Endoscopic Injection for Bleeding Peptic Ulcer: A Comparison of Adrenaline Alone With Adrenaline Plus Human Thrombin

GASTROENTEROLOGY 1996;111:623 628 Endoscopic Injection for Bleeding Peptic Ulcer: A Comparison of Adrenaline Alone With Adrenaline Plus Human Thrombin ADAM K. KUBBA, WILLIAM MURPHY, and KELVIN R. PALMER Gastrointestinal Unit, Western General Hospital, Edinburgh, Scotland Background & Aims: Endoscopic injection therapy im- nal bleeding lesions to stop bleeding and to prevent reproves outcome in bleeding peptic ulcer, but the opti- bleeding. 15 19 Although the use of thrombin is attractive mum regimen is unknown. The aim of this study was because it represents the best theoretical approach to to compare the efficacy of endoscopic therapy with causing thrombosis, trials have not shown that this is dilute adrenaline alone vs. adrenaline plus human better than injection using alternative injection soluthrombin in the treatment of patients with major peptic tions. 17 19 This may be because studies have involved ulcer hemorrhage. Methods: One hundred forty patients small numbers of patients, a suboptimal dose of thromwith significant peptic ulcer hemorrhage and active arbin, and inappropriate patient selection and because some terial bleeding or a nonbleeding visible vessel were ranhigh-risk patients have been excluded. domized to endoscopic injection with 1:100,000 adrenaline (70 patients; group 1) or to adrenaline plus 600 In this study, we have examined the efficacy of a com- 1000 IU of human thrombin (70 patients; group 2). bination of adrenaline plus thrombin extracted from The two groups were well matched for age, shock, he- pooled human plasma as injection therapy for bleeding moglobin concentration, comorbid disease, endo- peptic ulcer, comparing it with our gold standard of scopic findings, and consumption of nonsteroidal anti- 1:100,000 adrenaline injections. inflammatory drugs. Results: Fourteen patients from group 1 (20%) and 3 patients from group 2 (4.5%) re- Patients and Methods bled (P õ 0.005). Seven patients from group 1 (10%) but no patients from group 2 died within 30 days of Design and Inclusion Criteria admission (P õ 0.013). Patients from group 1 were Between October 1994 and November 1995, consecuadministered a total of 297 units of blood compared tive patients presenting with severe gastrointestinal hemorwith 219 units in group 2 (P õ 0.041). Conclusions: rhage were considered for this study. These patients were ad- Endoscopic injection using adrenaline plus human mitted to the four short-term receiving units in the Lothian thrombin is superior to injection with dilute adrenaline region of Scotland (Western General Hospital, Edinburgh; alone and may represent the best treatment for bleed- Edinburgh Royal Infirmary, Edinburgh; St. John s Hospital, ing peptic ulcers. Livingston; and Eastern General Hospital, Edinburgh). The admitting teams then contacted one of the investigators he prognosis of patients who present to a hospital (A.K.K.) using an air call pager. After resuscitation, endoscopy T because of major peptic ulcer hemorrhage is im- was performed in all patients by one of the investigators proved by endoscopic injection therapy. A range of soluscope (Keymed Ltd., Southend on Sea, England). Patients who (A.K.K.) using an Olympus XQ10 forward-viewing gastro- tions have been injected, and all have similar efficacy in reducing the rate of rebleeding, operation, and mortalbleeding or had a nonbleeding visible vessel were included in were found to have a peptic ulcer that was either actively ity. 1 4 In controlled trials, Choudari and Palmer 5 and Chung et al. 6 showed that diluted adrenaline (1:100,000 the study if they had one other clinical risk factor. These risk factors were as follows: older than 60 years of age; initial or 1:10,000) is at least as effective as alternative regihemoglobin concentration of õ10 g/dl; or shock, defined as mens, and in contrast to sclerosants injections, 7 10 adrena pulse rate of more than 100 beats/min, a systolic blood aline has not been associated with systemic or local compressure of õ100 mm Hg, or both. A history of nonsteroidal plications. anti-inflammatory drug (NSAID) use, previous history of pep- Unfortunately, 15% 20% of patients continue to bleed tic ulcer disease, and Helicobacter pylori status were recorded. despite technically successful injection therapy. 1 4,11 This Patients who were admitted to the hospital because of peptic may be because the bleeding artery is not thrombosed; ulcer hemorrhage (primary bleeders) and patients who bled indeed, experiments in animals fail to show arterial thrombosis after adrenaline injection. 12 14 1996 by the American Gastroenterological Association Bovine thrombin has been injected into gastroduode- 0016-5085/96/$3.00

624 KUBBA ET AL. GASTROENTEROLOGY Vol. 111, No. 3 after admission for other medical conditions (secondary bleed- success or otherwise of the endoscopic therapy was communicated ers) were recruited. to these doctors, but details of randomization were not Comorbidity was assessed in each patient. Cardiovascular shown. morbidity was defined as previous myocardial infarction with Decisions regarding blood transfusion and the need for recurrent angina attacks and/or the presence of congestive cardiac emergency surgery were made independently by those teams. failure. Respiratory comorbidity was defined as the pres- ence of chronic obstructive airways disease sufficient to limit End Points normal daily activity and/or pneumonia at the time of bleed- The following end points were determined: rebleeding, ing. Renal comorbidity comprised long-term elevation of se- defined as fresh hematemesis, melena, or both with either shock rum creatinine levels to ú500 mmol/l. Neurological disability (pulse rate of more than 100 beats/min and systolic blood was defined as central nervous system disease resulting in loss pressure of õ100 mm Hg) or a decrease of hemoglobin concenof physical independence; it encompasses long-term disability tration of at least 2 g/dl during a 24-hour period, with concaused by stroke, demylination, or other degenerative neuro- firmation of rebleeding by endoscopy or emergency surgery logical disease. The postoperative phase was defined as the in every patient; surgical operation (rebleeding was the only period in which a patient was in the hospital after a major indication for surgery); units of blood transfused; duration of operative procedure. Mentally and physically handicapped in- hospital stay; and 30-day mortality rate (from the time of dividuals were those institutionalized for cerebrovascular dis- admission). ease or cerebral palsy. During the study period, 289 patients underwent endoscopy, Repeat Endoscopy and 149 were excluded because they had bled from other Repeat endoscopy was performed in one of the follow- causes (n Å 91), because they did not have major stigmata of ing three situations: if, in the opinion of an investigator recent hemorrhage within an ulcer bed (n Å 55), because they (A.K.K.), the initial endoscopic injection was believed to be had significant liver disease (n Å 1), or because they were suboptimal because blood made therapy difficult or because of being treated with anticoagulant drugs (n Å 2). an awkward ulcer position, repeat endoscopy was then under- Patients were randomized by opening a sealed envelope to taken within 24 hours as a planned elective procedure; to be administered either injection with dilute adrenaline (group confirm rebleeding; and at the request of the admitting teams 1) or adrenaline plus human thrombin (group 2). The random- if there was doubt concerning rebleeding. ization was performed during the endoscopy, after examination of the peptic ulcer, and before injection of diluted adrenaline. Policy After Rebleeding In many patients, it was necessary to wash the ulcer bed using Rebleeding was treated either by surgical operation or endoscopically positioned catheters to show stigmata of hemorby endoscopic therapy (if requested by the admitting team). rhage. Endoscopically treated rebleeding patients were treated with Endoscopic Technique the same form of endoscopic therapy as at the time of admission. Injections were administered using a disposable 4-mm, 23-gauge injection needle (Keymed Ltd.) Patients from group Statistical Analysis 1 were treated with multiple injections (each 1 2 ml) of Differences in rebleeding and mortality rates were ana- 1:100,000 adrenaline into and around the bleeding vessel. lyzed using the x 2 test and Fisher s Exact Test, respectively. Patients from group 2 were injected with 1:100,000 adrenaline Differences in blood transfusion were performed using the Stuin exactly the same manner followed by at least a 2.8-mL (600 dent s t test. IU) injection of human thrombin injected into the vessel. Human thrombin used in this trial was manufactured from Ethical Consideration pooled human plasma and was donated on a named patient The study was approved by the Lothian Ethical Subbasis (as a component of a fibrin sealant kit) by the Scottish Committee of Medicine and Oncology. Written consent was National Blood Transfusion Service. obtained from all patients or their next of kin. The volume of injection was similar in groups 1 and 2. Patients from group 1 were administered 4 18 ml (median, Results 10 ml) of 1:100,000 adrenaline. Patients from group 2 were One hundred forty patients fulfilled the inclusion administered 4 12 ml (median, 7 ml) adrenaline plus 2.8 criteria and were randomized. Two patients (one in each 4.5 ml (median, 3.5 ml) of thrombin suspended in 40 mmol/ L calcium chloride, representing 600 1000 IU of human group) in whom initial treatment was thought by an thrombin. This dose of thrombin is similar to that which has investigator (A.K.K.) to be suboptimal underwent elec- been shown to reduce blood loss when injected into anasmaterials) within 36 hours. Both patients presented with tive repeat endoscopic therapy (using the same injection tamosis during vascular surgery. 20,21 Management after endoscopy was left in the hands of the active bleeding, and although initial hemostasis was efadmitting teams who were unaware of what was injected. The fected in both cases, anatomic distortion and the presence

September 1996 THROMBIN PLUS ADRENALINE FOR ULCER HEMORRHAGE 625 Table 1. Patient Characteristics Group 1 (n Å 70): adrenaline alone Group 2 (n Å 70): adrenaline / thrombin Active bleeding Nonbleeding vessel Active bleeding Nonbleeding vessel Characteristics (n Å 24) (n Å 46) (n Å 27) (n Å 43) Median age (yr) (range) 71.0 (42 90) 71.0 (26 91) 68.0 (27 83) 69.5 (33 92) Sex (M/F) 15/9 34/12 17/10 27/16 Mean admission hemoglobin concentration (g/dl) (SD) 8.4 (2.2) 8.5 (2.0) 8.9 (2.1) 8.5 (1.9) No. in shock 14 18 20 20 NSAID users 16 22 13 21 Esophageal ulcers 4 2 2 2 Gastric ulcers 7 20 9 16 Duodenal ulcers 13 24 16 25 No. with comorbid disease 18 27 17 21 H. pylori positive/negative 03/21 15/31 13/14 19/24 Primary/secondary bleeders 20/04 40/6 23/04 36/7 of blood resulted in uncertainty about the adequacy of A total of 8 patients from groups 1 and 2 who continued the procedure. to bleed underwent emergency surgery. All but 1 Table 1 shows the characteristics of the randomized patient recovered, although all of the patients had serious patients. The two groups were well matched with regard postoperative complications. One patient died of adult to age, sex, NSAID intake, ulcer type, and risk factors for respiratory distress syndrome 6 days after undergoing rebleeding (admission hemoglobin concentration, shock, emergency surgery. serious comorbid disease, and endoscopic stigmata). The Although median transfusion requirements were simi- nature of comorbid diseases is shown in Table 2. lar in both groups, significantly more total units were Table 3 shows outcome in the two groups. Fourteen transfused in group 1 compared with group 2 (297 vs. patients from group 1 (20%) and 3 patients from group 219; P õ 0.041). The duration of the hospital stay was 2 (4.5%) rebled during their hospital admission (P õ similar in both groups. Seven patients from group 1 died, 0.005). Those patients who rebled included a signifi- but there were no fatalities among patients from group cantly greater number of patients who presented with 2(Põ0.013). Deaths were restricted to patients who active bleeding than those who had a nonbleeding visible had severe comorbid disease, and three of these deaths vessel at endoscopy (11 vs. 6) (P õ 0.036). Ten patients occurred in secondary bleeders who had been admitted from group 1 who rebled were retreated with adrenaline to the hospital because of unrelated serious medical prob- alone, and permanent hemostasis was achieved in 5 pain lems. One of these patients died of bleeding that occurred tients. Permanent hemostasis was not achieved by endowith association with multiple complications associated scopic therapy in 7 patients from group 1. Five of these total cystectomy for invasive transitional cell carciscopic patients underwent urgent surgery: 1 patient died of noma. A second patient with congestive cardiac failure exsanguination before a surgical operation could be pergestive caused by ischemic heart disease rebled and died of con- formed, and another patient who was unfit for surgery heart failure. A third patient died of recurrent because of a recent myocardial infarction had no further bleeding in association with chronic renal failure and bleeding after an intravenous infusion of octreotide. disseminated intravascular coagulation. A fourth elderly patient was admitted because of upper gastrointestinal bleeding in association with lobar pneumonia and died Table 2. Type and Distribution of Comorbid Disease 7 days later of respiratory failure, despite successful treatment Group 1 Group 2 of her bleeding ulcer. A fifth patient was treated Disease (n Å 45) (n Å 38) with unsuccessful injection therapy for bleeding ulcer Cardiovascular 9 6 followed by two laparotomies for gangrenous appendici- Respiratory 6 5 tis and septicemia. One other death followed emergency Cardiovascular and respiratory 13 15 ulcer surgery. A final patient exsanguinated from ulcer Cardiovascular, respiratory, and renal 2 4 Neurological 3 1 bleeding before an operation could be performed. Renal 2 1 No complications followed endoscopic therapy in ei- Postoperative 6 4 ther group. The presence of H. pylori on biopsy specimens Mentally and physically handicapped 4 2 did not influence outcome.

626 KUBBA ET AL. GASTROENTEROLOGY Vol. 111, No. 3 Table 3. Results of Treatment Group 1 (n Å 70): adrenaline alone Group 2 (n Å 70): adrenaline / thrombin Active bleeding Nonbleeding vessel Active bleeding Nonbleeding vessel (n Å 24) (n Å 46) (n Å 27) (n Å 43) No. rebleeding a 9 5 2 1 No. retreated 6 4 1 0 Permanent hemostasis 20 43 25 42 Emergency surgery 3 2 2 1 Median units transfused (range) 2 (0 17) 3 (0 10) 4 (0 9) 3 (0 6) Median duration of hospital stay (range) 6 (2 37) 7 (3 65) 6 (2 25) 6 (4 35) Death b 3 4 0 0 a P õ 0.005. b P õ 0.013. Discussion In this trial of patients presenting with major peptic ulcer hemorrhage, combination injection treat- ment with dilute adrenaline plus human thrombin achieved significantly greater permanent hemostasis and improved mortality rate compared with patients treated by adrenaline alone. Previous studies had shown that conservatively managed patients who have major endoscopic stigmata of bleeding have a high risk of uncontrolled bleeding and rebleeding. 22 25 Others have shown that this risk is re- duced greatly by endoscopic therapy. 1 4 Consequently, we believed that it would be unethical to include a con- trol (no endoscopic treatment arm) in the current study. All patients who died had significant medical comorbidity, and three deaths occurred in patients who bled after admission to the hospital for illness unrelated to peptic ulcer. It is clearly possible that these patients would have inevitably died as a consequence of their cardiorespiratory or other diseases and that bleeding ulcer was an agonal incidental event and that, by chance, all of these were randomized to group 1. However, it is likely that uncontrolled bleeding in each patient contributed to decompensation of comorbid illness and was an important contributor to death. In this study, the severity of comorbid disease was not quantified; nevertheless, the definition of comorbidity was such that these patients had severe illness with significant physical disability. The distribution of comorbid disease, including the presence of multiple conditions, was very similar in the two groups, and differences in outcome after endoscopic therapy are unlikely to have been caused by unequal distribution of severity or the number of conditions present. Although permanent hemostasis was achieved more often with combination therapy than by injection of dilute adrenaline alone, the median blood transfusion requirements and median duration of hospital stay were similar in the two groups. The length of hospital stay is frequently dependent on social factors and the course of unrelated medical illness; therefore, it is not surprising that the duration of hospital stay was similar in the two groups. The observation that there were fewer total transfusion requirements in group 2 and that the median of transfused units was similar in both groups reflects the wide range of transfusions necessary in this study. In a clinical trial, Lin et al. 11 showed that hemostasis from bleeding ulcer can be as readily achieved by the injection of inert substances as by the injection of sclerosants or adrenaline. This suggested that a tamponade effect, in which a relatively large volume of fluid is injected into the rigid confines of a chronic ulcer, can stop bleeding by compressing the bleeding vessel. Others 13,26,27 have also referred to this possible mechanism. In the current study, the median total injected volume was similar in the two groups. Therefore, although a tamponade effect may have contributed to hemostasis in many patients, we have shown an additional specific ef- fect for human thrombin. The efficacy of injected bovine thrombin (alone or in combination with other agents) has been examined in several clinical trials with varying conclusions. Juszkie- wicz et al. 15 randomized 50 patients who presented with nonbleeding visible vessels within peptic ulcers to either no endoscopic therapy or injection with 200 IU of animal thrombin and reported a rebleeding rate of 4% in the treated group compared with 40% in the control group. Although this difference is statistically significant (P õ 0.01) and encouraging, there must be reservations about the conclusions of such a small study. This criticism also applies to the trial reported by Moreto et al. 16 who randomized 38 patients (with spurting, oozing, or nonbleeding visible vessel) to conservative therapy or injec- tion with ethanolamine plus 50 IU of bovine thrombin. Rebleeding in the treated group was 5.3% compared with 57.9% in the control group.

September 1996 THROMBIN PLUS ADRENALINE FOR ULCER HEMORRHAGE 627 The clinical trial that is most comparable to our own modalities but suggests that the combination of injection is that reported by Balanzo et al., 18 who randomized 64 and thermal therapy should be evaluated in a large clini- patients admitted to the hospital because of ulcer bleed- cal trial. ing and were found at endoscopy to have active bleeding The current study clarifies the value of adrenaline plus or a nonbleeding visible vessel to adrenaline alone thrombin injection therapy in patients who show major (1:10,000) or adrenaline plus 150 300 IU of bovine stigmata of hemorrhage. Unlike many previous studies, thrombin. Rebleeding rates (18.7% vs. 15.6%) and other the highest-risk patients were not excluded, an adequate end points in both groups were very similar. It is likely sample size was studied, and endoscopic therapy was that this trial eliminated some high-risk patients because undertaken by a single endoscopist using a standard technique patients older than 80 years of age and those unable to and a relatively high dose of human thrombin. give written consent were excluded. The number of pa- The injected dose of human thrombin used in the current tients included was less than in our study, and the study was at least 60% greater than that used in other amount of injected thrombin was approximately 25% of clinical trials of bleeding peptic ulcer. This dose of human that used in our study. Other studies of thrombin are thrombin was chosen because of experience in vascuthat not comparable with the current study and do not clarify lar surgery. When injected in combination with fibrinogen the value of thrombin. For example, Koyama et al. 19 as a fibrin sealant, 600 1000 IU thrombin reported a trial involving 62 patients who were found to significantly reduced bleeding from anastomosis fashioned have a nonbleeding visible vessel within a gastric ulcer. during carotid endarterectomy 20 and peripheral All ulcers were sprayed with a mixture of dilute adrena- vascular surgery. 21 Furthermore, although no formal, sequential line plus 20,000 IU of thrombin, and half were randomized studies of intravascular thrombosis have been to absolute alcohol injection. Perhaps not surpris- undertaken in these trials, neither local nor systemic ingly, the injected group had a lower rebleeding rate thrombosis nor embolization has occurred. than the group who were not injected (12% vs. 34.5%). It is possible that both bovine and human thrombin Benedetti et al. 17 studied 82 patients injected with either could provoke an anaphylactic reaction, particularly if 1% polidocanol (28 patients) or 100 IU bovine thrombin repeated injections are administered. However, such reactions (54 patients). They excluded patients with giant ulcers have not been reported, despite relatively extensive covered by a large blood clots, patients with firm scle- use of thrombin in vascular surgery 20,21,30 as therapy for rotic ulcers, and patients with deep ulcers. The remaining gastric and esophageal varices 31 and in the current and patients had ulcers that were either actively bleeding, other trials of peptic ulcer bleeding. had a nonbleeding visible vessel, were oozing, or had an The thrombin used in this trial was manufactured adherent clot. The rebleeding rate was similar in both from pooled human plasma. The extraction procedure groups (17.8% vs. 14.8%, respectively). includes a solvent-detergent virus step that has a good Other trials have compared the efficacy of adrenaline safety record. 32 In addition, this product has been used injection with that of adrenaline in combination with as part of fibrin sealant in a number of clinical trials, sclerosants. 5,6,28 These studies do not show advantages for and no evidence of virus transmission was found over a combination therapy using either ethanolamine, 5 sodium 6-month serological follow-up period. 20,32 Recombinant tetradecyl sulfate, 6 or polidocanol 28 with almost identical thrombin is currently being manufactured by the Scotrates of uncontrolled bleeding, rebleeding, surgical intering tish Blood Transfusion Service; therefore, concerns relatvention, and death. The approach to hemostasis of injecinated to viral transmission are likely to be completely elim- tion plus a thermal modality is theoretically attractive, in the near future. and many clinicians use combination injection with bipotion The current study encourages us to believe that injec- lar electrocautery or the heater probe. In a study reported with a combination of dilute adrenaline plus human by Loizou et al., 29 42 patients who presented with peptic thrombin is a significant therapeutic advance in the man- ulcer bleeding in association with visible vessels were agement of major peptic ulcer bleeding. This conclusion randomized to injection therapy with 1:10,000 adrenarelating must be confirmed by larger studies, and issues of toxicity line or to a combination of adrenaline injection plus to possible thrombotic complications and viral neodymium:yttrium-aluminum-garnet photocoagulation. transmission must be addressed. The initial hemostasis was similar in both groups, but 3 patients from the group treated by adrenaline alone References rebled compared with none who were being treated with 1. Panes J, Forne M, Marco C, Viver J, Garcia-Oliveres E, Garau J. Controlled trial of endoscopic sclerotherapy in bleeding peptic combination therapy. This study was clearly too small ulcers. Lancet 1987; 2:1292 1294. to show statistical differences between these treatment 2. Lazo MD, Andrade R, Medina MC, Gracia-Fernadez G, Sanchez-

628 KUBBA ET AL. GASTROENTEROLOGY Vol. 111, No. 3 Cantos AM, Franquelo E. Effect of injection sclerosis with alcohol K, Yamaguchi M, Sakai T, Hisatsugu T. Prevention of recurrent on the rebleeding rate of gastroduodenal peptic ulcers with non bleeding from gastric ulcer with a non bleeding visible vessel by bleeding visible vessels: a prospective controlled trial. Am J Gas- endoscopic injection of absolute ethanol: a prospective, controlled troenterol 1992; 87:843 846. trial. Gastrointest Endosc 1995; 42:128 131. 3. Balanzo J, Sainz S, Such J. Endoscopic haemostasis by local 20. Milne AA, Murphy WG, Reading SJ, Ruckley CV. Fibrin sealant injection of epinephrine and polidocanol in bleeding ulcers: a reduces suture line bleeding during carotid endarterectomy: a prospective randomised trial. Endoscopy 1988; 20:289 291. randomised trial. Eur J Vasc Endovasc Surg 1995; 10:91 94. 4. Rajgopal C, Palmer KR. Endoscopic injection sclerosis: effective 21. Milne AA, Murphy WG, Reading SJ, Ruckley CV. A randomised treatment for bleeding peptic ulcer. Gut 1991; 32:727 729. trial of fibri sealant in peripheral vascular surgery. Vox Sang (in 5. Choudari CP, Palmer KR. Endoscopic injection therapy for bleeding press). peptic ulcer; a comparison of adrenaline alone with adrena- 22. Chang-Chin C, Wu CS, Chen PC, Lin DY, Chu GM, Fang KM. line plus ethanolamine oleate. Gut 1994; 35:608 610. Different implications of stigmata of recent haemorrhage in gas- 6. Chung SCS, Leung JWC, Leong HT, Lo KK, Li AKC. Adding a tric and duodenal ulcers. Dig Dis Sci 1988; 33:400 404. sclerosant to endoscopic epinephrine injection in actively bleeding 23. Brearly S, Morris DL, Hawker PC, Dykes PW, Keighley MR. Predic- ulcers: a randomised trial. Gastrointest Endosc 1993; 39: tion of mortality at endoscopy in bleeding peptic ulcer disease. 611 615. Endoscopy 1985; 17:173 174. 7. Loperfid S, La Torre L, Patelli G. Extensive necrosis of gastric 24. Bornman PC, Theodorou NA, Shuttleworth RD, Essel HP, Marks mucosa following injection therapy of bleeding peptic ulcer (letter). IN. Importance of hypovalaemic shock and endoscopic signs in Endoscopy 1990; 22:285 286. predicting recurrent haemorrhage from peptic ulceration: a pro- 8. Pousset JP. Necrosis of the firs part of duodenum after endoscopic spective evaluation. BMJ 1985; 291:245 247. therapy (letter). Presse Med 1992; 21:994. 25. Clason AE, MacLeod DAD, Elton RA. Clinical factors in the predicspective 9. Roul JL, Rupert A, Siproudhis L. Gastric ulcer and cholestasis tion of further haemorrhage or mortality in acute upper gastroinfollowing injection therapy for bleeding duodenal ulcer (letter). testinal haemorrhage. Br J Surg 1986; 73:985 987. Endoscopy 1991; 23:351 352. 26. Leung JWC, Chung SCS. Endoscopic injection of adrenaline in 10. Levy J, Khakoo S, Barton R, Vicary R. Fatal injection sclerotherapy bleeding peptic ulcers. Gastrointest Endosc 1987; 33:73 75. of bleeding ulcer (letter). Lancet 1991; 337:504. 27. Lai KH, Peng SN, Guo WS, Lee FY, Chang FY, Malik U, Wang JY, 11. Lin HJ, Perng CL, Lee FY. Endoscopic injection for the arrest of Lo GH, Cheng JS, Lee SD, Tsai YT. Endoscopic injection for the peptic ulcer haemorrhage. Gastrointest Endosc 1993; 39:15 treatment of bleeding ulcers: local tamponade or drug effect? 19. Endoscopy 1994; 26:338 341. 12. Rajgopal C, Lessel A, Palmer KR. Mechanism of action of injection 28. Villaneuva C, Balanzo J, Espinos JC, Fabrega E, Sainz S, Gonzales therapy for bleeding peptic ulcer. Br J Surg 1992; 79:782 D, Vilardell F. Endoscopic injection therapy of bleeding ulcer: a 784. prospective and randomised comparison of adrenaline or with 13. Chung SCS, Leung FW, Leung JWC. Is vasoconstriction the mech- polidocanol. J Clin Gastroenterol 1993; 17:195 200. anism of haemostasis in bleeding ulcers injected with epinephrine? 29. Loizou LA, Bown SG. Endoscopic treatment for bleeding peptic A study using reflectanc spectrophotometry. Gastrointest ulcer: randomised comparison of adrenaline injection and adren- Endosc 1988; 34:174 175. aline injection plus Nd YAG laser photocoagulation. Gut 1991; 14. Rutgeerts P, Geboes K, Vantrappen G. Experimental studies of 32:1100 1103. injection therapy for severe non-variceal bleeding in dogs. Gastroenterology 30. Kram H, Nugent P, Reaben B, Shoemaker W. Fibrin glue sealant 1989; 97:610 621. of polytetrofluorethylen vascular graft anastomosis. J Vasc Surg 15. Juszkiewicz P, Wajda Z, Dobosz M, Babicki A, Marczewski R. 1988; 8:563 568. The role of endoscopic thrombin injections in the treatment of 31. Snoble J, Van Buuren HR, Van Blackestein M. Endoscopic injection gastroduodenal bleeding. S Afr J Surg 1993; 31:98 102. therapy using thrombin: an effective and safe method of 16. Moreto M, Zaballa M, Suarez MJ, Ibanez S, Ojembarrena E, Cas- controlling oesophago-gastric variceal bleeding (abstr). Gastroenterology tillo JM. Endoscopic local injection of ethanolamine oleate and 1992; 102:A891. thrombin as an effective treatment for bleeding duodenal ulcer: 32. Bennet B, Dewson AA, Gibson BS, Hepplestone A, Lowe GDO, a controlled trial. Gut 1992; 33:456 459. Ludlam CA, Mayne EE, Taylor T. Study of viral safety of Scottish 17. Benedetti G, Sablich R, Lacchin T. Endoscopic injection sclero- National Blood Transfusion Service factor VIII/IX concentrate. therapy in non variceal upper gastrointestinal bleeding. Surg Endosc Transfus Med 1993; 3:295 298. 1991; 5:28 30. 18. Balanzo J, Villanueva S, Sainz J, Espinos JC, Mendez C, Guarner C, Vilardell F. Injection therapy of bleeding peptic ulcer. Random- Received December 26, 1995. Accepted May 27, 1996. ised trial using adrenaline and thrombin. Endoscopy 1990; 20: Address requests for reprints to: Adam K. Kubba, M.D., Gastroin- 157 159. testinal Annexe, Western General Hospital, Crewe Road, Edinburgh, 19. Koyama T, Fujimoto K, Iwakiri R, Sakata H, Sakata Y, Yamaoka Scotland, EH4 2XU. Fax: (44) 31-537-1006.