William G. Bennett, MD, MSPH

1) Evaluate cost savings associated with the initial eradication of H. pylori as compared (H2) to traditional acid supression

2) determine if urease-based treatment (e.g. Clo-test) is cost-effective compared to empiric treatment.

**Patients and methods**

Simulated cohort of otherwise healthy patients having endoscopically-proven duodenal ulcer (DU) without specific risk factors for DU (e.g., NSAIDS use, inflammatory bowel disease, critical illness, and cancer). Simple decision tree utilizing probabilities of benefit (e.g., ulcer cure) and risks (e.g, recurrence, adverse drug reaction) and estimated direct costs of care.

**Results**

Average direct costs (about one-half of charges) of cure at the end of one year for DU were as follows:

Treatment Cost (range) H2RA $398 (360-768) HpRx $284 (267-468) Urease $294 (274-486)

Results were sensitive to the prevalence of H. pylori in DU (base=95%). Urease-based testing was preferred below an H. pylori prevalence of 66% (rare except in a population where NSAID use is high).

An interesting simulation and analytic evaluation that needed to be done. The study provides good evidence that H2RA therapy alone for DU is passe. This result came through despite several biases against HpRx treatment (e.g., duration of analysis was short, only 1 year, precluding the occurrence of more and multiple DU recurrences with H2RA alone and the natural history of DU was not included and avoided additional costs of recurrent DU). One bias toward HpRx was the absence of amoxicillin/omperazole combination which has a risk, albeit low, of C. difficile and amoxicillin-induced diarrhea with additional costs for stool studies.

I think we'll see additional cost-effectiveness analyses on this subject. They should avoid several problems highlighted in this study:

1) there is no need to consider H2RA but, rather, the mode of
evaluation and treatment for someone with peptic symptoms should be modeled,

2) the natural history of DU should also be modeled,

3) extend the model beyond one year to capture the cost
of multiple recurrences, and

4) use quality-adjusted life years (QALYS) as an outcome rather
than cost per cure to allow comparison to other medical interventions.

Additionally, several errors missed by reviewers:

1) Table 3 used average cost versus marginal cost which inflates the true cost per cure.

2) Figure 2 is a one-way sensitivity analysis not two-way (only variable modified is prevalence since cost per cure is an outcome, not an independent variable).

3) Figure 3 does not label outcome being compared (e.g., cures vs cost per cure).

1) construction of a decision tree that explicitly shows each possible strategic choice and possible outcome (see Fig. 1 in the article).

2) assignment of a probability to each outcome (based on review of the literature or other source of knowledge) and a $ cost for each procedure or outcome (based on reasonable cost estimates from various sources).

3) once all of this knowledge is entered into the decision tree, a computer analysis is done, which yields the estimated average cost for each possible strategy, and thus determines a most cost-effective strategy. Obviously, this determination will depend on each of the assumptions about probability and cost that have been made, thus, an important fourth step is:

4) a sensitivity analysis, where the results are recalculated, varying the previously determined parameters (cost and probability) within reasonable limits, to see if the most cost-effective choice remains the same within all reasonable limits of the parameters.

In this article, the most cost-effective initial approach to treating endoscopically documented duodenal ulcer turned out to be treatment with H. pylori antibiotics (triple therapy with amoxicillin, metronidazole and bismuth subsalicylate) plus H2 receptor antagonists. This is not surprising, considering that the authors decided to assume that patients had no other known risk factors for DU, and thus that the prevalence of H. pylori would be extremely high (95% in their estimate).

As Dr. Bennett points out, it will be more interesting to see what such a cost analysis shows for the frequent patient presenting with dyspepsia. Is endoscopy the most cost-effective approach?

In my opinion, a very real benefit of working through this not-so-easy article is to understand how such an analysis is performed, since it will certainly be seen frequently in the medical literature in the years to come.

Subject: Re: A Cost Analysis of Alternative Treatments for Duodenal

As a practicing primary care provider in the U.S. these cost analyses seem to skip too many steps to me. I may have missed part of the information provided, and any clarification would be appreciated.

As I understand it, H. pylori eradication is clearly indicated for patients with documented duodenal (and probably gastric) ulcer but not for patients with simple dyspepsia or dyspepsia/gastritis. H2 blockers do seem helpful for non-ulcer dyspepsia, although perhaps only as placebo.

For the patient, younger than 50, presenting with symptoms of dyspepsia, it seems to me I have 3 choices: treat empirically for non-ulcer dyspepsia with H2 blockers, treat empirically for ulcer with H. pylori eradication, or do a diagnostic procedure (UGI series or endoscopy). For cost and comfort reasons, my patients are reluctantant to pursue diagnosis and I find compelled to do so only in patients with risk of malignancy. But then which treatment course to pursue? Most cost analysis seems to occur after diagnostic endoscopy and not place this procedure into the equation. Any further perspectives or strategies?

Jim Walsh, MD

Community Health Center of Snohomish County (Washington)

cojim@wolfenet.com

Although only more studies/analyses can answer these questions, one point can be made based on the current study: urease testing prior to therapy with H. Pylori antibiotics was only slightly more expensive than empiric therapy alone, and only as long as the prevalence of H. Pylori was greater than 66%. Looking at all-comers with dyspepsia, the prevalence of H. Pylori is likely to be much lower than the 95% (assumed by the authors) in patients with documented duodenal ulcer. Thus, empiric treatment with antibiotics in addition to H2 blockers for symptoms of dyspepsia does not seem to be justified. Whether urease-test-based antibiotic therapy for H Pylori is a reasonable approach cannot be answered from the data presented here.

*
The authors themselves, in their discussion, state: "Because we assume in this model that duodenal ulcer disease is diagnosed by endoscopy, the results do not apply to patients with suspected ulcer disease or dyspepsia ..."-- mj*

**Journal Club on the Web home page **