Literature search performed herein. PPI indicates proton pump inhibitor.
Effect of high-dose vs non–high-dose proton pump inhibitors (PPIs) on rebleeding. Odds ratios (ORs) and 95% confidence intervals (CIs) were computed by the Mantel-Haenszel method.
Effect of high-dose vs non–high-dose proton pump inhibitors (PPIs) on surgical intervention. Odds ratios (ORs) and 95% confidence intervals (CIs) were computed by the Mantel-Haenszel method.
Effect of high-dose vs non–high-dose proton pump inhibitors (PPIs) on mortality. Odds ratios (ORs) and 95% confidence intervals (CIs) were computed by the Mantel-Haenszel method.
Wang C, Ma MH, Chou H, Yen Z, Yang C, Fang C, Chen S. High-Dose vs Non–High-Dose Proton Pump Inhibitors After Endoscopic Treatment in Patients With Bleeding Peptic UlcerA Systematic Review and Meta-analysis of Randomized Controlled Trials. Arch Intern Med. 2010;170(9):751–758. doi:10.1001/archinternmed.2010.100
High-dose proton pump inhibitors (PPIs) (80-mg bolus, followed by 8-mg/h continuous infusion for 72 hours) have been widely studied and used. However, to date no concrete evidence has shown that high-dose PPIs are more effective than non–high-dose PPIs.
We performed a literature search for randomized controlled trials that compared the use of high-dose PPIs vs non–high-dose PPIs in patients with bleeding peptic ulcer and determined their effects on rebleeding, surgical intervention, and mortality. Outcomes data were combined in a meta-analysis and were reported as odds ratios (ORs) with 95% confidence intervals (CIs).
A total of 1157 patients from 7 high-quality randomized studies were included in this meta-analysis. High-dose PPIs and non–high-dose PPIs did not differ in their effects on the rates of rebleeding (7 studies and 1157 patients; OR, 1.30; 95% CI, 0.88-1.91), surgical intervention (6 studies and 1052 patients; 1.49; 0.66-3.37), or mortality (6 studies and 1052 patients; 0.89; 0.37-2.13). Post hoc subgroup analyses revealed that summary outcomes measures were unaffected by severity of signs of recent hemorrhage at initial endoscopy, route of PPI administration, or PPI dose.
Compared with non–high-dose PPIs, high-dose PPIs do not further reduce the rates of rebleeding, surgical intervention, or mortality after endoscopic treatment in patients with bleeding peptic ulcer.
Bleeding peptic ulcer is a common cause of emergency admissions to hospitals. It is associated with high morbidity, mortality, and health care costs.1- 3 Primary management of bleeding peptic ulcer includes prompt fluid replacement, treatment of comorbidities, administration of acid-suppressing agents, endoscopic therapy, and surgery. Despite advances in pharmacologic options and endoscopic techniques, morbidity and mortality have remained around 5% to 10% over the past decade.2,3 Endoscopic intervention has significantly reduced the rates of rebleeding, surgical intervention, and mortality.4 However, about 10% to 20% of patients undergoing endoscopic intervention experience rebleeding and have almost a 4% mortality rate.5 Medical treatment is useful to stabilize clots after endoscopic treatment.
In vitro studies6- 8 have shown that coagulation and stable platelet aggregation do not occur at pH levels below 6. Moreover, when pH is elevated above 5, clot lysis by pepsin is largely reduced.9 Histamine2 receptor antagonists (H2RAs) and proton pump inhibitors (PPIs) are widely used to suppress gastric acid secretion. Several studies10- 13 have shown that PPIs are superior to H2RAs in gastric acid suppression and in control of rebleeding episodes. However, no consensus exists about the optimal dose of intravenous PPIs.
One study14 revealed that intragastric pH can be maintained above 6 using high-dose PPIs. Several randomized controlled studies15- 18 compared high-dose PPIs vs placebo. Those investigators concluded that high-dose PPIs were effective in reducing rebleeding rates. Nevertheless, a meta-analysis19 demonstrated that non–high-dose PPIs were also superior to placebo or H2RAs in terms of rebleeding rates. In other words, there was little solid evidence demonstrating that high-dose PPIs were more effective than non–high-dose PPIs in stopping bleeding peptic ulcer. Because high-dose PPIs are significantly more expensive, we systematically reviewed the literature and performed a meta-analysis to determine whether high-dose PPIs were superior to non–high-dose PPIs in treating bleeding peptic ulcer.
Two of us (C.-H.W. and Z.-S.Y.) performed a comprehensive literature search of several databases, including BioMedCentral, CINAHL, MEDLINE, EMBASE, and Cochrane Central Register of Controlled Trials. We searched Micromedex for available PPIs and found 5 agents (omeprazole, sodium/magnesium, pantoprazole sodium, lansoprazole, rabeprazole sodium, and esomeprazole sodium/magnesium). The following strategy keywords were proton pump inhibitor, omeprazole, pantoprazole, lansoprazole, rabeprazole, eosmeprazole, hemorrhage, and bleeding. The full search strategies are available in the Appendix (available on request from the corresponding author). No language limits were imposed. The search was performed independently by two of us (C.-H.W. and Z.-S.Y.) in August 2009. We used backward snowballing (ie, scanning of references of retrieved articles and pertinent reviews) for further studies.
References acquired from databases and the literature search were first examined at the title and abstract level. High-dose PPIs were defined as a dose equivalent to an 80-mg bolus of omeprazole or pantoprazole, followed by a continuous intravenous infusion of drug at 8 mg/h for 72 hours.5,19- 21 Continuous infusion doses exceeding 192 mg/d were also considered high-dose PPIs. Other doses were considered non–high-dose PPIs.
We used the following inclusion criteria to select pertinent studies: (1) randomized controlled trials, (2) comparison of high-dose vs non–high-dose PPIs, (3) endoscopically confirmed bleeding peptic ulcer, (4) administration of PPIs after endoscopic intervention, and (5) outcomes reported as rates of rebleeding, surgical intervention, or mortality.
Baseline and outcomes data were independently abstracted by two of us (C.-H.W. and Z.-S.Y.), with divergences resolved by consensus. Specifically, we extracted study design, patient characteristics (age, sex, and endoscopic findings), population, number of randomized patients, and interventions (route, dose, and frequency of administration of PPIs). On the basis of endoscopic appearance of ulcers, patients were categorized as being at high risk for further bleeding (spurting, oozing, nonbleeding visible vessel, or adherent clot) or at low risk for further bleeding (flat pigmented spot or clean ulcer base). Our primary outcome was recurrent ulcer bleeding within 30 days of randomization. Rebleeding should be confirmed by endoscopy. Secondary outcomes included surgical intervention and mortality from any cause (30-day mortality or “in-hospital” mortality).
Rebleeding rate was difficult to define because there was significant variation in the definition of rebleeding among the studies. The distinction between continuing bleeding (active bleeding with failed endoscopic intervention) and rebleeding (recurrent bleeding after endoscopically confirmed hemostasis spontaneously or due to endoscopic intervention) was not clearly made. Four studies22- 25 excluded patients from study enrollment if their ulcer bleeding did not stop spontaneously or was promoted by endoscopy. Therefore, bleeding after the index endoscopy could be deemed recurrent bleeding. Another study26 excluded patients with peptic ulcers that were actively bleeding. Two other studies27,28 did not specifically exclude patients whose ulcer was bleeding and in whom it could not be terminated by endoscopic intervention. Udd et al27 defined rebleeding as rebleeding found on endoscopy or persistent bleeding necessitating an emergency operation. Udd et al excluded patients with immediate failure of endoscopic therapy and operation from analysis. Bajaj et al28 excluded patients with profuse hemorrhage leading to persistent shock who were unable to be resuscitated without interventional radiology or surgery.
All studies set some clinical criteria for suspicion of rebleeding, including hemodynamic changes (hypotension or tachycardia), decline in hemoglobin level, or overt bleeding (hematemesis or melena). Suspected rebleeding was subsequently confirmed by endoscopy. The time from suspicion to confirmation was not stated. Udd et al27 performed routine control endoscopy 72 hours after the index endoscopy. In their study, 14 patients had rebleeding, 3 of whom were found to have signs of rebleeding or ongoing bleeding on routine endoscopic examination, without any clinically detectable bleeding signs or symptoms. Andriulli et al24 performed selective successive endoscopy in high-risk patients (≥6 points at admission according to the scoring system by Rockall et al29). Timing of assessment for rebleeding varied across studies. Udd et al27 and Yüksel et al25 did not clearly specify the timing. Andriulli et al24 recorded rebleeding within 7 days and did not record rebleeding episodes after patients were discharged. Cheng,22 Yılmaz,26 Bajaj,28 and Hung23 and colleagues assessed episodes of rebleeding within 30 days. In the present meta-analysis, the pooled rebleeding analysis referred to rebleeding reported from the time the patient was randomized up to 30 days later.
There were no clearly stated surgical indications in most studies except for that by Yüksel et al.25 The most common indication for surgery was unsuccessful endoscopic treatment. However, it was not mentioned how many times endoscopic treatment could be performed before endoscopic intervention was deemed a failure and radiologic or surgical intervention began. Only Yüksel et al specified that, if rebleeding occurred after the second endoscopic treatment, surgery was indicated. It was not clearly stated in all studies whether radiologic intervention was counted as a surgical intervention. In our meta-analysis, the pooled surgical intervention analysis referred to surgical intervention indicated at the discretion of the primary care physicians when medical treatment was deemed a failure.
Of 6 studies reporting mortality, Hung et al23 and Yüksel et al25 did not state the timing of assessment of mortality. Udd,27 Yılmaz,26 and Bajaj28 and colleagues reported mortality within 30 days. Andriulli et al24 reported only in-hospital mortality. Regarding the causes of mortality, all studies described them in the text. Causes included unsuccessful hemostasis, pneumonia, myocardial infarction, and unrelated operations. Only Andriulli et al specifically reported mortality rates according to the cause of death. In this meta-analysis, the pooled mortality analysis referred to all-cause mortality.
Two of us (C.-H.W. and Z.-S.Y.) assessed the methods quality of each included study using the system described by Jadad et al.30 Each study was evaluated using a 5-point scale, with 1 point being awarded for each of the following criteria: randomized controlled trial, details of randomization methods provided, double-blind study, details of blinding method provided, and information on study withdrawals provided. Studies scoring 3 points or higher were included. Discrepancies in ratings were resolved by discussion between the 2 of us.
All data were entered and analyzed using available software (Review Manager [RevMan], version 5.0; Copenhagen, Denmark The Nordic Cochrane Centre, The Cochrane Collaboration, 2008). Binary outcomes from individual studies were collected to compute individual odds ratios with 95% confidence intervals by the Mantel-Haenszel method. A fixed-effects model was used to calculate the pooled summary effect. A random-effects model was used if we detected moderate (50%) or higher statistical inconsistency.31
To evaluate statistical heterogeneity and inconsistency of treatment effects across studies, Cochrane Q tests and I2 statistics, respectively, were used.32Statistical significance was set at 0.10 for Cochrane Q tests. I2 statistics measured the extent of inconsistency among results of the studies. The outcome is expressed as a percentage of total variation across studies that is due to statistical heterogeneity rather than chance. A value of 0% indicates that all variability in effect estimates is due to chance rather than statistical heterogeneity. A value exceeding 50% indicates substantial statistical heterogeneity.
Post hoc subgroup analyses were performed to examine the effect of severity of signs of recent hemorrhage at initial endoscopy, route of PPI administration, and PPI dose. This study was performed in compliance with The Cochrane Collaboration and the Quality of Reporting of Meta-analyses statement.33
Using our search strategy, we initially found 7243 citations (Figure 1). These included heterogeneous studies. The treatment groups differed in the dose, route, and frequency of administration of PPIs. The control groups also varied, with the use of placebo, H2RAs, somatostatin, or the same PPI via different doses, routes, and frequencies of administration. After examining these studies at the title and abstract level, we found 9 citations through database searches that might compare high-dose with non–high-dose PPIs. Backward snowballing identified another relevant study. We retrieved the full texts of these 10 articles and assessed them using our inclusion criteria. Two studies34,35 were excluded because the PPI doses in the treatment groups were slightly lower than our predetermined high-dose PPI level. One study34 used omeprazole, 80-mg bolus injection, followed by 160-mg continuous infusion for 24 hours; the other study35 used pantoprazole, 40-mg bolus injection, followed by 8-mg/h continuous infusion for 72 hours. Thereafter, we evaluated the quality of the remaining 8 studies. A third study36 was excluded because it achieved only 1 point using the scoring system by Jadad et al.30 The remaining 7 studies22- 28 were consistent with the inclusion criteria and achieved at least 3 points using the scoring system (Table 1 and Table 2). These 7 studies included 1157 patients. The study by Hung et al23 compared treatment effects among high-dose, non–high-dose, and placebo groups. We extracted only the results of the high-dose and non–high-dose groups for analysis.
The odds ratios (95% confidence intervals) for rebleeding, surgical intervention, and mortality were 1.30 (0.88-1.91), 1.49 (0.66-3.37), and 0.89 (0.37-2.13), respectively (Figures 2, 3, and 4). Statistical heterogeneity values for rebleeding, surgical intervention, and mortality were P = .90, P = .89, and P = .74, respectively. I2 statistic was 0% for all 3 effect measures. Therefore, there was no significant statistical heterogeneity or inconsistency across included studies.
Results of post hoc subgroup analyses are summarized in Table 3. No significant difference was noted in rebleeding, surgical intervention, or mortality between high-dose and non–high-dose groups among subgroups categorized by severity of signs of recent hemorrhage at initial endoscopy, route of PPI administration, or PPI dose. Outcomes data among high-risk patients were retrieved from 5 articles.23- 25,27,28Among those studies, Udd et al27 did not classify adherent clots as high-risk lesions. Therefore, in that study we used outcomes data from patients with F1A, F1B, and F2A lesions (based on the Forrest classification37). Low-dose PPIs were defined as 40 mg/d or less of intravenous or oral omeprazole or pantoprazole.24 Intermediate-dose PPIs were defined as those between high and low doses of intravenous or oral omeprazole or pantoprazole.
The most important result of this meta-analysis is the finding among patients with bleeding peptic ulcer that high-dose PPIs are not superior to non–high-dose PPIs in reducing the rates of rebleeding, surgical intervention, or mortality after endoscopic treatment. While a plethora of studies have compared high-dose PPI treatment with placebo or H2RAs, few studies have compared different doses of PPIs. A meta-analysis by Leontiadis et al5 included 24 randomized controlled studies in which patients received high-dose and non–high-dose PPIs. Rates of rebleeding and surgical intervention were significantly reduced in the high-dose and non–high-dose PPI groups compared with patients who received placebo or H2RAs. However, meta-regression analysis revealed no association of PPI dose with treatment effects. Our meta-analysis further focused on head to head randomized controlled studies and surprisingly identified only 7 such studies. The results of this meta-analysis might be considered counterintuitive in that higher doses of PPIs had no advantage over lower doses of PPIs.
There may be several explanations for this. The first is related to pharmacogenetic and geographic factors. Enhanced efficacy of PPI therapy for bleeding peptic ulcer has been noted among Asians.5 Two of our 7 reviewed trials studied Asians. One was performed in Taiwan22 and the other in Hong Kong.23 Other investigators have suggested that Chinese populations may have smaller parietal cell mass compared with populations of white race/ethnicity.38 High-dose and non–high-dose PPIs may have similar ability to saturate parietal cells. Other studies39,40 identified different metabolic rates of omeprazole by cytochrome P450s, with Chinese populations metabolizing omeprazole more slowly than persons of white race/ethnicity.39,40 The higher prevalence of Helicobacter pylori infection among Asian populations would also be consistent with a greater acid-suppressive effect of a given PPI dose.41 In the meta-analysis by Leontiadis et al,5 including 8 studies in Asia and 16 studies mainly in Europe, treatment effects of PPIs were consistently higher among Asians than non-Asians in rates of rebleeding, surgical intervention, and mortality. Therefore, it would be less likely that high-dose PPI treatment would exhibit significant benefit over non–high-dose PPI treatment in head to head studies among Asians.
A second explanation for our findings may be that some studies enrolled patients with less severity of bleeding peptic ulcer. The studies by Udd,27 Cheng,22 Yılmaz,26 and Bajaj28 and colleagues included patients having ulcers with lower potential for rebleeding (ie, Forrest classification types IIC and III).37 The study by Bajaj et al excluded patients who could not take oral medications, which might exclude seriously ill patients from study enrollment. Inclusion of patients with lower potential for rebleeding or exclusion of seriously ill patients might dilute the effect of high-dose PPI treatment. Post hoc subgroup analyses performed herein among high-risk patients demonstrated that high-dose PPIs were no more effective than non–high-dose PPIs. However, only 5 studies were included in the subgroup analysis, with one of them (by Andriulli et al24) providing most of the patients, which might have affected the pooled results. More studies were needed to have adequate power in such an analysis.
Third, these studies each recruited a small population, rendering it difficult to detect significant differences in head to head studies. In the study by Andriulli et al,24 an estimated 8% difference in rebleeding rates between high-dose and non–high-dose PPIs was used to calculate the required sample size. However, the actual difference might be even smaller, if a difference existed. Larger populations are needed to assess any benefit of high-dose over non–high-dose PPIs.
Theoretically, maintenance of intragastric pH above 6 in patients with bleeding peptic ulcer should result in better clinical outcomes, including reduced rates of rebleeding, surgical intervention, and mortality. This was indirectly corroborated in the retrospective study by Simon-Rudler et al,42 who reviewed patients with high-risk bleeding peptic ulcer (ie, spurting and oozing hemorrhage or nonbleeding visible vessels) from 1997 to 2004. Compared with non–high-dose omeprazole (40 mg/d intravenously), high-dose omeprazole reduced the occurrence of rebleeding, need for surgery, and mortality due to hemorrhagic shock in this patient group. Other studies11,18,23 recorded the fraction of time during which the intragastric pH was above 6 using 24-hour pH monitors in patients with bleeding peptic ulcer receiving PPIs. The reported fractions of time differed among these studies, ranging from 50% to 85%. However, the rebleeding rate was similar or lower in the study23 with a short period during which intragastric pH was kept above 6. Such indirect evidence suggests that PPI treatment effect may not be directly related to intragastric pH. Another study by Udd et al43 further compared the effects of high-dose and non–high-dose omeprazole on intragastric acidity among patients with bleeding peptic ulcer treated endoscopically. They found a significant difference in intragastric pH between the 2 groups on the first and second days of treatment. However, the difference in rebleeding rates between the 2 groups was not significant. Therefore, whether intragastric pH can serve as a reliable proxy for adequate treatment is questionable.
Since publication of a 2000 article by Lau et al,17 the use of high-dose PPIs has been widely studied. Two consensus documents in 2002 and 2003 endorsed the use of high-dose PPI treatment.44,45 However, because of the cost of high-dose PPIs, more evidence is needed to recommend this therapy. A 2008 review article by Gralnek et al37 recommended the use of high-dose PPIs in high-risk patients. However, the authors questioned the goal of “intragastric pH above 6” as a reliable proxy for adequate treatment. Our meta-analysis provides insufficient evidence to support the use of high-dose over non–high-dose PPIs. Only 7 randomized controlled studies compared high-dose with non–high-dose PPIs. No study showed any significant difference in the associated rates of rebleeding, surgical intervention, or mortality. Despite the absence of statistical heterogeneity, there was much clinical heterogeneity across these 7 trials in study design, patient inclusion, endoscopic treatment, route and dose of PPI administration in control groups, and outcomes assessment. We anticipate the performance of more high-quality randomized controlled trials (such as the recent study by Sung et al20) that will include diverse racial/ethnic groups, standardized endoscopic diagnosis and treatment, double-blind treatment design, and outcomes assessment using specific criteria over set periods. In this way, the issue of high-dose PPIs can be explored, and the effect of H2RAs can be reexamined.21
There were 2 limitations to our study. First, data synthesis in this meta-analysis was not performed according to the intent-to-treat principle, with violations of this principle in the studies by Udd,27 Hung,23 Andriulli,24 and Yüksel25 and colleagues. We strove to reconstruct the results using the intent-to-treat principle. However, detailed data were lacking, and assumptions were difficult to make. Per-protocol analysis tended to overestimate the treatment effect. In studies analyzed using per-protocol analysis, high-dose PPIs showed no benefit. Therefore, we conservatively assumed that, even if these studies had been performed using the intent-to-treat principle, they would not have shown better treatment effects, and the results of this meta-analysis would not have changed. Second, we did not perform funnel plots to detect publication biases. Our study included only 7 studies, most of which enrolled few patients. It had limited power to detect publication bias by visual inspection of funnel plots. Instead, we tried to avoid such biases by adhering to the most stringent guidelines of The Cochrane Collaboration and the Quality of Reporting of Meta-analyses statement. The results of our meta-analysis indicate that, compared with non–high-dose PPIs, high-dose PPIs do not further reduce the rates of rebleeding, surgical intervention, or mortality after endoscopic treatment in patients with bleeding peptic ulcer.
Correspondence: Zui-Shen Yen, MD, MPH, Department of Emergency Medicine, National Taiwan University Hospital and National Taiwan University College of Medicine, 7 Chung Shan S Rd, Taipei, Taiwan 100 (firstname.lastname@example.org).
Accepted for Publication: December 23, 2009.
Author Contributions:Study concept and design: Wang, Ma, Chou, Yen, Yang, Fang, and Chen. Acquisition of data: Wang and Yen. Analysis and interpretation of data: Wang, Ma, and Yen. Drafting of the manuscript: Wang, Ma, Chou, and Yen. Critical revision of the manuscript for important intellectual content: Wang, Ma, Chou, Yen, Yang, and Fang. Statistical analysis: Wang and Yen. Administrative, technical, and material support: Ma, Chou, Yen, Yang, Fang, and Chen. Study supervision: Ma, Yen, Yang, Fang, and Chen.