Outcome of endoscopic treatment for peptic ulcer bleeding: Is a second look necessary? A meta-analysis

Transcription

1 Outcome of endoscopic treatment for peptic ulcer bleeding: Is a second look necessary? A meta-analysis Riccardo Marmo, MD, Gianluca Rotondano, MD, Maria Antonia Bianco, MD, Roberto Piscopo, MD, Antonio Prisco, MD, Livio Cipolletta, MD Torre Del Greco, Italy Background: Endoscopic therapy for GI bleeding is highly effective. Nevertheless, bleeding recurs in 10% to 25% of cases, irrespective of the method of treatment used. Whether a second-look endoscopy with retreatment after initial hemostasis is of clinical value is controversial. A metaanalysis was performed to assess whether systematic second-look endoscopy with retreatment reduces the risks of recurrent bleeding, salvage surgery, and death in patients with peptic ulcer bleeding. Methods: A systematic review was performed of randomized controlled studies of the value of second-look endoscopy in patients with peptic ulcer bleeding published between 1990 and Four studies were selected according to predefined criteria. Two investigators extracted the data independently. Pooled risk estimates and number need to treat were calculated for each procedure. Heterogeneity of effects was tested. Results: The absolute risk reduction in clinical recurrent bleeding was 6.2% (p < 0.01). Absolute risk reduction for surgery and mortality were, respectively, 1.7% and 1.0% (not significant). The second look with retreatment significantly reduced the risk of recurrent bleeding compared with control patients (OR 0.64; 95% CI [0.44, 0.95]; p < 0.01), with a number needed to treat of 16. There was no heterogeneity among studies. The risk of surgery as well as the risk of death were not significantly influenced by the second-look endoscopy with retreatment (number needed to treat, respectively, 58 and 97). Conclusions: Systematic second-look endoscopy with retreatment significantly reduces the risk of recurrent bleeding in patients with peptic ulcer bleeding compared with control patients, but it does not substantially reduce the risk of salvage surgery or mortality. (Gastrointest Endosc 2003;57:62-7.) Received January 4, For revision March 26, Accepted September 18, Current affiliations: U.O. di Gastroenterologia Ospedale Maresca- Torre Del Greco, Italy. Presented in part as oral presentation at the annual meeting of Digestive Disease Week, May 20-23, 2001, Atlanta, Georgia (Gastrointest Endosc 2001;53:AB66). Reprint requests: Livio Cipolletta, MD, Via S. Domenico al Vomero 24, I-80126, Naples, Italy. Copyright 2003 by the American Society for Gastrointestinal Endoscopy /2003/$ doi: /mge Endoscopic treatment has advanced the management of severe peptic ulcer bleeding (PUB) over the past 2 decades. Injection therapy, thermal coagulation, or application of hemoclips are all superior to conservative therapy in high-risk patients with PUB who have major stigmata of recent hemorrhage at index endoscopy. 1-4 Recurrent bleeding, arising in 10% to 30% of cases, remains the single most important determinant of poor prognosis, irrespective of the method of treatment used. 3 Because hemostasis is initially achieved in more than 90% of cases with several endoscopic modalities, the question is no longer whether to offer endoscopic hemostasis but rather, what is optimal management after hemostasis has been achieved? Whether a secondlook endoscopy with retreatment after initial hemostasis is of clinical value is still controversial. Data from prior investigations are inadequate in relation to sample size or they indicate contradictory outcomes A systematic review was carried out to determine whether second-look endoscopy with retreatment reduces the risk of recurrent bleeding, salvage surgery, and death in patients with PUB. PATIENTS AND METHODS A MEDLINE search was conducted by using the following terms: gastrointestinal bleeding, peptic ulcer bleeding, endoscopic treatment and endoscopic re-treatment (both overall and for specific procedures), treatment outcome, second look, and recurrent bleeding. A similar search was done with the same key words using EMBASE/Excerpta Medica and Current Contents. A decision was made to start the search after 1989, the year of publication of the National Institutes of Health Consensus Conference on therapeutic endoscopy and bleeding ulcers, 1 the bibliography of which contained adequate information about results of endoscopic therapy in patients with PUB up to The search, therefore, covered the period from January 1990 to December It 62 GASTROINTESTINAL ENDOSCOPY VOLUME 57, NO. 1, 2003

2 Outcome of endoscopic treatment for peptic ulcer hemorrhage: meta-analysis R Marmo, G Rotondano, M Bianco, et al. Table 1. Characteristics of the patients enrolled in the randomized controlled trials Villanueva 5 Saeed 6 Rutgeerts 7 Messmann 8 RT No RT RT No RT RT No RT RT No RT No. patients Mean age (y) Treatment EPI HP ± EPI EPI ± FG EPI ± FG Time to 2nd endoscopy (h) NSAID use (%) Gastric/duodenal ulcer (%) 29/65 23/63 32/58 57/43 42/58 45/55 Adjuvant drug therapy Ranitidine IV Ranitidine IV Ranitidine IV Omeprazole IV Follow-up (d) 30 Until discharge from hospital RT, Second look with retreatment; No RT, second look only if clinical suspicion of recurrent bleeding; EPI, epinephrine injection; HP, heater probe; FG, fibrin glue. was restricted to human studies in adults and was not limited to the English language. Second-look endoscopy was defined as a scheduled endoscopy within 24 hours of an index endoscopy with retreatment of the bleeding stigmata at the ulcer base by using the same modality previously used. To be included in the analysis, articles had to report results in full for studies designed as prospective randomized, controlled trials on the efficacy of second-look endoscopy versus no systematic retreatment (i.e., repeat endoscopic evaluation only for clinical suspicion of recurrent bleeding) in patients with PUB. Abstracts, editorials, letters, comments, or preliminary reports were excluded. It was permissible that retreatment at second-look endoscopy may have been applied routinely in all patients or only those with visible bleeding stigmata. Explicit data had to be reported for the rate of recurrent bleeding, need for surgery, and mortality associated with the tested treatments (i.e., presence or absence of the aforementioned outcomes) in either text or tables. Articles had to provide enough data to estimate the relative risk comparing the treatments. There were 126 citations identified concerning endoscopic treatment of PUB, and their abstracts were examined. If no clear reason for exclusion was found in the abstract, the full publication was obtained. If multiple reports of a single trial had been published, the most recent version was obtained. Reference citations for the selected articles and clinical reviews were studied. Two investigators (R.M., G.R.) independently extracted data by using a standardized form. Decisions regarding inclusion of articles and data extraction were reached by consensus between the 2 reviewers. If there was disagreement, the publications were jointly evaluated to resolve the inconsistency. The studies were examined with respect to the following criteria: study design (including randomization), inclusion and exclusion criteria, patient characteristics (including risk assessment for bleeding and indications for endoscopic therapy), technical details of endoscopic therapy (e.g., for injection, agent used, volume injected; for thermal therapy, device, total energy delivered, pulse duration), and definition of study outcomes and monitoring methods. Mortality was defined as any death within 30 days of the index endoscopy. There were no differences in follow-up among the 4 studies considered. Information on objective quality-related characteristics was also collected and a quality score calculated, according to the method of Cook et al. 2 (ranging from 2 to 22). A total of 7 original reports assessing second-look endoscopy versus no endoscopic retreatment were identified. Three studies were rejected for the following reasons: 2 were published only as abstracts 9,10 and one was not randomized. 11 Hence, the number of articles suitable for inclusion in the meta-analysis was Study investigators were not contacted for additional data. The pooled odds ratio (OR) and 95% confidence intervals (CI) were calculated from the raw study data by using the Mantel-Haenszel method (fixed effect model) as modified by Robbins et al. 12 and the Der Simonian and Laird method (random effect model). 13 Data from the various studies were combined by using appropriate weights. 13 The test of heterogeneity was used to determine whether the fixed effect model or random effects model was appropriate for calculation of the weights, used in the meta-analytic estimate, and the 95% CI for the meta-analytic estimate. A statistical evaluation of heterogeneity by the chisquare test was used to assess whether the variation in treatment effect within trials of the same group was greater than might be expected. Heterogeneity was considered to be present if p < In the absence of statistically significant heterogeneity, OR and 95% CI by the fixed effect model only are given in the results. The Peto fixed model was used, which produced a Q statistic with a chi-squared distribution with n-1 degrees of freedom, where n is equal to the number of trials. Calculations of risks ratios in which a correction factor was required were made with statistical software (SAS, version 6.12, SAS Institute Inc., Carey, N.Y.). All other calculations were made with another statistical software package (SPSS, version 7.1, M.J. Norusis, Chicago, Ill.). For studies in which the constructed 2 2 tables contained cells with zero events, a standard correction factor of 0.5 was added to each cell. The absolute risk reduction (ARR) and number needed to treat (NNT) with 95% CIs were calculated for each of the trials included in the analysis, based on the raw data presented in each publication. The ARR is the absolute arith- VOLUME 57, NO. 1, 2003 GASTROINTESTINAL ENDOSCOPY 63

3 R Marmo, G Rotondano, M Bianco, et al. Outcome of endoscopic treatment for peptic ulcer hemorrhage: meta-analysis Figure 1. Risk of recurrent bleeding: cumulative meta-analysis of second-look endoscopy with retreatment versus no retreatment. Odds ratio = 0.64 (95% CI [0.44, 0.95]). Number needed to treat = 16 (95% CI [9, 75]). Figure 2. Risk of surgery: cumulative meta-analysis (odds ratio) of second-look endoscopy with retreatment versus no retreatment. Odds ratio = 0.68 (95% CI [0.35, 1.3]). Number needed to treat = 58 (95% CI [ 64, 28]). metic difference in the event rate (ARR = experimental event rate [ERR] control event rate [CER]). The NNT is a quantitative estimate of the therapeutic gain: it expresses the number of patients to be treated with the experimental treatment (in this case second-look endoscopy with retreatment) so that at least undesired event(s) are avoided, such as recurrent bleeding, surgery, or death, compared with control group (in this case observation and second-look endoscopy only if needed). It is therefore evident that the lower the NNT, the more effective is the procedure. 14 For a significant meta-analysis finding, publication bias was defined as the number of additional nonsignificant studies that, if included in the meta-analysis, would have removed the significant meta-analysis finding. RESULTS The endoscopic modality used in 3 randomized trials of second-look endoscopy was injection therapy with epinephrine or fibrin glue 5,7,8 ; in 1 study the heat probe was used. 6 The presence of malignancy or coagulopathy were considered exclusion criteria in all studies. All patients in the studies had highrisk stigmata of recent hemorrhage (active bleeding, visible vessels, or adherent clot). Other patient background variables, including mean age, percentage taking nonsteroidal anti-inflammatory drugs (NSAIDs) or antisecretory drugs, time to scheduled second-look endoscopy, and length of follow-up were similar (Table 1). No data were available on prior peptic ulcer disease, prior endoscopic therapy, or Helicobacter pylori status. Median quality score of the trials was 17.5 (range from 15 to 21). Data for occurrence of key events reported for the individual trials included in the meta-analysis are shown in Table 2. A total of 785 patients were evaluated (393 submitted to systematic second-look endoscopy, 392 observed unless signs of recurrent bleeding). Overall, 120 patients had recurrent bleeding, of whom 38 required emergency surgery and 38 died. The null hypothesis of homogeneity was not rejected. The p value for the Q statistic for all 3 outcomes considered was, respectively, 0.15, 0.56, and From a fixed effect model, bleeding recurred more commonly in the control group than in the retreatment group (18.2% vs. 12%; ARR 6.2%). The plot in Figure 1 shows that second look with retreatment significantly reduced the risk of recurrent bleeding compared with control patients (OR 0.64; 95% CI [0.44, 0.95]; p < 0.01). The calculated NNT for this outcome was 16, that is, for every 16 patients managed with systematic second-look endoscopy and retreatment, 1 patient avoids a recurrence of hemorrhage that would occur with expectant management. The only study in which omeprazole was administered intravenously found no difference in recurrent bleeding between 64 GASTROINTESTINAL ENDOSCOPY VOLUME 57, NO. 1, 2003

4 Outcome of endoscopic treatment for peptic ulcer hemorrhage: meta-analysis R Marmo, G Rotondano, M Bianco, et al. Figure 3. Risk of death: cumulative meta-analysis (odds ratio) of second-look endoscopy with retreatment versus no retreatment. Odds ratio = 0.80 (95% CI [0.42, 1.54]). Number needed to treat = 97 (95% CI [ 25, 53]). Table 2. Occurrance of key events for the outcomes considered Recurrent bleeding Surgery Mortality RT No RT RT No RT RT No RT n (%) n (%) n (%) n (%) n (%) n (%) Villanueva 5 11 (21) 15 (29) 4 (8) 8 (15) 1 (2) 2 (4) Saeed 6 0 (0) 5 (24) 0 (0) 0 (0) 1 (5) 2 (11) Rutgeerts 7 27 (10) 42 (16) 9 (3) 12 (5) 12 (4) 15 (6) Messmann 8 11 (21) 9 (17) 3 (6) 2 (4) 3 (6) 2 (4) Overall 49 (12) 71 (18.2) 16 (4) 22 (5.7) 17 (4.3) 21 (5.2) ARR [95% CI] 6.2 [1.3, 11.1] 1.7 [ 1.6, 4.0] 1.0 [ 4, 1.9] NNT [95% CI] 16 [9, 75] 58 [ 64, 28] 97 [ 25, 53] p Value <0.01 NS NS RT, Second look with retreatment; No RT, second look only if clinical suspicion of recurrent bleeding; ARR, absolute risk reduction; NNT, number needed to treat; NS, not significant. patients who underwent second-look endoscopy and those who did not. 8 The risk of surgery was not significantly influenced by second-look endoscopy with retreatment (Fig. 2), nor was the risk of death (Fig. 3). The calculated values of NNT for these outcomes were, respectively, 58 and 97 (Table 2). Publication bias, calculated only for reduction in recurrent bleeding, was 1. Therefore, the significant meta-analysis finding for recurrent bleeding could be removed by including only 1 additional study with nonsignificant results. DISCUSSION Peptic ulcer is the most common cause of acute upper GI hemorrhage, accounting for about 60% of the cases. 15 Active bleeding can be stopped initially and visible vessels treated in virtually all patients with the use of several endoscopic modalities; however, bleeding recurs in a significant number of patients. 16 Most endoscopists adopt an expectant management strategy after initial endoscopic control of ulcer hemorrhage. Nonetheless, if bleeding recurs there is disquietude that an earlier opportunity for prevention might have been missed. Available data suggest that scheduled endoscopic retreatment may confer a small benefit in terms of a reduction in recurrence of bleeding. Whether this benefit can be shown to be significant in a single randomized controlled trial depends not only on the study sample size, but also on the type of endoscopic treatment and the risk factors for recurrence of bleeding among patients entered. The overall mortality and the need for surgery were extremely low in all studies; thus, a benefit of second-look endoscopy in terms of significantly reducing mortality or the need for surgery was neither expected nor found, given the sample size of each study. To prove that second-look endoscopy significantly reduces the need for surgery or mortality would require enormous studies of over, respectively, 5000 and 14,000 patients. Such a trial is unlikely to be conducted. The appropriate selection of patients for second-look endoscopy is crucial to the design of these trials. To study whether endoscopic retreatment reduces recurrent bleeding, the population at increased risk for bleeding after initial endoscopic hemostasis must be defined with precision. Although 3 studies randomized all patients undergoing endoscopic treatment for hemorrhage, 5,7,8 Saeed et al. 6 randomized only those patients who were prospectively identified as being at high risk for recurrent bleeding after initial endoscopic hemostasis. The avoidance of the dilution effect resulting from retreatment of low-risk patients at a second-look endoscopy may be one factor that enabled VOLUME 57, NO. 1, 2003 GASTROINTESTINAL ENDOSCOPY 65

5 R Marmo, G Rotondano, M Bianco, et al. Outcome of endoscopic treatment for peptic ulcer hemorrhage: meta-analysis these investigators to achieve significant results with a smaller sample size. In the studies that used injection therapy, the decision to render retreatment at the second-look endoscopy was based on the perceived risk of recurrence of bleeding according to the appearance of the ulcer. Although the ulcer stigmata that predict recurrence of bleeding before endotherapy have been studied extensively, 3,17 few studies have specifically investigated stigmata that would predict recurrent bleeding after initial endoscopic treatment If patients who are at increased risk for recurrent bleeding can be reliably identified (e.g., a persistently positive endo-doppler signal after treatment 21 ), secondlook endoscopy and focused retreatment for this selected high-risk group may be warranted. Factors that predict failure after endoscopic treatment have been identified. 18,19 In a study of 1144 consecutive patients who underwent combined endoscopic therapy, the rate of recurrent bleeding was 8.2%. Hypotension (OR 0.21), hemoglobin less than 10 g/dl (OR 1.87), fresh blood in stomach (OR 2.15), active bleeding (OR 1.65), and large ulcers (OR 1.80) were independent predictors of therapeutic failure and might be appropriate indications for second-look endoscopy. 20 The present study represents the first quantitative analysis of the efficacy of systematic second-look endoscopy versus expectant management. Pooled data showed a clinically relevant (ARR 6.2%, NNT 16) and significant benefit of endoscopic retreatment in terms of recurrent bleeding. However, the benefit is likely small and a large number of patients would undergo second-look endoscopy unnecessarily. No cost effectiveness data have been reported. Future studies should focus on pharmacologic and economic issues, such as the impact of a reduction in recurrent bleeding in terms of decreased hospital length of stay or number of patients who require transfusion. Note that endoscopic retreatment also has risks. Current evidence indicates that combination treatment with injection followed by thermal coagulation or fibrin glue injection achieves consistently lower rates of recurrent bleeding in high-risk patients compared with single-modality therapy Hemoclip application may provide a significant therapeutic gain over injection or heat probe therapy, resulting in a significantly lower rate of recurrent bleeding. 4,25 Intravenous administration of high doses of omeprazole has been demonstrated to significantly decrease the frequency of recurrent bleeding after initially successful endoscopic treatment of PUB. 26 A meta-analysis has confirmed that the use of proton pump inhibitors versus H2 receptor antagonists is associated with a reduced likelihood of recurrence of bleeding. 27 Therefore, improved prima- ry endoscopic treatments as well as effective adjuvant pharmacotherapy will likely further reduce the need/benefit of second-look endoscopy. In conclusion, the present quantitative analysis indicates that second-look endoscopy with systematic retreatment significantly reduces the risk of recurrence of bleeding caused by peptic ulcer compared with control patients undergoing expectant management, but it does not substantially reduce the risk of salvage surgery or mortality. Based on the present meta-analysis and other evidence, a second-look endoscopy may be recommended only for patients at high risk for recurrent bleeding. Further study is needed before any change in routine clinical practice can be recommended. REFERENCES 1. NIH Consensus Conference: Therapeutic endoscopy and bleeding ulcers. JAMA 1989;262: Cook DJ, Guyatt GH, Salena BJ, Laine L. Endoscopic therapy for acute non variceal upper gastrointestinal hemorrhage: a meta-analysis. Gastroenterology 1992;102: Laine L, Peterson WL. Bleeding peptic ulcer. N Engl J Med 1994;331: Cipolletta L, Marmo R, Bianco MA, Rotondano G, Piscopo R. Clinical impact of endoscopic hemostasis for non variceal bleeding: an evidence-based meta-analysis [abstract]. Gastrointest Endosc 1998;47: Villanueva C, Balanzo J, Torras X, Soriano G, Sainz S, Vilardell F. Value of second-look endoscopy after injection therapy for bleeding peptic ulcer: a prospective and randomized trial. Gastrointest Endosc 1994;40: Saeed ZA, Cole RA, Ramirez FC, Schneider FE, Hepps KS, Graham DY. Endoscopic retreatment after successful initial hemostasis prevents ulcer rebleeding: a prospective randomized trial. Endoscopy 1996;28: Rutgeerts P, Rauws E, Wara P, Swain P, Hoos A, Solleder E, et al. Randomised trial of single and repeated fibrin glue compared with injection of polidocanol in treatment of bleeding peptic ulcer. Lancet 1997;350: Messmann H, Schaller P, Andus T, Lock G, Vogt W, Gross V, et al. Effect of programmed endoscopic follow-up examinations on the rebleeding rate of gastric or duodenal peptic ulcers treated by injection therapy: a prospective, randomized controlled trial. Endoscopy 1998;30: Lin CK, Lai KH, Lo GH, Cheng JS, Huang RL, Hsu PI, et al. The value of second-look endoscopy after endoscopic injection therapy for bleeding peptic ulcer [abstract]. Gastroenterology 1996;110:A Ell C. and the members of the German Ulcer Bleeding study group (DUS II). Scheduled endoscopic re-treatment vs. single injection therapy in bleeding gastroduodenal ulcers: results of a multicenter study [abstract]. Gastrointest Endosc 1998;47: AB Kapetanakis AM, Kyprizlis EP, Tsikrikas TS. Efficacy of repeated therapeutic endoscopy in patients with bleeding ulcer. Hepatogastroenterology 1997;44: Robbins J, Brenslow N, Greenland S. Estimators of the Mantel-Haenszel variance consistent in both sparse data and large strata limiting models. Biometrics 1986;42: Der Simonian R, Laird N. Meta-analysis in clinical trials. Controlled Clin Trials 1986;7: GASTROINTESTINAL ENDOSCOPY VOLUME 57, NO. 1, 2003

6 Outcome of endoscopic treatment for peptic ulcer hemorrhage: meta-analysis R Marmo, G Rotondano, M Bianco, et al. 14. Sackett LD, Strauss S, Richardson WS, Rosenberg W, Haynes RB. Evidence based medicine. 2nd ed. New York: Churchill Livingstone; p Longstreth GF. Epidemiology of hospitalization for acute upper gastrointestinal hemorrhage: a population-based study. Am J Gastroenterol 1995;90: Savides TJ, Jensen DM. Therapeutic endoscopy for non variceal gastrointestinal bleeding. Gastroenterol Clin North Am 2000;29: Lau JY, Sung JJ, Chan AC, Lai GW, Lau JT, Ng EK, et al. Stigmata of hemorrhage in bleeding peptic ulcers: an interobserver agreement study among international experts. Gastrointest Endosc 1997;46: Chung IK, Kim EJ, Lee MS, Kim HS, Park SH, Lee MH, et al. Endoscopic factors predisposing to rebleeding following endoscopic haemostasis in bleeding peptic ulcers. Endoscopy 2001;33: Saeed ZA, Ramirez FC, Hepps KS, Cole RA, Graham DY. Prospective validation of the Baylor bleeding score for predicting the likelihood of rebleeding after endoscopic hemostasis of peptic ulcers. Gastrointest Endosc 1995;41: Wong SIC, Yu LM, Lau JY, Lam YH, Chan AC, Ng EK, et al. Prediction of therapeutic failure after adrenaline injection plus heater probe treatment in patients with bleeding peptic ulcer. Gut 2002;50: Riemann JF, Rosenbaum A. The role of Doppler ultrasound in gastrointestinal bleeding. Baillieres Best Pract Res Clin Gastroenterol 2000;14: Kubba A, Murphy W, Palmer KR. Endoscopic injection for bleeding peptic ulcer: a comparison of adrenaline alone with adrenalin plus human thrombin. Gastroenterology 1996;111: Chung SCS, Lau JYW, Sung JJY, Chan AC, Lai CW, Ng EK, et al. A randomized comparison between adrenaline injection alone and adrenaline injection, plus heat probe treatment for actively bleeding ulcers. BMJ 1997;314: Machicado GA, Jensen DM. Thermal probes alone or with epinephrine for the endoscopic hemostasis of ulcer haemorrhage. Baillieres Best Pract Res Clin Gastroenterol 2000;14: Cipolletta L, Bianco MA, Rotondano G, Marmo R, Piscopo R. Endoclips vs. heater probe in preventing early recurrent bleeding from peptic ulcer: a prospective and randomized trial. Gastrointest Endosc 2001;53: Lau JYW, Sung JJY, Lee KK, Yung MY, Wong SK, Wu JC, et al. Effect of intravenous omeprazole on recurrent bleeding after endoscopic treatment of bleeding peptic ulcers. N Engl J Med 2000;343: Gisbert JP, Gonzalez L, Calvet X, Roque M, Gabriel R, Pajares JM. Proton pump inhibitors versus H2-antagonists: a metaanalysis of their efficacy in treating bleeding peptic ulcer. Aliment Pharmacol Ther 2001;15: Submission of ACCEPTED manuscript on diskette Gastrointestinal Endoscopy strongly encourages the submission of final manuscripts on disk. Although files created with WordPerfect are preferred, please send your final manuscript in any electronic format. On your disk, please indicate computer system (e.g., IBM, MacIntosh) and word processing software used (e.g., WordPerfect 6.1). Please refer to complete Instructions to Authors in the most recent January or July issue of the Journal. VOLUME 57, NO. 1, 2003 GASTROINTESTINAL ENDOSCOPY 67