Is routine second-look endoscopy effective after endoscopic hemostasis in acute peptic ulcer bleeding? A meta-analysis

ORIGINAL ARTICLE: Clinical Endoscopy

Is routine second-look endoscopy effective after endoscopic hemostasis in acute peptic ulcer bleeding? A meta-analysis Sara El Ouali, MD,1 Alan N. Barkun, MD, MSc,1,2 Jonathan Wyse, MD,3 Joseph Romagnuolo, MD, MSc,4 Joseph J. Y. Sung, MD, PhD,5 Ian M. Gralnek, MD, MSHS, FASGE,6 Marc Bardou, MD, PhD,7,8 Myriam Martel, BSc1 Montreal, Quebec, Canada; Charleston, South Carolina, USA; Shatin, Hong Kong, The People’s Republic of China; Haifa, Israel; Dijon, France

Background: Routine second-look endoscopy in modern-era peptic ulcer bleeding (PUB) remains controversial. Objective: To assess the effectiveness of routine second-look endoscopy in patients with PUB exhibiting high-risk stigmata after standard medical care and endoscopic therapy. Design: Comprehensive literature searches (1990-2011) were performed, seeking randomized trials comparing a routine with an as-needed second endoscopy. Main Outcome Measurements: The main outcome was rebleeding. Secondary outcomes were surgery and mortality. Subanalyses assessed the influence of study quality, rebleeding definitions, endoscopic hemostasis modality, and proton pump inhibitor (PPI) therapies. Analyses were performed with Revman 5.1. Results are shown as odds ratios (ORs) and 95% confidence intervals (CIs). Results: Only 4 published articles completely reporting studies and 4 abstracts (of 577 citations) were included (938 patients). Rebleeding was significantly decreased by a routine second-look endoscopy (OR 0.55; 95% CI, 0.37-0.81), as was surgery (OR 0.43; 95% CI, 0.19-0.96), but not mortality (OR 0.65; 95% CI, 0.26-1.62). Results remained robust with varying definitions of rebleeding, but not with varying endoscopic hemostasis modalities and PPI therapies; the only trial in which high-dose PPI was used did not show a benefit of a second-look endoscopy. When removing the 2 trials that included patients at highest risk of rebleeding, no significant benefit attributable to a second-look endoscopy was noted (OR 0.65; 95% CI, 0.42-1.00). Limitations: The small number of trials and patients in each of these studies. Conclusions: In the absence of high-dose PPI, especially in patients at very high risk (eg, active bleeding), routine second-look endoscopy appears effective in these selected patients with PUB. However, the generalizability of these results to the era of high-dose PPI and otherwise unselected patients with high-risk stigmata is unclear. (Gastrointest Endosc 2012;76:283-92.)

Peptic ulcer bleeding (PUB) remains the most common cause of acute nonvariceal upper GI bleeding, accounting for 250,000 to 300,000 hospital admissions yearly in the United States, and in older patients, it is still associated

with a 7% to 10% mortality rate.1,2 Despite recent advances in endoscopic hemostatic techniques, recurrent bleeding after primary endoscopy still remains a clinical problem in 13% to 20% of patients.3,4

Abbreviations: CI, confidence interval; HRS, high-risk stigmata; IV, intravenous; OR, odds ratio; PPI, proton pump inhibitor; PUB, peptic ulcer bleeding; WMD, weighted mean difference.

bec, Canada, Division of Gastroenterology (3), Jewish General Hospital, McGill University, Montreal, Quebec, Canada, Division of Gastroenterology and Hepatology (4), Medical University of South Carolina, Charleston, South Carolina, USA, Department of Medicine and Therapeutics (5), The Chinese University of Hong Kong, Hong Kong, China, Department of Gastroenterology and GI Outcomes Unit (6), Rambam Health Care Campus, Bruce and Ruth Rappaport Faculty of Medicine, Technion-Israel Institute of Technology, Haifa, Israel, INSERM CIC-P 803 (7), CHU de Dijon, France, Université de Bourgogne (8), Dijon, France.

DISCLOSURE: The authors disclosed no financial relationships relevant to this publication. Copyright © 2012 by the American Society for Gastrointestinal Endoscopy 0016-5107/$36.00 http://dx.doi.org/10.1016/j.gie.2012.04.441 Received February 1, 2012. Accepted April 5, 2012. Current affiliations: Divisions of Gastroenterology (1) and Clinical Epidemiology (2), McGill University Health Center, McGill University, Montreal, Que-

www.giejournal.org

Reprint requests: Alan N, Barkun, MD, MSc, Room D16.137, Division of Gastroenterology, The McGill University Health Centre, Montreal, Qc, Canada, H3G 1A4.

Volume 76, No. 2 : 2012 GASTROINTESTINAL ENDOSCOPY 283

Second-look endoscopy in peptic ulcer bleeding

Second-look endoscopy, defined as scheduled repeat endoscopy 16 to 24 hours (in some studies as long as 48 hours) after primary endoscopic hemostasis,1 even when rebleeding is not suspected, has been considered. Its goal is to identify patients with persistent high-risk stigmata (HRS) to provide an opportunity to re-treat them when needed, thereby further improving clinical outcomes. It is potentially a high-cost intervention and might even add unnecessary anesthesia risk in high-risk patients or lead to reclassifying patients inappropriately as lower risk; yet its effectiveness remains unclear, particularly with modern medical therapy and more effective initial endoscopic hemostasis techniques. Several studies looked at the efficacy of routine secondlook endoscopy in patients at high risk of rebleeding, but trials have yielded conflicting results. Our aim was to evaluate the effectiveness of a policy of routine secondlook endoscopy in PUB in patients with HRS and to guide expert recommendations at a recent international consensus meeting.1

METHODS Search strategy A comprehensive literature search was performed of several databases including OVID MEDLINE (January 1990 to July 2011), EMBASE (January 1990 to July week 30), the Cochrane Library, Central Register of Controlled Trials, and ISI Web of Knowledge 5.3. All databases were searched using validated search terms specified for PUB and second-look endoscopy. Abstracts presented at major conferences in the past 8 years or before were also handsearched (Digestive Disease Week, United European Gastroenterology Week). Additional relevant studies were identified from cross-referencing of the retrieved articles. All human adult studies published in French and English were considered.

Study selection We selected all randomized, controlled trials comparing routine second-look endoscopy to observation (with or without high-dose proton pump inhibitors [PPIs]), with as-needed endoscopy, for rebleeding. Fully reported published studies and abstracts were included. Second-look endoscopy was defined as a preplanned repeated endoscopy performed 16 to 48 hours after the index endoscopy. We included studies in which (1) second-look endoscopy was assessed, (2) only patients with PUB exhibiting HRS seen at index endoscopy were included, (3) successful hemostasis was achieved at the index endoscopy, and (4) recorded outcomes included rebleeding. The primary outcome of this study was rebleeding. Secondary outcomes included surgical intervention and mortality. Units of blood transfused and duration of hospital stay could not be evaluated because the data were not consistently reported or extractable. 284 GASTROINTESTINAL ENDOSCOPY Volume 76, No. 2 : 2012

El Ouali et al

Take-home Message ●

●

No contemporary meta-analysis has assessed the existing literature to provide best-evidence guidance in determining the optimal role of second-look endoscopy. Second-look endoscopy after endoscopic hemostasis in patients with acute peptic ulcer bleeding is not routinely indicated in contemporary practice and may be more optimally performed in selected patients at very high risk of rebleeding.

Validity assessment and data abstraction Two reviewers (S.O., M.M.) independently identified and evaluated the relevant studies. A third independent reviewer (A.N.B.) resolved disagreements on specific studies. The quality of each study was assessed using Jadad criteria.5 In addition to the Jadad score, a generic specific quality score was also applied to each trial. This scale was created for a national consensus conference1 and provides methodological quantification of additional content-specific components of the study design. Comparative qualitative analyses evaluated the homogeneity of study characteristics such as patient populations, interventions, and outcomes across studies, guiding possible subgroup analyses. In particular, the risk spectrum of patients included in the study and the medical/ endoscopic treatment provided in the control groups (and its relationship to current standard of care) were assessed.1

Data synthesis and analysis Effect measures from each comparative study were reported as the odds ratio (OR) and 95% confidence interval (CI) for dichotomized outcomes. The Mantel-Haenszel method for fixed-effect models were applied to all comparisons to determine corresponding overall effect sizes and their CIs. A random-effects model was used if statistical heterogeneity was noted, using the DerSimonian and Laird method.6 Weighted mean difference (WMD) was used for continuous variables by using the inverse variance model. The Higgins I2 statistic was calculated to quantify the proportion of variation in treatment effects attributable to between-study heterogeneity,7 for which values of 25%, 50%, and 75% indicate low, moderate, and high statistical heterogeneity, respectively. Sensitivity analysis was performed to explore potential sources of heterogeneity by removing 1 study at a time as done previously. As already discussed, subgroup analyses were also performed to assess the influence of rebleeding definitions, endoscopic hemostasis methods, HRS spectrum, PPI use on rebleeding as well as the study quality score. Publication bias was evaluated by examining funnel plots for asymmetry. For all comparisons, publication bias was evaluated using the Begg adjusted rank correlation www.giejournal.org

El Ouali et al

Second-look endoscopy in peptic ulcer bleeding

Figure 1. Quorum diagram.

test and the Egger regression asymmetry test.8,9 Zero-event trials may be a potential source of bias.10 Review Manager (RevMan, Version 5.1 Copenhagen: The Nordic Cochrane Centre, The Cochrane Collaboration, 2008) handles single zero-event trials by adding a correction coefficient of 0.5 to treatment and control groups, but it cannot handle double zero-event trials. To confirm that double zero-event trials do not significantly affect heterogeneity or P values, the Meta package in R was used, and correcting double zeroevents by adding a coefficient of 0.5.10 We also verified the pooled effect with the correction coefficient of the reciprocal of the opposite treatment arm size11 because studies may have an imbalance in group size. Statistical analyses were done using Revman and R (version 2.4.0, R Foundation for Statistical Computing, Vienna, Austria, 2008).

RESULTS Study identification We initially identified 577 citations. After review, a total of 569 studies were excluded because of the following reasons: nonrandomized, not PUB, or not assessing second-look endoscopy (Fig. 1). Five fully reported published studies and 6 abstracts of randomized trials evaluated second-look endoscopy. Among the 5 published studies, 1 was excluded12 because it included second-look endoscopy in both arms. Among the 6 abstracts identified, 2 were excluded because routine repeat endoscopy was performed in both allocation groups.13,14 Therefore, a total of 8 randomized trials comparing routine second-look endoscopy with alternate strategies were included. These studies (4 fully reported15-18 www.giejournal.org

and 4 abstracts19-22) published between 1994 and 2011, ranged in sample size from 40 to 194 subjects and yielded a total of 938 patients (472 of whom were randomly allocated to routine second-look endoscopy).

Study characteristics Characteristics of the included studies are summarized in Table 1.

Study quality, the presence of heterogeneity, and publication bias The quality scores attributed to each trial are displayed in Table 2. They ranged from 4 to 14 on a maximum of a 16-point scale, with a mean of 8.3 (standard deviation 3.6, median 8.5). There was no observed statistical heterogeneity among the relevant comparisons; therefore, fixedeffects models were created. The funnel plots showed that the studies were reasonably well scattered and did not suggest any publication bias for any of the study outcomes; furthermore, the Begg adjusted rank correlation and Egger regression asymmetry tests did not suggest publication bias for any outcomes (data available on request).

Patient populations All studies included patients presenting with bleeding gastric or duodenal ulcers, ranging from Forrest type Ia to IIb, except for the abstract of the study reported by Lin et al,22 in which the ulcer bleeding stigmata were not specified (we assumed for purposes of the analysis that they were HRS because the patients were treated with endoscopic hemostasis). Two studies recruited particularly high-risk subgroups. The first was based on the Baylor Volume 76, No. 2 : 2012 GASTROINTESTINAL ENDOSCOPY 285

Second-look endoscopy in peptic ulcer bleeding

El Ouali et al

TABLE 1. Characteristics of the included studies

Chiu et al, 200315

Messmann et al, 199816

Saeed et al, 199617

Type of publication

Full report

Full report

Full report

Setting, study type

Single center

Multicenter (n ⫽ 5)

Single center

Total no. of patients (second look vs control)

194 (100 vs 94)

105 (52 vs 53)

40 (19 vs 21)

Patients (Forrest class)*

Ia-IIb sick patients (47% in shock and 41% with active bleeding)

Ia-IIb

Ia-IIb with high risk Baylor bleeding score

Time to first endoscopy

Within 24 h of admission

Within 4 h of admission

Within 24 h of admission

Endoscopic treatment at primary and second-look endoscopy

Epinephrine injection ⫹ heater probe

Epinephrine injection ⫾ fibrin glue

Heater probe ⫾ epinephrine injection

Treatment in the control group

No second-look endoscopy

No second-look endoscopy

No second-look endoscopy

Other treatment for both groups

IV omeprazole 40 mg q12h after initial endoscopy for 3 consecutive days

IV omeprazole 40 mg bid followed by 40 mg PO once daily after 48 h; repeat endoscopy after 1 wk to take biopsy samples for Helicobacter pylori and control ulcer healing after 4 wk

IV ranitidine until oral intake was resumed (dose not specified)

Definition of rebleeding

Clinical with endoscopic confirmation

Clinical or endoscopic

Clinical

Follow-up Quality score

30 d

4 wk

11

14

Until discharge 10

*Forrest classification: Ia (spurting hemorrhage), Ib (oozing hemorrhage), IIa (visible vessel), IIb (adherent clot), IIc (hematin on ulcer base), III (lesions without signs of recent hemorrhage). (table continued on next page)

bleeding score17; this scoring system uses a preendoscopic score based on age, number and severity of comorbidities, and endoscopic findings, with high-risk defined as a pre-endoscopy score higher than 5 or postendoscopy score higher than 10. This same trial enrolled men only, whereas other studies included both sexes equally. The second trial included particularly high-risk subjects based on a high prevalence of active bleeders (41%) and patients in shock (47%).15

Setting Four included studies were single-center trials15,17,18,20; in 1 trial, patients were enrolled at 5 different hospitals.16 There was no mention of the type of setting in 3 trials.19,21,22 Two studies specified that the trial took place at an academic center.16,20 286 GASTROINTESTINAL ENDOSCOPY Volume 76, No. 2 : 2012

Performance of endoscopy The index endoscopy was described as being performed within 4 to 24 hours of admission. A second-look endoscopy was scheduled within 16 to 48 hours after the index endoscopy. Treatment, if clinically indicated, at the second-look endoscopy included monotherapy injection of epinephrine18 (considered suboptimal therapy in a recent consensus guideline1) or other unspecified product, thermal coagulation, clips, or a combination of injection and thermal coagulation or clips. In 1 study, epinephrine injection followed by fibrin glue was used.16

Comparator control group and adjuvant therapy In the control groups, patients did not undergo a second-look endoscopy and were closely observed for signs of rebleeding (specific criteria are listed in Table 1). www.giejournal.org

El Ouali et al

Second-look endoscopy in peptic ulcer bleeding

TABLE 1. (continued) Villanueva et al, 199418

Park et al, 201121

Chiu et al, 200619

Lee et al, 200520

Lin et al, 199622

Full report

Abstract

Abstract

Abstract

Abstract

Single center

Not specified

Not specified

Single center

Not specified

104 (52 vs 52)

73 (39 vs 34)

164 (80 vs 84)

143 (70 vs 73)

115 (60 vs 55)

I-IIa

I-IIb

Ia-IIb

Ia-IIb

Bleeding ulcers

Within 4 h of admission

Not specified

Not specified

Not specified

Within 48 h

Epinephrine injection

Epinephrine injection ⫾ electric coagulation or hemoclipping

Epinephrine ⫹ heater probe ⫾ hemoclipping

Epinephrine injection ⫹ hemoclipping

Injection therapy

No second-look endoscopy

No second-look endoscopy

High-dose omeprazole (infusion)

No second-look endoscopy

No second-look endoscopy

Ranitidine (50 mg IV q6h or 150 mg PO bid)

Intravenous PPI 40 mg q12h

Intravenous omeprazole high-dose 80-mg bolus followed by a 72-h infusion at 8 mg/h in the control group; 40 mg q12h intravenously in the second-look group

Not specified

Ranitidine, dose not specified

Clinical with occasional endoscopic confirmation

Not specified

Clinical with endoscopic confirmation

Not specified

Not specified

Until discharge

30 d

30 d

30 d

10

7

6

4

Both second-look endoscopy and control groups in all included studies were comparable to each other with respect to age, sex distribution (except in Saeed et al17), comorbidities, intake of nonsteroidal anti-inflammatory drugs, ulcer characteristics, and severity of bleeding. One study compared second-look endoscopy with highdose omeprazole infusion (80-mg bolus followed by a 72hour infusion at 8 mg/h).19 In all other trials, both treatment and control arms received similar adjuvant therapies, in most cases a single or double daily dose of a PPI15,16,21 or intravenous H2 receptor antagonists.17,18,22 Dosing or information was missing in 1 study.22 Only the highdose intravenous (IV) PPI (bolus plus infusion) regimen was recommended by recent consensus guidelines.1

Definition of rebleeding The definition of rebleeding varied among the different trials. Most studies used clinical symptoms as evidence of www.giejournal.org

Not specified 4

rebleeding, with endoscopic confirmation.15,16,19 One study required both clinical and endoscopic evidence of rebleeding.16 Two studies used clinical evidence of recurrent bleeding, without consistent endoscopic confirmation.17,18 Definition of “clinical” also varied among trials and was not specified in 3 studies.20-22

Outcomes The primary outcome in all trials was recurrent bleeding. Other outcomes were mortality, transfusion requirements (except the study by Lin et al22), surgery (except the study by Lee et al20), and length of hospital stay (except in the studies by Saeed et al17 and Lin et al22).

Follow-up Patients were monitored until their discharge from hospital or up to a 30-day follow-up (not mentioned by Lin et al22) (Table 1). In 1 study, all patients underwent a repeat endosVolume 76, No. 2 : 2012 GASTROINTESTINAL ENDOSCOPY 287

Second-look endoscopy in peptic ulcer bleeding

El Ouali et al

TABLE 2. Methodological quality assessment Score

Points allocated

Jadad scale ⫹1

Study described as randomized

⫹1

Description of the randomization method

⫺1

Inappropriate method of randomization

⫹1

Double-blind study

⫹1

Description of double-blinding method

⫺1

Inappropriate method of blinding

Withdrawal

⫹1

Description of withdrawal

Specific score

⫹1

Estimation of sample size

⫹1

Description of patient selection

⫹1

Description of patient characteristics

⫹1

Patient selection criteria for second-look clearly stated

⫹1

Concomitant therapy applied equally to both arms

⫹1

Adequate definition of rebleeding

⫹1

Indication for surgery

⫹1

Description of cause of death

⫹1

30-day time period for outcome assessment

⫹1

Endoscopically noted, clinically suspected rebleeding

⫹1

Endoscopic therapy performed

Randomization

Blinding

copy 1 week after the initial endoscopy to evaluate for Helicobacter pylori and to take biopsy specimens of gastric ulcers to rule out malignancy.16 In 4 studies, the 30-day outcomes (rebleeding or mortality) were stated.15,19-21

Qualitative presentation of results and sensitivity analysis Rebleeding. Routine second-look endoscopy significantly reduced rebleeding (odds ratio [OR] 0.55; 95% confidence interval [CI], 0.37-0.81) (Fig. 2). Results remained robust when performing sensitivity analysis by removing each trial in turn.

Secondary outcomes Second-look endoscopy significantly reduced the need for surgery (OR 0.43; 95% CI, 0.19-0.96) (Fig. 3). This was not the case when the trials by Chiu et al15 (OR 0.56; 95% CI, 0.23-1.37), Chiu et al19 (OR 0.47; 95% CI, 0.2-1.13), and Villanueva et al18 (OR 0.41; 95% CI, 0.14-1.16) were individually removed. Mortality was not significantly reduced by second-look endoscopy (OR 0.65; 95% CI, 0.26-1.62) (Fig. 4). For both outcomes, the conclusions remained unchanged when correcting for double-zero events (see Methods). 288 GASTROINTESTINAL ENDOSCOPY Volume 76, No. 2 : 2012

Length of hospital stay was reported in 4 studies,15,16,18,20 but only Villanueva et al18 included analyzable data, and there was no significant difference (WMD 2.50; 95%CI, ⫺6.25, 1.25). Five studies reported the number of units of blood transfused,15-18,21 with data analyzable in only 2.15,18 Blood transfusion was not significantly different between the 2 groups (WMD ⫺0.39; 95%CI, ⫺0.86, 0.09).

Subgroup analyses Study population: definition of high-risk patients. All trials included patients with HRS based on the Forrest classification. Saeed et al,17 however, as stated previously, included patients at an even higher risk of rebleeding. Although the results of this trial are reported as being “significant” these authors reported a 1-sided nonexact test of significance; the decrease in rebleeding with secondlook endoscopy failed to achieve significance with the more conservative analysis (OR 0.08; 95% CI, 0.00-1.50). Chiu et al15 enrolled very sick patients including 47% who were in shock and 41% with active bleeding and a significant decrease in rebleeding (OR 0.33; 95% CI, 0.11-0.96). When considering only these 2 studies together, the dewww.giejournal.org

El Ouali et al

Second-look endoscopy in peptic ulcer bleeding

Figure 2. Forrest plot: rebleeding outcome. CI, confidence interval; Fixed, fixed-effects; M-H, Mantel-Haenszel.

Figure 3. Forrest plot: surgery outcome. CI, confidence interval; Fixed, fixed-effects; M-H, Mantel-Haenszel.

monstrable efficacy was greater (OR 0.26; 95% CI, 0.100.69). However, excluding these studies because of their different higher-risk patient populations led to a loss in demonstrable improvement in rebleeding attributable to the second-look strategy (OR 0.65; 95% CI, 0.42-1.00). Endoscopic hemostasis methods. In view of current consensus recommendations, we compared rebleeding outcomes in trials in which injection of epinephrine alone18 (or an unspecified product22 or fibrin glue16) was administered with those in which thermal treatment alone or combination treatment was administered. Rebleeding was not significantly reduced by second-look endoscopy when the endoscopic therapy (initial and repeat) was injection alone (OR 0.66; 95% CI, 0.37-1.15). PPI use. The only trial that used high-dose PPI infusion, as currently recommended, compared its use in the control group with second-look endoscopy (with patients in this latter group receiving twice-daily dosing instead).19 There was no decrease in rebleeding (OR 0.63; 95% CI, 0.2-2.02) in the second-look endoscopy group. Rebleeding definitions (endoscopic confirmation). Analysis for rebleeding in only the trials using systematic endoscopic confirmation in the definition of primary outcome (ie, excluding the others) yielded a summary OR of 0.44 (95% CI, 0.20-0.96).15,19 Unfortunately, other risk-group stratified analyses were not possible because the results of studies were not rewww.giejournal.org

ported according to very high risk (eg, active bleeders or shock at presentation) versus all subjects with HRS.

Study quality scores The median quality score (score ⫽ 8.5) was used as a threshold with which to analyze the study quality subgroup. When removing articles or abstracts with a quality score less than 8, 4 studies remained15-17,23; there was no significant decrease in rebleeding with second-look endoscopy (OR 0.59; 95% CI, 0.35-1.00). When including the 4 studies with the lowest quality score,19,21,22,24 rebleeding was significantly reduced by second-look endoscopy (OR 0.50; 95% CI, 0.28-0.89).

DISCUSSION The efficacy of routine second-look endoscopy in PUB has been evaluated in several randomized, controlled trials that yielded disparate results. Our meta-analysis including 938 patients thus reviewed the existing evidence published on the topic in the hope of clarifying a clinically relevant summary conclusion. Rebleeding was found to be decreased with second-look endoscopy in PUB in patients at high risk according to the Forrest criteria (as well as the Baylor bleeding score in 1 trial17). The need for surgery was decreased as well, but without any associated effect on mortality. The clinical applicability of these results to Volume 76, No. 2 : 2012 GASTROINTESTINAL ENDOSCOPY 289

Second-look endoscopy in peptic ulcer bleeding

El Ouali et al

Figure 4. Forrest plot: mortality. CI, confidence interval; Fixed, fixed-effects; M-H, Mantel-Haenszel.

current-day practice, however, requires discussion, especially in light of the paucity of trials using standard of care cotherapies (eg, noninjection-alone endoscopic therapy, high-dose IV PPI). Because of the small numbers of trials and patients, we attempted to get as much information as possible and to initially include in the analysis as many studies as possible, including those by Lee et al20 and Lin et al22 for which there was much missing information (quality scores of 4 for each, Table 1). Also, although no statistical heterogeneity was noted, there was significant clinical heterogeneity among the available trials; the small sample size of very high-risk cohorts studied (n ⫽ 40, Saeed et al17) limited the power of this heterogeneity testing. Indeed, they differed in their definition of eligible patients, the type of endoscopic hemostatic technique, use of pharmacotherapy, and rebleeding definitions, all of which can confound outcomes and limit generalizability to present-day practice, which may involve different endoscopic therapy initially and/or different medical adjuvant therapy. Additionally, in many studies, it is unclear whether patients who rebled and therefore clinically needed repeat endoscopy anyway were included or excluded from the preallocated routine second-endoscopy group. If included, this may have led to a bias favoring second-look endoscopy. Additionally, this supposition also depends on whether a true intention-to-treat analysis was performed, which is not always clearly described. Only 3 trials individually reported significant reductions in rebleeding.15,17 Of these 3 studies, the sole abstract by Lin et al22 lacked much information, limiting interpretation. In the trial by Saeed et al,17 a total of only 40 patients at extremely high risk of rebleeding were enrolled, using the aforementioned Baylor score criteria and 1-tailed inferential P testing).17 This trial found the greatest decrease in rebleeding (OR 0.08; 95% CI, 0.00-1.50) with routine second-look endoscopy. In the third, conducted by Chiu et al,15 a significant decrease in rebleeding (OR 0.33; 95% CI, 0.11-0.96) was noted without, however, an effect on surgery or mortality. The enrolled patients were very sick, including 47% who were in shock and 41% with active bleeding. The results of these trials are likely not general290 GASTROINTESTINAL ENDOSCOPY Volume 76, No. 2 : 2012

izable to patients with HRS. Indeed, when both trials by Saeed et al17 and Chiu et al15 were removed in sensitivity analysis, no benefit attributable to second-look endoscopy was shown. The trials by Saeed et al17 and Chiu et al15 used combination endoscopic therapy (an unknown proportion of patients in the trial by Park et al21 also received such hemostasis), whereas neither used the currently recommended high-dose PPI treatment.25-29 The more recent trial by Chiu et al19 also included the use of clips. However, this trial failed to show a significant decrease in rebleeding compared with a control group that received continuous high-dose IV PPI (P.Y. Chiu PY, personal communication, 2011), as recommended by contemporary guidelines1; only twice-daily PPI dosing was used in the second-look endoscopy group. This implies that if secondlook endoscopy is effective (without high-dose infusions of PPI), high-dose infusion of PPI might be equally effective. Whether second-look endoscopy would have an incremental benefit over high-dose PPI infusions has not been studied. The issue of the type of endoscopic therapy applied is also interesting, with lower effects seen if injection alone was used. This might imply that, just as initial hemostasis using injection appears less effective,1 injection as a follow-up therapy at second-look endoscopy might also not be effective enough to make this approach efficacious. In trying to interpret the aforementioned results, we postulate that a second-look policy might be most useful when reserved for patients at a very high risk of rebleeding, including those with initial hemodynamic instability and active bleeding,30 with noninjection-alone therapies available. The recommended high-dose bolus plus a continuous IV PPI regimen may attenuate any benefit of a routine second-look endoscopy approach, however, even in a high-risk group, although the incremental effect remains unstudied. At best, the benefit attributable to second-look endoscopy might be as high as that approximated by pooling the 3 most favorable studies,15,17,22 which translates to a number needed to treat of 8.5-17 This effect would presumably be tempered by contemporary use of the high-dose IV PPI regimen, which itself is assowww.giejournal.org

El Ouali et al

ciated with a number needed to treat of 17,31 but may be less costly and less risky than a second-look endoscopy and has been more extensively studied in the context of ideal endoscopic therapy and patients with HRS. When reviewing all randomized, controlled trial evidence, clinicians must also assess the applicability of the adopted interventions to current standards of care (ie, affecting the external validity or generalizability of the conclusions). These can be questioned for most trials in this meta-analysis because of both endoscopic and pharmacological cotherapies; the conclusions appear most generalizable to scenarios that do not reflect the best standard practice today. A few reviews attempted to summarize the efficacy of second-look endoscopy. In 2003, an initial meta-analysis by Marmo et al32 reviewed the available data at the time from 4 studies12,16-18 and concluded that routine secondlook endoscopy with retreatment significantly decreased the risk of recurrent bleeding in patients with PUB. This study, however, exhibited some methodological shortcomings limiting its interpretation.33 In 2009, a review article by Tsoi et al34 followed by a meta-analysis by the same group35 examined the available evidence on second-look endoscopy and compared trials using thermal coagulation and injection therapy. It was found that repeat endoscopy with thermal coagulation decreased rebleeding rates, whereas injection therapy alone did not. The lower effect of injection-alone therapy has also been noted in meta-analyses outside the context of second-look endoscopy,36-39 so the conclusion seems plausible. However, Tsoi et al34 only included 5 of the 8 studies that we included. The cost-effectiveness of a routine approach of secondlook endoscopy has also been studied. Spiegel et al40 evaluated the cost-effectiveness of 4 approaches in the management of PUB: (1) clinical observation after primary hemostasis; (2) intravenous PPI after hemostasis; (3) use of second-look endoscopy at 24 hours in all patients; and (4) use of selective second-look endoscopy at 24 hours. They concluded that, compared with the other aforementioned strategies, performing selective second-look endoscopy in high-risk patients may prevent more cases of rebleeding, surgery, or death at a lower overall cost. However, IV PPI was likely to reduce the need for second-look endoscopy, and IV PPI was the preferred choice in a clinical setting in which the rebleed rate was lower (ⱕ9%) or in which the costs of IV PPIs were low (⬍$10/day) and/or the cost of endoscopy was high (⬎$1225). In addition, a major limitation, in our opinion, given our results, is that this analysis assumed a uniform relative risk reduction among risk strata, and this remains to be proven (ie, second-look endoscopy might not just be less absolutely effective [same relative risk] in lower-risk patients; it might, in fact, have no reduction [or a much lower relative risk reduction]). www.giejournal.org

Second-look endoscopy in peptic ulcer bleeding

In conclusion, although existing data support the notion that routine second-look endoscopy in PUB might be effective in high-risk patients without high-dose PPI infusion, the external validity of these results as they relate to current clinical practice are questionable. This is why the internal consensus meeting participants concluded that routine second-look endoscopy was not recommended in patients with bleeding and HRS, but that this approach be instead reserved for selected patients at particularly increased risk of rebleeding (eg, hemodynamic instability, active bleeding, large ulcers).1 However, even this latter notion remains unproven in the era of high-dose IV PPI, and the cost-effectiveness may not be favorable in that setting.

REFERENCES 1. Barkun AN, Bardou M, Kuipers EJ, et al, International Consensus Upper Gastrointestinal Bleeding Conference Group. International consensus recommendations on the management of patients with nonvariceal upper gastrointestinal bleeding. Ann Intern Med 2010;152:101-13. 2. Gralnek IM, Barkun AN, Bardou M. Management of acute bleeding from a peptic ulcer. N Engl J Med 200;359:928-37. 3. Barkun A, Sabbah S, Enns R, et al. The Canadian Registry on Nonvariceal Upper Gastrointestinal Bleeding and Endoscopy (RUGBE): endoscopic hemostasis and proton pump inhibition are associated with improved outcomes in a real-life setting. Am J Gastroenterol 2004;99:1238-46. 4. Hearnshaw SA, Logan RF, Lowe D, et al. Acute upper gastrointestinal bleeding in the UK: patient characteristics, diagnoses and outcomes in the 2007 UK audit. Gut 2011;60:1327-35. 5. Jadad AR, Moore RA, Carroll D, et al. Assessing the quality of reports of randomized clinical trials: is blinding necessary? Control Clin Trials 1996; 17:1-12. 6. DerSimonian R, Laird N. Meta-analysis in clinical trials. Control Clin Trials 1986;7:177-88. 7. Higgins JP, Thompson SG. Quantifying heterogeneity in a metaanalysis. Stat Med 2002;21:1539-58. 8. Begg CB, Mazumdar M. Operating characteristics of a rank correlation test for publication bias. Biometrics 1994;50:1088-101. 9. Egger M, Davey Smith G, Schneider M, et al. Bias in meta-analysis detected by a simple, graphical test. BMJ 1997;315:629-34. 10. Friedrich JO, Adhikari NK, Beyene J. Inclusion of zero total event trials in meta-analyses maintains analytic consistency and incorporates all available data. BMC Med Res Methodol 2007;7:5. 11. Sweeting MJ, Sutton AJ, Lambert PC. What to add to nothing? Use and avoidance of continuity corrections in meta-analysis of sparse data. Stat Med 2004;23:1351-75. 12. Rutgeerts P, Rauws E, Wara P, et al. Randomised trial of single and repeated fibrin glue compared with injection of polidocanol in treatment of bleeding peptic ulcer. Lancet 1997;350:692-6. 13. Babicki A, Marczewski R, Wajda Z. Second-look endoscopy with repeated injection of thrombin with adrenaline vs fibrin glue in bleeding peptic ulcers. Br J Surg 1998;85:113. 14. Ell C. Scheduled endoscopic retreatment vs. single injections therapy in bleeding gastroduodenal ulcers: results of a multicenter study [abstract]. Gastrointest Endosc 1998;47:AB83. 15. Chiu PW, Lam CY, Lee SW, et al. Effect of scheduled second therapeutic endoscopy on peptic ulcer rebleeding: a prospective randomised trial. Gut 2003;52:1403-7. 16. Messmann H, Schaller P, Andus T, 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;30:583-9.

Volume 76, No. 2 : 2012 GASTROINTESTINAL ENDOSCOPY 291

Second-look endoscopy in peptic ulcer bleeding

17. Saeed ZA, Cole RA, Ramirez FC, et al. Endoscopic retreatment after successful initial hemostasis prevents ulcer rebleeding: a prospective randomized trial. Endoscopy 1996;28:288-94. 18. Villanueva C, Balanzo J, Torras X, et al. Value of second-look endoscopy after injection therapy for bleeding peptic ulcer: a prospective and randomized trial. Gastrointest Endosc 1994;40:34-9. 19. Chiu PW, Joeng H, Choi C, et al. The effect of scheduled second endoscopy against intravenous high dose omeprazole infusion as an adjunct to therapeutic endoscopy in prevention of peptic ulcer rebleeding-a prospective randomized study [abstract]. Gastroenterology 2006;130: A121. 20. Lee SY, Lee MK, Cho CM, et al. The effect of second look endoscopy in patients with bleeding peptic ulcers [abstract]. Gastroenterology 2005;128:A639. 21. Park SJ, Park H, Lee YC, et al. Effect of scheduled second-look endoscopy on peptic ulcer bleeding: interim analysis of a prospective randomized multicenter trial in Korea [abstract]. Gastrointest Endosc 2011;73(Suppl):AB223. 22. Lin CK, Lai KH, Lo GH, et al. The value of second-look endoscopy after endoscopic injection therapy for bleeding peptic ulcer [abstract]. Gastroenterology 1996;110:A177. 23. Villanueva C, Balanzo J, Torras X, et al. Value of second-look endoscopy after injection therapy for bleeding peptic ulcer: a prospective and randomized trial. Gastrointest Endosc 1994;40:34-9. 24. Lee SY, Lee LM, Cho CM , et al. The effect of second look endoscopy in patients with bleeding peptic ulcers. Gastroenterology. 2005;128:A639. 25. Andriulli A, Loperfido S, Focareta R, et al. High- versus low-dose proton pump inhibitors after endoscopic hemostasis in patients with peptic ulcer bleeding: a multicentre, randomized study. Am J Gastroenterol 2008;103:3011-8. 26. Barkun AN, Kuipers EJ, Sung JJ. It is premature to recommend lowdose intravenous proton pump inhibition after endoscopic hemostasis in patients with bleeding ulcers. Am J Gastroenterol 2009;104: 2120-1. 27. Calvet X, Barkun A, Bardou M, et al. High-dose vs non-high-dose PPIs after endoscopic treatment in patients with bleeding peptic ulcer: current evidence is insufficient to claim equivalence. Arch Intern Med 2010; 170:1698-9; author reply 700. 28. Chan WH, Khin LW, Chung YFA, et al. Randomized controlled trial of standard versus high-dose intravenous omeprazole after endoscopic

292 GASTROINTESTINAL ENDOSCOPY Volume 76, No. 2 : 2012

El Ouali et al

29.

30.

31.

32.

33. 34.

35.

36.

37.

38.

39.

40.

therapy in high-risk patients with acute peptic ulcer bleeding. Br J Surg 2011;98:640-4. Wang CH, Ma MH, Chou HC, et al. High-dose vs non-high-dose proton pump inhibitors after endoscopic treatment in patients with bleeding peptic ulcer: a systematic review and meta-analysis of randomized controlled trials. Arch Intern Med 2010;170:751-8. Elmunzer BJ, Young SD, Inadomi JM, et al. Systematic review of the predictors of recurrent hemorrhage after endoscopic hemostatic therapy for bleeding peptic ulcers. Am J Gastroenterol 2008;103: 2625-32; quiz 33. Barkun AN, Adam V, Sung JJ, et al. Cost effectiveness of high-dose intravenous esomeprazole for peptic ulcer bleeding. Pharmacoeconomics 2010;28:217-30. Marmo R, Rotondano G, Bianco MA, et al. Outcome of endoscopic treatment for peptic ulcer bleeding: is a second look necessary? A metaanalysis. Gastrointest Endosc 2003;57:62-7. Romagnuolo J. Routine second look endoscopy: ineffective, costly and potentially misleading. Can J Gastroenterol 2004;18:401-4. Tsoi KK, Chiu PWY, Sung JJY. Endoscopy for upper gastrointestinal bleeding: is routine second-look necessary? Nat Rev Gastroenterol Hepatol 2009;6:717-22. Tsoi KK, Chan HC, Chiu PW, et al. Second-look endoscopy with thermal coagulation or injections for peptic ulcer bleeding: a meta-analysis. J Gastroenterol Hepatol 2010;25:8-13. Calvet X, Vergara M, Brullet E, et al. Addition of a second endoscopic treatment following epinephrine injection improves outcome in highrisk bleeding ulcers. Gastroenterology 2004;126:441-50. Barkun AN, Martel M, Toubouti Y, et al. Endoscopic hemostasis in peptic ulcer bleeding for patients with high-risk lesions: a series of metaanalyses. Gastrointest Endosc 2009;69:786-99. Laine L, McQuaid KR. Endoscopic therapy for bleeding ulcers: an evidence-based approach based on meta-analyses of randomized controlled trials. Clin Gastroenterol Hepatol 2009;7:33-47; quiz 1-2. Sung JJ, Tsoi KK, Lai LH, et al. Endoscopic clipping versus injection and thermo-coagulation in the treatment of non-variceal upper gastrointestinal bleeding: a meta-analysis. Gut 2007;56:1364-73. Spiegel BM, Ofman JJ, Woods K, et al. Minimizing recurrent peptic ulcer hemorrhage after endoscopic hemostasis: the cost-effectiveness of competing strategies. Am J Gastroenterol 2003;98:86-97.

www.giejournal.org