Correspondence to: Dr R. A. Enns, #300–1144 Burrard St., St. Paul's Hospital, Vancouver, BC, Canada, V6K 2A5. E-mail: email@example.com
Background : The administration of proton pump inhibitors intravenously after endoscopic treatment of peptic ulcers significantly reduces the recurrence of bleeding.
Aim : To evaluate the incremental cost-effectiveness in Canada of intravenous proton pump inhibitor before endoscopic therapy to patients presenting with acute upper gastrointestinal bleeding, compared with endoscopic treatment alone.
Methods : From a third-party payer perspective, we modelled the costs and effectiveness over 60 days of the two approaches using decision analysis. The probabilities of various outcomes, such as re-bleeding and the need for surgery, were taken from the published literature. We included the costs of intravenous proton pump inhibitor, therapeutic endoscopy, surgical procedures and hospitalizations, all expressed in 2001 Canadian dollars.
Results : In a hypothetical cohort of 1000 patients, the intravenous proton pump inhibitor approach resulted in mean savings of $20 700 with 37 re-bleeding episodes averted. The investigation of uncertainty resulted in a likelihood of intravenous proton pump inhibitor being cost-effective of at least 0.73.
Conclusion : It is common in Canada to administer intravenous proton pump inhibitors to patients with upper gastrointestinal bleeding even before endoscopic confirmation of bleeding peptic ulcers. Our results suggest that this approach has a high likelihood of being cost-effective.
Upper gastrointestinal bleeding accounts for approximately 150 000 hospital admissions each year in the USA and about one-half originate from peptic ulcer disease.1 Although equivalent statistics in Canada are not available, based on the ratios of the populations, a crude estimate of approximately 15 000 per year would probably be close to the actual number. The high incidence of upper gastrointestinal bleeding therefore represents a substantial economic burden on the Canadian health care system.
In patients who present to the emergency room with signs of upper gastrointestinal bleeding from peptic ulcer disease or another cause, therapy with acid suppression has become one of the mainstays of treatment.2–6 Although endoscopic therapy of peptic ulcers is well accepted,7–11 until recently there was no compelling evidence to support the pharmacological suppression of acid for acute bleeding. Proton pump inhibitors have improved the management of peptic ulcer disease and the symptoms of other upper gastrointestinal diseases.12–18 However, their use in treating acute haemorrhagic peptic ulcers is in the process of being defined.19,20 In the setting of upper gastrointestinal bleeding, the use of continuous infusions yields profound acid suppression with improved clinical outcomes.21–25
The literature suggests that approximately 50% of upper gastrointestinal bleeding is caused by peptic ulcer disease.26 Of patients with bleeding due to peptic ulcer disease, approximately one-third will have endoscopic features associated with a high risk of re-bleeding (e.g. active haemorrhage, visible vessel, overlying clot).9 Therefore, less than one-half of patients with acute upper gastrointestinal bleeding are likely to benefit from intravenous proton pump inhibitor.20,22–25,27
After peptic ulcer with high-risk stigmata had been confirmed endoscopically, one study from Hong Kong showed that intravenous proton pump inhibitor was able to significantly reduce the incidence of re-bleeding and the need for surgical intervention, as well as the length of stay in hospital.20 Therefore, the use of intravenous proton pump inhibitor in conjunction with standard endoscopic treatment of peptic ulcer would have an impact on the costs of treating this condition.
In the Hong Kong study, all patients underwent endoscopy within 24 h after admission. Additionally, the trial was designed so that all enrolled patients were shown to have endoscopically confirmed gastric or duodenal ulcers with stigmata that had been documented to be at high risk for re-bleeding.20 In practice, however, the sequence of treatments for patients presenting with upper gastrointestinal bleeding is different from that of the Hong Kong protocol. First, 80% of bleedings cease spontaneously, so that an emergency endoscopy is often delayed.28 Second, many physicians may elect to start patients on intravenous proton pump inhibitors whilst awaiting endoscopic confirmation of the diagnosis, as is common practice with intravenous H2-receptor antagonists.29 Because current practice is often to start acid suppressants without endoscopic confirmation,2,5,6,26 we hypothesized that it might be feasible from an economic standpoint to give intravenous proton pump inhibitor to all patients with acute upper gastrointestinal bleeding before endoscopy. We constructed a model to determine the cost-effectiveness of giving intravenous proton pump inhibitor to all patients presenting to emergency with upper gastrointestinal bleeding.
The purpose of this study was to determine whether the administration of intravenous proton pump inhibitor to all patients presenting with symptoms of upper gastrointestinal bleeding in Canadian emergency rooms would be cost-effective.
Decision analytical model
To investigate the costs, effectiveness and incremental cost-effectiveness, we constructed a decision analytical model comparing two different treatment approaches for patients presenting to the emergency department with an acute undifferentiated cause of upper gastrointestinal bleeding.30,31 The first approach was to administer intravenous proton pump inhibitor to all patients presenting to the emergency department with signs (e.g. melaena or haematemesis) of upper gastrointestinal bleeding. The second approach was to withhold intravenous proton pump inhibitor from all patients and provide standard treatment with endoscopic therapy after confirmation of a peptic ulcer with high-risk stigmata.
The study was conducted from a third-party payer perspective, such as a provincial Ministry of Health, with a 60-day time horizon. Because of the short time horizon of the study, discounting of costs and outcomes was not applied. We assumed that all patients would undergo endoscopy on average at 24 h of presentation. Those that had peptic ulcer disease with active bleeding, visible vessel or overlying clot would continue with intravenous proton pump inhibitor for 72 h. For patients with upper gastrointestinal bleeding due to other causes, we assumed that intravenous proton pump inhibitor would be stopped immediately after endoscopy.
The outcomes for individuals undergoing endoscopic treatment followed one of two paths: successful treatment or a recurrence of bleeding requiring further treatment. Depending on the severity, a re-bleed was treated either with a repeat endoscopic treatment or surgically. Suture ligations of the ulcer site were the most common surgical procedures, with vagotomy and partial gastrectomies being used less commonly. Patients treated successfully (e.g. no recurrent bleed) were assumed to be dispensed an oral proton pump inhibitor daily for 30 days. The probabilities used to populate the tree and the relative risks of major outcomes have been reported elsewhere (Gagnon YM, et al, 2002, submitted).
Resource utilization and unit costs of treatment
Hospitalization costs were estimated using: (i) year 1999–2000 patient-level discharge data from eight Canadian provinces on lengths of stay and resource intensity weights from the Canadian Institute for Health Information;32 and (ii) the average cost per weighted case for Canada published by the Institute of Health Economics.33 Other categories of resource utilization in Canada are shown in Table 1.
Table 1. Values for health resource utilization in Canada
Health resource utilization item
Values for the length of stay were estimated from year 1999–2000 hospital discharge data from the Canadian Institute for Health Information (CIHI). Requirements for anaesthesia were derived from expert opinion elicited at St. Paul's Hospital, Vancouver, BC, Canada. The recommended number of days for oral proton pump inhibitor (PPI) course of treatment was obtained from the manufacturer's product monograph for pantoprazole (Pantoloc™); s.d., standard deviation; N, number of procedures.
Unit costs for the resources were derived from a number of sources, shown in Table 2. All costs reported are in 2001 Canadian dollars. Currently, pantoprazole (Pantoloc™) is the only intravenous proton pump inhibitor available for routine use in Canada and the cost was obtained from one large teaching hospital in Vancouver.
Table 2. Costs of resources for treating upper gastrointestinal bleeding in Canada
CIHI RIW™, Canadian Institute for Health Information resource intensity weight; IHE, Institute of Health Economics; IMS HEALTH, Intercontinental Marketing Services.
All costs [except oral proton pump inhibitor (PPI)] rounded to their nearest dollar and are reported in 2001 Canadian dollars. Cost of intravenous proton pump inhibitor (IV PPI) is based on that of intravenous pantoprazole.
IV PPI (first 24 h: 80-mg bolus + 8 mg/h continuous infusion)
St. Paul's Hospital, Vancouver, BC, Canada
IV PPI (2nd day: 8 mg/h continuous infusion)
St. Paul's Hospital, Vancouver, BC, Canada
IV PPI (3rd day: 8 mg/h continuous infusion)
St. Paul's Hospital, Vancouver, BC, Canada
Per diem hospitalization for endoscopic treatment
Derived using CIHI RIW™ and publication from IHE33
Per diem hospitalization for suture ligation
Derived using CIHI RIW™ and publication from IHE33
Per diem hospitalization for partial gastrectomy
Derived using CIHI RIW™ and publication from IHE33
Consistent with the clinical objective, the effectiveness measure in our study was the expected number of bleeding recurrences in a hypothetical cohort of 1000 patients. The primary outcome was the incremental cost-effectiveness of the intravenous proton pump inhibitor treatment strategy vs. no intravenous proton pump inhibitor. To calculate the expected costs for each strategy, we first multiplied the mean resource use values by their corresponding unit costs for every item and summed over each outcome event. Then, we weighted the total dollar amount with the corresponding conditional probability for a patient to follow a particular sequence of events. Finally, summing over all weighted values assigned to the different sequence of events yielded the expected cost of each treatment strategy.
One-way sensitivity analysis was conducted on all variables included in the model by varying the original values by plus or minus 20%. Because each variable in the model typically does not vary over a range of plausible values whilst all the others stay at a constant level, one-way sensitivity analysis is likely to underestimate the uncertainty associated with the results of the cost-effectiveness analysis.36–38 To undertake a more realistic analysis of uncertainty, we also performed a probabilistic sensitivity analysis in which uncertainty in all variables is investigated simultaneously by randomly choosing values from pre-specified distributions over a large number of times.36,39–42 Further details of this procedure can be found elsewhere (Gagnon YM, et al, 2002, submitted). All analyses were performed in Excel with add-on software Crystal Ball.
Decision analytical model
In a hypothetical cohort of 1000 patients presenting with upper gastrointestinal bleeding to emergency, it was estimated that approximately 216 would have peptic ulcers with high-risk stigmata (Figure 1).
Figure 2 shows the decision analytical model for treating upper gastrointestinal bleeding with or without intravenous proton pump inhibitor. For example, the uppermost branch of the figure shows a patient given intravenous proton pump inhibitor undergoing endoscopic treatment (probability 0.22), failing treatment (probability 0.07), having a serious re-bleed requiring surgical repair (probability 0.25) and finally dying (probability 0.12). The conditional probability of a patient dying via this pathway was therefore 0.000462 and the total resource cost was $13 530. In comparison, the conditional probability for a patient not given intravenous proton pump inhibitor following the same sequence was 0.000911 with a total resource cost of $13 220.
The baseline results indicate that the universal approach using intravenous proton pump inhibitor yields lower costs and better outcomes in a cohort of 1000 patients (Table 3). The incremental saving favouring the intravenous proton pump inhibitor treatment was $37 840, with 40 re-bleeds prevented, resulting in a negative incremental ratio of ($946) per re-bleed averted. Therefore, using intravenous proton pump inhibitor universally would be considered a ‘dominant’ therapy when compared with endoscopic treatment alone.
Table 3. Results of base case cost-effectiveness analysis of intravenous proton pump inhibitor vs. no intravenous proton pump inhibitor in a hypothetical cohort of 1000 patients
Expected costs ($)
Expected no. of re-bleeds
ICER ($/re-bleed averted)
ICER, incremental cost-effectiveness ratio; IV PPI, intravenous proton pump inhibitor.
Monetary values within parentheses indicate savings.
IV PPI started prior to endoscopy + endoscopic treatment
The base case results were robust to changes of plus or minus 20% of the original values specified for the variables in the model (Table 4). Only when the baseline probability of a first re-bleed without intravenous proton pump inhibitor was lowered by 20% did the incremental cost-effectiveness ratio become positive at $105 per re-bleed averted, indicating increased costs for health benefit. Whilst this scenario results in savings, this value of the ratio would still be consistent with the intravenous proton pump inhibitor approach being cost-effective.
Table 4. Results of one-way sensitivity analysis on selected variables (± 20% of original values in base case analysis)
Variables were varied by ± 20% of their original values. ICERs with incremental costs in parentheses indicate that intravenous proton pump inhibitor is cost saving with better outcomes, i.e. a dominant therapy.
Baseline probability of a first re-bleed
Proportion of patients with confirmed peptic ulcer bleed
Proportion of patients with ulcer amenable to endo Tx (high-risk stigmata)
Hospital length of stay for endo Tx
Per diem cost hospitalization for endo Tx
Relative risk of first re-bleed
Baseline probability of second re-bleed
Proportion of patients undergoing suture ligation
Hospital length of stay for partial gastrectomy
Per diem cost hospitalization for partial gastrectomy
Hospital length of stay for suture ligation
Per diem cost hospitalization for suture ligation
Relative risk of second re-bleed
Relative risk of need for surgery after first re-bleed
Baseline probability of first re-bleed requiring surgical procedure
The results from the probabilistic cost-effectiveness analysis indicate that the intravenous proton pump inhibitor approach yields lower costs and better outcomes in a cohort of 1000 patients. The incremental saving favouring intravenous proton pump inhibitor treatment was $20 700, with a mean of 37 re-bleeds averted, resulting in a negative incremental ratio of ($560) per re-bleed averted. On the other hand, the 95% confidence interval (CI) had a lower limit indicating a lower cost and better outcome with intravenous proton pump inhibitor and an upper limit showing a higher cost with better outcome [95% CI: ($2490); $6710]. The likelihood that intravenous proton pump inhibitor would be cost saving was relatively high, with a probability of 0.63 (data not shown).
Intravenous proton pump inhibitors have been shown to be efficacious in reducing bleeding recurrence when used as an adjunct to endoscopic treatment. In practice, however, other acid suppressants and intravenous proton pump inhibitor are often initiated before endoscopic confirmation of the lesion.26,29 The results presented here indicate that, based on available evidence, universal administration of intravenous proton pump inhibitor to all patients presenting with upper gastrointestinal bleeding in Canada is likely to be cost-effective.
Our model has several limitations. First, we assumed that the administration of intravenous proton pump inhibitor before endoscopic therapy would have the same benefit as that found by Lau et al. when it was given after endoscopic treatment.20 Second, we assumed that intravenous proton pump inhibitor was discontinued in all patients without high-risk endoscopic stigmata and that all patients received endoscopy on average 24 h after presentation. In many circumstances, intravenous proton pump inhibitor may be inadvertently continued or endoscopy may be delayed beyond 24 h. This would serve to increase the costs in the intravenous proton pump inhibitor arm. When using a mean time to endoscopy of 36 h, instead of 24 h, we found that intravenous proton pump inhibitor was cost neutral with better outcomes. Third, we assumed that all patients with lesions other than peptic ulcer disease with high-risk endoscopic stigmata would gain no benefit from acid suppression. Fourth, this study did not assess indirect costs from work absenteeism which would add costs to a higher frequency of re-bleeds, making the universal application of intravenous proton pump inhibitor more attractive.
Comparing the cost-effectiveness of intravenous proton pump inhibitor and other pharmacological agents before and after endoscopy could address the issue of the optimal timing of endoscopic therapy. The recent medical literature suggests that delayed endoscopy is associated with poorer outcomes compared to early endoscopy (i.e. less than 24 h).43,44 Therefore, the investigation of the effectiveness and cost-effectiveness of early acid suppression vs. its initiation after endoscopy would be useful for gaining insight into whether the benefits of intravenous acid suppression would be comparable. Although it is not known whether there would be any additional medical benefit of early intravenous proton pump inhibitor, our results suggest that this treatment approach is likely to be cost-effective when used in this patient population.
To further investigate cost-effectiveness, we calculated a cost-effectiveness acceptability curve45–48 (data not shown). This curve indicates the probability of the cost-effectiveness of intravenous proton pump inhibitor for different monetary values that an individual would be willing to pay for a treatment to prevent a bleeding recurrence. When an individual is not prepared to disburse anything, we found that the treatment has a probability of 0.63 of being cost-effective. This probability rises sharply with an increasing willingness to pay, so that it reaches 0.72 when the hypothetical value is set at $500 and 0.95 when it is $3700.
Because there are no published studies estimating the willingness to pay for a treatment to prevent an upper gastrointestinal bleed, analyses investigating similar conditions may give an indication of the possible monetary value that would be assigned to it. One study evaluating complete symptom relief in gastro-oesophageal reflux disease reported values of US$182 per month,49 whilst another, for the treatment of von Willebrand's disease, reported an annual willingness to pay of US$2178.50 Thus, we may estimate a conservative lower bound that the probability of intravenous proton pump inhibitor being cost-effective is approximately 0.73 or higher. More insight would be gained by investigating the willingness to pay for a treatment to prevent the recurrence of peptic ulcer bleeding.
We have previously shown, in the USA, that the early administration of intravenous proton pump inhibitor was unambiguously cost saving with better outcomes (Gagnon YM, et al, 2002, submitted). The difference between the earlier USA study (i.e. unambiguously cost saving) and the current Canadian study (i.e. likely to be cost-effective) originates from the marked difference between the two health care systems.51,52 Because of the much higher costs in the USA, a treatment approach such as intravenous proton pump inhibitor, that prevents further hospital admissions and procedures, will have a substantially larger impact on the expected treatment costs than that in Canada. With the expected costs behaving distinctively in the USA and in Canada, the cost-effectiveness of the intravenous proton pump inhibitor approach also differs, stressing the need for country-specific studies.
In conclusion, based on its cost-effectiveness, we suggest that the routine administration of intravenous proton pump inhibitor to all patients presenting to Canadian emergency rooms with upper gastrointestinal bleeding should be considered as standard practice, particularly for patients in whom urgent endoscopy is delayed.