Without invasive diagnostic testing, patients with ulcer disease who would benefit from H. pylori eradication are difficult to distinguish clinically from persons with upper abdominal symptoms not caused by ulcer [21-29]. Moreover, it remains controversial whether antibiotic therapy and eradication of H. pylori provides clinical benefit in persons without active ulcer disease. Therefore, to minimize inappropriate use of antibiotic therapy, guidelines such as those of the NIH Consensus Panel recommend objective documentation of an active ulcer and H. pylori infection before prescription of antibiotic treatment.

Such a recommendation makes two assumptions: 1) that diagnostic confirmation of both ulcer disease and H. pylori infection is the optimal strategy after the benefits, risks, and costs of available alternatives are carefully considered and 2) that community physicians, who are accustomed to using initial empiric therapy in patients with ulcer-like symptoms, will find this guideline instructive in their clinical practice, given that several alternative options do not rely on initial invasive diagnostic testing.

Accordingly, we used decision analysis to estimate the clinical and economic effects of two immediate endoscopy and three initially noninvasive diagnostic and treatment strategies for patients with symptoms suggestive of peptic ulcer disease: 1) immediate endoscopy for peptic ulcer and biopsy for H. pylori; 2) immediate endoscopy without biopsy; 3) serologic testing for H. pylori infection; 4) empiric treatment with antisecretory therapy; and 5) empiric treatment with antisecretory and antibiotic therapy.

Methods

Top Methods Results Discussion Conclusion Author & Article Info References

Decision Analytic Model

Using a decision analysis software program, Decision Maker 7.0 (Pratt Medical Group, Boston, Massachusetts), we constructed a computer simulation to predict the natural history of peptic ulcer disease, its interaction with H. pylori infection, and the effects of various diagnostic and therapeutic medical interventions. The analysis started with a cohort of 1000 hypothetical patients presenting to a physician with symptoms suggestive of peptic ulcer disease who were not concurrently taking NSAIDs. On entry into the simulation, all patients had symptoms severe enough to justify an empiric course of antisecretory agents and had no previously documented peptic ulcer disease.

Patients whose symptoms had causes other than peptic ulcer disease were captured in the simulation up to and including the point at which they were objectively identified as not having an ulcer on endoscopic evaluation. Until a diagnosis was made endoscopically, resources were accrued for all patients, including the consumption of pharmaceutical agents, physician visits, and diagnostic tests (including endoscopy and biopsy). Exclusion of treatment costs after the demonstration of a nonulcer cause was consistent with our objective to measure the clinical and economic effect of alternative management strategies for suspected peptic ulcer disease. This exclusion was also in concordance with the assumption that no clinical benefit resulted from the eradication of H. pylori in patients without ulcer disease.

Initial Diagnostic and Therapeutic Interventions

Immediate Endoscopy Strategies

In the first invasive diagnostic strategy, strategy 1, all patients had immediate endoscopy and biopsy to determine the presence of H. pylori. Patients with an active ulcer on endoscopy were prescribed antisecretory therapy. Antibiotic therapy was limited to persons with both an ulcer and evidence of H. pylori infection found on biopsy.

With strategy 2, patients had immediate endoscopy but did not have a biopsy to determine the presence of H. pylori. Persons with active ulcer disease confirmed by endoscopy were presumed to be infected with H. pylori and were prescribed both antisecretory and antibiotic therapy.

Noninvasive Strategies

The noninvasive treatment strategies did not require diagnostic confirmation of an ulcer before the initiation of either antisecretory or antibiotic therapy. Every patient having each of the noninvasive strategies was prescribed an empiric course of antisecretory agents at the initial medical encounter. With strategy 3, all patients also had a qualitative serologic test for H. pylori. Only patients with evidence of previous or current H. pylori infection, as shown by a positive serologic test result, received antibiotic therapy in addition to the antisecretory agent. Persons with a negative serologic test result received only antisecretory therapy.

The remaining two noninvasive strategies were purely empiric, with no associated diagnostic testing before therapy was started. With strategy 4, all patients were prescribed antisecretory therapy alone; with strategy 5, all patients received both antisecretory and antibiotic therapy.

The model captured resource use, such as physician visits, pharmaceutical use, procedures, and hospitalizations. Direct medical expenditures were calculated from these data. After the initial management encounter, patients moved among different states of health determined by the likelihood of particular clinical events and the effect of certain medical interventions on the natural history of peptic ulcer disease. Each patient was evaluated in the model at 6-week intervals for 1 year for the presence or absence of three clinical conditions, on which all further interactions with the medical care system were based: recurrent symptoms, H. pylori infection, and active ulcer disease.

Clinical Conditions and Subsequent Intervention

Recurrent Symptoms

It was assumed that after the initial intervention, each patient with persistent or recurrent ulcer-like symptoms would return for medical evaluation regardless of physician or patient knowledge of an underlying diagnosis. Patients with recurrent symptoms who were initially managed noninvasively (strategies 3, 4, and 5) had endoscopy with biopsy on the first return visit to determine the presence of active ulcer disease and H. pylori infection. All subsequent treatments were based on the endoscopy results and biopsy findings. Thus, if a single 6-week course of empiric therapy did not completely relieve symptoms for the entire study period, ensuing treatment was always directed by objective diagnostic data. Patients in whom an ulcer was diagnosed by endoscopy at any time were treated with high-dose antisecretory agents. Antibiotic therapy was prescribed when objective evidence suggested H. pylori infection, except for patients having only immediate endoscopy. In this group, patients with ulcers automatically received H. pylori eradication therapy (strategy 2). Patients with ulcers who remained symptomatic after three complete courses of antisecretory therapy had a second endoscopic evaluation to assess ulcer healing and H. pylori status.

It was assumed that patients who became asymptomatic at any time after the initial encounter did not visit a physician, regardless of their underlying diagnosis, unless symptoms were to recur.

Helicobacter pylori Infection

Helicobacter pylori status affected the likelihood of ulcer recurrence but not the development or resolution of symptoms. Pathologic evaluation of the biopsy specimen was assumed to be a perfect test for H. pylori infection. With strategy 3, the qualitative serologic test was presumed to have 95% sensitivity and 95% specificity [30, 31] for identifying previous or current H. pylori infection. If endoscopy and biopsy were not to be done for a particular symptomatic encounter in patients who had previously received antibiotic therapy, a urea breath test was administered to evaluate how effectively the organism was being eliminated [32-37]. Persons who did not clear the infection after three courses of antibiotics, as shown by infection on repeat biopsy specimens or a positive urea breath test result, were assumed to be infected for the rest of the study period.

It was assumed that once H. pylori had cleared, infection did not recur [10, 38, 39]. Successful eradication of H. pylori in patients with ulcers associated with NSAIDs or with diagnoses other than active ulcer disease (for example, gastritis) was assumed to offer no benefit in terms of reducing symptom severity or preventing future ulcer development [40].

Active Ulcer Disease

Ulcer status determined the likelihood that therapy would relieve symptoms and thus determined the need for future physician visits and related medical interventions. Endoscopy was assumed to be a perfect test for diagnosing ulcer disease and was presumed to have no associated adverse events. Ulcer recurrence not associated with NSAID use was related to H. pylori status and concurrent use of antisecretory therapy [8, 17, 18, 41-49]. Spontaneous ulcer development was assumed not to occur in patients whose initial symptoms were secondary to nonulcer causes [50]. Ulcers that did not completely heal after three courses of antisecretory therapy required maintenance antisecretory therapy for the duration of the study period. Resource use related to persistent ulcer symptoms was calculated for the remaining weeks for 1 year.

At the end of each 6-week interval, patients were newly distributed among different health states depending on the likelihood of symptoms, H. pylori infection, and ulcer disease and the effect of the prescribed diagnostic and therapeutic interventions on the natural history of the disease.

Data Sources

Clinical Probabilities

To obtain pertinent clinical data for the simulation model, we searched the MEDLINE database for English-language articles and reviewed bibliographies of selected articles. We also searched current issues of peer-reviewed general medicine, infectious disease, and gastroenterology journals to identify additional reports not included in the computerized database. In cases in which data were unavailable or uncertain, we used unpublished data presented at the NIH Consensus Development Meeting (7 to 9 February 1994) [20].

Clinical input probabilities and ranges used in the simulation are shown in Table 1. Not all patients who presented with symptoms suggestive of peptic ulcer disease had active ulcers on entry into the model. The hypothetical cohort was presumed to have a mix of clinical conditions as drawn from the literature. Studies of open-access endoscopy units showed that peptic ulcers not associated with NSAID use were found in at least 20% of symptomatic patients presenting for medical attention [21-29].

Helicobacter pylori infection was seen more frequently when an ulcer caused the symptoms (95%) than when it did not. Ulcer recurrence was related to the presence of H. pylori infection and recent antisecretory therapy. Active ulcer disease was predominantly symptomatic. However, symptoms rarely developed after an ulcer had healed. After a course of empiric antisecretory therapy, 30% of patients whose initial symptoms were not ulcer-related had recurrent symptoms that necessitated a physician visit over the study year. At that encounter, endoscopy and biopsy were done; these patients were identified as having no ulcer disease and were no longer followed in the simulation. If symptoms did not recur, the patients had no further contact with the medical care system.

Multiple antibiotic and antisecretory regimens of varying efficacy, cost, and compliance were evaluated. The relatively small differences in ulcer healing and H. pylori eradication rates among available regimens were investigated using sensitivity analysis. Compliance and outcomes related to specific medication use, including antibiotic-related adverse events, were incorporated into these analyses.

Cost Inputs

The perspective of the economic analysis was that of the payer. Therefore, only third-party expenditures were included in the cost calculations of medical resource use. Actual payments, not charges, for ambulatory services, inpatient care, and physicians fees were obtained from a large private third-party payer covering approximately 3 million persons in the Eastern United States who were enrolled in various indemnity and managed care insurance plans (Table 1). Given that actual payments for similar services vary by geographic region and delivery system, we used the national average of charges allowed by the Health Care Financing Administration for Medicare reimbursement to determine the lower bound of the cost estimates.

Direct medical costs were measured only for patients with confirmed peptic ulcer disease or patients with symptoms suggestive of ulcer disease before invasive diagnostic testing. Out-of-pocket expenses were collected only for pharmaceutical agents; deductibles and copayments made by patients for medical services other than drugs were not estimated. Because we included only a small portion of the patients' financial burden of direct medical costs and did not include indirect costs related to lost productivity, the cost estimates are conservative.

Costs of pharmaceutical agents were estimated by actual payments made by patients at seven retail pharmacies in three eastern states (Table 1). Average wholesale pharmaceutical prices plus a $5 dispensing fee were used to establish the lower ranges of the costs of medications.

In light of published guidelines that advocated diagnostic certainty over empiric therapy, we wanted to recommend an initial noninvasive strategy only if we could do so convincingly on both clinical and economic grounds. Therefore, when confronted with uncertainty in the estimation of clinical probabilities or cost inputs, we explicitly made assumptions that favored the initially invasive diagnostic strategies.

Sensitivity Analysis

We did sensitivity analyses to evaluate the effect of variations in single clinical probabilities and cost estimates on the results. These reflect the ranges reported in the literature (Table 1). We only investigated the input values whose inclusion into the model would enhance the cost-effectiveness of the immediate endoscopy strategies or diminish the cost-effectiveness of initial noninvasive strategies. We did additional analyses in which two inputs were simultaneously evaluated over their respective ranges. We also examined a worst-case scenario using the lowest rate of antibiotic efficacy, the lowest ulcer healing rate, and the highest pharmaceutical costs for each noninvasive strategy. Discounting of costs and benefits was not necessary given the 1-year duration of the model.

Results

Top Methods Results Discussion Conclusion Author & Article Info References

Baseline Analysis

Average Cost per Ulcer Cured

The estimated treatment costs per ulcer cured by strategy were as follows: endoscopy and biopsy for H. pylori (strategy 1), $8045; endoscopy only (strategy 2), $6984; serologic test for H. pylori (strategy 3), $4541; empiric antisecretory therapy (strategy 4), $4835; and empiric antisecretory and antibiotic therapy (strategy 5), $4155 (Figure 1). Given the high rate of symptoms in patients with active ulcer disease, nearly all active ulcers and associated H. pylori infections were objectively diagnosed and treated by the end of the study period in each strategy. As a result, the percentage of patients who had a persistent active ulcer in each strategy was low (1% to 2% of the entry population) at the end of the study year.

View larger version (22K): [in this window] [in a new window]   Figure 1. Cost per ulcer cured and cost per patient treated by management strategy.

Average Cost per Patient Treated

A similar cost-effectiveness advantage for the initial noninvasive management strategies was predicted by the simulation when cost per patient treated was calculated as the primary outcome of interest (Figure 1). However, because active ulcer disease was seen in only 20% of patients in the entry cohort who had ulcer-like symptoms, the estimated treatment costs per patient by strategy were lower: endoscopy and biopsy for H. pylori (strategy 1), $1584; endoscopy only (strategy 2), $1375; serologic tests for H. pylori (strategy 3), $894; empiric antisecretory therapy (strategy 4), $952; and empiric antisecretory and antibiotic therapy (strategy 5), $818.

Sensitivity Analysis

Immediate Endoscopy Strategies Compared with Initial Noninvasive Strategies

Examination of clinical and cost inputs over broad ranges showed that the cost-effectiveness advantage of noninvasive strategies (strategies 3, 4, and 5) relative to immediate endoscopy strategies (strategies 1 and 2) was sensitive to only two variables: 1) the cost of endoscopy and 2) the probability of recurrent symptoms in patients not initially treated with an invasive procedure and in whom ulcer disease was not the underlying cause of symptoms.

As the cost of endoscopy decreased, the cost per patient treated of strategies using immediate endoscopy approached that of the noninvasive strategies (Figure 2). Endoscopy costs, including payments for professional fees and suite charges, must decrease to less than $500 for an equivalent cost-effectiveness ratio to result.

View larger version (19K): [in this window] [in a new window]   Figure 2. Sensitivity analysis: cost of endoscopy.

The other variable found to be important in establishing an advantage for the noninvasive strategies was the probability of recurrent symptoms in patients without ulcer disease. Patients without ulcer disease who still have recurrent symptoms after a course of empiric therapy incurred the cost of invasive diagnostic testing (mandatory for recurrent symptoms in all patients not treated with an invasive procedure) in addition to the cost of initial therapy. Therefore, as the probability of recurrent symptoms increased in this population, the potential savings of initial noninvasive management diminished as the use of endoscopy increased. As the annual recurrent symptom rate approached 80%, the cost per patient treated of the noninvasive strategies approached that of the immediate invasive diagnostic strategies.

We did a sensitivity analysis that simultaneously varied the cost of endoscopy and the probability of recurrent symptoms in patients without an ulcer. This analysis showed that the choice of immediate endoscopy and biopsy (strategy 1) versus empiric antisecretory and antibiotic therapy (strategy 5) can be made at different values of these clinical and economic inputs (Figure 3).

View larger version (16K): [in this window] [in a new window]   Figure 3. Two variable sensitivity analyses: cost of endoscopy and probability of recurrent symptoms in patients without ulcer disease. Area above the solid line favors empiric therapy; area below the line favors immediate endoscopy.

Sensitivity analyses of the worst-case scenarios, which were designed to maximize the relative cost-effectiveness of the immediate invasive diagnostic strategies, found that the cost-effectiveness advantage of the noninvasive strategies remained despite the use of clinical and economic inputs that further biased the models against the noninvasive strategies.

Immediate Endoscopy with Biopsy Compared with Immediate Endoscopy without Biopsy

We also used sensitivity analyses to compare the two immediate endoscopy strategies. The combination of the high likelihood of H. pylori infection in patients with active ulcers, incremental biopsy cost, and relatively low cost of antibiotic therapy precluded the identification of a single clinical scenario in which the performance of a biopsy for H. pylori was justified in patients with an endoscopically diagnosed peptic ulcer. Adherence to a strategy requiring pathologic proof of H. pylori infection necessitated an additional $21 000 in biopsy-related costs to identify a single patient with an ulcer who was not simultaneously infected with H. pylori. If the presence of H. pylori must be confirmed, the sensitive, low-cost, Campylobacter-like-organism (CLO) test should replace pathologic testing of biopsy specimens.

Empiric Antisecretory Therapy Compared with Empiric Antisecretory and Antibiotic Therapy

We also investigated the relative cost-effectiveness among empiric strategies. Sensitivity analysis showed that there was not a single instance in which the cost-effectiveness of empiric antisecretory therapy only (strategy 4) was superior to that of the combined empiric regimen (strategy 5). This may be explained by the dramatic reduction of ulcer recurrence in patients with active ulcer disease in whom H. pylori was successfully eradicated.

Serologic Testing Compared with Empiric Treatment

When we compared the combined empiric regimen (strategy 5) with a strategy of initial serologic testing (strategy 3), the optimal choice was less clear. The operating characteristics and the cost of the qualitative serologic test for H. pylori infection were the most important determinants in this decision. If the cost of the H. pylori serologic test were to decrease to $12, an equivalent cost-effectiveness ratio would result.

Discussion

Top Methods Results Discussion Conclusion Author & Article Info References

For nearly two decades, the safety and efficacy of histamine-2-receptor (_antagonists) and, more recently, proton-pump inhibitors have led many clinicians to prescribe an initial empiric course of pharmaceutical treatment before they do invasive diagnostic testing in patients with upper abdominal symptoms. The profound effects of antisecretory agents on ulcer healing and symptom relief have led to a marked worldwide reduction of patient morbidity and societal use of resources [65-67]. For patients with symptoms suggestive of peptic ulcer disease, this innovation also led to a change in the venue of treatment from an inpatient to an ambulatory setting and in the source of care from the gastroenterologist to the generalist [68].

In a position paper [69], the American College of Physicians supported the empiric approach, recommending endoscopy only for persons not responding to an appropriate course of antisecretory therapy. For patients with active ulcer disease not related to NSAIDs, the realization that acid suppression must be accompanied by H. pylori eradication to prevent ulcer recurrence necessitates reevaluating the existing treatment algorithm. Any recommended changes in this widely accepted treatment strategy must be evaluated against likely clinical alternatives for their effects on patients, providers, and payers, while incorporating new information on the role of H. pylori.

The NIH Consensus Panel Statement implicitly endorsed a strategy of documenting the presence of both an ulcer and H. pylori infection before antibiotic medications intended to eradicate the organism are prescribed [20]. The recommendation requiring confirmation of both diagnoses has raised considerable uncertainty among clinicians who have widely accepted an empiric first-line management of patients with symptoms suggestive of peptic ulcer disease.

Because persons with ulcer disease are difficult to identify from the large pool of patients with upper-abdominal symptoms [70-73], compliance with the NIH Consensus Panel recommendations requires increased invasive diagnostic testing before treatment. The provision of perfect diagnostic information necessitates increased health system and patient-related costs.

Our cost-effectiveness analyses support the continued use of a noninvasive treatment strategy at the first symptomatic episode (similar to the empiric histamine-2 therapy currently used), adapted to address the possibility of H. pylori infection. In each of the three noninvasive empiric strategies evaluated, a single course of therapy with a duration of no longer than 6 weeks was attempted, and endoscopy was reserved for patients who remained symptomatic or whose symptoms recurred. Because of the high rate of symptoms in patients with active ulcers, almost all patients with ulcers unresponsive to a single course of therapy had their ulcer later identified endoscopically. Consequently, the percentage of the cohort free of active ulcer disease at the end of the study period was the same in each of the five strategies evaluated (>98%).

We limited the duration of the model to 1 year. Because the sequence of interventions administered to symptomatic patients was similar in all management strategies after the initial medical encounter, clinical outcomes were almost identical across treatment strategies at the end of the study period. Thus, future costs and clinical effects beyond this period were expected not to deviate across strategies.

A concern raised with the use of empiric treatment is the potential delay in diagnosing gastric cancer, a rare but aggressive malignant disorder found in approximately 1% of patients presenting with dyspepsia [74, 75]. Because symptoms persist in nearly all persons with gastric cancer [76], the underlying condition in these patients would eventually be identified by endoscopy in any of the treatment strategies evaluated. No available data suggest that a 6- to 18-week delay in diagnosis would have any adverse consequences on patient mortality and morbidity [77]. Perhaps more important is that such a delay in diagnosis is typical in current clinical practice, where empiric use of antisecretory agents is well established [69, 78, 79]. Therefore, if our recommendation was adopted and initial empiric therapy was merely adapted to address the possibility of H. pylori infection, no additional delay would result.

Although the cost-effectiveness ratios included antibiotic-related adverse events, the model did not include the nondrug costs related to the overprescribing of antibiotics, which, in theory, increase the likelihood of H. pylori resistance. The issue of bacterial resistance is a real one in discussions of the broad use of antibiotics, especially when an important percentage of patients for whom these agents are prescribed may not receive clinical benefit. Data are currently being collected to allow quantification of antibiotic resistance. This calculation is complicated by the fact that all antibiotic agents currently recommended (but not approved by the Food and Drug Administration) for treatment of H. pylori are prescribed for other clinical indications.

Promotion of antibiotic resistance is not frequently considered in the individual patient–physician episode of care but is an important societal issue. Nonetheless, the potential to maintain effective treatments in the future is of considerable importance [80]. Antibiotic resistance to H. pylori may be viewed differently by clinicians; compared with other pathogens (for example, Staphylococcus aureus), H. pylori has a relatively low level of infectiousness, and the resultant infection usually does not cause clinically significant illness.

To allow clinicians to make their own decisions about these nonmonetary "costs" of unnecessary antibiotic prescribing, our model allowed the calculation of the incremental cost in diagnostic testing that would be necessary to avoid an inappropriate course of antibiotic therapy. For example, in comparing serologic testing (strategy 3) with the combination empiric therapy (strategy 5), the extra testing costs that were incurred to avoid one unnecessary course of antibiotic therapy was $195. However, this Figure would be reduced if the cost of serologic testing decreased. The incremental costs necessary to avoid inappropriate antibiotic therapy were substantially higher when immediate endoscopy strategies were compared with the noninvasive alternatives.

Why Is It Not Cost-Effective To Do Endoscopy Initially?

Our study showed that because invasive diagnostic tests can be costly, treatment strategies that economize on invasive testing may be preferred. Strategy 1 (immediate endoscopy and biopsy) provided the earliest and best information but was also the most expensive strategy. Total costs were reduced when less invasive strategies were chosen, but this resulted in less diagnostic certainty.

The cost-effectiveness advantage of the initial noninvasive strategies (strategies 3, 4, and 5) relative to the invasive ones (strategies 1 and 2) reflects the high cost of endoscopy compared with the cost of a course of antibiotic therapy. This finding was confirmed when multiple available antibiotic and antisecretory regimens were used. Immediate endoscopy may save the relatively small monetary costs of inappropriate antibiotic use in the large subset of patients found not to have active ulcer disease. However, because empiric antisecretory treatment alleviated symptoms and healed ulcers (at least temporarily) in many cases and because a course of antibiotic therapy to eradicate H. pylori was successful most of the time, the high cost of invasive diagnostic testing was avoided in approximately two thirds of all patients.

Our model assumed that antibiotic therapy had no benefit for patients with NSAID-associated ulcers and symptoms caused by conditions other than ulcer disease (for example, gastritis). If one were to incorporate a benefit of H. pylori eradication by preventing future ulcer development [50, 81-83], reducing the risk for gastric malignancy [84-89], or enhancing symptom relief [90-94] in patients without ulcer disease as postulated in the literature, the clinical and economic advantages of initial noninvasive treatment with antibiotic therapy would far surpass those reported here.

It is unlikely that the superiority of noninvasive treatment would disappear if nonmonetary aspects of patient utility (such as pain relief and ability to function normally) were added to our model. Any utility gained (by patient and physician) by early diagnosis would probably be offset by the disutility associated with the invasive endoscopy itself (for example, risk, discomfort, and inconvenience). This is particularly true because invasive testing for patients who remain symptomatic is delayed for no more than 6 weeks in the noninvasive strategies. Thus, the value of early diagnostic information is low because the delay is short in diagnosing alternatives to ulcer disease, such as gastritis, in symptomatic patients relative to the ultimate resolution of the patients' illness. However, one drawback of noninvasive management is that it may postpone the determination of an objective diagnosis; in cases of successful empiric treatment, the cause of symptoms may never be confirmed.

The cost-effectiveness advantage for the noninvasive strategies relative to a strategy of immediate endoscopy was sensitive to two inputs: the cost of endoscopy and the probability of recurrent symptoms in patients in whom ulcer disease was not an underlying cause. Lower payments for endoscopy, such as those found in competitive managed care markets or delivery systems outside the United States, may lead to a different strategy choice. Independent of endoscopy cost, the savings achieved by initially avoiding this diagnostic test was lost as the likelihood of eventual endoscopy increased. In a recent prospective trial from Denmark [95] comparing empiric histamine-2-blocker therapy and immediate endoscopy, 65% of empirically treated patients with dyspepsia eventually had endoscopy, but eradication of H. pylori was not attempted. At this rate of recurrent symptoms, the cost-effectiveness advantage of the initial noninvasive strategies was minimal (Figure 3).

Initial Serologic Testing or Empiric Antisecretory and Antibiotic Therapy?

As was the case with immediate endoscopy and noninvasive strategies, the trade-off between serologic testing for H. pylori (strategy 3) and combination empiric therapy (strategy 5) was increased testing costs for improved diagnostic certainty. With serologic testing, 50% fewer patients were exposed to unnecessary antibiotic treatment, reducing the likelihood of adverse effects and nondrug costs associated with antibiotic use, such as the development of antibiotic resistance. This choice depended on the cost of the serologic test itself and the value the clinician placed on the nonmonetary "costs" of unnecessary antibiotic prescribing.

Although the cost used for the qualitative H. pylori serologic test in our analysis was $80, if we included the new, office-based serologic test kits (reimbursed at $25), the difference in the cost-effectiveness of testing and empiric combination therapy would be negligible. If these rapid, easy-to-use tests have performances in physicians' offices similar to those of the more expensive assays, a strategy that includes serologic testing to guide antibiotic use may be increasingly attractive to clinicians. The number of patients unnecessarily treated with antibiotic therapy would be reduced, and, thus, the likelihood of H. pylori resistance to available antibiotic agents would diminish.

Conclusion

Top Methods Results Discussion Conclusion Author & Article Info References

Concerns about resource use and health care costs have led all stakeholders (patients, providers, and payers) into new relationships, such as managed care, and have resulted in new demands for participation in the development of clinical guidelines. Thus, expert guidelines, such as those of the NIH Consensus Development Program, must be analyzed and reviewed from multiple perspectives. Clinical algorithms may no longer be derived solely from safety and clinical efficacy data or driven by desires for perfect diagnostic information [96]. Information on additional outcomes on relief of symptoms, functional status, and resource use is required to guide decision making; improvements in these outcomes may ultimately be preferred to complete diagnostic certainty.

We limited our analysis to five treatment strategies that are likely alternatives in community practice, a setting in which reasonable and reproducible data were available. Additional alternative strategies could be constructed, including those that rely on differing durations of empiric treatment, the use of upper gastrointestinal radiography, or delayed invasive diagnostic testing. The strategies chosen reflect reasonable clinical practice and prudent patient management.

Endoscopy, although costly, precisely guides diagnosis and treatment and thus potentially reduces the number of patients treated unnecessarily with antibiotics. However, the safety, effectiveness, and availability of less expensive, less invasive treatment strategies support the initial noninvasive management of patients with ulcer-like symptoms in many clinical situations. The cost-effectiveness advantage of initial noninvasive therapy was not guaranteed in all circumstances; if market forces lower the cost of endoscopy or if the likelihood of endoscopy increases significantly in patients receiving empiric treatment, the more desirable strategy may change to one of immediate invasive diagnostic testing.

Although our findings support the continued practice of initial noninvasive management of patients with symptoms suggestive of peptic ulcer disease, the best way to achieve the benefits of H. pylori eradication—testing to guide antibiotic use or treating with empiric antibiotic therapy—was less clear. As the price of a noninvasive test to confirm H. pylori infection was reduced, the incremental resources devoted to diagnostic testing appeared to be a worthwhile investment to decreasing the indiscriminate use of antibiotic agents.

Portions of this article were presented at Digestive Disease Week, San Diego, California, 14 May 1995.