Meta-analysis: the efficacy of intravenous H2-receptor antagonists in bleeding peptic ulcer

Authors


Correspondence to: Dr C. W. Howden, Division of Gastroenterology and Hepatology, Northwestern University Medical School, 676 N. St. Clair Avenue, Suite 880, Chicago, IL 60611, USA. E-mail: c.howden@northwestern.edu

Summary

Background : Although a previous meta-analysis found that intravenous H2-receptor antagonists were only weakly beneficial in bleeding gastric ulcer and of no benefit in bleeding duodenal ulcer, patients with ulcer bleeding continue to receive such treatment.

Aim : To re-evaluate the efficacy of intravenous H2-receptor antagonists in ulcer re-bleeding, surgery and mortality by updating the previous meta-analysis.

Methods : After two independent literature searches, randomized, placebo-controlled trials of intravenous H2-receptor antagonists in bleeding ulcer published between 1984 and 2000 were added to those from the initial meta-analysis. Pooled rates of re-bleeding, surgery and death were re-calculated, together with the relative risk reduction, absolute risk reduction, number needed to treat and Mantel–Haenszel odds ratio.

Results : Intravenous H2-receptor antagonists did not significantly reduce re-bleeding, surgery or death in bleeding duodenal ulcer. There were small but significant reductions in re-bleeding, surgery and death in bleeding gastric ulcer; the absolute risk reductions were 7.2%, 6.7% and 3.2%, respectively.

Conclusions : Intravenous H2-receptor antagonists are of no value in bleeding duodenal ulcer, although they may be mildly beneficial in bleeding gastric ulcer. Because proton pump inhibitors have a greater inhibitory effect on gastric acid secretion than H2-receptor antagonists, they may be more effective in ulcer bleeding and should be further evaluated for that indication.

Introduction

The treatment of actively bleeding peptic ulcers remains a challenge. The possible benefit of the intravenous administration of H2-receptor antagonists (H2RA) has stimulated interest for some time. In a 1985 meta-analysis, Collins & Langman studied randomized controlled trials that had compared intravenous H2RA with placebo in bleeding duodenal ulcer and gastric ulcer.1 The clinical end-points assessed werere-bleeding, surgical intervention and mortality. They were unable to demonstrate any statistically significant benefit for intravenous H2RA over placebo for any of these end-points in bleeding duodenal ulcer. However, they found small, but statistically significant, benefits in bleeding gastric ulcer. Since the publication of that meta-analysis, further randomized controlled trials have compared intravenous H2RA with placebo in patients with bleeding ulcers. These have included a large, multicentre, randomized controlled trial from the UK and Ireland2 which used a high dose of intravenous famotidine that had previously been shown to elevate intragastric pH to > 6 for up to 24 h in patients with actively bleeding ulcers.3 However, the randomized controlled trial failed to demonstrate significant reductions in any of the end-points studied.2

Despite the apparently weak level of supportive evidence, patients with undiagnosed upper gastrointestinal bleeding and those with known bleeding from peptic ulcer disease are likely to receive intravenous H2RA treatment. With the inclusion of more recent randomized controlled trials into the database originally compiled by Collins & Langman,1 we aimed to re-evaluate the evidence for intravenous H2RA in ulcer bleeding.

Materials and methods

Using the MEDLINE and PREMEDLINE databases, two of the authors (JEL, GIL) performed independent, fully recursive literature searches for randomized controlled trials comparing intravenous H2RA with placebo in actively bleeding peptic ulcers. The search was confined to the period 1984–2000 and any relevant randomized controlled trials identified were added to those that had been included by Collins & Langman.1 The search strategy employed is described in the Appendix. We also searched for abstracts presented at major meetings.

For a trial to be included in the meta-analysis, adequate information had to be extractable concerning the numbers of patients randomized to intravenous H2RA and placebo, and the rates of re-bleeding and/or surgical intervention and/or mortality in both the treatment and control groups. Dual publications were excluded; if more than one publication of the same data was retrieved, only the more recent version was included. Pharmacodynamic studies of H2RAs, uncontrolled studies of H2RAs in bleeding ulcer and studies of H2RAs in upper gastrointestinal tract bleeding from unspecified sources or sources other than ulcer disease were also excluded. Each author reviewed all trials independently to assess their suitability for inclusion in the meta-analysis; any differences were resolved by consensus.

We assessed trial homogeneity according to the Breslow–Day method.4 For each of the three pre-determined end-points — re-bleeding, surgical intervention and mortality — the rate on intravenous H2RA was defined as the ‘experimental event rate’ (EER) and the rate on placebo was defined as the ‘control event rate’ (CER).5 The relative risk reduction (RRR) was calculated as (CER − EER)/CER, the absolute risk reduction (ARR) was calculated as CER − EER, and the number needed to treat (NNT) was calculated as the inverse of ARR.6 The 95% confidence interval (CI) for the ARR was calculated for each randomized controlled trial and for the pooled data. The upper and lower limits of the 95% CI for the NNT were defined, respectively, as the inverse of the lower and upper limits of the 95% CI of the ARR. When the 95% CI of the ARR encompassed zero, indicating a statistically non-significant result, no NNT value was computed. As an overall summary statistic of the treatment effect, the pooled Mantel–Haenszel odds ratio (ORM–H) was determined for each outcome of interest using ‘MA 2 × 2’ statistical software (Department of Clinical Epidemiology and Biostatistics, McMaster University, Canada).

Results

Following our literature searches, we were only able to identify five randomized controlled trials2, 7–10 that had not been included in the initial meta-analysis of Collins & Langman.1 The study of Walt et al., which used high-dose intravenous famotidine, reported re-bleeding and surgical intervention rates for peptic ulcer at all sites, but listed mortality rates individually for duodenal ulcer and gastric ulcer.2 A small Hungarian study of cimetidine only reported re-bleeding rates, but did not subdivide these for duodenal ulcer and gastric ulcer.7 An Italian study randomized patients to placebo, ranitidine or somatostatin and reported rates of re-bleeding, surgical intervention and death, but not according to ulcer site.8 An Australian study of intravenous ranitidine reported rates of re-bleeding, surgical intervention and mortality separately for duodenal ulcer and gastric ulcer.9 Another Italian study randomized patients to intravenous ranitidine, somatostatin or placebo, but only reported re-bleeding rates that were not subdivided according to ulcer site.10

Bleeding peptic ulcer (duodenal and gastric ulcer combined)

Thirty randomized controlled trials reported re-bleeding rates on intravenous H2RA and placebo. There was significant heterogeneity among these randomized controlled trials (P = 0.001) due presumably to the different proportions of duodenal ulcer and gastric ulcer included. This makes their pooling for meta-analysis questionable. There were, however, 1901 patients randomized to intravenous H2RA and 1885 to placebo. Pooled rates of re-bleeding were 21.4% on intravenous H2RA and 24.0% on placebo. The RRR was 11% and ARR was 2.6% (95% CI, 0.0–5.3%). The pooled ORM–H was 0.86 (95% CI, 0.74–1.00; P = 0.053).

Twenty-five randomized controlled trials reported rates of surgical intervention on intravenous H2RA treatment and placebo. There was no significant heterogeneity among these randomized controlled trials (P = 0.34). There were 1694 patients randomized to intravenous H2RA treatment and 1706 to placebo. Pooled rates of surgical intervention were 13.4% on intravenous H2RA and 15.8% on placebo. The RRR was 15%, ARR was 2.4% (95% CI, 0.1–4.8%) and NNT was 41 (95% CI, 21–1880). The pooled ORM–H was 0.83 (95% CI, 0.68–1.00; P = 0.057).

Twenty-seven randomized controlled trials reported mortality rates on intravenous H2RA and placebo. There was no significant heterogeneity among these randomized controlled trials (P = 0.57). There were 1865 patients randomized to intravenous H2RA treatment and 1857 to placebo. Pooled mortality rates were 5.6% and 6.8%, respectively. The RRR was 18% and ARR was 1.2% (95% CI, − 0.3–2.8%). The pooled ORM–H was 0.81 (95% CI, 0.62–1.06; P = 0.15).

Figure 1 depicts the pooled ORM–H and 95% CI for each of the three pre-determined end-points of interest in bleeding ulcer (duodenal ulcer and gastric ulcer combined).

Figure 1.

Pooled Mantel–Haenszel odds ratio (ORM–H) and 95% confidence interval (CI) for re-bleeding, surgical intervention and mortality from a meta-analysis of randomized controlled trials comparing intravenous H2-receptor antagonists with placebo in bleeding ulcer (duodenal and gastric ulcer combined).

Bleeding duodenal ulcer

Twenty-three randomized controlled trials reported duodenal ulcer re-bleeding rates on intravenous H2RA and placebo. There was no significant heterogeneity among these randomized controlled trials (P = 0.25). There were 469 patients randomized to intravenous H2RA and 446 to placebo. Pooled rates of re-bleeding were 23.9% on intravenous H2RA and 25.3% on placebo. The RRR was 6% and ARR was 1.5% (95% CI, − 4.1–7.0%). The pooled ORM–H was 0.90 (95% CI, 0.66–1.22; P = 0.48).

Eighteen randomized controlled trials reported rates of surgical intervention between intravenous H2RA and placebo in bleeding duodenal ulcer. There was no significant heterogeneity among these randomized controlled trials (P = 0.82). There were 392 patients randomized to intravenous H2RA treatment and 378 to placebo. Pooled rates of surgical intervention were 15.8% for intravenous H2RA and 19.8% for placebo. The RRR was 20.0% and ARR was 4.0% (95% CI, − 1.4–9.4%). The pooled ORM–H for surgical intervention in bleeding duodenal ulcer with intravenous H2RA was 0.76 (95% CI, 0.52–1.11; P = 0.16).

Twenty-one randomized controlled trials reported rates of mortality in bleeding duodenal ulcer managed by intravenous H2RA or placebo. There was no significant heterogeneity among these randomized controlled trials (P = 0.92). There were 727 patients randomized to intravenous H2RA and 717 to placebo. Pooled mortality rates were 5.0% for intravenous H2RA treatment and 4.3% for placebo. The RRR was − 15.0% and ARR was − 0.6% (95% CI, − 2.8–1.5%). The pooled ORM–H for mortality in bleeding duodenal ulcer with intravenous H2RA was 1.15 (95% CI, 0.72–1.82; P = 0.57).

Figure 2 depicts the pooled ORM–H and 95% CI for each of the three pre-determined end-points of interest in bleeding duodenal ulcer.

Figure 2.

Pooled Mantel–Haenszel odds ratio (ORM–H) and 95% confidence interval (CI) for re-bleeding, surgical intervention and mortality from a meta-analysis of randomized controlled trials comparing intravenous H2-receptor antagonists with placebo in bleeding duodenal ulcer.

Bleeding gastric ulcer

Twenty-four randomized controlled trials reportedre-bleeding rates from bleeding gastric ulcer treated with intravenous H2RA or placebo. There was no significant heterogeneity among these randomized controlled trials (P = 0.77). There were 313 patients randomized to intravenous H2RA treatment and 347 to placebo. Pooled re-bleeding rates were 20.4% for intravenous H2RA and 27.7% for placebo. The RRR was 26%, ARR was 7.2% (95% CI, 0.7–13.7%) and NNT was 14 (95% CI, 7–137). The pooled ORM–H was 0.66 (95% CI, 0.46–0.95; P = 0.02).

Twenty randomized controlled trials reported rates of surgical intervention for bleeding gastric ulcer treated with intravenous H2RA or placebo. There was no significant heterogeneity among these randomized controlled trials (P = 0.47). There were 291 patients randomized to intravenous H2RA and 312 to placebo. Pooled rates of surgical intervention were 14.1% for intravenous H2RA and 20.8% for placebo. The RRR was 32%, ARR was 6.7% (95% CI, 0.7–12.8%) and NNT was 15 (95% CI, 8–139). The pooled ORM–H for surgical intervention in bleeding gastric ulcer managed with intravenous H2RA was 0.61 (95% CI, 0.39–0.93; P = 0.02).

Twenty-two randomized controlled trials reported mortality rates for bleeding gastric ulcer managed by either intravenous H2RA or placebo. There was no significant heterogeneity among these randomized controlled trials (P = 1.00). There were 516 patients randomized to intravenous H2RA treatment and 546 to placebo. Pooled mortality rates were 5.8% for intravenous H2RA and 9.0% for placebo. The RRR was 35%, ARR was 3.2% (95% CI, 0.0–6.3%) and NNT was 32 (95% CI, 16–3844). The pooled ORM–H for mortality from bleeding gastric ulcer managed with intravenous H2RA was 0.68 (95% CI, 0.43–1.06; P = 0.09).

Figure 3 depicts the pooled ORM–H and 95% CI for each of the three pre-determined end-points of interest in bleeding gastric ulcer.

Figure 3.

Pooled Mantel–Haenszel odds ratio (ORM–H) and 95% confidence interval (CI) for re-bleeding, surgical intervention and mortality from a meta-analysis of randomized controlled trials comparing intravenous H2-receptor antagonists with placebo in bleeding gastric ulcer.

Discussion

The management of patients with bleeding peptic ulcer remains a challenge and, in some respects, controversial. Intravenous H2RA treatment is widely used for patients presenting with upper gastrointestinal haemorrhage before diagnostic endoscopy has demonstrated the site or cause of bleeding. Whether this contributes anything to the overall management is unclear and cannot be addressed by this meta-analysis, which focuses on randomized controlled trials among patients known to be bleeding from peptic ulcer.

What is clear from the results of this meta-analysis is that intravenous H2RA treatment provides no additional benefit for patients found to be bleeding from duodenal ulcer. Such patients should receive appropriate endoscopic treatment whenever it is clinically indicated. The remainder can be safely started on oral proton pump inhibitor treatment as soon as it is considered to be clinically appropriate. At least two randomized controlled trials have demonstrated the safety and efficacy of oral proton pump inhibitor treatment in this context.11, 12 Patients who require endoscopic treatment of bleeding duodenal ulcer might benefit from intravenous proton pump inhibitor treatment for up to 72 h, as has been demonstrated in a large randomized controlled trial from Hong Kong.13 Proton pump inhibitors can maintain intragastric pH > 6 for prolonged periods if given in an adequate dose.14, 15 Although a high-dose intravenous infusion of famotidine has been shown to maintain intragastric pH > 6 in patients with actively bleeding ulcers,3 this effect was only observed for up to 24 h. After this period, it is likely that the antisecretory effect would lessen because of the development of pharmacological tolerance — as has previously been shown for ranitidine when given by high-dose intravenous infusion.16 However, tolerance does not develop to proton pump inhibitors during a few days of continuous intravenous infusion,14–17 supporting their potential usefulness in bleeding peptic ulcer. At present, pantoprazole is the only proton pump inhibitor available in the USA as an intravenous formulation. To date, there are limited data available concerning the efficacy of intravenous pantoprazole in bleeding peptic ulcer,18, 19 although further trials are being conducted.

There may, however, be a useful but limited role for intravenous H2RA treatment in patients found to be bleeding from gastric ulcer. We found that the pooled OR for re-bleeding was 0.66, suggesting a 34% reduction attributable to intravenous H2RA treatment. This is very close to the original estimate of Collins & Langman of 0.68, or a 32% reduction in risk.1 Whether or not intravenous H2RA treatment has any significant effect on mortality from bleeding gastric ulcer is less clear. Collins & Langman reported a pooled OR of 0.53 (95% CI, 0.30–0.96), suggesting that intravenous H2RA treatment might reduce mortality from bleeding gastric ulcer by around 47%. Our revised estimate is 0.68, suggesting a reduction of around 32%. However, the 95% CI on that OR encompassed unity (0.43–1.06) and was therefore not statistically significant (P = 0.09). The results of the large randomized controlled trial by Walt et al.,2 which failed to show any benefit of high-dose intravenous famotidine on mortality from bleeding gastric ulcer (or duodenal ulcer), were largely responsible for the non-significant result in our analysis. However, reduced rates of re-bleeding and surgery among patients with bleeding gastric ulcer would still be important clinical benefits even without any reduction in mortality.

In conclusion, we found no evidence to support the continued use of intravenous H2RA treatment for patients with bleeding duodenal ulcer, but moderate evidence for patients with bleeding gastric ulcer. Further study of proton pump inhibitors in bleeding ulcer seems logical and appropriate because of their greater antisecretory effect and because, unlike H2RAs, there is no development of pharmacological tolerance to their use as an intravenous infusion.

Acknowledgements

This work was not financially supported by any outside agency or pharmaceutical company.

Appendix

The search string we used was (‘Peptic Ulcer Hemorrhage’[MESH] OR ‘peptic ulcer hemorrhage’ OR ‘peptic ulcer haemorrhage’ OR ‘bleeding peptic ulcer’ OR ‘bleeding duodenal ulcer’ OR ‘bleeding gastric ulcer’ OR ‘upper gastrointestinal bleeding’ OR ((bleeding OR hemorrhage OR haemorrhage OR haemorrhagic OR hemorrhagic) and (‘upper gastrointestinal’ OR gastroduodenal OR duodenal OR gastric OR ‘peptic ulcer’ OR ‘duodenal ulcer’ OR ‘gastric ulcer’))) and (‘Histamine H2 Antagonists’[MESH] OR ‘histamine H2 antagonists’ OR ‘histamine H2 antagonist’ OR H2RAs OR H2RA OR Cimetidine[MESH] OR cimetidine OR Famotidine[MESH] OR famotidine OR Ranitidine[MESH] OR ranitidine OR Nizatidine[MESH] OR nizatidine).

Ancillary