Recognition of the role of Helicobacter pylori in the pathogenesis of peptic ulcer disease has revolutionized ulcer therapy and precipitated a reevaluation of the clinical approach to dyspepsia. Helicobacter pylori is a spiral, gram-negative rod that is the major cause of chronic superficial gastritis [7] and most duodenal (80% to 95%) and gastric (70% to 90%) ulcers that are not caused by nonsteroidal anti-inflammatory drugs [8-10]. Although a consensus has been reached about the role of anti-H. pylori therapy in peptic ulcer disease [11], the role of anti-H. pylori therapy in the management of nonulcer dyspepsia is uncertain for two reasons. First, the epidemiologic evidence of an association between H. pylori gastritis and nonulcer dyspepsia is equivocal [12-14]. Second, antibiotic treatment has not consistently alleviated dyspeptic symptoms [15], although a recent meta-analysis [16] highlighted serious methodologic problems with all of the randomized, placebo-controlled trials that have addressed this question.

Practitioners in the United States have largely adopted the American College of Physicians' guidelines, which recommend 6 to 8 weeks of empirical antisecretory therapy for the initial management of uncomplicated dyspepsia [6]. However, the H. pylori revolution in ulcer therapy, along with the increasing availability of noninvasive H. pylori testing, has given rise to alternative strategies for the initial management of dyspepsia. These options, outlined in Figure 1, include initial endoscopy, empirical anti-H. pylori therapy without testing for H. pylori, and noninvasive testing for H. pylori followed by either anti-H. pylori therapy or endoscopy in persons with positive test results.

View larger version (9K): [in this window] [in a new window]   Figure 1. Alternative strategies for the initial management of dyspepsia. The bold-faced limbs of the tree are the subject of this analysis. H. pylori = Helicobacter pylori.

Should the management of uninvestigated dyspepsia be guided by knowledge of H. pylori status? Although serologic and other noninvasive tests for H. pylori are now being investigated as case-finding tools for uninvestigated dyspepsia, no data are available to guide clinicians in the management of patients who have positive results on serologic testing. In many European countries, endoscopy plays a prominent role in the initial evaluation of dyspepsia. "Screening" dyspeptic patients in these countries by using serologic testing for H. pylori and reserving endoscopy for persons with positive test results reduces the endoscopy workload by avoiding endoscopy, which is more likely to yield negative results [17-20]. The alternative to initial endoscopy in H. pylori-seropositive patients with dyspepsia is initial anti-H. pylori therapy, but the costs and benefits of these alternative strategies have not been explored. We therefore used decision analysis to compare the potential costs and clinical consequences of initial endoscopy with those of initial anti-H. pylori therapy in the management of patients with simple, uninvestigated dyspepsia who have a positive result on noninvasive testing for H. pylori.

Methods

Top Methods Results Discussion Author & Article Info References

Decision analysis is a quantitative method for estimating the financial costs and clinical outcomes of alternative management strategies under conditions of uncertainty. We developed our decision model using DATA decision analysis software (TreeAge Software, Inc., Boston, Massachusetts) [21]. Our analysis does not consider patients with complicated dyspepsia, history of peptic ulcer disease, previous anti-H. pylori therapy, use of nonsteroidal anti-inflammatory drugs within the past month, or typical symptoms of gastroesophageal reflux disease. Because no prospective data are available on the prevalence of underlying disease in patients receiving over-the-counter, intermittent, or long-term antisecretory therapy without a diagnosis, we did not exclude such patients. We describe the decision analysis using base-case estimates of the most likely clinical scenarios, and we then evaluate the strategies over a range of cost and probability estimates in a process known as sensitivity analysis. The model estimates financial and clinical outcomes for 1 year after each initial management strategy.

Data Sources

Probability estimates for the model, as shown in Table 1, were derived from published reports listed in the MEDLINE database. We selected English-language articles that were published from 1975 to the present and reviewed the bibliographies of these articles. When the literature offered a range of probabilities, we chose estimates that biased the model against the initial anti-H. pylori strategy.

Cost Estimates

The cost analysis was done from the perspective of a third-party payer. Drug costs were obtained from a large health maintenance organization pharmacy in California, and costs for physician services and hospitalizations were obtained from the 1995 Medicare Resource-Based Relative Value Scale, the 1995 American Association of Anesthesiologists payment schedules, and 1993-1994 Diagnostic Related Groups hospital payment schedules. Table 2 lists the Diagnostic Related Group codes and a description of the service and cost to Medicare. Current Procedural Terminology codes and the cost to Medicare are listed in Table 3. Institutional charges were assumed to be reimbursed at the customary Medicare cost-to-charge ratio for internal medicine of 0.50.

Decision Model

As shown in Figure 2, the decision model compares initial endoscopy with initial anti-H. pylori therapy. The strategy of initial endoscopy in the base-case analysis includes biopsy and a rapid urease test of all ulcers; histologic testing is also done for gastric ulcers. Patients with peptic ulcer disease receive anti-H. pylori therapy. Patients with gastric ulcers also have repeated endoscopy to document ulcer healing and the absence of an underlying malignant condition. Patients with esophagitis are prescribed omeprazole, 20 mg twice daily, for 12 weeks. Patients found to have gastric cancer are treated surgically. Patients with nonulcer dyspepsia receive a trial of ranitidine, 150 mg twice daily, for 8 weeks. Patients who have ulcers and do not respond to anti-H. pylori therapy or who have recurrent symptoms within 1 year receive an alternative anti-H. pylori regimen. Because each management option is controversial, we also used sensitivity analysis (described later in this article) to test alternative management options.

View larger version (20K): [in this window] [in a new window]   Figure 2. The decision tree. Each terminal branch with complications is only shown once in each limb. DU = duodenal ulcer; GU = gastric ulcer; H. pylori = Helicobacter pylori; NUD = nonulcer dyspepsia. A plus sign indicates the presence of a hidden branch similar to the adjacent branch.

The second strategy in the decision model is initial anti-H. pylori therapy. Patients whose symptoms either persist or recur within 1 year after initial therapy are referred for endoscopy and, if found to have an ulcer, receive an alternative anti-H. pylori regimen. In these patients, biopsy and a rapid urease test are done; in patients who have gastric ulcers, histologic testing is also done. Patients with gastric ulcers have repeated endoscopy to document ulcer healing. Patients found to have esophagitis or nonulcer dyspepsia receive the same treatment used in the initial endoscopy strategy.

Clinical Inputs and Probability Estimates

A supporting literature review is listed in the Appendix.

Helicobacter pylori Status

All dyspeptic patients entered into the model have a positive result on serologic testing for H. pylori. The model assumes that serologic testing has a sensitivity of 91% and a specificity of 91%. Therefore, in a population that has a 65% prevalence of H. pylori infection, serologic testing has a positive predictive value of 95% for active infection [44]. One office-based serologic test had a sensitivity of 87% and a specificity of 100% in a primary care practice that treated dyspeptic patients and had a prevalence of H. pylori infection of 46% [45]. Therefore, our model assumes that a positive result on H. pylori testing indicates an active case of H. pylori infection.

Effectiveness of Initial Endoscopy Compared with That of Initial Helicobacter pylori Therapy

We found no studies that directly compared ulcer healing, mortality, quality of life, or patient satisfaction between patients with uninvestigated dyspepsia receiving initial anti-H. pylori therapy and those receiving initial endoscopy. In our analysis, we assume that relevant long-term outcomes are not critically affected by the initial management strategy and that they are equivalent in both strategies. Therefore, cost differences represent differences in cost-effectiveness. The short-term effectiveness of initial anti-H. pylori therapy depends on the specific cause of dyspeptic symptoms (Table 1). The ranges of estimates found in the medical literature for the four most prevalent underlying conditions are 15% to 35% for ulcer, 36% to 87% for nonulcer dyspepsia, 0.7% to 11% for esophagitis, and 0% to 2% for gastric cancer.

Age

The prevalence of underlying gastric cancer [46] and peptic ulcer disease is greater in patients older than 60 years of age [32, 47, 48]. Although our model excludes patients who have warning signs of serious disease, the base-case analysis does not stratify the cohort by age and assumes that all age groups share the same probability of disease, complications of therapy, and costs. To account for the higher prevalence of disease in the elderly (age >70 years), we tested an upper-bounds model in which the probabilities of peptic ulcer disease, gastric cancer, and complications of therapy are at the upper bounds of our range of estimates (Table 1). To account for the fact that elderly persons are at risk for greater illness from complications, we assume that the cost of mild and moderate antibiotic complications is twice that of our base-case estimate.

Endoscopy Practices

We tested several alternative assumptions about endoscopy and its related costs. The base-case model adds the cost of referral to a subspecialist to the cost of each endoscopy. Many countries have a system of "open access," in which subspecialty consultation is not needed to obtain endoscopy; this minimizes costs. We also estimated the costs and outcomes of the alternative management strategies, assuming a system of open access.

Because patients with ulcer in this cohort are seropositive and have a high probability of active H. pylori infection, the added value of antral biopsy with a rapid urease test or biopsy alone is uncertain. A recent cost analysis [49] found that significant savings resulted from using anti-H. pylori therapy to treat duodenal ulcers without first documenting H. pylori infection [49]. We tested the effect of the assumption that duodenal ulcers do not need biopsy confirmation of H. pylori infection before anti-H. pylori therapy. To further reduce the estimates of endoscopy-related costs, we also tested the assumption that gastric ulcers do not require repeated endoscopy to document ulcer healing [50]. To simplify the number of models reported, all of these assumptions are incorporated into the open access endoscopy model.

Complications of Endoscopy

The most common complications of endoscopy are cardiorespiratory; such complications generally require only additional observation. Our model assumes a 0.05% probability of severe endoscopic complications requiring hospitalization and surgery. The costs of severe endoscopic complications are modeled after the surgical repair of a perforation.

Side Effects of Antibiotics

We modeled three categories of adverse effects. The first consists of mild, self-limited side effects, such as those that occur in patients who complete therapy despite side effects and those that occur in patients who discontinue therapy and are re-treated. We assume that no additional costs are incurred unless mild side effects result in the discontinuation of therapy and re-treatment. The second category includes moderate side effects, such as pseudomembranous enterocolitis, treated on an outpatient basis. The final category, a "worst-case" scenario for complications of antibiotic therapy, is pseudomembranous enterocolitis requiring hospitalization and surgery.

Downstream Costs for Patients without Ulcers

Our analysis does not consider downstream costs that accumulate after 1 year in patients with nonulcer dyspepsia, gastric cancer, and esophagitis. The model assumes that all such patients who receive initial anti-H. pylori therapy and subsequently require endoscopy do so in the first year. After patients have endoscopy and receive therapy, any future diagnostic procedures, therapeutic interventions, and associated costs are assumed to be independent of the initial management strategy.

Sensitivity Analyses

One-way sensitivity analyses were done to evaluate the effect of varying individual cost and probability estimates over ranges greater than the degree of uncertainty that would be expected on the basis of published reports. Table 1, Table 2, and Table 3 list the ranges of these variable estimates. Two-way sensitivity analyses were done on the most clinically significant and potentially influential variables.

Results

Top Methods Results Discussion Author & Article Info References

Initial anti-H. pylori therapy yields an average cost savings of $456 per patient compared with initial endoscopy. Although the initial anti-H. pylori strategy results in greater use of antibiotics than the initial endoscopy strategy, the cost of this is more than offset by a reduction in the number of endoscopies (Table 4).

Compared with initial endoscopy, the savings produced by initial anti-H. pylori therapy were robust across the entire range of cost and probability estimates used in one-way sensitivity analyses. We used sensitivity testing to examine the threshold limits beyond which initial anti-H. pylori therapy would no longer be more cost-effective. The cost of initial anti-H. pylori therapy must exceed $674 before early endoscopy becomes the preferred strategy; this cost is nearly 10 times our base-case estimate. We also found extreme threshold levels for other key clinical variables (Table 5).

A 96% reduction in the base-case cost of endoscopy is required for the initial endoscopy strategy to be as cost-effective as the initial anti-H. pylori strategy (Table 6). This large reduction in endoscopy cost would be necessary to offset two factors. First, the cost of endoscopy is roughly 12-fold greater than that of anti-H. pylori therapy. Second, the initial endoscopy strategy requires more than twice as many endoscopies as does the initial anti-H. pylori strategy. Furthermore, if subspecialty consultation is not obtained, infection is not documented by biopsy, and the practice of documenting gastric ulcer healing with a repeated endoscopy is eliminated, the cost savings of the initial anti-H. pylori strategy would still be an average of $279 per patient treated. In this scenario, the cost of endoscopy must still be reduced by an additional 93% before the two strategies are equally cost-effective (Table 5).

In our base-case analysis, anti-H. pylori therapy is assumed to be 80% effective. To account for the potential effects of resistant H. pylori and lack of patient compliance with treatment regimens, we assumed in the sensitivity analysis that the rate of H. pylori eradication would be only 50%. Even at this level of effectiveness, initial anti-H. pylori therapy saves an average of $367 per patient (Table 6). The model was not sensitive to variations in the probability that mild side effects of antibiotics would require discontinuation of therapy and re-treatment or to variations in the probability of developing pseudomembranous enterocolitis. Even if 90% of patients develop mild side effects from antibiotics, the initial anti-H. pylori strategy still saves an average of $389 per patient treated.

Our model assumes that 70% of patients have nonulcer dyspepsia. Of these, only 45% will respond to initial anti-H. pylori therapy; the remainder will return with symptoms. Because much uncertainty surrounds the efficacy of anti-H. pylori therapy in nonulcer dyspepsia, we tested the extreme assumption that all patients with nonulcer dyspepsia would return with symptoms and ultimately have endoscopy. Initial anti-H. pylori therapy still resulted in an average cost savings of $200 per patient. The savings persist in this scenario because of our assumption that 80% of the patients with peptic ulcer disease (16% of the study cohort) respond to initial anti-H. pylori therapy and do not incur the costs of endoscopy.

Elderly patients with dyspepsia are at greater risk for gastric cancer, peptic ulcer disease, complications of endoscopy, and more severe illness from antibiotic complications. In sensitivity testing, we biased our model against initial anti-H. pylori therapy by assuming that older patients are not at greater risk for endoscopic complications and by assuming that complications of antibiotics are more costly. In this scenario, initial anti-H. pylori therapy still saves an average of $417 per patient.

We also estimated the effect on outcomes when the values for pairs of the most important clinical variables are changed conjointly. The values necessary to change the preferred strategy from initial anti-H. pylori therapy to early endoscopy are shown in Figure 3 and Figure 4. For example, for initial endoscopy to become the preferred strategy, the cost of initial anti-H. pylori therapy would have to be more than $234 and endoscopy-related cost would have to be decreased by 70% (Figure 3). The prevalence of peptic ulcer disease must be less than 6%, and the response to initial anti-H. pylori therapy in nonulcer dyspepsia would have to be less than 8% before initial endoscopy would be the preferred strategy (Figure 4).

View larger version (36K): [in this window] [in a new window]   Figure 3. Two-way sensitivity analysis: the cost of anti-Helicobacter pylori therapy and the fraction of base-case endoscopy costs. The values in the area marked by grid lines are the values of the variables under which initial anti-H. pylori therapy is preferred. The values in the area marked by horizontal lines are the values of the variables under which initial endoscopy is preferred.

View larger version (51K): [in this window] [in a new window]   Figure 4. Two-way sensitivity analysis: the probability of having peptic ulcer disease and the probability that patients with nonulcer dyspepsia will respond to initial anti-Helicobacter pylori therapy. The values in the area marked by grid lines are the values of the variables under which initial anti-H. pylori therapy is preferred. The values in the area marked by horizontal lines are the values of the variables under which initial endoscopy is preferred.

Discussion

Top Methods Results Discussion Author & Article Info References

Our analysis of alternative strategies for the initial management of H. pylori-seropositive patients with simple, uninvestigated dyspepsia suggests that initial anti-H. pylori therapy could result in substantial cost savings compared with initial endoscopy. The values assigned to the model variables are based on the best data available. However, when data were equivocal or absent, we assigned values that tended to bias the model in favor of initial endoscopy. Despite this bias, our model consistently indicates that initial anti-H. pylori therapy is more cost-effective. Moreover, the cost advantage of initial medical therapy persists even when key clinical assumptions are changed over a wide range of values.

Our analysis does not consider all possible clinical outcomes, such as reduction in the use of anti-secretory drugs or antacids, days of work lost because of illness or visits to physicians, personal discomfort as a result of an invasive procedure or a complication of therapy, or anxiety over the lack of a confirmed diagnosis. Proponents of endoscopy contend that a negative result has a beneficial effect on outcomes because it reduces both the utilization of services and patient anxiety by providing the patient and physician with needed reassurance. However, evidence to support this claim is sparse and conflicting [51-53]. Because few data are available on patient preferences for initial anti-H. pylori therapy or initial endoscopy, we did not address this issue. Well-designed prospective trials are needed to determine the effect of either initial anti-H. pylori therapy or initial endoscopy on functional status, quality of life, and patient satisfaction.

Two recent decision analyses address a related issue of the value of noninvasive H. pylori testing in patients with dyspepsia. Fendrick and colleagues [54] compared initial endoscopy, empirical antisecretory therapy, and a "test and treat" strategy of serologic testing for H. pylori followed by empirical antibiotic therapy in patients who were presumed to have peptic ulcer disease but had unknown H. pylori status. Their results suggest that the "test and treat" strategy is more cost-effective than initial endoscopy or empirical antisecretory therapy in all patients presenting with simple dyspepsia. However, Fendrick and colleagues did not consider initial endoscopy as a potential strategy reserved only for H. pylori-positive patients with dyspepsia. Silverstein and colleagues [55] compared initial management strategies for all patients presenting with a first episode of dyspepsia, including patients with reflux-like symptoms. They found that the decision between initial endoscopy and empirical antisecretory therapy was essentially a "toss-up." However, in a subset of dyspeptic patients found to be positive for H. pylori, initial antibiotic therapy was less costly than initial endoscopy.

Our analysis differs from these previous studies in that we present a detailed assessment of the alternative initial management strategies for patients with simple dyspepsia who are seropositive for H. pylori. Our analysis considers a wide range of clinical and economic consequences, including complications of endoscopy and antibiotic therapy, and assumptions that underlie various health care systems, alternative endoscopy practices, and the potential effect of different levels of effectiveness of and resistance to antibiotics.

Previous Recommendations

The panelists of the National Institutes of Health Consensus Development Conference recommended antimicrobial therapy in patients with documented ulcers, but they did not support the "routine detection and treatment of H. pylori infection in the absence of ulcers." Because of the lack of conclusive data on the effectiveness of anti-H. pylori therapy in nonulcer dyspepsia [11], a strategy of initial anti-H. pylori therapy would result in many patients receiving antimicrobial therapy without proven benefit. However, our analysis shows that the economic advantage of initial anti-H. pylori therapy is not substantially affected by variations in the rate of response in patients with nonulcer dyspepsia.

A related concern is that the widespread use of initial anti-H. pylori therapy might contribute to the increasing microbial resistance that has emerged as one of the most substantial public health issues in this decade [56]. The potential risks include both the general induction of resistance to antimicrobial agents and the induction of H. pylori resistance [57, 58]. However, because anti-H. pylori regimens include antibiotics that are widely prescribed for unrelated conditions, it is difficult to determine the overall contribution of anti-H. pylori therapy to microbial resistance.

Our analysis suggests that the increased costs and decreased effectiveness of anti-H. pylori regimens that are expected to result from microbial resistance and poor patient compliance must be substantial to overcome the economic benefits of initial anti-H. pylori therapy. Conversely, the relative insensitivity of our decision model to the cost of anti-H. pylori therapy suggests there could be a large economic incentive for the pharmaceutical industry to develop new regimens to mitigate the problems of microbial resistance and lack of patient compliance.

Policy Implications

If health policy decision makers believe that more harm than good might result from administering antibiotics to large numbers of patients with nonulcer dyspepsia, then the endoscopic evaluation of all H. pylori-seropositive patients with dyspepsia would be necessary to ensure that only patients with peptic ulcer actually received anti-H. pylori therapy. Such a policy would result in higher rates of endoscopy and subspecialty referral and increased costs to payers.

Can the expense of identifying the subgroup of patients who have ulcers and an unequivocal indication for anti-H. pylori therapy be justified? In making such a determination, two points deserve consideration. The first is the medical opportunity costs of the alternative allocation of such resources. For example, if a health care organization could save $456 000 for every 1000 patients who received initial anti-H. pylori therapy rather than initial endoscopy, this money could be used to pay for more than 5000 adolescent hepatitis B virus vaccinations [59]. The benefits of providing this service might outweigh concerns over the lack of proven benefit of anti-H. pylori therapy in nonulcer dyspepsia and the risk for increasing microbial resistance. Second, what percentage of patients with nonulcer dyspepsia must derive benefit from initial anti-H. pylori therapy before the individual benefits outweigh the risks for increased microbial resistance? Randomized, controlled trials of the efficacy of anti-H. pylori therapy for long-term outcomes of nonulcer dyspepsia should be done to definitively establish benefit in these patients. In addition, health care decision makers should evaluate patient preferences for various diagnostic and treatment options.

Conclusions

The results of our analysis show that the most cost-effective strategy for H. pylori-seropositive patients with simple dyspepsia is initial anti-H. pylori therapy. Such a clinical approach requires two critical steps: 1) a thorough clinical history and examination to identify risk factors for gastric cancer and to exclude markers of serious underlying organic disease and 2) reassessment after 4 to 8 weeks for persistence or recurrence of symptoms. Patients with recurrent or persistent symptoms should then have endoscopy. Because empirical antisecretory therapy precludes H. pylori-seropositive patients with ulcers from receiving initial curative therapy, this option should be reserved for H. pylori-seronegative patients with dyspepsia or for the initial management of dyspepsia in the absence of H. pylori testing.

If a strategy of initial anti-H. pylori therapy for patients with dyspepsia cannot be endorsed at this time, we should 1) await the results of trials designed to determine the benefit of anti-H. pylori therapy in nonulcer dyspepsia and 2) establish patient preferences for alternative strategies before endorsing the use of noninvasive H. pylori testing to screen patients with dyspepsia. If policymakers and researchers wait too long to study the unresolved questions highlighted in our analysis, the unregulated use of noninvasive H. pylori testing, driven by the powerful economic advantage of initial anti-H. pylori therapy, may soon make it impossible to conduct the trials that could provide the data needed to inform the important clinical and policy issues.

Appendix: Literature Review and Rationale Supporting the Model Estimates

Effectiveness of Initial Endoscopy Compared with Initial Anti-Helicobacter pylori Therapy

No prospective studies have directly compared initial endoscopy with initial anti-H. pylori therapy in uninvestigated dyspepsia. In studies of patients with dyspepsia and unknown H. pylori status, the effect on outcomes of initial investigation (endoscopy or radiography) compared with empirical antisecretory therapy has not been conclusively established [60-62]. Moreover, there is conflicting evidence about the effect of a negative endoscopy result on the utilization of services, patient anxiety, and the provision of needed reassurance to the patient and physician [51-53]. In studies that compared dyspeptic patients who had negative endoscopy results with those who had positive results, the investigators found that prescribing rates and consultation for gastrointestinal symptoms were significantly reduced after a negative test result [52, 53]. In contrast, a questionnaire study found that a negative test result actually increased uncertainty and patient anxiety [51]. In a study that compared antacid therapy and reassurance with upper gastrointestinal radiography in patients with dyspepsia, no differences were found in patient satisfaction, days of disability, symptom resolution, or functional status [62]. Finally, empirical antisecretory therapy and initial endoscopy in unselected patients with dyspepsia produced indistinguishable effects on symptom assessment and quality of life [60]. The empirical therapy arm, however, had nearly twice as many office visits, significantly greater drug costs, more sick-leave days, and lower rates of satisfaction with care. These results are difficult to interpret because patients in the endoscopy group who did not have ulcer received specific reassurance but patients in the empirical therapy group who had no ulcer did not.

Effectiveness of Anti-Helicobacter pylori Therapy

Patients with Underlying Ulcer

Between 15% and 35% of H. pylori-positive patients with dyspepsia will be found to have peptic ulcer disease [17-1922, 23]. We assume that 25% of H. pylori-positive patients with dyspepsia have ulcers. For these patients, anti-H. pylori therapy is optimal management [11] and, compared with antisecretory therapy, dramatically decreases recurrences and complications [63, 64]. Recent evidence indicates that 1-week regimens of dual antibiotics plus either colloidal bismuth or a proton-pump inhibitor are as much as 93% effective in eradicating H. pylori and healing ulcers [30, 65]. In our model, several alternative regimens were considered and subjected to sensitivity analyses. An optimal first-line initial anti-H. pylori regimen would be of short duration to maximize patient compliance. We selected a 1-week regimen such as MOC (metronidazole, omeprazole, and clarithromycin, all given twice daily), which has demonstrated eradication rates of about 90% [30, 65]. These short-term regimens are slightly more expensive than our second-line regimen, 14-day triple therapy (bismuth, tetracycline, and metronidazole with antisecretory therapy). The second-line regimen has demonstrated eradication rates of approximately 90% [58]. Our model assumes that cure of H. pylori infection in patients with peptic ulcer disease results in symptom resolution and ulcer healing. Because eradication rates vary extensively among studies and often are not reproducible in clinical practice, our model assumes an 80% rate of H. pylori eradication.

Patients with Nonulcer Dyspepsia

Between 30% and 70% of H. pylori-positive patients with dyspepsia have nonulcer dyspepsia [17, 19]. Our model assumes that 70% of H. pylori-positive patients have nonulcer dyspepsia. Controversy surrounds the question of whether anti-H. pylori therapy enhances responses of symptoms above the rates achieved with placebo in these patients [15, 28]. In 30% to 60% of patients with nonulcer dyspepsia, symptoms resolve with empirical antisecretory therapy; the placebo effect may account for much of this resolution [66]. Our model assumes that, like therapy with antisecretory agents, initial anti-H. pylori therapy will be associated with a robust placebo response and that some of the patients who respond to treatment will not require further evaluation. In the base-case analysis, we estimated that 45% of patients with nonulcer dyspepsia will respond to initial anti-H. pylori therapy and require no further evaluation.

Patients with Esophagitis

There is no clear association between H. pylori infection and esophagitis [67]. The prevalence of esophagitis in studies of H. pylori-positive patients with dyspepsia ranged from 0.7% to 11% [18, 19, 24]. Patients with esophagitis have a rate of placebo response that approaches 30% [68], but the likelihood that patients with esophagitis will become asymptomatic several weeks after receiving anti-H. pylori therapy that includes a proton-pump inhibitor is uncertain. Our model assumes that after patients with typical reflux symptoms (heartburn and acid regurgitation) are excluded, 5% of H. pylori-positive patients may have underlying esophagitis as the cause of their dyspepsia. If these same patients initially receive anti-H. pylori therapy, 85% will return with symptoms during the first year. We further assume that the delay in the endoscopic diagnosis of esophagitis resulting from a trial of anti-H. pylori therapy will not result in serious complications.

Patients with Underlying Gastric Cancer

The prevalence of gastric cancer in H. pylori-positive patients with simple dyspepsia is unknown. Several studies suggest that in patients with dyspepsia seen in a primary care practice, the rate is approximately 1% to 2% [31-34]. Many of these studies, however, were not done in the United States, which has a lower prevalence of gastric cancer than most countries, and did not exclude patients with warning signs of serious disease [46]. Our base-case analysis incorporates a prevalence of 1%, and this estimate was tested with sensitivity analyses over a range of 0% to 3%.

Although no evidence suggests that a delayed diagnosis of gastric cancer resulting from an empirical trial of therapy adversely affects outcomes, this remains a concern [31, 66]. A recent decision analysis explored various assumptions about the effect of a delayed diagnosis of gastric cancer on life expectancy in patients with dyspepsia and found little or no difference between initial endoscopy and empirical antisecretory therapy in overall life expectancy [55]. Our model assumes that a delayed diagnosis does not affect outcomes. In sensitivity analyses, we tested the assumption that a delayed diagnosis is associated with increased financial costs.

Complications of Endoscopy

Serious complications occur most frequently in elderly persons who have several medical problems and are reported in 0.10% to 0.54% of patients [35, 37, 69]. In the ambulatory setting in the United States, the true incidence of serious endoscopic complications is unknown, but one audit reported that diagnostic endoscopy is associated with a perforation rate of 0.05% [36]. Our model assumes that severe endoscopic complications requiring hospitalization and surgical intervention are rare, and the base-case estimate is 0.05%.

Side Effects of Antibiotics

The first category of antibiotic side effects includes self-limited side effects that require the discontinuation of therapy and re-treatment. Although as many as 50% of patients report mild side effects, only 5% require discontinuation of therapy [38, 39]. Our estimate of the costs of minor side effects that require discontinuation of therapy include the cost of an additional office visit and the cost of an additional anti-H. pylori regimen. The second category considers moderate side effects, such as pseudomembranous enterocolitis treated on an outpatient basis. Data from individual trials suggest that pseudomembranous enterocolitis may occur in 0.1% to 0.5% of patients [40, 41]. The costs associated with this complication in the model include the costs associated with mild side effects plus two additional office visits; a stool test for culture, ova, and parasites; Clostridium difficile toxin; and a course of metronidazole. The final category is a worst-case scenario for antibiotic complications: pseudomembranous enterocolitis requiring hospitalization and surgery. Epidemiologic studies suggest that the attack rate of community-acquired, antibiotic-associated enterocolitis requiring hospitalization is 0.5 to 1.0/100 000 persons [43]. The base-case estimate of 0.001% is extremely conservative because we assumed that all of these patients will require surgery and incur the associated costs.

Dr. Etchason: Atlanta Veterans Affairs Medical Center, HSR&D (151), 1670 Clairmont Road, Decatur, GA 30033.

Dr. Kahn: University of California, Los Angeles, School of Medicine, Division of General Internal Medicine, B542 Factor Building, Los Angeles, CA 90024-1736.