Peptic ulcers remain the most common cause of upper

CLINICAL GASTROENTEROLOGY AND HEPATOLOGY 2009;7:33 47 STATE OF THE ART Endoscopic Therapy for Bleeding Ulcers: An Evidence-Based Approach Based on Meta-Analyses of Randomized Controlled Trials LOREN LAINE* and KENNETH R. MCQUAID, *Division of Gastrointestinal and Liver Diseases, Keck School of Medicine, University of Southern California, Los Angeles, California; Veterans Affairs Medical Center, San Francisco, California; and the Department of Medicine, University of California San Francisco, San Francisco, California See CME exam on page 2. See Cooper G et al on page 108 for companion article in the January 2009 issue of Gastroenterology. The aim of this study was to determine appropriate endoscopic treatment of patients with bleeding ulcers by synthesizing results of randomized controlled trials. We performed dual independent bibliographic database searches to identify randomized trials of thermal therapy, injection therapy, or clips for bleeding ulcers with active bleeding, visible vessels, or clots, focusing on results from studies without second-look endoscopy and re-treatment. The primary end point was further (persistent plus recurrent) bleeding. Compared with epinephrine, further bleeding was reduced significantly by other monotherapies (relative risk [RR], 0.58 [95% CI, 0.36 0.93]; number-needed-to-treat [NNT], 9 [95% CI, 5 53]), and epinephrine followed by another modality (RR, 0.34 [95% CI, 0.23 0.50]; NNT, 5 [95% CI, 5 7]); epinephrine was not significantly less effective in studies with second-look and re-treatment. Compared with no endoscopic therapy, further bleeding was reduced by thermal contact (heater probe, bipolar electrocoagulation) (RR, 0.44 [95% CI, 0.36 0.54]; NNT, 4 [95% CI, 3 5]) and sclerosant therapy (RR, 0.56 [95% CI, 0.38 0.83]; NNT, 5 [95% CI, 4 13]). Clips were more effective than epinephrine (RR, 0.22 [95% CI, 0.09 0.55]; NNT, 5 [95% CI, 4 9]), but not different than other therapies, although the latter studies were heterogeneous, showing better and worse results for clips. Endoscopic therapy was effective for active bleeding (RR, 0.29 [95% CI, 0.20 0.43]; NNT, 2 [95% CI, 2 2]) and a nonbleeding visible vessel (RR, 0.49; [95% CI, 0.40 0.59]; NNT, 5 [95% CI, 4 6]), but not for a clot. Bolus followed by continuous-infusion proton pump inhibitor after endoscopic therapy significantly improved outcome compared with placebo/no therapy (RR, 0.40 [95% CI, 0.28 0.59]; NNT, 12 [95% CI, 10 18]), but not compared with histamine 2 -receptor antagonists. Thermal devices, sclerosants, clips, and thrombin/fibrin glue appear to be effective endoscopic hemostatic therapies. Epinephrine should not be used alone. Endoscopic therapy should be performed for ulcers with active bleeding and nonbleeding visible vessels, but efficacy is uncertain for clots. Bolus followed by continuous-infusion intravenous proton pump inhibitor should be used after endoscopic therapy. Peptic ulcers remain the most common cause of upper gastrointestinal bleeding, accounting for approximately 30% to 60% of patients hospitalized for upper gastrointestinal bleeding. 1 5 Upper endoscopy is crucial in the diagnosis, stratification, and management of patients with bleeding ulcers, and endoscopic hemostatic therapy is the mainstay of treatment in patients with endoscopic findings that predict an increased risk of further bleeding. A variety of modalities currently are used for endoscopic therapy of ulcer bleeding, including thermal (eg, bipolar electrocoagulation, heater probe, argon plasma coagulation), injection (eg, epinephrine, sclerosants, thrombin/fibrin glue), mechanical (eg, clips), and combinations of these modalities. Numerous randomized controlled trials of endoscopic treatment for bleeding ulcers have been published. However, integrating the results of these studies is made difficult by variations in the populations enrolled, differences in methods, variations in definitions and end points, evaluation of multiple hemostatic modalities or combination of modalities, and use of mandatory second-look endoscopies with repeat therapy in some studies. Nevertheless, making sense of this large body of literature is extremely important for gastroenterologists because treatment of upper gastrointestinal bleeding remains the most common and important emergency procedure for our subspecialty. We therefore sought to develop an evidence-based approach to the endoscopic treatment of bleeding ulcers by performing a systematic review and meta-analyses to address clinically relevant questions regarding endoscopic treatment of bleeding ulcers. Our aims were to assess currently used techniques for the following: (1) to determine which techniques are better than no endoscopic therapy; (2) to compare different standard therapies; (3) to compare single and dual therapies; (4) to determine efficacy related to stigmata of hemorrhage; (5) to determine Abbreviations used in this paper: BPEC, bipolar electrocoagulation; CI, confidence interval; HP, heater probe; H2RA, histamine 2 -receptor antagonist; NNT, number-needed-to-treat; PPI, proton pump inhibitor; RR, relative risk. 2009 by the AGA Institute 1542-3565/09/$36.00 doi:10.1016/j.cgh.2008.08.016

34 LAINE AND MCQUAID CLINICAL GASTROENTEROLOGY AND HEPATOLOGY Vol. 7,. 1 separately results of studies that did not use routine secondlook endoscopy with re-treatment (because these studies allow a better assessment of the efficacy of primary endoscopic therapy and reflect the current practice in the United States); and (6) to determine the utility of proton pump inhibitor (PPI) therapy as an adjunct to endoscopic therapy. Methods The study type we defined for inclusion was randomized controlled trials that compared one of the included interventions with no therapy or with a second included intervention. Studies with pseudorandomization (eg, day of the week, odd vs even patient number) were not included. The population defined for inclusion was patients presenting with bleeding from a gastric or duodenal ulcer and found to have active bleeding, a nonbleeding visible vessel, or a clot at endoscopy. Because patients with flat spots and clean-based ulcers have a very low rate of clinically important recurrent bleeding 6 and no investigators or publications suggest that endoscopic therapy be provided to such patients, we excluded studies including such patients unless data on patients with higher-risk stigmata could be separated from those with lowerrisk findings. We also excluded studies with bleeding lesions other than gastric or duodenal ulcers unless data for the gastric and duodenal ulcers were provided separately. Studies including only patients in a selected population (eg, elderly, inpatients, Helicobacter pylori infection) also were excluded. The interventions defined for inclusion were those in current use for bleeding ulcers: thermal contact devices (heater probe [HP] and bipolar electrocoagulation [BPEC]), argon plasma coagulation, injection (epinephrine, sclerosants including alcohol, thrombin, and fibrin glue), and clips. We did not include laser therapy, monopolar electrocoagulation, microwave, injection of water or saline, or injection of tissue adhesives. Studies that allowed the endoscopist to choose more than one technique (eg, clips or bands, clips epinephrine) in a treatment arm were not included. To assess the utility of PPI therapy after endoscopic therapy, we also included the intervention of PPI therapy versus control (no therapy, placebo, or histamine 2 - receptor antagonist [H2RA]) in patients with bleeding ulcers and active bleeding, visible vessels, or clots who all received one of our defined endoscopic therapies. The outcomes defined for inclusion were initial hemostasis in active bleeding, further bleeding, need for surgery, need for urgent intervention (subsequent endoscopic treatment with the same or different therapy, surgery, or interventional radiology), mortality, and complications (perforation, induction of bleeding that could not be controlled with the study-mandated endoscopic therapy). The primary end point was further bleeding, which included persistent bleeding and recurrent bleeding. We believe further bleeding provides the best overall assessment of a modality s hemostatic efficacy because it assesses both its initial hemostatic efficacy and its ability to prevent subsequent bleeding. Bleeding required clinical manifestations; identification of bleeding on scheduled second-look endoscopy without clinical evidence of bleeding was not included. We performed separate analyses for studies that did and did not include scheduled second-look endoscopies at which re-treatment was allowed or provided; if second-look endoscopies were performed but no indication of re-treatment at the second-look Table 1. Methodologic Quality Scoring System for Endoscopic Hemostatic Randomized Trials Methodologic Quality Criteria (1 point given for each) Randomized stated in article Randomization method described Concealed allocation Full accounting of all randomized subjects Clear, objective definitions of initial hemostasis or rebleeding Calculations performed for sample size determination Blinded observer managing/assessing patient after endoscopic therapy procedure was reported, the study was not considered a secondlook trial for purposes of our analysis. L.L. and K.R.M. independently performed searches of 2 bibliographic databases: MEDLINE (1950 March, 2008) and the Cochrane Central Register of Controlled Trials. The MeSH search terms gastrointestinal hemorrhage or hemostasis, endoscopic were used for both databases. For MEDLINE, the search was limited to the MeSH term clinical trial or randomized controlled trial, and also to the MeSH term metaanalysis or the non-mesh term systematic review to check these sources for randomized controlled trials missed by our search strategy. language restriction was incorporated. Independent review of titles and abstracts was performed by the 2 authors to identify potentially relevant articles for full review. Any article considered potentially relevant by one of the reviewers was reviewed in full. The authors independently reviewed the articles to identify all studies meeting inclusion criteria. Any disagreements were resolved by discussion and consensus. A data abstraction sheet to record information on study methodology, patient characteristics, and predefined outcomes was developed, pilot tested on known articles, and revised before formal use. The two authors independently completed a data abstraction sheet for each included article and then reviewed the other s sheet. Any disagreements were resolved by discussion and consensus. Studies were assessed for methodologic quality using a 7-point scoring system that incorporated the Jadad et al 7 criteria plus additional criteria (Table 1). We did not include double-blind in our scale because endoscopic treatment trials have not blinded the endoscopist performing the procedure. This is a methodologic shortcoming so a high score with our scale still falls short of the highest methodologic quality because of the absence of double-blinding. We did include one point if providers following up the patient after endoscopic treatment were blinded to the treatment. This methodologic quality score has not been previously used or validated. For studies comparing PPI with no PPI therapy after endoscopic therapy the quality score was modified to an 8-point scale: the blinded observer criterion was replaced by double-blind and mechanism of blinding described. All included studies were randomized, per our inclusion criteria, and therefore had a minimum score of 1. When only one study provided head-to-head comparison for an end point, the Fisher exact test was used for statistical analysis. Meta-analysis was performed when at least 2 studies provided head-to-head comparisons of the therapies of interest. A summary relative risk (RR) with 95% confidence intervals (CIs) was calculated with Review Manager software, version

January 2009 ENDOSCOPIC THERAPY FOR BLEEDING ULCERS 35 Table 2. Meta-Analyses for Epinephrine Monotherapy Versus Other Monotherapy and Versus Epinephrine Dual Therapy Comparison End point Number of comparisons RR (95% CI) NNT (95% CI) Monotherapy (BPEC, fibrin glue, clip) Further bleeding 3 50,54,60 0.58 (0.36 0.93) 9 (5 53) v epinephrine (no second-look studies) Surgery 3 50,54,60 0.44 (0.20 0.98) 10 (7 250) Urgent intervention 2 50,60 0.73 (0.44 1.21) Mortality 3 50,54,60 0.37 (0.10 1.37) Epinephrine second therapy (BPEC, Further bleeding 7 38,45,54,58 61 0.34 (0.23 0.50) 5 (5 7) sclerosant, thrombin, clip Surgery 6 38,45,54,59 61 0.33 (0.17 0.66) 13 (10 25) v epinephrine (no second-look studies) Urgent intervention 4 45,59 61 0.39 (0.24 0.62) 6 (5 9) Mortality 6 38,45,54,59 61 0.52 (0.23 1.16) Epinephrine second therapy (HP, BPEC, Further bleeding 7 52,53,55 57,62,64 0.89 (0.65 1.21) sclerosant, thrombin, fibrin glue) Surgery 7 52,53,55 57,62,64 0.81 (0.57 1.15) v epinephrine (second-look studies) Urgent intervention 2 53,64 1.24 (0.61 2.50) Mortality 7 52,53,55 57,62,64 0.71 (0.39 1.30) 4.2.7 (Cochrane Collaboration, Oxford, UK). Heterogeneity was calculated using the chi-square test with n-1 degrees of freedom, where n represented the number of studies contributing to the meta-analysis. Significant heterogeneity was defined as a P value of.10 or less. A fixed-effect model was used when significant heterogeneity was absent, and a random-effects model was used when significant heterogeneity was noted. When a significant difference in an end point was identified, we determined the pooled incidence of end points in the control groups of the individual meta-analysis and multiplied this incidence by the relative risk reduction to estimate the absolute risk reduction and determine the number-needed-to-treat (NNT). Results The MEDLINE search produced 2044 citations and the Cochrane Central Registry search produced 1432 citations. A review of titles and abstracts led to 92 articles considered to be potentially relevant for full review. A review of the 92 full articles revealed 74 articles (75 studies) for inclusion. 8 81 Ten studies of endoscopic therapy were excluded for the following reasons: inability to distinguish outcomes in patients with bleeding from peptic ulcers versus bleeding from other sources, 82 85 inability to distinguish outcomes in patients with high-risk versus low-risk stigmata, 83,84,86,87 lack of 2 standard endoscopic treatment arms, 88 lack of uniform/standardized endoscopic therapy in one arm, 89,90 and not a clinical trial. 91 Eight studies assessing the efficacy of PPIs after endoscopic therapy were excluded for the following reasons: preliminary report of another study, 92 not studies of endoscopic therapy, 93 lack of inclusion of a standard endoscopic treatment arm, 94 inability to distinguish outcomes in patients who did and did not receive endoscopic therapy, 95 98 and nonstandard randomization scheme. 99 A review of the references from 16 systematic reviews/ meta-analyses identified in our search 91,100 114 yielded 2 additional articles that were reviewed, but neither met inclusion criteria. 115,116 In addition, communication with investigators in the field of ulcer bleeding revealed 2 studies in press or preparation. 117 119 Additional clarifying information was provided by Drs H. J. Lin, J. R. Saltzman, L. Cipolletta, L. Laine, and C. van Rensburg. Details on the component studies are provided in Appendices 1 and 2. Epinephrine Versus Therapy One trial reported 100% initial hemostasis of active bleeding with epinephrine injection. 15 The RR of further bleeding leading to surgery with epinephrine versus medical therapy was 0.36 (95% CI, 0.14 0.88); however, this assessment was confounded by the fact that patients in the epinephrine group had a second-look endoscopy with re-treatment of active bleeding (6 of 34 were re-treated). A trial in 40 patients with nonbleeding visible vessels did not show a significant reduction in further bleeding (RR, 0.50; 95% CI, 0.23 1.07) or surgery (RR, 0.29; 95% CI, 0.07 1.21) with epinephrine. 45 Epinephrine Versus Other Monotherapy Six trials assessed epinephrine versus other monotherapy with HP, 14,63 BPEC, 60 fibrin glue, 31,50 or clip. 54 significant difference was found in initial hemostasis of active bleeding between monotherapy and epinephrine (RR, 0.93; 95% CI, 0.85 1.01). Assessment of subsequent outcomes was confounded in 3 of these trials owing to second-look with retreatment. 14,31,63 Meta-analysis of the studies without secondlook 50,54,60 revealed significantly less further bleeding and surgery with monotherapy without a significant decrease in urgent intervention or mortality (Table 2). The one second-look study that allowed assessment of further bleeding 14 found no significant difference between HP and epinephrine (further bleeding: RR, 1.29; 95% CI, 0.69 2.40). Epinephrine Versus Epinephrine Second Modality In 15 trials epinephrine was compared with epinephrine plus a second modality: sclerosant, 45,53,55,56,58,61,63,64 BPEC, 60 HP, 57 thrombin, 52,59 fibrin glue, 62 or clips (but with epinephrine injected after rather than before the clips). 38,54 Assessment was confounded by second look plus re-treatment in 8 trials. 52,53,55 57,62 64 Meta-analyses of the 7 studies without second look plus re-treatment 38,45,54,58-61 showed a significant benefit of adding the second modality for further bleeding, surgery, and urgent intervention (Table 2). Significant benefit was seen with dual therapy in those with active bleeding (further bleeding: RR, 0.40; 95% CI, 0.23 0.68) 45,54,59,61 and those with nonbleeding visible vessels (further bleeding: RR, 0.35; 95% CI, 0.14 0.89). 45,54,59 Results of the 7 second-look studies allowing assessment of outcomes beyond initial hemosta-

36 LAINE AND MCQUAID CLINICAL GASTROENTEROLOGY AND HEPATOLOGY Vol. 7,. 1 Table 3. Meta-Analyses for Endoscopic Hemostatic Therapies Versus Endoscopic Therapy (Excluding Trials With Second- Look Endoscopies Specifying Re-Treatment) Comparison End point Number of comparisons RR (95% CI) NNT (95% CI) Thermal contact (HP, BPEC) Further bleeding 15 11,18,20,21,23,27,28,33 35,39,41,77 0.44 (0.36 0.54) 4 (3 5) v no endoscopic therapy Surgery 14 11,18,20,21,23,27,28,33 35,41,77 0.39 (0.27 0.55) 8 (6 10) Urgent intervention 14 11,18,20,21,23,27,28,33 35,41,77 0.50 (0.37 0.66) 8 (6 11) Mortality 13 11,18,20,21,23,27,28,33 35,41 0.58 (0.34 0.98) 33 (21 1000) Epinephrine thermal contact Further bleeding 3 65,71,72 0.10 (0.03 0.40) 6 (5 9) v no endoscopic therapy Surgery 3 65,71,72 0.71 (0.19 2.62) Urgent intervention 2 71,72 0.16 (0.04 0.69) 6 (5 16) Mortality 3 65,71,72 0.80 (0.27 2.37) Sclerosant Further bleeding 3 25,34,44 0.56 (0.38 0.83) 5 (4 13) v no endoscopic therapy Surgery 3 25,34,44 0.24 (0.09 0.64) 7 (6 15) Urgent intervention 3 25,34,44 0.65 (0.43 0.98) 7 (5 125) Mortality 3 25,34,44 0.18 (0.05 0.68) 9 (8 24) Epinephrine sclerosant Further bleeding 6 8,42,44,45,66,67 0.60 (0.36 1.00) a v no endoscopic therapy Surgery 6 8,42,44,45,66,67 0.61 (0.22 1.73) a Urgent intervention 6 8,42,44,45,66,67 0.62 (0.45 0.86) 7 (5 20) Mortality 6 8,42,44,45,66,67 0.80 (0.40 1.61) a Statistical heterogeneity (P.10) and random-effects model used. sis 52,53,55 57,62,64 showed no benefit of adding a second modality to epinephrine (Table 2). Thermal Contact Therapy Versus Therapy Thermal contact therapy with HP or BPEC versus no endoscopic therapy was assessed in 15 comparisons (10 HP, 5 BPEC) in 14 studies 11,18,20,21,23,27,28,33 35,39,41,77 (1 article comprised 2 studies 20 ); none reported second-look with re-treatment. Initial hemostasis for active bleeding, assessed in 4 trials, 18,21,28,34 was significantly better with thermal contact (RR, 11.70; 95% CI, 5.15 26.56). Further bleeding, surgery, urgent intervention, and mortality also were decreased significantly (Table 3). HP and BPEC were directly compared in 3 trials 20,22,32 : RR for further bleeding with HP versus BPEC was 1.01 (95% CI, 0.57 1.80). Epinephrine plus thermal therapy versus no endoscopic therapy was compared in 3 studies 65,71,78 : combination therapy was significantly better than no therapy for further bleeding and urgent intervention, but not for surgery or mortality (Table 3). Injection of Sclerosants Versus Therapy Sclerosant alone (absolute alcohol) was compared with no therapy in 4 trials, 25,34,44,72 one of which included secondlook endoscopy and re-treatment. 72 Initial hemostasis for active bleeding was evaluated in only 1 study, 34 with initial hemostasis in 13 (46%) of 28 treated patients versus 2 (8%) of 24 control patients (RR, 5.57; 95% CI, 1.39 22.26). In the 3 trials without second-look and re-treatment, absolute alcohol injection was superior to no therapy for further bleeding, surgery, urgent intervention, and mortality (Table 3). The 6 trials that compared epinephrine plus sclerosant versus no therapy 8,42,44,45,66,67 showed a nonsignificant trend in favor of epinephrine plus sclerosant, although significant heterogeneity was present (Table 3). Urgent intervention was significantly less frequent with epinephrine plus sclerosant (Table 3). trials compared sclerosant alone with epinephrine alone. Thermal Contact Versus Sclerosant Injection Thermal contact was compared with sclerosant alone 26,34,36,49,51 or epinephrine plus sclerosant 13,20,21,37,63 in 10 studies; 3 were confounded by second look with re-treatment. 13,37,63 trial compared epinephrine plus thermal contact versus a sclerosant-containing regimen. Results for initial hemostasis with thermal versus sclerosant alone showed no significant difference (RR, 1.27; 95% CI, 0.93 1.75), but were heterogeneous (P.0001). 26,34,36,49,51 Among other outcomes, only urgent intervention was significantly different, with thermal contact superior to sclerosant (Table 4). When thermal contact was compared with sclerosants with and without epinephrine, initial hemostasis results were again Table 4. Meta-Analyses Comparing Thermal Contact With Sclerosants (Excluding Trials With Second-Look Endoscopies Specifying Re-Treatment) Comparison End point Number of comparisons RR (95% CI) NNT (95% CI) Thermal contact Further bleeding 5 26,34,36,49,51 0.69 (0.47 1.01) v sclerosant Surgery 4 26,34,49,51 1.74 (0.83 3.67) Urgent intervention 3 26,34,36 0.52 (0.31 0.88) 7 (5 29) Mortality 4 26,34,49,51 1.60 (0.57 4.52) Thermal contact Further bleeding 7 20,21,26,34,36,49,51 0.72 (0.51 1.02) v sclerosant epinephrine Surgery 6 20,21,26,34,49,51 1.57 (0.77 3.20) Urgent intervention 5 20,21,26,34,36 0.60 (0.37 0.97) 10 (7 143) Mortality 6 20,21,26,34,49,51 1.03 (0.43 2.47)

January 2009 ENDOSCOPIC THERAPY FOR BLEEDING ULCERS 37 heterogeneous (P.0001) without a significant difference (RR, 1.24; 95% CI, 0.97 1.58). 21,26,34,36,37,49,51,63 Among other outcomes in studies without second-look and re-treatment, only urgent intervention was significantly different (Table 4). The 2 evaluable studies with second-look 13,37 had nearly identical outcomes for thermal versus epinephrine plus sclerosant (further bleeding: RR, 1.14; 95% CI, 0.65 2.02). Injection Plus Thermal Contact Versus Thermal Contact Meta-analysis of the 2 studies comparing epinephrine injection preceding BPEC with BPEC monotherapy 10,60 revealed the following RRs: initial hemostasis, 1.31 (95% CI, 1.02 1.68); further bleeding, 0.35 (95% CI, 0.18 0.71); surgery, 0.33 (95% CI, 0.07 1.57); urgent intervention, 0.37 (95% CI, 0.18 0.74); and mortality, 0.49 (95% CI, 0.09 2.60). The rates of initial hemostasis in the thermal monotherapy arms were 68% and 78%, and the rates of further bleeding in the thermal monotherapy arms were 25% and 34%. A large study of 247 patients, comparing HP plus thrombin versus heater probe plus placebo, was the only double-blind study of endoscopic therapy and the highest methodologic quality study we reviewed. However, it was not included in the meta-analysis because it included 9 patients with esophageal ulcers. 82 This comparison of injection plus thermal contact versus thermal showed no suggestion of a difference (further bleeding: RR, 1.03; [95% CI, 0.61 1.77]). Clips study was identified that compared clip with no endoscopic therapy. Two studies compared clips with epinephrine, 38,54 with the latter study comparing epinephrine with clip followed by epinephrine. Meta-analysis of these studies revealed that further bleeding and surgery were decreased with clips, but mortality was not significantly different (Table 5). Another randomized trial comparing clips with injection of water was not included in our meta-analysis based on our inclusion criteria (injection of water not included), 88 but showed clips superior to water injection (further bleeding: RR, 0.34; 95% CI, 0.12 0.97). Six studies compared clips with other standard therapies: HP, 16,29,30 epinephrine plus BPEC, 46 or sclerosants with 19 or without 47 epinephrine. Two of these were confounded by second-look and re-treatment. 19,47 Meta-analysis of studies without second-look and re-treatment showed that clips achieved initial hemostasis less frequently than other therapies (RR, 0.78; 95% CI, 0.64 0.95). 16,29,30 Other outcomes are shown in Table 5. Although the summary estimate revealed no significant difference in further bleeding, marked heterogeneity was present. Clips were significantly better in one trial, 16 significantly worse in 2 trials, 29,30 and not significantly different in the fourth trial, 46 making aggregation of the individual studies problematic. Among the 2 studies with second-look and re-treatment, one showed no significant difference for clips versus alcohol (further bleeding: RR, 0.67; 95% CI, 0.20 2.19) 47 whereas the other showed poorer outcome with clips versus epinephrine plus polidocanol (further bleeding: RR, 2.53; 95% CI, 1.01 6.32). 19 Three studies assessed injection plus clips versus clips alone. The dual therapies were clips followed by epinephrine in hypertonic saline, 54 injection of epinephrine and polidocanol followed by clips, 19 or injection of alcohol either before or after clip application. 47 Meta-analysis of the 2 studies without second-look and re-treatment 19,54 revealed no significant difference in further bleeding (Table 5); other outcomes were not provided in both studies. The study with second-look and re-treatment 47 also showed no significant difference in further bleeding for combined versus monotherapy (RR, 0.75; 95% CI, 0.18 3.15). Thrombin/Fibrin Glue Thrombin or fibrin glue was compared with no therapy in one trial. 24 Thrombin significantly decreased further bleeding and urgent intervention (RR, 0.10; 95% CI, 0.01 0.72) without significant decrease in surgery or death. Fibrin glue was compared with epinephrine injection in 2 studies, 31,50 although one study had second look with re-treatments at different times in each group so it could be evaluated only for initial hemostasis. 31 Initial hemostasis with fibrin glue was reported in 27 of 28 50 and 15 of 15 31 versus 27 of 29 and 18 of 18 with epinephrine injection. Further bleeding was not significantly lower with fibrin glue (RR, 0.49; 95% CI, 0.23 1.07) in the one evaluable study. 50 Epinephrine plus thrombin was significantly better than epinephrine in the single trial without second-look and retreatment (further bleeding: RR, 0.21; 95% CI, 0.06 0.71). 59 Meta-analysis of the 2 second-look trials comparing epinephrine plus thrombin or fibrin glue versus epinephrine alone showed no significant difference (further bleeding: RR, 1.11; 95% CI, 0.62 1.99). 52,62 Fibrin glue was not significantly different than polidocanol in a study 9 with second-look and retreatment (further bleeding: RR, 0.65; 95% CI, 0.17 2.53), nor was epinephrine plus fibrin glue significantly different from Table 5. Meta-Analyses Assessing Clips (Excluding Trials With Second-Look Endoscopies Specifying Re-Treatment) Comparison End point Number of comparisons RR (95% CI) NNT (95% CI) Clips Further bleeding 2 38,54 0.22 (0.09 0.55) 5 (4 9) v epinephrine a Surgery 2 38,54 0.22 (0.06 0.83) 11 (9 50) Mortality 2 38,54 1.68 (0.23 12.45) Clips Further bleeding 4 16,29,30,46 1.31 (0.36 4.75) b v other standard therapies Surgery 4 16,29,30,46 0.85 (0.32 2.26) Urgent intervention 2 16,30 1.17 (0.09 14.46) b Mortality 4 16,29,30,46 1.16 (0.38 3.52) Clips injection v clips alone Further bleeding 2 19,54 0.92 (0.48 1.77) a One trial compared clips followed by epinephrine versus epinephrine. b Statistical heterogeneity (P.10) and random-effects model used.

38 LAINE AND MCQUAID CLINICAL GASTROENTEROLOGY AND HEPATOLOGY Vol. 7,. 1 Table 6. Meta-Analyses Assessing Endoscopic Therapy Versus Endoscopic Therapy Related to Stigmata of Hemorrhage (Excluding Trials With Second-Look Endoscopies Specifying Re-Treatment) Stigmata End point Number of comparisons RR (95% CI) Pooled rate: no endoscopic therapy n/n (%) NNT (95% CI) Active bleeding Further bleeding 7 8,18,21,28,39,41 0.29 (0.20 0.43) 77/95 (81%) 2 (2 2) Surgery 4 8,21,28 0.25 (0.13 0.50) 33/47 (70%) 2 (2 3) Urgent intervention 4 8,21,28 0.31 (0.17 0.58) 33/47 (70%) 2 (2 3) Mortality 5 8,18,21,28 1.28 (0.26 6.21) 2/68 (3%) nbleeding visible vessel Further bleeding 20 20,23 25,27,35,39,41,42,44,45,66,67,77,78 0.49 (0.40 0.59) 224/579 (39%) 5 (4 6) Surgery 18 20,23 25,27,35,42,44,45,66,67,77,78 0.41 (0.24 0.71) a 91/508 (18%) 9 (7 19) Urgent intervention 18 20,23 25,27,35,42,44,45,66,67,77,78 0.49 (0.38 0.63) 150/508 (30%) 7 (5 9) Mortality 16 20,23 25,27,35,42,44,45,66,67 0.62 (0.36 1.06) 30/441 (7%) Clot Further bleeding 5 39,41,65,71,78 0.31 (0.06 1.77) a 26/118 (22%) Surgery 3 65,71,78 0.47 (0.10 2.26) 6/76 (8%) Mortality 2 65,71 0.90 (0.23 3.58) 5/52 (10%) a Statistical heterogeneity (P.10) and random-effects model used. epinephrine plus polidocanol in a study with second-look and re-treatment (further bleeding: RR, 0.87; 95% CI, 0.61 1.24). 43 Argon Plasma Coagulation Argon plasma coagulation was compared with epinephrine plus polidocanol 48 with no significant differences noted (further bleeding: RR, 0.80; 95% CI, 0.23 2.81). Argon plasma coagulation was not significantly different from HP (further bleeding: RR, 0.76; 95% CI, 0.24 2.44) in a second-look study 17 and epinephrine plus argon plasma coagulation was not significantly different from epinephrine plus HP (further bleeding: RR, 0.72; 95% CI, 0.42 1.24) in a second-look study. 12 Endoscopic Therapy Versus Endoscopic Therapy Based on Stigmata of Active Bleeding, nbleeding Visible Vessel, or Adherent Clot in Studies Without Second- Look Endoscopy and Re-Treatment Specified Endoscopic therapy was significantly better than no therapy for active bleeding in all outcomes except mortality (Table 6). Initial hemostasis was significantly better in all 7 comparisons and the summary RR was 8.80 (95% CI, 4.93 15.71). 8,18,21,28,34 Endoscopic therapy also was significantly better than no therapy for nonbleeding visible vessels in all outcomes except mortality (Table 6). Comparison of endoscopic therapy versus no endoscopic therapy for clots showed no significant benefit in any outcome (Table 6). However, these studies had markedly variable results and there was significant heterogeneity for further bleeding (P.03). One study had only 5 patients; of the other 4 studies, 2 favored endoscopic therapy (1 of 21 vs 12 of 35; RR, 0.14; 95% CI, 0.02 0.99 65 ; and 0 of 15 vs 6 of 17; RR, 0.09; 95% CI, 0.01 1.42; P.02 71 ), one study found no significant difference for endoscopic therapy versus control (4 of 18 vs 5 of 39; RR, 1.73; 95% CI, 0.53 5.70), 41 and 1 study, in which medical therapy included bolus followed by continuous infusion of intravenous PPI (after vigorous irrigation of the clot), revealed no further bleeding during hospitalization with or without endoscopic therapy (0 of 15 vs 0 of 24). 78 A study 72 included in a prior meta-analysis of clots 101 was not included here because the number used for rebleeding in that meta-analysis included patients with only endoscopic evidence of rebleeding (without clinical manifestations). Complications of Endoscopic Therapy Analysis of controlled trials revealed that complications of perforation and induced bleeding were more common with endoscopic therapy than no endoscopic therapy (RR, 2.12; 95% CI, 0.79 5.70), 11,15,18,20,21,23 25,27,28,33 35,39 42,44,45,52,65 67,71,72,77,78 although the difference was not statistically significant. The pooled rates for complications in this comparison were 8 of 1044 (0.8%; 95% CI, 0.3 1.5) versus 1 of 931 (0.1%; 95% CI, 0.003 0.6). Pooled complications rates from the studies included in the systematic review for different modalities are epinephrine alone: 2 induced bleeding episodes in 958 cases (0.2%; 95% CI, 0.03 0.8); sclerosant epinephrine: 6 perforations and 1 induced bleeding in 1339 cases (0.5%; 95% CI, 0.2 1.1); HP epinephrine: 9 perforations (4 on second treatment and 2 others in second-look studies that did not state if these cases occurred on re-treatment) and 2 induced bleeding in 1070 cases (1.0%; 95% CI, 0.5 1.8); BPEC epinephrine: 2 perforations and 1 induced bleeding in 580 cases (0.5%; 95% CI, 0.1 1.5); thrombin or fibrin glue epinephrine: 3 perforations in 553 cases (0.5%; 95% CI, 0.1 1.6); clips epinephrine: no complications in 373 cases (0; 95% CI, 0 1.0). Pooling of complications for all of these modalities combined revealed a rate of 26 of 4873 (0.5%; 95% CI, 0.4 0.8). Proton Pump Inhibitor Therapy as an Adjunct to Endoscopic Therapy Four studies comparing bolus followed by continuous infusion of intravenous PPI with placebo or no treatment 68,74,81,118,119 showed a significant benefit in further bleeding, surgery, urgent intervention, and mortality (Table 7). An additional 3 studies compared this PPI regimen with H2RA therapy, 70,75,117 with no significant differences seen (Table 7). Five studies evaluated intermittent PPI therapy given as intravenous and/or oral boluses versus no therapy or placebo. 49,68,69,73,80 We included intravenous and oral bolus studies together because pharmacodynamic studies indicated no significant difference in antisecretory effect beyond the first hour of

January 2009 ENDOSCOPIC THERAPY FOR BLEEDING ULCERS 39 Table 7. Meta-Analyses for PPI Therapy as an Adjunct After Endoscopic Hemostatic Therapy Comparison End point Number of comparisons RR (95% CI) NNT (95% CI) Intravenous PPI: bolus plus continuous infusion v placebo Further bleeding 4 68,74,81,118,119 0.40 (0.28 0.59) 12 (10 18) Surgery 3 74,81,118,119 0.43 (0.24 0.76) 28 (21 67) Urgent intervention 3 68,74,81 0.31 (0.18 0.53) 8 (7 12) Mortality 4 68,74,81,118,119 0.41 (0.20 0.84) 45 (33 167) Intravenous PPI: bolus plus continuous infusion v H2RA Further bleeding 3 70,75,117 0.63 (0.37 1.08) Surgery 2 75,117 1.11 (0.49 2.49) Urgent intervention 2 75,117 0.37 (0.04 3.36) a Mortality 3 70,75,117 0.62 (0.20 1.96) PPI: oral or intermittent intravenous bolus v placebo Further bleeding 5 49,68,69,73,80 0.53 (0.35 0.78) 10 (7 21) Surgery 4 49,69,73,80 0.62 (0.25 1.53) Urgent intervention 2 68,80 0.43 (0.12 1.62) Mortality 5 49,68,69,73,80 0.61 (0.18 2.04) a Statistical heterogeneity (P.10) and random-effects model used. administration. 120 Further bleeding was decreased but significant differences were not seen in surgery, urgent intervention, or mortality (Table 7). An additional 2 studies evaluated intermittent PPI therapy versus H2RA control. 76,79 Addition of these trials did not appreciably alter the results for further bleeding (RR, 0.56; 95% CI, 0.42 0.75), surgery (RR, 0.63; 95% CI, 0.35 1.11), 49,69,73,76,79,80 or mortality (RR, 0.65; 95% CI, 0.27 1.55), but did lead to a significant benefit for urgent intervention (RR, 0.63; 95% CI, 0.42 0.94). 68,76,79,81 Discussion Meta-analyses dating back to the early l990s documented that endoscopic therapy overall was effective for patients with bleeding ulcers. 91,100 However, those analyses were driven by results from modalities that are not generally used at present (laser and monopolar electrocoagulation). Many more studies comparing endoscopic hemostatic therapies with no therapy or with one another have been published in the intervening years. We sought to broadly address the major clinically relevant questions regarding endoscopic hemostatic therapy for the treatment of bleeding ulcers. In addition, we wanted to avoid pitfalls encountered in some prior meta-analyses: failure to account for studies with routine second-look endoscopies and re-treatment, inclusion of rebleeding without clinical manifestations that is only identified at mandatory second-look endoscopy, assessment of rebleeding without noting initial hemostasis, use of terms such as rebleeding and further bleeding loosely and interchangeably, and unwarranted combinations of studies of different types of therapies together in one analysis. We also sought to provide clinically relevant end points, such as NNT, to assist the practicing endoscopist in assessing the results of endoscopic therapy. Limited data indicate that epinephrine is effective for initial hemostasis. However, epinephrine appears less effective in preventing further bleeding than other monotherapies (NNT, 9) and definitely is less effective than epinephrine followed by a second modality such as sclerosant or a thermal contact device (NNT, 5). When the analysis was restricted to studies that used routine second-look endoscopy with re-treatment of high-risk stigmata, epinephrine was not less effective than other monotherapies or epinephrine followed by a second modality (Table 2). Thus, routine second-look endoscopy with re-treatment can improve the results of endoscopic therapies and must be taken into consideration when reviewing trials of endoscopic therapy. Prior meta-analyses 103,107 also found that combined therapy was superior to epinephrine alone. Vergara et al 103 reported that combined therapy was significantly better in the actively bleeding group but not in those with nonbleeding visible vessels. However, their meta-analysis differed from ours as a result of different inclusion criteria, and they included studies with and without second-look endoscopy and re-treatment in their analysis related to stigmata. We showed that in trials without second-look and re-treatment, combined therapy was superior to epinephrine alone in patients with either active bleeding or nonbleeding visible vessels (Table 2). In contrast to Vergara et al, 103 our analysis showed no benefit of combined therapy versus epinephrine alone in studies that used a second-look endoscopy with re-treatment. Our results suggest again that performing second-look endoscopy with re-treatment can improve the results for a less-effective therapy and that such studies must be assessed separately from studies that did not include a second-look and re-treatment. A large number of studies documented the benefit of thermal contact devices, BPEC and HP, in improving outcomes of further bleeding, surgery, urgent intervention, and mortality. significant differences were identified between the 2 modalities. Results of 2 small studies did suggest a benefit of combination epinephrine plus BPEC as compared with BPEC alone, 10,60 but results with thermal monotherapy were poorer in these trials than in many other studies. A larger high-quality study found that injection of thrombin plus HP was not better than HP alone. 82 Thus, although limited information suggests that epinephrine followed by thermal contact therapy is efficacious, data are insufficient to recommend that thermal contact devices should not be used alone as monotherapy. Fewer studies compare sclerosants (often combined with epinephrine) with no therapy and data are somewhat more heterogeneous. Alcohol alone was clearly beneficial for all outcomes as compared with no therapy; however, when used alone it may not be ideal for initial hemostasis of actively bleeding ulcers, especially given the volume limitations because of concerns about tissue injury. In these situations, the use of epinephrine to decrease or stop bleeding before injection of a

40 LAINE AND MCQUAID CLINICAL GASTROENTEROLOGY AND HEPATOLOGY Vol. 7,. 1 sclerosant makes clinical sense, although data to support this benefit are not present in randomized trials. Comparisons between thermal contact and sclerosant therapy show a trend to less further bleeding and significantly fewer urgent interventions (NNT, 7) with thermal therapy (Table 4). Thus, thermal therapy may hold a slight advantage over sclerosant therapy. However, given the wide variation in study designs and methods used, we conclude that both forms of therapy should be considered appropriate in patients with bleeding ulcers. We do not believe the results are compelling enough to conclude that thermal contact devices are more effective than sclerosant therapy. Clips have been shown to be superior to therapy with injection of epinephrine (NNT, 5) (Table 5) or water. Although we did not find a difference in efficacy between clips and other standard therapies such as thermal contact or sclerosant, the marked heterogeneity in the results of different trials with clips significantly better and significantly worse suggests that success with clips may vary from endoscopist to endoscopist. In addition, most studies with clips used devices, now obsolete, which required repeated loading of clips onto a clip applicator. Clips have the theoretical advantage of not causing tissue damage, which is induced by thermal modalities and sclerosants. More trials with currently available clips are necessary to better define the role of clips in endoscopic hemostasis of ulcers. Injection of thrombin or fibrin glue (which includes thrombin and fibrinogen) produces hemostasis by formation of a fibrin clot. These modalities have the theoretical advantage of not causing tissue damage, although other safety concerns include viral transmission and allergic reactions to bovine thrombin. Although available, these products are not approved for injection therapy in the United States. Nevertheless, available data do suggest that these agents are effective for endoscopic therapy and can be considered roughly comparable with therapy with sclerosants. Argon plasma coagulation is a thermal modality that does not use tamponade as do BPEC and HP. Limited data from 3 trials suggest it is similar in efficacy to sclerosant or thermal contact therapy. We did not include 2 thermal therapies, monopolar electrocoagulation and laser photocoagulation, because they generally are not used for ulcer hemostasis at present owing to concerns about tissue injury and, for laser, expense and lack of portability. However, these therapies were shown to be effective in a prior meta-analysis 100 in reducing further bleeding and mortality. As the risk level of the stigmata of hemorrhage decreases so does the benefit of endoscopic therapy. Endoscopic therapy is definitely effective when used to treat ulcers with active bleeding or nonbleeding visible vessels, with greater benefit for active bleeding (NNT, 2) than nonbleeding visible vessels (NNT, 5) for prevention of further bleeding (Table 6). Current studies do not clearly document a benefit in patients with adherent clots, although results are quite heterogeneous, as previously noted by Kahi et al. 101 Two small studies reported high rates of recurrent bleeding with standard oral antisecretory therapy versus nearly 100% success at preventing recurrent bleeding with endoscopic therapy. Many patients in these studies were enrolled at tertiary care hospitals and had other comorbidities (eg, more than half in the medical therapy arm in one study 71 were already hospitalized for another problem when bleeding started). In contrast, a randomized trial revealed no rebleeding among 24 patients with clots resistant to vigorous irrigation who received bolus plus constant-infusion intravenous PPI and no endoscopic therapy. Results from this study were supported by a double-blind trial from Kashmir of oral omeprazole 40 mg twice daily with no endoscopic therapy, in which 0 of 64 patients with adherent clots had recurrent bleeding with the high-dose PPI regimen versus 13 (21%) of 61 receiving placebo. 121 Thus, endoscopic therapy may be beneficial in patients with a high risk of recurrent bleeding with clots (eg, serious concurrent illness), but intensive PPI therapy without endoscopic treatment may be sufficient in patients with adherent clots resistant to vigorous irrigation. Even after successful endoscopic hemostatic therapy, antisecretory therapy with PPI does appear beneficial (Table 7). The data appear strongest for bolus plus constant-infusion intravenous PPI given for 72 hours, which showed significant benefit after endoscopic therapy for all outcomes from further bleeding (NNT, 12) to mortality (NNT, 45). Constant-infusion PPI therapy was significantly more effective than placebo/no therapy but not more than H2RA controls. Prior meta-analysis has suggested that H2RA may have modest efficacy in reducing rebleeding and surgery in patients with bleeding gastric, but not duodenal ulcers, 122 potentially explaining our findings. Boluses of either oral or intravenous PPI also showed benefit in reducing further bleeding but not other outcomes. These studies are variable in the doses used and the populations studied. The theoretical goal of PPI therapy to reduce ulcer rebleeding is maintenance of intragastric ph greater than 6. However, no clinical data have documented the need for an intragastric ph greater than 6 to reduce rebleeding, and it is conceivable that lower ph thresholds (eg, 4 5 to prevent pepsin-induced clot lysis) may be sufficient. In addition, frequent oral dosing appears to provide a pharmacodynamic effect similar to constant infusion, 123 and initial data from randomized comparisons of bolus followed by constant-infusion PPI versus intermittent boluses of intravenous 68,124 or oral PPI 125,126 show no suggestion of a difference. We await further clinical trial data to confirm that boluses of oral or intravenous PPI will provide efficacy comparable with constant-infusion intravenous PPI before recommending intermittent dosing as first-line adjunctive therapy. The pooled complication rate for endoscopic therapies in our review was 0.5%, and the 95% CIs of the rates for individual modalities all overlapped. Clips and epinephrine generally had the lowest rates, with no perforations in either group. Approximately half of the HP perforations occurred in patients receiving 2 consecutive HP treatments, suggesting that HP therapy, and perhaps any modality that causes tissue injury, should not be used when endoscopic therapy is repeated a second time during a hospitalization. When treating recurrent bleeding in this situation, a therapy with limited tissue injury, such as clips, makes intuitive sense, although no studies provide direct information to inform this decision. Summary of Recommendations Epinephrine injection alone should not be used. Thermal therapy, sclerosant therapy, clips, and thrombin/ fibrin glue all appear to be effective endoscopic hemostatic therapies.

January 2009 ENDOSCOPIC THERAPY FOR BLEEDING ULCERS 41 Epinephrine injection before these therapies may be beneficial, especially for the actively spurting ulcer, but data supporting such a benefit are extremely limited. Endoscopic therapy should be used for ulcers with active bleeding and nonbleeding visible vessels. The role of endoscopic therapy for ulcers with adherent clots is uncertain. Endoscopic therapy may be considered, although intensive PPI therapy alone may be sufficient. PPIs should be given after endoscopic therapy for ulcers. Results are most consistent for bolus followed by continuous-infusion intravenous PPI for 72 hours. References 1. van Leerdam M, Vreeburg E, Rauws E, et al. Acute upper GI bleeding: did anything change? Time trend analysis of incidence and outcome of acute upper GI bleeding between 1993/1994 and 2000. Am J Gastroenterol 2003;98:1494 1499. 2. Thomopoulos K, Vagenas K, Vagianos C, et al. Changes in aetiology and clinical outcome of acute upper gastrointestinal bleeding during the last 15 years. Eur J Gastroenterol Hepatol 2004;16:177 182. 3. Di Fiore F, Lecleire S, Merle V, et al. Changes in characteristics and outcome of acute upper gastrointestinal haemorrhage: a comparison of epidemiology and practices between 1996 and 2000 in a multicentre French study. Eur J Gastroenterol Hepatol 2005;17:641 647. 4. Jensen D, Mawas I, Lousuebsakul V, et al. Changes in the prevalence of different diagnoses for UGI hemorrhage in the last two decades. Gastrointest Endosc 2003;57:AB147. 5. Enestvedt B, Gralnek I, Mattek N, et al. An evaluation of endoscopic indications and findings related to nonvariceal upper-gi hemorrhage in a large multicenter consortium. Gastrointest Endosc 2008;67:422 429. 6. Laine L, Peterson W. Bleeding peptic ulcers. N Engl J Med 1994;331:717 727. 7. Jadad A, Moore R, Carroll D. Assessing the quality of reports of randomized clinical trials: is blinding necessary? Control Clin Trials 1996;17:1 12. 8. Balanzó J, Sainz S, Such J, et al. Endoscopic hemostasis by local injection of epinephrine and polidocanol in bleeding ulcer. A prospective randomized trial. Endoscopy 1988;20:289 291. 9. Berg P, Barina W, Born P. Endoscopic injection of fibrin glue versus polidocanol in peptic ulcer hemorrhage: a pilot study. Endoscopy 1994;26:528 530. 10. Bianco M, Rotondano G, Marmo R, et al. Combined epinephrine and bipolar probe coagulation vs. bipolar probe coagulation alone for bleeding peptic ulcer: a randomized, controlled trial. Gastrointest Endosc 2004;60:910 915. 11. Brearley S, Hawker P, Dykes P, et al. Per-endoscopic bipolar diathermy coagulation of visible vessels using a 3.2 mm probe a randomised clinical trial. Endoscopy 1987;19:160 163. 12. Chau C, Siu W, Law B, et al. Randomized controlled trial comparing epinephrine injection plus heat probe coagulation versus epinephrine injection plus argon plasma coagulation for bleeding peptic ulcers. Gastrointest Endosc 2003;57:455 461. 13. Choudari C, Rajgopal C, Palmer K. Comparison of endoscopic injection therapy versus the heater probe in major peptic ulcer haemorrhage. Gut 1992;33:1159 1161. 14. Chung S, Leung J, Sung J, et al. Injection or heat probe for bleeding ulcer. Gastroenterology 1991;100:33 37. 15. Chung S, Leung J, Steele R, et al. Endoscopic injection of adrenaline for actively bleeding ulcers: a randomized trial. Br Med J (Clin Res Ed) 1988;296:1631 1633. 16. Cipolletta L, Bianco M, Marmo R, et al. Endoclips versus heater probe in preventing early recurrent bleeding from peptic ulcer: a prospective and randomized trial. Gastrointest Endosc 2001; 53:147 151. 17. Cipolletta L, Bianco M, Rotondano G, et al. Prospective comparison of argon plasma coagulator and heater probe in the endoscopic treatment of major peptic ulcer bleeding. Gastrointest Endosc 1998;48:191 195. 18. Fullarton G, Birnie G, Macdonald A, et al. Controlled trial of heater probe treatment in bleeding peptic ulcers. Br J Surg 1989;76:541 544. 19. Gevers A, De Goede E, Simoens M, et al. A randomized trial comparing injection therapy with hemoclip and with injection combined with hemoclip for bleeding ulcers. Gastrointest Endosc 2002;55:466 469. 20. Gralnek I, Jensen D, Gornbein J, et al. Clinical and economic outcomes of individuals with severe peptic ulcer hemorrhage and nonbleeding visible vessel: an analysis of two prospective clinical trials. Am J Gastroenterol 1998;93:2047 2056. 21. Gralnek I, Jensen D, Kovacs T, et al. An economic analysis of patients with active arterial peptic ulcer hemorrhage treated with endoscopic heater probe, injection sclerosis, or surgery in a prospective, randomized trial. Gastrointest Endosc 1997;46: 105 112. 22. Hui W, Ng M, Lok A, et al. A randomized comparative study of laser photocoagulation, heater probe, and bipolar electrocoagulation in the treatment of actively bleeding ulcers. Gastrointest Endosc 1991;37:299 304. 23. Jaramillo J, Carmona C, Gálvez C, et al. Efficacy of the heater probe in peptic ulcer with a non-bleeding visible vessel. A controlled, randomised study. Gut 1993;34:1502 1506. 24. Juszkiewicz P, Wajda Z, Dobosz M, et al. The role of endoscopic thrombin injections in the treatment of gastroduodenal bleeding. S Afr J Surg 1993;31:98 102. 25. Koyama T, Fujimoto K, Iwakiri R, et al. Prevention of recurrent bleeding from gastric ulcer with a nonbleeding visible vessel by endoscopic injection of absolute ethanol: a prospective, controlled trial. Gastrointest Endosc 1995;42:128 131. 26. Laine L. Multipolar electrocoagulation versus injection therapy in the treatment of bleeding peptic ulcers. A prospective, randomized trial. Gastroenterology 1990;99:1303 1306. 27. Laine L. Multipolar electrocoagulation in the treatment of peptic ulcers with nonbleeding visible vessels. A prospective, controlled trial. Ann Intern Med 1989;110:510 514. 28. Laine L. Multipolar electrocoagulation in the treatment of active upper gastrointestinal tract hemorrhage. A prospective controlled trial. N Engl J Med 1987;316:1613 1617. 29. Lin H, Perng C, Sun I, et al. Endoscopic haemoclip versus heater probe thermocoagulation plus hypertonic saline-epinephrine injection for peptic ulcer bleeding. Dig Liver Dis 2003;35:898 902. 30. Lin H, Hsieh Y, Tseng G, et al. A prospective, randomized trial of endoscopic hemoclip versus heater probe thermocoagulation for peptic ulcer bleeding. Am J Gastroenterol 2002;97:2250 2254. 31. Lin H, Hsieh Y, Tseng G, et al. Endoscopic injection with fibrin sealant versus epinephrine for arrest of peptic ulcer bleeding: a randomized, comparative trial. J Clin Gastroenterol 2002;35: 218 221. 32. Lin H, Wang K, Perng C, et al. Heater probe thermocoagulation and multipolar electrocoagulation for arrest of peptic ulcer bleeding. A prospective, randomized comparative trial. J Clin Gastroenterol 1995;21:99 102. 33. Lin H, Wang K, Perng C, et al. Octreotide and heater probe thermocoagulation for arrest of peptic ulcer hemorrhage. A prospective, randomized, controlled trial. J Clin Gastroenterol 1995;21:95 98.