Helicobacter pylori eradication decreases ulcer recurrence and should prevent recurrent ulcer haemorrhage.
Helicobacter pylori eradication decreases ulcer recurrence and should prevent recurrent ulcer haemorrhage.
By meta-analysis, to compare treatment of H. pylori infection with other approaches to prevent recurrent ulcer haemorrhage and, by cost minimization analysis, to determine the least costly strategy.
We searched for randomized, controlled trials comparing treatment of H. pylori infection with ulcer healing alone or with maintenance therapy in preventing recurrent ulcer haemorrhage. We calculated the relative and absolute risk reductions and numbers needed to treat.
Treatment of H. pylori infection decreased recurrent bleeding by 17% (numbers needed to treat=6) compared with ulcer healing treatment alone. Compared with ulcer healing treatment followed by maintenance therapy, recurrent bleeding was decreased by 4% (numbers needed to treat=25). Decision model-based cost minimization analysis demonstrated that treatment of H. pylori infection was the least costly strategy unless the incidence of complicated recurrences after treatment was over 6%, or the cost of confirming eradication was over $741.
Treatment of H. pylori infection is superior to ulcer healing treatment with or without maintenance therapy in preventing recurrent ulcer haemorrhage. All patients with ulcer bleeding should be tested for H. pylori infection and appropriately treated if positive.
Helicobacter pylori causes chronic active gastritis and is a major aetiological factor in peptic ulcer disease.1, 2 Haemorrhage is one of the most frequent complications of peptic ulcer disease and is associated with substantial morbidity, mortality and health-care costs.3, 4 Patients with recurrent haemorrhage, particularly the chronically ill and elderly, have excessive morbidity and mortality.4 The prevention of recurrent ulcer haemorrhage is therefore a highly desirable clinical goal. An expert committee convened by the National Institutes of Health recommended long-term maintenance therapy with gastric acid inhibitory agents in patients who had experienced an episode of bleeding peptic ulcer, even after treatment of H. pylori infection.5 However, these recommendations were promulgated before the results of studies examining re-bleeding rates after treatment of the infection were known. The eradication of H. pylori infection in patients with uncomplicated peptic ulcer disease decreases ulcer recurrence and is cost-effective.6–10 Eradication should also prevent recurrent bleeding from peptic ulcers. However, whether the effect of H. pylori eradication is superior to long-term maintenance therapy with an antisecretory drug is not known. We therefore conducted a meta-analysis of randomized, controlled trials to evaluate the role of treatment of H. pylori infection in the prevention of recurrent peptic ulcer haemorrhage. Using data obtained from the meta-analysis, a decision-based cost minimization analysis was performed to determine whether treatment of H. pylori infection was the least costly strategy.
We conducted a fully recursive literature search using the MEDLINE database for publications from January 1966 until December 2000. We searched for randomized, controlled trials that compared the treatment of H. pylori infection with initial ulcer healing treatment (without treatment of H. pylori infection), with or without subsequent maintenance therapy to prevent recurrent peptic ulcer haemorrhage. Keywords for the search were ‘Helicobacter pylori’ or ‘Campylobacter pylori’ or ‘pylori’ (text word), combined with ‘gastrointestinal bleeding’ or ‘peptic ulcer haemorrhage’ or ‘ulcer haemorrhage’ (text word). We searched for publications in abstract form using the article references and official proceedings of all major North American and European meetings. Dual publications were excluded. If more than one version of the same trial had been retrieved, only the most recent data were used.
To be included in the meta-analysis, publications had to contain information about the diagnosis of H. pylori infection, the therapy used to treat it, the therapy used in the control group, the compliance with medications, the frequency of use of NSAIDs, the duration of follow-up and the rates of eradication and recurrent bleeding. Treatment regimens in the active treatment and control groups were noted, together with documented H. pylori eradication rates. The rates of recurrent bleeding in each treatment arm of the individual trials were noted.
All authors independently reviewed each study. The homogeneity of trials was measured by the Breslow–Day method before pooling for subsequent meta-analysis.11 Individual trials were then pooled and the overall re-bleeding rates, together with the 95% confidence intervals (CI) in each treatment arm, were calculated. We also calculated the relative risk reduction (RRR), absolute risk reduction (ARR) and numbers needed to treat (NNT) to prevent one episode of recurrent bleeding for individual trials and for the pooled data.12 The NNT values were rounded to the nearest whole number.
A decision model was designed using Data 3.5 decision analysis software (Treeage Inc., Williamstown, MA, USA) to compare the direct costs related to the management of ulcer disease over a 12-month period for three strategies. These were: (i) treatment of H. pylori infection; (ii) initial ulcer healing treatment followed by subsequent maintenance therapy with an H2-receptor antagonist; and (iii) ulcer healing treatment only (without treatment of H. pylori infection or subsequent maintenance therapy; Figure 1). The model incorporated the costs of drugs for initial ulcer management, diagnostic endoscopy and interventions required in the case of symptomatic ulcer recurrence. Cost inputs were estimates for direct hospital costs obtained from our institutions (Table 1). For sensitivity analysis, costs were varied between 50% and 1000% of the initial estimate. Initial medical management options included treatment of H. pylori infection (a 2-week course of two antibiotics and a proton pump inhibitor), ulcer healing therapy alone with an H2-receptor antagonist for 8 weeks, or ulcer healing therapy followed by 10 months of maintenance therapy with an H2-receptor antagonist at a maintenance dose. Symptomatic ulcer recurrence was considered to be complicated (that is, associated with bleeding) or uncomplicated (that is, associated with pain but without bleeding). We considered that an uncomplicated recurrence would require repeat out-patient endoscopy and ulcer healing therapy with an H2-receptor antagonist for 8 weeks, and that a complicated recurrence would require repeat endoscopy, ulcer healing therapy and hospitalization for 5–10 days. The cost of managing a complication that proved fatal was assumed to be 10 times the cost of managing a non-fatal ulcer complication. One-way sensitivity analysis and threshold analysis were performed on all variables to determine which most affected the total costs of management and whether the optimal strategy was likely to change. We derived the probability estimates for ulcer recurrences and death and the ranges for the sensitivity analysis from the meta-analysis.
We initially identified 10 randomized, controlled trials that appeared to be appropriate for inclusion in the meta-analysis.13–22 Six had compared treatment of H. pylori infection with ulcer healing treatment alone.13–18 Four had compared it with initial ulcer healing treatment and subsequent maintenance therapy with an H2-receptor antagonist or omeprazole.19–22 All were published in English. Seven were full peer-reviewed publications,13–16, 19–21 three of which had also been published as abstracts.17, 18, 22 The abstracts were excluded from the meta-analysis.17, 18, 22 There was complete agreement among participating authors regarding trial inclusion and exclusion.
Table 2 lists details of the four trials comparing treatment of H. pylori infection with ulcer healing treatment alone. Three trials were restricted to patients with duodenal ulcer haemorrhage;14–16 one included patients with either gastric or duodenal ulcer haemorrhage.13 There was no significant heterogeneity among the trials (P=0.36), which were therefore combined in the meta-analysis. One hundred and thirty-three patients received some form of treatment aimed at eradicating H. pylori infection; 129 control patients received ulcer healing treatment without subsequent maintenance therapy (Table 3). Of the six patients in the treatment group who had recurrent haemorrhage, four had failed eradication of H. pylori infection, one was reported to have recurrence of H. pylori infection and one was using an NSAID. The pooled RRR for recurrent bleeding was 79%, with a range for individual trials of 50–100%. The pooled ARR was 17% (95% CI, 9–25%). Figure 2 displays the ARR and 95% CI for individual trials and the pooled data. Given a pooled ARR of 17%, the NNT in order to prevent one recurrent ulcer haemorrhage was six (95% CI, 4–11).
Three of the four trials reported rates of recurrent ulcer with pain but without bleeding.14–16 Five patients in the treatment group (4%) had a recurrent ulcer with pain, compared with 34 (30%) in the control group. The pooled RRR for recurrent ulcer with pain but without bleeding was 85% (range, 69–94%) and the pooled ARR was 25% (95% CI, 16–34%); the NNT to prevent one painful recurrence was four (95% CI, 3–6). All treated patients who developed recurrent ulcer with pain but without bleeding had evidence of failed eradication of H. pylori infection.
When all ulcer recurrences with bleeding, or with pain but without bleeding, within 12–24 months of follow-up, were combined, there were 11 (10%) patients in the treatment group and 58 (50%) patients in the control group. The pooled RRR for recurrent ulcer with bleeding or pain was 81% (range, 75–88%) and the pooled ARR was 41% (95% CI, 30–52%); the NNT was two (95% CI, 2–3).
Table 4 lists details of the three trials comparing treatment of H. pylori infection with initial ulcer healing treatment followed by maintenance therapy. Each trial included patients with either gastric or duodenal ulcer haemorrhage. There was no significant heterogeneity among the trials (P=0.81), which were therefore combined in the meta-analysis. Two hundred and fifty-seven patients received some form of treatment aimed at eradicating H. pylori infection; 213 control patients received ulcer healing treatment followed by maintenance therapy with either an H2-receptor antagonist19–21 or omeprazole21 (Table 5). Of the four patients with recurrent haemorrhage in the treatment group, two had failed eradication of H. pylori infection and the other two had used NSAIDs. The pooled RRR for recurrent ulcer bleeding was 72%, with a range for individual trials of 49–100%. The pooled ARR was 4.1% (95% CI, 0.6–7.5%); the NNT to prevent one recurrent ulcer haemorrhage was 25 (95% CI, 13–167). Figure 3 displays the ARR and 95% CI for individual trials and the pooled data.
All three trials also reported rates of recurrent ulcer with pain but without bleeding. Five patients in the treatment group (2%) had a recurrent painful ulcer without bleeding, compared with 20 (9%) in the control group. One of the five patients in the treatment group had persistent infection with H. pylori.19 Data regarding H. pylori status and NSAID use for the other four patients were not reported.20 The pooled RRR for ulcer recurrence with pain but no bleeding was 79% (range, 76–100%) and the pooled ARR was 7.4% (95% CI, 3–12%); the NNT to prevent one painful ulcer recurrence was 13 (95% CI, 8–33).
When all ulcer recurrences with bleeding, or with pain but without bleeding, within 12–24 months of follow-up, were combined, there were nine (4%) patients in the treatment group and 32 (15%) patients in the control group. The pooled RRR for ulcer recurrence with bleeding or pain was 77% (range, 77–89%) and the pooled ARR was 12% (95% CI, 6–17%); the NNT to prevent one ulcer recurrence with bleeding or pain was nine (95% CI, 6–17).
Based on the initial model inputs (Table 6), treatment of H. pylori infection was the least costly means of decreasing the overall 1-year costs associated with symptomatic ulcer recurrence. The estimated 1-year costs were $620 for treatment of H. pylori infection, $1260 for initial ulcer healing treatment followed by subsequent maintenance therapy, and $4280 for initial ulcer healing treatment alone. Sensitivity analysis showed that the overall cost was most dependent on the estimated cost associated with the management of complications that ultimately would have proved fatal and the cost associated with treating non-fatal recurrent ulcer complications, and the probability of a complicated recurrence. Threshold analysis revealed possible changes in the optimal management strategy from treatment of H. pylori infection to initial ulcer healing treatment followed by maintenance therapy if the risk of a complicated ulcer recurrence after H. pylori eradication was higher than 6%, or if the cost of treating H. pylori infection (including the cost of confirming its eradication) was over $741 (Figures 4 and 5).
Peptic ulcer haemorrhage is a common problem, accounting for over 250 000 hospital admissions annually in the USA.4 Mortality rates vary from 8% to 29% depending upon the presentation of bleeding,23 and despite advances in endoscopic treatment and the widespread use of potent gastric acid inhibitory agents. Long-term maintenance antisecretory treatment after the initial ulcer bleed has been widely accepted as the standard of care.5 However, there are only few controlled data on the efficacy of maintenance therapy with an H2-receptor antagonist in preventing recurrent ulcer bleeding. Murray et al. found no significant difference in the rates of recurrent ulcer bleeding between patients treated with ranitidine, 150 mg, and those treated with placebo for up to 3 years.24 Jensen et al. also performed a randomized, placebo-controlled trial of ranitidine, 150 mg, in 55 patients who had had an episode of duodenal ulcer haemorrhage.25 The mean follow-up was 61 weeks. Recurrent ulcer bleeding occurred in three of the 32 patients (9%) on ranitidine and in 12 of the 33 patients (36%) on placebo. A crude re-analysis of these figures gives an ARR of 27% (NNT=4). However, the recurrent bleeding rate, despite ranitidine, of 9% in this study25 is still unacceptably high. Therefore, it is illuminating to examine the effects of treatment of H. pylori infection on subsequent ulcer bleeding. Eradication of the infection may eliminate ulcer recurrence. It therefore seems logical to assume that it would also decrease the rate of recurrent ulcer bleeding. Three trials have found a statistically significant decrease in the rate of recurrent ulcer haemorrhage,13–15 while one16 has shown an ARR for recurrent ulcer haemorrhage whose 95% CI encompasses zero (Figure 2). The pooled data show a clear beneficial effect of treating H. pylori infection in decreasing recurrent ulcer haemorrhage.
When treatment of H. pylori infection is compared with initial ulcer healing and subsequent maintenance therapy with an H2-receptor antagonist or omeprazole, no single trial demonstrates a statistically significant benefit of either treatment approach over the other.19–21 The pooled ARR for each trial has a 95% CI that encompasses zero (Figure 3). However, by pooling the results of these three non-heterogeneous trials, there is a small but statistically significant advantage in treating H. pylori infection over maintenance therapy. That the effect is quantitatively small is reflected in the comparatively large NNT of 25. Effective treatment of H. pylori infection may also be less costly than indefinite maintenance therapy, because the financial outlay for medication in the former approach is not cumulative as with continuous maintenance therapy. As evident from our cost minimization analysis, the small advantage in preventing ulcer recurrence results in a substantial health-care cost saving. Furthermore, there is increasing evidence that continued administration of an H2-receptor antagonist leads to pharmacological tolerance, with a decrease in its effect in controlling gastric acid secretion.26 In addition, patient compliance with maintenance H2-receptor antagonist treatment may not be sustained.
As with any meta-analysis, there are possible limitations in combining results from separate trials. Certain factors may explain some of the variability in the results of the different trials included. These include minor variation in inclusion and exclusion criteria, different therapies for initial ulcer healing and subsequent maintenance, compliance rates with medication and duration of follow-up. However, our pooled results are probably appropriate and accurate as they are based on the combination of non-heterogeneous trials that, taken together, comprise a large number of patients.
The cost minimization analysis confirmed that effective treatment of H. pylori infection is the least costly strategy for the prevention of recurrent ulcer haemorrhage, and that total costs are most affected by the costs associated with managing recurrent complications, some of which would be fatal. The H. pylori treatment strategy remains less costly than maintenance therapy even if the cost of confirming H. pylori eradication is as high as $741. However, if treatment of H. pylori infection does not decrease the incidence of complicated recurrences to below 6%, maintenance therapy may be a more reasonable option.
The eradication of H. pylori infection following an ulcer haemorrhage is already strongly recommended on the grounds of clinical efficacy.27 Therefore, it is particularly disappointing that relatively few patients admitted to hospital with ulcer complications appear to be tested for the infection or to be treated for it if found positive. A retrospective study of 1580 such patients admitted to US hospitals found that only 56% were tested for H. pylori infection or appropriately treated for it.28
Re-infection with H. pylori is infrequent in adults.29 Therefore, effective treatment of the infection as soon after the initial bleeding episode as is feasible is likely to confer long-term protection from further episodes of bleeding. However, this assumes that the patient is adequately compliant with a regimen of proven and acceptably high efficacy and is not subsequently exposed to another ulcerogen, such as aspirin or an NSAID. In the trials included here, most recurrent haemorrhages occurred in patients who had persistent (or, conceivably, recurrent) H. pylori infection or who had been taking aspirin or an NSAID. Current national guidelines suggest that patients who have had an ulcer complication should be tested after treatment of H. pylori infection to confirm its eradication.30 By detecting those who have failed H. pylori eradication therapy and treating them with another appropriate antibiotic regimen, this should further decrease ulcer recurrence and decrease health-care costs. The most appropriate post-treatment test to employ should be determined by local availability and relative cost.