Cost effectiveness of screening for and eradication of Helicobacter pylori in management of dyspeptic patients under 45 years of age

MODELLING THE TREATMENT OF PEPTIC ULCER PATIENTS

A Markov model was used to represent the costs and outcomes associated with the empirical treatment of recurrent ulcer with cimetidine versus a standard triple eradication treatment based on bismuth. Cimetidine was chosen as the cheapest antisecretory agent. The methods for using Markov models in health services research are well documented.16 17 In brief, Markov models entail splitting a patient's possible prognoses into a number of states. The probabilities defining transition between each of these states are specified over a specific time frame such as a month or a year, and the model is then run over a large number of periods to see how a patient would move between states. The probabilities can be redefined for each treatment being evaluated, and the costs and effects of treatments can be estimated according to the time spent in each state. Figure 1 presents the structure of the Markov model, with ovals representing the possible disease states and arrows indicating possible transitions between those states. A four week time frame was taken for the model: equivalent to the standard duration of a single treatment with H2-receptor antagonists.

Fig 1

Markov model characterising the empirical treatment of dyspepsia with cimetidine. A state of cure is also shown for successful eradication treatment. Patients who receive unsuccessful Helicobacter pylori eradication treatment or who become reinfected are treated conventionally

• Download figure
• Open in new tab
• Download powerpoint

The model was structured with the following assumptions. Under conventional treatment we assumed that patients presenting with peptic type dyspepsia are given a healing course of cimetidine (800 mg a day for four weeks). If after this course the ulcer has not healed (that is, if the symptoms have not resolved), a second healing course is prescribed. Healing rates at four weeks for cimetidine are about 80%18 with almost every ulcer having healed after 12 weeks of treatment.19 For the purposes of this model we assumed that a third healing course would heal all ulcers which persist beyond eight weeks. Once their ulcers heal most patients will enter a “remission” state when they have no ulcer and receive no medication. Each month there is a chance that patients will experience a recurrence of their ulcer, in which case they begin with the first course of ulcer healing drugs once more. Those patients whose ulcers persist beyond eight weeks after the first occurrence or four weeks after the first recurrence and all those patients who experience a second recurrence were assumed to receive long term maintenance treatment once their ulcers had healed. This treatment entails a maintenance dose of cimetidine (about half the healing dose) taken each day for the duration of the model. Ulcers can recur while patients are on maintenance treatment, in which case patients receive healing doses again followed by maintenance doses once the ulcer has healed.

The Markov model presented in figure 1 also shows a state of “cure” for those patients who have received successful eradication treatment. Success of eradication treatment is determined by using a carbon-13 (13C) labelled urea breath test conducted at hospital. Although no test for H pylori infection is perfect, this test has been shown to produce consistently accurate results,20 21 and for the sake of simplicity in this model the 13C-urea breath test is assumed to have a sensitivity and specificity of 100%. When eradication of H pylori is unsuccessful, patients receive a second course of eradication treatment. This second course is assumed to be less effective than the first because of increased antimicrobial resistance. Those patients who become reinfected with H pylori are assumed to experience a recurrence of their ulcer and to receive conventional empirical treatment with H2-receptor antagonists.

The Markov model was used to compare patients with peptic ulcer treated conventionally (who start with the first healing dose state of the model) with ulcer patients receiving eradication treatment (who begin the model in the “cure” state if H pylori was successfully eradicated, otherwise they start with the first healing dose of cimetidine as with conventional treatment). The success rate of eradication treatment depends on the regimen used. A standard triple treatment formulation based on bismuth was used in this model; dosages, costs, and success rates are presented in table 2. The Markov model was run for 10 years to show how the costs of conventional versus eradication treatment accumulate over time for patients with peptic ulcer disease.

CLASSIFYING DYSPEPTIC PATIENTS

To extend the cost effectiveness analysis to a strategy of combined screening for and eradication of H pylori in patients with peptic ulcer, it is necessary to estimate the proportion of patients with the potential to benefit from such a strategy from among the dyspeptic population. Therefore, the second part of the model classified dyspeptic patients by the presence or absence of H pylori infection and by whether or not peptic ulcer is the underlying cause of their symptoms. Figure 2 presents these four categories of dyspepsia in the form of a probability tree. The probabilities at each of the circular “chance nodes” represent the prevalence of H pylori infection among patients with dyspepsia, the proportion of seropositive patients with ulcer disease, and the proportion of seronegative patients with ulcer disease.

Fig 2

Probability tree showing underlying disease of patients presenting with dyspepsia

• Download figure
• Open in new tab
• Download powerpoint

The community based prevalence in the United Kingdom of H pylori in patients under 45 years of age with dyspepsia is unknown. A lower estimate, however, equivalent to the prevalence of H pylori in a randomly selected sample of a community based population aged under 45 years in the United Kingdom ranges between 19% and 36%22; while in patients who have been investigated in hospital and found to have non-ulcer dyspepsia, prevalence figures range from about 40% to 58%.23 We assumed in the base case analysis that the prevalence of H pylori in dyspeptic patients under 45 years of age presenting to their general practitioner was 40%, a midpoint between these two extremes. Variation in this estimate was explored in the sensitivity analysis.

To identify the appropriate patients to receive eradication treatment, all dyspeptic patients receive a serology test for H pylori, and those with positive results receive further investigation by endoscopy to determine whether they have peptic ulcer disease. For the sake of simplicity the results of endoscopy were assumed to be 100% sensitive and specific; this assumption was then tested in the sensitivity analysis. Under the screening and eradication strategy, some patients without H pylori related peptic ulcer disease will receive inappropriate eradication treatment. As their symptoms will persist, however, they will be returned to conventional empirical treatment. Hence, the difference in costs between the two management strategies for these patients is simply the additional cost of the inappropriate eradication treatment.

COSTS

Only health service costs were considered, the bulk of which comprise the drug costs of eradication treatment or cimetidine, but which also include the cost of consultations with general practitioners for symptoms, entailing prescription of healing H2-antagonists or eradication treatment, and a 13C-urea breath test to check the success of the eradication treatment. Patients in the “cure” or “remission” states of the model were assumed not to consume any resources. All costs were discounted at the Treasury rate of 6% a year to allow for differential timing24 and were adjusted to 1995 prices when appropriate.

SYMPTOMS

Data on symptoms experienced by patients with peptic ulcer disease receiving H2-antagonist treatment are sparse. In this model, health outcomes were measured in terms of the proportion of time patients spend without a recurrent ulcer—that is, in a state of “cure,” in remission, or on maintenance cimetidine.

Results

CONVENTIONAL VERSUS ERADICATION TREATMENT FOR PEPTIC ULCER

The results of the Markov model showed that the higher initial cost of H pylori eradication treatment is quickly offset by the avoidance of long term cimetidine treatment, indicating that H pylori eradication treatment is cost saving compared with H2-antagonist treatment, with the magnitude of these savings increasing over time. The average cost per patient of conventional treatment after 10 years was £812 compared with a total 10 year cost of £209 for those patients who initially receive eradication treatment (fig 3).

Fig 3

Ten year accumulation of costs for conventional v Helicobacter pylori eradication treatment for patients with confirmed infection and peptic ulcer disease

• Download figure
• Open in new tab
• Download powerpoint

The average proportion of time spent without ulcer (and therefore without symptoms) for patients in the model was 99% for triple treatment eradication compared with 95% for conventional treatment. These figures reflect the fact that H2-antagonists are an effective treatment for controlling symptoms of peptic ulcer but that eradication of H pylori infection effectively cures the condition for most patients. As the difference in effectiveness is small (and clearly in favour of eradication treatment), the subsequent analysis focuses on costs.

SCREENING AND ERADICATION STRATEGIES

Figure 2 shows that only 10% of dyspeptic patients have both H pylori infection and peptic ulcer. The results of comparing a screening strategy to identify these patients from among the dyspeptic population with conventional empirical cimetidine treatment are presented in figure 4. The cost of screening increases expenditure in the first year. Only as the costs associated with long term maintenance treatment accrue does the screening strategy finally result in savings, after almost eight years.

Fig 4

Difference in costs between conventional cimetidine treatment and strategy of identifying appropriate patients to receive Helicobacter pylori eradication treatment

• Download figure
• Open in new tab
• Download powerpoint

SENSITIVITY ANALYSIS

Sensitivity analysis examines the implications for the results of an evaluation of varying its parameters.25 Table 3 presents the results of such an analysis in relation to the comparison of the screening and eradication strategy with conventional management. Given the focus on costs, the sensitivity analysis concentrated on the payback period for the initial investment in screening and eradication strategy—that is, the point in time at which the costs of both strategies are equal.

It is clear from table 3 that the prevalence of H pylori infection among dyspeptic patients is a key parameter of the model. If an upper estimate of 60% is used the payback period is seven years. A lower estimate of 20% produces a payback period of 12 years. The prevalence of ulcer among those seropositive for H pylori is also a key parameter, although variation in the prevalence of ulcer among dyspeptic patients negative for H pylori has little effect on the payback period. The cost and accuracy of the serology test also have important effects on the payback period, as does the cost of an endoscopy, because of their effect on the cost of identifying patients for H pylori eradication treatment. By contrast, the accuracy of endoscopy is not an important parameter. A plausible variation in the cost of H pylori eradication treatment has little overall effect on the period until cost savings are apparent. Improving the effectiveness of eradication treatment to 100% will have little effect on the payback period, although this period is much more sensitive to assumptions of reduced effectiveness. Variation in the costs of a 13C-urea breath test and in outpatient attendance have little effect as these affect only the overall cost of eradication treatment. The recurrence rates for ulcer will affect the overall cost of conventional treatment and are therefore key parameters for this model.

The most important parameter with respect to the payback period was the cost of antisecretory medication. In the base case analysis, the proprietary price of cimetidine was used as this was thought to be the most common treatment for peptic-type dyspepsia. Increasingly, however, many general practitioners favour proton pump inhibitors over cimetidine, which have similar ulcer healing properties but which cost almost twice as much. With the price of a proton pump inhibitor as an upper estimate, the sensitivity analysis showed that if general practitioners favoured inhibitors over H2-receptor antagonists for the treatment of peptic type dyspepsia, it would take five years for an H pylori screening and eradication strategy to yield cost savings. Generic H2-receptor antagonists, however, have recently become available. With the price of generic cimetidine as the lower estimate, the sensitivity analysis suggests that when general practitioners are prescribing generic H2-receptor antagonists the payback period for a screening and eradication strategy could be as long as 18 years.

Discussion

Other economic evaluations of the use of H pylori eradication treatment as an alternative to maintenance treatment with H2-receptor antagonists have tended to focus their analysis on the group of seropositive patients with confirmed peptic ulcer. It is clear from the results presented in figure 4, however, that this is potentially misleading. The implication of these earlier results is that the NHS could save money in the first year of treatment if general practitioners were to switch to an H pylori eradication programme for appropriate dyspeptic patients. In fact, if the recent proposals for screening dyspeptic patients, with endoscopy for patients whose tests yield positive results, were to be implemented8 26 27 it may be almost eight years before any cost savings are realised. Furthermore, the initial costs in the first year may be over £50 per patient in excess of the costs of empirical treatment with cimetidine. Clearly, such a disparity between the length of time before the NHS can expect to see cost savings as a result of a switch to an H pylori screening strategy could have important implications for both fundholders and purchasing authorities who try to balance their budgets on a year by year basis. Even by allowing for a reasonable level of uncertainty, it will be at least five years before cost savings are realised, although it may take as long as 18 years.

There are several problems associated with modelling the course of H pylori eradication treatment in the way we have presented. The underlying assumptions of the model may not reflect clinical practice. In particular, the assumption that the treatment of patients with non-ulcer dyspepsia will not change as a result of implementing a screening strategy entailing testing for H pylori is likely to bias the results against eradication treatment. In theory, general practitioners might interpret a negative test result as a sign that patients are unlikely to have peptic ulcer and so prescribe fewer antisecretory drugs, although in fact preliminary data suggest this does not occur.28 Further research is required to assess the prescribing habits of general practitioners after the results of H pylori testing.

The long payback period associated with adopting a screening and eradication strategy in general practice is partly due to the low proportion of patients with both H pylori infection and peptic ulcer disease for whom eradication treatment is appropriate. It is also due to the fact that the screening strategy entails all patients testing positive for H pylori undergoing endoscopy, an investigation which would not have been undertaken under conventional management. If eradication treatment were shown to be of benefit in non-ulcer dyspepsia, then a policy of eradication in all patients with positive results could be justified, thereby avoiding the need for endoscopy. Current evidence suggests, however, that there is no proved benefit of eradicating the infection in these patients.12 Moreover, one of the most common causes of non-ulcer dyspepsia is gastro-oesophageal reflux, which is not helped by eradication treatment but is improved by antisecretory drugs.

CONCLUSIONS

Although H pylori eradication treatment has been shown to be cost saving in the treatment of seropositive patients with peptic ulcer disease, considerable caution should be exercised when extrapolating such results to the screening of dyspeptic patients for H pylori in routine clinical practice. Cost savings will eventually accrue, but this may take some years, which contrasts with the immediate cost savings implied in the literature. The actual cost effectiveness of strategies for H pylori screening and eradication may depend crucially on the effect of eradication treatment on non-ulcer dyspepsia rather than for peptic ulcer where the benefits have been established.