Given what we know about H. pylori-related diseases, would it not be best from a societal perspective to eliminate them altogether? Although we are currently unable to prevent infection, results of animal studies suggest that it may be possible to develop a vaccine [2]. Lacking an effective vaccine, but given our ability to identify and cure infected persons, perhaps we should consider widespread noninvasive testing followed by treatment of all infected persons. Clearly, it is not yet be prudent to advocate such an approach because beyond issues of feasibility and cost, we lack information on possible untoward outcomes associated with this approach. Measurement of outcomes related to peptic ulcer and gastric cancer is relatively straightforward. However, we also need to consider the many potential adverse outcomes associated with widespread use of antimicrobial agents, such as illnesses related to alterations in the normal human bacterial flora that protect mucosal surfaces or have a role in the metabolism of drugs and carcinogens. Another important adverse effect may be increased resistance to antimicrobial agents in bacteria that are not the intended targets of therapy. The number of human pathogens that have acquired resistance to antimicrobial agents continues to increase, to the extent that our continued ability to successfully treat important infections, such as bacterial pneumonia, has been questioned [3].

As public awareness about the links among H. pylori infection, peptic ulcer, and gastric cancer increases, physicians will be under increasing pressure from patients to test for infection. However, we believe that it is inappropriate to test patients for infection unless the physician is prepared to discuss the potential consequences of infection and offer therapy. The physician must tell the patient that the infection damages the structure and function of the stomach, can lead to peptic ulcer and gastric cancer, can potentially be transmitted to others in the community, and in fact has no known beneficial effects for the host or for society. Thus, our guiding principle is "Do not test if you are not willing to treat." Whom, then, should we be willing to treat? On the basis of the available evidence, we should treat all patients who have or have had active peptic ulcer disease [4]. The second group of patients may include those with dyspepsia.

Dyspepsia consists of episodic or persistent upper abdominal pain or discomfort that is thought by the physician to arise in the upper gastrointestinal tract [5-8]. Uncomplicated dyspepsia refers to dyspepsia that is not signaled by coexisting "warning" features, such as anemia and unexplained weight loss, that indicate the need for prompt diagnostic work-up. In most patients with uncomplicated dyspepsia who have endoscopy, no abnormality is detected. In three large series, for example, only one third of patients had an abnormality that explained their symptoms [9-11]. The main findings were erosive esophagitis, peptic ulcer disease, and gastric cancer. A varying proportion of patients had gastric or duodenal erosions of uncertain relation to their symptoms, and some patients with normal endoscopic results undoubtedly had gastroesophageal reflux disease. Clearly, however, many patients with dyspepsia have no detectable organic disease. The critical question is how best to manage patients with uncomplicated dyspepsia in the H. pylori era.

One approach investigators have used to tackle this question is to develop and prospectively test algorithms, based on H. pylori serologic testing and clinical factors (such as age and use of nonsteroidal anti-inflammatory drugs), that are intended to identify patients with organic disease [12, 13]. Although the use of these algorithms can reduce the volume of endoscopies in patients referred to open-access units, the algorithms do not guide primary care physicians in the use of noninvasive testing or antimicrobial therapy in overall patient management.

Another approach is to use decision analysis models that are based on data obtained from the published literature. One model [14], which compared two invasive strategies (prompt endoscopy with or without biopsies to detect H. pylori) with three noninvasive strategies (serologic H. pylori testing and treatment of infected patients, empirical antisecretory therapy, and empirical antisecretory and antimicrobial therapy), found that the noninvasive strategies were associated with lower costs per ulcer cured at 1 year. A second model indicated a "toss-up" in terms of cost and life expectancy [15]. These discrepant results were attributed primarily to differences in the data (cost and probability estimates) used in the two models [15]. A third model [16] found that at 1 year, costs for empirical antisecretory therapy were lower than costs for serologic H. pylori testing and prompt endoscopy in infected patients. Cost savings for the serologic testing and endoscopy strategy did not begin to accrue for at least 8 years [16].

In this issue, Ofman and colleagues [17] report the results of a decision analysis model that compared prompt endoscopy with empirical antisecretory and antimicrobial therapy in a hypothetical cohort of patients who had dyspepsia and were known to be infected with H. pylori on the basis of serologic testing. The investigators found that the empirical (nonendoscopic) antisecretory and antimicrobial strategy for dyspeptic patients with H. pylori infection was associated with lower costs at 1 year. The structure of Ofman and colleagues' decision tree is clinically sensible, and the authors incorporated complications of endoscopy and antimicrobial therapy into the model. In addition, the data on which the model was based are reasonable. Further, the sensitivity analyses provide fresh insights. For example, they indicate that the financial benefits of the empirical strategy do not depend on the response to H. pylori therapy in patients without organic disease. On the other hand, largely to cope with existing gaps in the available data, the investigators assumed that the clinical outcomes for the two strategies were equivalent. What Ofman and colleagues gained by using this assumption is a simplification of the analysis to a comparison of costs. However, they gave up their ability to provide information on important clinical outcomes, including pain from dyspepsia and satisfaction with dyspepsia-related health, that had not previously been fully addressed [14, 15].

The conceptual revolution in the pathogenesis of peptic ulcer disease and gastric cancer has raised the larger question of how to integrate noninvasive H. pylori tests and antimicrobial therapy into the management of patients in primary practice settings. Taken together, the evidence obtained from decision analyses indicates that antimicrobial and antisecretory therapy may be the preferred way to manage persons who have dyspepsia and are shown by noninvasive testing to have H. pylori infection. Now that this indirect evidence is emerging, it is time for a randomized clinical trial comparing alternate strategies. Such a trial is needed to directly measure key symptom-related outcomes, such as pain from dyspepsia, for which these patients seek health care.