reports may be due to different aetiologies of cirrhosis and the stringent criteria adopted by us. Additionally, prior use of a vasoconstrictor has been shown to reduce the incidence of active bleeding during endoscopy.2 30 Our study showed that patients with no active bleeding at endoscopy receiving a combination of terlipressin and EVL apparently had a lower rate of very early rebleeding and treatment failure than patients treated with terlipressin alone. Since both treatment groups had similar baseline characteristics, the lower rebleeding rate could be ascribed to the association of EVL and terlipressin infusion. It has been demonstrated that the haemostatic rate was similar between high dose and low dose terlipressin.23 31 Escorsell also showed that the effects of a reduced dose (1 mg) were almost as pronounced and prolonged as those of a dose of 2 mg.31 Thus, terlipressin treatment at low dosage (1 mg) for several days was recommended to reduce the risk of early rebleeding.26 31 In our study, the maintenance dose of terlipressin was 1 mg every 6 h. The control of acute bleeding achieved by simply terlipressin infusion was 91% and the very early rebleeding rate was 15%. These figures corresponded to those found in previous studies.9 24 Probably, using the a higher dose of 2 mg/4 h may have slightly increased the efficacy of terlipressin administered alone and the difference in rebleeding is no more evident. When analysing the relationship between Child–Pugh class and rebleeding, there was no significant difference in the Combined group. This may imply that association of EVLwith terlipressin is applicable to those of different Child–Pugh class.

The optimal duration of use of a vasoconstrictor in the treatment of acute oesophageal variceal haemorrhage remains an unresolved issue. Using vasoactive drugs for up to 5–6 days to prevent early rebleeding has been recommended by hepatology experts.8 24 25 Our results can be interpreted as showing that with combination with EVL, 2 days rather than 5 days of terlipressin infusion is sufficient to prevent very early rebleeding in patients with inactive variceal bleeding at endoscopy.

Due to a lower rebleeding rate, the requirement for blood transfusion was significantly reduced in the Combined group compared with the Terlipressin group. It is of note that the complications were not appreciably enhanced in patients receiving EVL. Only 2 patients (4%) experienced postligation ulcer bleeding and both could be managed conservatively. The majority of other complications except for bacteraemia could possibly be ascribed to the use of terlipressin in both treatment groups. Our trial did not support the notion that endoscopic treatment may enhance the complications without enhancing the effect.11 This should be attributed to substantially fewer complications associated with EVL compared with sclerotherapy.32 33 34 Because of the low mortality rates in both groups, the survival rate was not significantly different. Our study showed that on the premise that there is no comorbid disease, the mortality was greatly reduced in patients with cirrhosis with acute inactive variceal bleeding at endoscopy when treated with terlipressin rescued by EVL or a combination of EVL and terlipressin for 2 days. Due to advances in treatment for acute oesophageal variceal haemorrhage, the 42-day mortality was recently estimated to be about 20%.13 The remarkably low mortality rate in the current trial was similar to that found in our previous report,15 and akin to the figure of 8.9% reported by Escorsell et al.6

In conclusion, patients with no active variceal bleeding at endoscopy still encountered appreciably high very early rebleeding rates if treated with vasoconstrictor alone. The combination of immediate banding ligation during endoscopy could significantly reduce the rebleeding rate without enhanced complications. Since our study was aimed at a subset of patients with acute bleeding, interpretation of these results should be limited to this subpopulation of patients with cirrhosis.