Flowchart of process for recruiting subjects for the study of Helicobacter pylori (HP) test-and-treat strategy in patients receiving long-term acid suppression therapy for physician-diagnosed peptic ulcer disease (PUD). Asterisk indicates electronic health plan systems–identified patients with long-term acid suppression therapy and a physician diagnosis of PUD or symptoms. Members were then invited to participate by mail with telephone follow-up. NSAID indicates nonsteroidal anti-inflammatory drug; PPI, proton pump inhibitor; and H2RA, histamine2 receptor antagonist.
Allison JE, Hurley LB, Hiatt RA, Levin TR, Ackerson LM, Lieu TA. A Randomized Controlled Trial of Test-and-Treat Strategy for Helicobacter pyloriClinical Outcomes and Health Care Costs in a Managed Care Population Receiving Long-term Acid Suppression Therapy for Physician-Diagnosed Peptic Ulcer Disease. Arch Intern Med. 2003;163(10):1165-1171. doi:10.1001/archinte.163.10.1165
Guidelines recommend Helicobacter pylori (HP) testing and treatment for patients with a history of peptic ulcer disease (PUD), assuming that PUD has been documented and that successful HP eradication would eliminate the need for further therapy and medical utilization.
An open-label, randomized controlled trial in a managed care setting evaluated the clinical outcome and costs of an HP test-and-treat (T & T) strategy in 650 patients receiving long-term acid suppression therapy for physician-diagnosed PUD. Patients were randomized to T & T for HP (n = 321) or to usual care (n = 329). Outcome measures included presence and severity of PUD symptoms, use of acid-reducing medication, and acid-peptic–related health care costs during 12-month follow-up.
Only 17% of study participants had PUD confirmed by radiography or endoscopy; only 38% of the T & T group tested positive for HP. At 12 months, patients in the T & T group were less likely to report ulcerlike dyspepsia or use of acid-reducing medication; however, 75% of the T & T group used acid-reducing medication during the second half of the 12-month follow-up. In the 12 months after randomization, the T & T group had higher total acid-peptic–related costs than the usual care group.
Most patients receiving long-term acid suppression therapy for physician-diagnosed PUD in community practice settings are likely to have HP-negative, uninvestigated dyspepsia. Routine testing and treating for HP will not reduce acid-peptic–related costs and have only a modest (though statistically significant) effect in reducing clinical symptoms and use of acid-reducing medications.
PUBLISHED GUIDELINES1,2 recommend antibiotic treatment of patients with Helicobacter pylori (HP)–positive peptic ulcer disease (PUD) who receive long-term acid suppression therapy. These guidelines assume PUD has been documented and that successful HP eradication would eliminate the need for further therapy and thereby reduce medical utilization and costs. Health care systems have considered implementing this policy as a cost savings and quality improvement strategy. Further support for this strategy comes from decision analysis models.3 The prevalence of HP infection and the effectiveness of testing and treatment have been well studied in selected patient populations who have documented PUD, nonulcer dyspepsia, or uninvestigated dyspepsia.4- 11 They have not been studied in the population of patients receiving long-term (≥1 year) acid suppression therapy for treatment of chronic or recurrent symptoms that a clinician has attributed to PUD. In the primary care setting, it is not uncommon for such patients to be treated for HP infection without seeking documentation of the physician diagnosis of PUD by radiography or endoscopy or the patient's HP status by serologic testing or urea breath test.12 If most patients receiving long-term acid suppression therapy for chronic PUD diagnosed by a physician have neither PUD nor HP infection, many will be treated unnecessarily and be exposed to undesirable outcomes (eg, multidrug-resistant bacteria or antibiotic-induced adverse effects, including colitis) with little likelihood of benefit. The Study of Management and Costs of H pylori Infection, the STOMACH study, was developed to evaluate these issues.
The STOMACH study was a randomized controlled trial of a test-and-treat (T & T) intervention for HP infection in outpatients receiving long-term acid-reducing therapy for chronic PUD diagnosed (but not necessarily confirmed with esophagogastroduodenoscopy or radiography) by a physician. The primary objective was to determine whether the T & T intervention would decrease PUD symptoms and acid-peptic–related health care costs in the intervention group when compared with the control group receiving usual care (UC), ie, acid suppression medication continued as directed by their physician. The secondary objectives were to determine the prevalence of HP infection in the intervention group and to assess to what extent a physician diagnosis of PUD had been verified by endoscopy, radiography, or both in the 2 study groups.
The Kaiser Foundation Research Institute's institutional review board approved this study, and the study was funded in May of 1996. Study subjects were drawn from the Kaiser Permanente (KP) Medical Care Program, a Northern California group model health maintenance organization with an ethnically and socioeconomically diverse membership representative of the greater San Francisco Bay Area population.13
The KP databases were accessed to identify patients aged 18 to 80 years receiving histamine2 receptor antagonists (H2RAs) or proton pump inhibitors (PPIs) for at least three 30-day periods within a 12-month time frame. Those identified were then linked to an outpatient database of physician diagnoses. Northern California KP physicians are required to diagnose and document the conditions that brought the patient to the clinic. A prior study showed that in 90% of cases where the physician indicated a diagnosis of PUD on an encounter form, a chart entry with that diagnosis also was present.14
Patients receiving H2RA or PPI therapy and having an outpatient diagnosis of duodenal ulcer, gastric ulcer, gastritis, or unspecified PUD were sent an invitation by mail to participate in the study. Those who responded were first screened by telephone and then excluded if they were found to have upper gastrointestinal tract malignancy, medical conditions that precluded endoscopy, or allergies to the study medications. In addition, persons were excluded if they were pregnant, were not currently enrolled in the health plan, or were using nonsteroidal anti-inflammatory drugs (NSAIDs) regularly (ie, >5 times per week). Patients who had gastroesophageal reflux disease symptoms (substernal burning, regurgitation, or both—worse after meals and on lying down and relieved with acid-reducing medication) were also excluded unless they also had documented PUD, ulcerlike dyspepsia, or both. Miscellaneous further exclusions included patients whose HP infection had already been treated, those with no documented PUD or ulcerlike dyspepsia, and those not receiving long-term acid suppression therapy as defined.
Eligible respondents were then invited to attend a group session at which information about the relation between HP infection and PUD, and details about the study requirements were presented. The medical records of attendees who gave a signed informed consent were reviewed to confirm eligibility. Chart records were also reviewed at this time to determine whether a physician diagnosis of PUD had been verified by endoscopy or radiography, or in the absence of these, whether a physician had made a diagnosis of PUD or had recorded a history of ulcerlike dyspepsia (defined as fasting or postprandial epigastric pain relieved by antacids, food, H2RA, or PPI).
Eligible subjects were randomly assigned to either the T & T group or the UC group in blocks of 8 by the recruitment coordinator using a computerized random assignment algorithm provided by the project analyst. The T & T group participants were tested for HP and treated if they tested positive. The T & T participants were not blinded to their HP status or to the success of their anti-HP therapy because blinding would have required producing and administering multiple placebo drugs and because we believed it unethical to withhold information about a positive test result from the patient or his or her physician. Control subjects were asked to continue acid suppression therapy as directed by their personal health care practitioner. The first subject was randomized in September 1996, and the last was randomized in August 1998.
At baseline, 6 months, and 12 months, all T & T and UC participants were asked to complete questionnaires that included demographic information (asked only at baseline), history of chronic medical conditions, medication use, symptoms of PUD, risk factors for ulcer disease (including the use of alcohol, tobacco, NSAIDs, and aspirin); and previous PUD treatment. Functional status assessment was done using the Gastrointestinal Symptom Rating Scale (GSRS). The GSRS (a disease-specific, validated instrument developed and based on reviews of gastrointestinal symptoms and clinical experience) is used to evaluate common symptoms of gastrointestinal disorders.15 Symptoms described by the GSRS questions were rated on a 5-point response scale (1
= no symptoms, 5 = severe symptoms).
At baseline (as part of a process to verify eligibility) and at 15 months after randomization, the medical records of consenting participants were reviewed to capture utilization of services in the prior 12 and 15 months, respectively. Services captured included use of acid-reducing medications, outpatient visits and diagnoses, notes relating to PUD, hospitalizations, gastrointestinal procedures, and any evidence of HP testing, eradication, or both.
Pharmacy databases were used to identify patients receiving long-term acid suppression therapy and to document all prescriptions filled in the 12 months prior to and the 15 months after randomization. Outpatient diagnosis databases were used to identify potential participants with physician-diagnosed PUD and to document use of outpatient medical services. Inpatient databases were used to capture hospitalizations. Cost data were extracted from the KP Northern California Cost Information Management System database that assigns a cost to every health service provided, including patient visits, laboratory tests, medical procedures, and prescriptions filled. Acid-peptic–related costs were those associated with outpatient visits or inpatient stays where the International Classification of Diseases, Ninth Revision (ICD-9) code assigned to the service provided was within the range: 530-537, 555-558, 560-569, 578.9, 787, 789.0. Costs used were those measured from the health care payer's perspective, were fully loaded (including facility overhead), and were amended to include costs of the study's T & T intervention.
Subjects in the T & T group were tested at baseline for HP infection with the PYtest (Tri-Med Specialties Inc, Charlottesville, Va), a carbon C 14 urea breath test (UBT), the study's criterion standard for HP infection. Subjects in the T & T group who tested negative for HP were encouraged to continue their acid-reducing medications as directed by their personal physician and were surveyed at 6 and 12 months after randomization. Subjects in the T & T group who tested positive for HP received 20 mg of omeprazole, 500 mg of clarithromycin, and 1 g of amoxicillin twice a day for 10 days except for those allergic to penicillin, where 500 mg of metronidazole was substituted for the amoxicillin.
Detailed instructions outlining the correct medication regimen as well as possible adverse effects were given to all treated participants. The T & T group participants treated for HP infection were retested for HP by UBT 4 weeks after therapy. Those who were UBT negative 4 weeks after therapy for HP discontinued acid-reducing medication and were observed for as long as 1 year. Those who were UBT positive after therapy for HP were treated with omeprazole, metronidazole, tetracycline, and bismuth (quadruple therapy). Participants in the T & T group who had negative UBT results after therapy for HP infection and who called and reported recurrent symptoms, and participants who did not call but who were identified by questionnaire as having recurrent symptoms after discontinuing acid suppression medication, were offered endoscopy. Patients consenting to endoscopy had a gastric biopsy and had urease activity tested by the CLOtest (Tri-Med Specialties Inc). Biopsy specimens were also analyzed in the pathology department (KP Medical Center, Oakland) for the presence of HP. At 1 year after randomization, UC participants were offered UBT.
The sample size was chosen to ensure 90% power to detect a difference in self-reported symptom rates of 15%, eg, 65% symptom rate in the control group and 50% rate in the T & T group, assuming a dropout rate as high as 20%. All analyses were done by intention to treat. As standardized, the individual symptom scores for the GSRS were derived by averaging the nonmissing responses for each symptom category. In addition, we determined for each subject 3 nonstandardized symptom scores: the mean, minimum, and maximum of all their responses to symptom measures. All statistical tests were 2-sided, and significance was defined as P<.05.
The results of the surveys at each time point (baseline, 6 months, and 12 months) were summarized using frequencies (for nominal and ordinal variables) and means, medians, and standard errors of the mean (for continuous variables). Between-group comparisons at baseline, 6 months, and 12 months were done using 2-sample t tests (for continuous variables), the Wilcoxon rank sum test (for ordinal variables), and the χ2 test (for nominal variables). When the frequencies of the nominal variables were very small (expected values <5), the Fisher exact test was used instead of the χ2 test.
Changes within a group across time were treated as continuous variables so significance was tested with paired t tests. The change scores were summarized using means, medians, and SEMs. A 2-sample t test was used to compare the within-group changes over time among T & T group subjects to changes over time among UC group subjects.
Per-person costs of care were calculated for the 12-month period before randomization and for the 12-month period after randomization. Costs were categorized by type, ie, outpatient services (which include office visits and office-based procedures); laboratory testing; emergency department visits and procedures; inpatient stays; and pharmacy prescriptions and services. Costs were also categorized by whether the cost was acid-peptic related, when this could be determined. Costs were summarized using means, medians, and SEMs. Comparisons of costs across time were done first within a study group using the Wilcoxon rank sum test (data not shown), and then between groups using the Wilcoxon signed rank test. These 2 tests are the nonparametric analogue of the paired and 2-sample t tests and are used because of the nonnormality of the cost data.
The process of study enrollment and reasons for exclusion are shown in Figure 1. The mean age of all 650 randomized participants was 57 years with approximately 66% being 50 years or older, and 48% of participants were male. Non-Hispanic whites represented 63% of the randomized participants; 15% were African American, 9% Latino/Hispanic, and 9% Asian. No statistically significant differences were seen between the 2 randomized groups in any of the following baseline variables: age, sex, race, ulcer documentation, PUD risk factors, country of birth, use of antibiotics in the 12 months before randomization, all self-reported symptom measures, GSRS and SF-12 Health Survey measures, or costs (with the exception of baseline hospital costs, which are discussed below).
Of the 650 study participants, only 17% had PUD confirmed by radiography or endoscopy. This was true for equal proportions of participants in the UC and T & T groups, 15% and 18%, respectively. Of the 58 participants in the T & T group with confirmed PUD, 47% had gastric ulcers, 50% had duodenal ulcers, and 3% had both duodenal and gastric ulcers. Thirty-four percent of this group had PUD confirmed by radiography and 66% had PUD confirmed by endoscopy. The HP prevalence of the T & T group participants with confirmed duodenal ulcer was 58%. For those with confirmed gastric ulcer it was 40%.
Of the 321 participants randomized to the T & T group, 288 (90%) completed the baseline clinic visit. Of these 288, 108 (38%) tested positive for HP by UBT (95% confidence interval, 34%-43%). In 2 persons, UBT results were indeterminate, and these 2 persons refused further testing. The HP-positive T & T group participants (n = 108) were significantly more likely than the T & T HP-negative participants (n = 179) to be nonwhite, non-Hispanic (64% vs 24%), to have documented PUD (28% vs 18%), to be 50 years or older (75% vs 62%), and not to have been exposed to antibiotics in the year before randomization (60% vs 42%) (all at P<.05).
The combined efficacy of a 10-day course of the omeprazole, amoxicillin, and clarythromycin regimen or the metronidazole, omeprazole, and clarithromycin regimen(for penicillin-allergic participants) in eliminating HP was 78% as analyzed by intention to treat and 84% as analyzed per protocol.
Symptoms of PUD in the T & T group decreased from baseline compared with the UC group (Table 1). At 6 months and 12 months, the proportion of participants reporting ulcerlike dyspepsia in the preceding 6 months was significantly lower in the T & T group than in the UC group (P = .01). Similarly, as measured by GSRS, the score for abdominal pain was statistically significantly lower in the T & T group than in the UC group at 6 months (P<.001) and again at 12 months (P = .02). The mean score of all GSRS measures was statistically significantly lower at 6 months (1.8 vs 2.0; P = .04) among T & T group subjects compared with UC subjects but was not different at 12 months (1.9 vs 2.0; P = .14). By all other GSRS indexes (reflux, constipation, diarrhea, and indigestion), the 2 groups did not differ at 6 or 12 months.
Although the T & T and UC groups were similar at baseline with respect to use of acid-reducing medication, use was significantly lower in the T & T group than the UC group at 6 months and 12 months (Table 1). At 12 months, 75% of the patients in the T & T group had used acid-reducing medication in the preceding 6 months compared with 83% of UC group participants (P = .02).
Of the 93 successfully treated participants in the T & T group, 38 (41%) reported recurrent symptoms after stopping their acid-reducing therapy, and 15 of these subjects agreed to have endoscopy. All 15 had negative rapid urease test (CLO Test) results at endoscopy, but 2 had HP identified by histopathologic examination. No ulcers were identified in any of these 15 participants. One participant had antral erosions and another had duodenal erosions, but the others had nonerosive gastritis, duodenitis, nonerosive esophagitis (diagnosed by biopsy, by endoscopist observation [mucosal erythema], or by both). One subject's endoscopy results were completely normal. Only 2 of the 15 had PUD documented by radiography at baseline.
Pharmacy, outpatient, inpatient, and total acid-peptic–related costs in the baseline period (the 12 months before randomization) were not significantly different between the 2 groups. Hospital costs were higher, although not significantly so, in the UC group in the baseline period as the result of an extended hospital stay for 1 patient. Hospital costs in general are highly variable and are probably not meaningful in cohorts the size of ours. Pharmacy, outpatient, and total acid-peptic–related costs after randomization were significantly higher (P<.001) in the T & T group compared with the UC group (Table 1). Consequently, because the 2 groups were similar at baseline, no acid-peptic–related cost savings were realized as a result of the T & T intervention in the 12 months after randomization. The higher costs among T & T group participants after randomization relative to UC subjects can be largely attributed to the costs of UBT and HP treatment. Inpatient costs did not differ between the 2 groups after randomization.
Our data represent the results of community-based practice in a well-defined population with access to medical care. We found only minor symptom relief and no cost savings from the T & T intervention in this population after 12 months of follow-up. Two reasons for the unexpected minimal response are apparent. First, a relatively small proportion had documented PUD. This low proportion of patients with documented PUD was unexpected given that all patients studied had ulcerlike dyspepsia, were receiving long-term acid suppression therapy, and had been diagnosed as having chronic PUD. These patients were, for the most part, patients with uninvestigated dyspepsia and presumed nonulcer dyspepsia. Our data suggest that, as a group, such patients are not likely to benefit from HP eradication. Other randomized controlled trials have shown no statistically significant decrease in dyspeptic symptoms following HP eradication even among HP-positive study patients with nonulcer dyspepsia.16- 18 A recent meta-analysis of 10 randomized controlled trials of therapy for HP in patients with nonulcer dyspepsia19 provided little evidence of efficacy for this treatment strategy in these patients. Thus, it is not surprising that we were unable to show a meaningful reduction in the use of acid-suppressing medication or in health care costs in our T & T group compared with controls.
We did a subset analysis of the T & T group and compared outcomes of those who at baseline were HP positive with those who were HP negative. At baseline, the HP-positive group was more likely to report ulcerlike dyspepsia, abdominal pain on their GSRS, or use of acid-reducing medications than the HP-negative group. It is reasonable to expect that HP-positive patients would show some benefit from treatment for HP infection. In the subset of patients who were HP positive at baseline, 85% of whom tested negative for HP after treatment, there was a greater reduction in symptoms and in use of acid-reducing medications at 12 months after randomization compared with T & T subjects who were HP negative at baseline. The differences between these 2 groups were greater than the differences noted between the whole T & T group and the control group and might have been greater if there had been 100% compliance with the HP treatment regimen. Nonetheless, at 12 months after randomization, 41% of the T & T group who were HP positive at baseline and successfully treated still had ulcerlike dyspepsia and were using or had used acid reduction therapy within the past 6 months.
The second reason for a minimal response in the T & T group was that the prevalence of HP in our study population was unexpectedly low. This finding has important implications for the clinical utility and cost-effectiveness of both T & T strategies, and empirical treatment strategies for HP in patients receiving long-term acid suppression therapy for physician-diagnosed PUD.
Until now, few data have been available on the prevalence of HP in a population already receiving treatment for physician-diagnosed PUD. The prevalence of HP in patients with documented PUD was estimated in 1995 to be between 60% and 100%.7 More recent observations, however, indicate that HP prevalence in patients with documented PUD varies with geography, for example, rates in the United States are generally lower than those found in Western Europe.20- 23 Community studies have reported HP prevalence to be 25% to 90% in patients with duodenal ulcer and 33% to 90% in patients with gastric ulcer.20- 23 For patients with nonulcer dyspepsia and patients with uninvestigated dyspepsia, the reported HP prevalence ranges from 26% to 41%,24,25 which is consistent with our observed HP prevalence of 38% in patients among whom only 17% had documented PUD.
It is not likely that the low HP prevalence in our T & T subjects is spuriously low. It is well documented that UBT offers 95% sensitivity and specificity for HP in the absence of bismuth and PPI use.26,27 At the time of UBT, study patients had not been receiving either antibiotics or bismuth compounds for 4 weeks and had not been receiving PPIs for at least 1 week. Further, it is unlikely that the documented PUD was HP-negative, NSAID-induced PUD. Patients with a history of regular NSAID use were ineligible for the study. NSAID use was determined by a questionnaire administered before randomization and by interview when accurate history of use was not obtainable from the questionnaire.
The reduction in symptoms noted in the T & T group could be a placebo effect that resulted because these participants were not blinded to the results of their HP testing or to their treatment assignment. However, if lack of blinding were important in this design, one would expect to find a certain amount of the benefit that was observed to be spuriously attributed to the intervention. So, the small benefit observed might be overstated.
The response rate to questionnaires administered to the study population at 6 months and 12 months was approximately 70% and 65%, respectively, in both study arms. Those who did respond tended to be older (58 vs 54 years) and have fewer symptoms (as measured by GSRS score at baseline) than those who did not respond (data not shown). However, these differences do not alter our interpretation of the findings because when self-reported symptoms were augmented by evidence of acid-peptic–related symptoms (collected on all patients by chart review) and by evidence of acid-reducing medication use (collected on all patients from administrative pharmacy databases), the results were similar to findings from self-reported data only.
There was no difference in the number of acid-peptic–related visits between T & T and UC groups after randomization. This finding is consistent with our analysis of cost data and supports our conclusion that no cost savings were derived from the T & T intervention. The majority of participants receiving long-term acid suppression therapy were receiving H2RA therapy, which is inexpensive compared with the cost of 2 branded drugs used in our HP eradication regimen. Potential decrease or increase in the cost factor of a T & T intervention over time could affect conclusions about cost savings. Further, cost-effectiveness ratios could become favorable to T & T if patients receive follow-up for more than 12 months.
Out of concern that volunteer bias might have influenced our findings, we reviewed questionnaire and chart review data from 306 potential participants who were not enrolled in the study. The only difference was that the excluded group had more people with documented PUD. Because of heightened physician awareness of HP infection and its association with PUD at the time our subjects were enrolled, it was not uncommon for patients with physician-diagnosed PUD (some documented and others not documented by radiography or endoscopy) to be tested for HP and treated if positive, thus making them ineligible for enrollment in our study. This also may explain to some degree the low prevalence of HP in our eligible population.
Our study showed that in this managed care setting, the majority of patients receiving long-term acid suppression therapy for chronic, physician-diagnosed PUD may have neither PUD nor HP infection. Only 17% of the participants in our study had radiographic or endoscopic confirmation of PUD, and only 38% of the T & T group tested positive for HP. The T & T intervention resulted in a modest reduction of some symptoms, but most T & T group participants continued or resumed some degree of acid suppression therapy within 1 year after the intervention. The intervention did not lead to peptic-acid–related cost savings in the short term. Clinical guidelines and cost-benefit analyses developed for the management of such patients should consider these findings. Peptic ulcer disease must be documented by radiography or endoscopy before adopting a strategy of routine testing for and treating of HP infection in patients similar to those in our study. Otherwise, expectations of dramatic symptom reduction and cost savings in these patients as a group may be unfounded. In settings where the proportion of patients with documented PUD is greater than 17% and where HP prevalence is higher than 38%, a T & T strategy may prove more beneficial than this study suggests.
Corresponding author and reprints: James E. Allison, MD, Division of Research, Kaiser Permanente Medical Care Program, 2000 Broadway, Oakland, CA 94612 (e-mail: email@example.com or firstname.lastname@example.org).
Accepted for publication August 29, 2002.
This research was supported by grant 9444 from Astra Zeneca, Wilmington, Del.
This study was presented at the 101st Annual Meeting of the American Gastroenterological Association, San Diego, Calif, May 21-24, 2000, and at the Annual Meeting of the American Gastroenterological Association and Digestive Disease Week, New Orleans, La, May 17-20, 1998, and has been published in abstract form.
We thank the following members of the Scientific Advisory Group: David A. Peura, MD, Julie Parsonnet, MD, and A. Mark Fendrick, MD, who assisted with study design, resolution of problems, and interpretation of results; Kenneth McQuaid, MD, and Joe Selby, MD, who reviewed the manuscript; and research associates and assistants for help in conducting the study. Lyn Wender and the Medical Editing Department of Kaiser Permanente, Northern California Region, provided editorial assistance.