Helicobacter pylori Test and Treat or Endoscopy for Managing Dyspepsia: An Individual Patient Data Meta-analysis

Transcription

1 GASTROENTEROLOGY 2005;128: Helicobacter pylori Test and Treat or Endoscopy for Managing Dyspepsia: An Individual Patient Data Meta-analysis ALEXANDER C. FORD,* MICHELLE QUME, PAUL MOAYYEDI, NICOLAAS L. A. ARENTS, ANNMARIE T. LASSEN, RICHARD F. A. LOGAN, # KENNETH E. L. MCCOLL,** PAUL MYRES, and BRENDAN C. DELANEY *Centre for Digestive Diseases, Leeds General Infirmary, Leeds, United Kingdom; Department of Primary Care and General Practice, University of Birmingham, Birmingham, United Kingdom; Gastroenterology Division, McMaster University, Health Sciences Center, Hamilton, Ontario, Canada; Department of Medical Microbiology, University Hospital Groningen, Groningen, The Netherlands; Department of Internal Medicine, Odense University Hospital, Odense, Denmark; # Division of Epidemiology and Public Health, Queens Medical Centre, University of Nottingham, Nottingham, United Kingdom; **Section of Medicine, Western Infirmary, Glasgow, United Kingdom; and Cymru Alliance of Primary Care Orientated Research Networks, Wales, United Kingdom Background & Aims: Helicobacter pylori test and treat has been recommended for the management of young dyspeptic patients without alarm symptoms, and trials have suggested that it is as effective as endoscopy. However, none of these trials have had sufficient sample size to confirm that test and treat costs less or to detect small differences in effect. A collaborative group has prospectively registered trials comparing prompt endoscopy with a test and treat approach, with the aim of performing an individual patient data meta-analysis of both effect and resource utilization data. Methods: Researchers provided data for meta-analysis, pooling effects of interventions on individual dyspepsia symptoms. Standardized unit costs were applied to resource utilization, and net benefit was calculated at patient level. Effects, costs, and net benefit were then pooled at study level. Results: Five trials were identified, containing 1924 patients (946 endoscopy [mean age, 40 years], 978 test and treat [mean age, 41 years]). The relative risk (RR) of remaining symptomatic after 1 year was reduced with endoscopy compared with test and treat (RR 0.95; 95% confidence interval [CI]: ). Test and treat cost $389 less per patient (95% CI: $275 $502). Using the net benefit approach, at no realistic level of willingness to pay per patient symptom-free did prompt endoscopy become cost-effective. Conclusions: Prompt endoscopy confers a small benefit in terms of cure of dyspepsia but costs more than test and treat and is not a cost-effective strategy for the initial management of dyspepsia. The management of dyspepsia is an important issue for both primary care physicians and specialists when the initial approach may dictate both patient outcome and future consumption of health care resources. Guidelines in the United States, 1 Canada, 2 and Europe 3 have recommended an empirical approach involving testing for Helicobacter pylori and treating the positive patients ( test and treat ), rather than an initial upper gastrointestinal endoscopy, for patients without symptoms suggestive of malignancy. However, until recently, evidence for the cost-effectiveness of this approach has been based on modeling studies, which have suggested that outcomes would be similar but costs reduced by fewer endoscopies. 4 7 In the past 3 years, several randomized controlled trials (RCTs) based in both primary and secondary care have made comparisons between these 2 approaches These trials are pragmatic in that they measure the benefit the intervention produces in routine clinical practice, their patient selection aims to produce a sample that is generalizable to a real population, they compare 2 strategies and not placebo, and they report on outcomes that are clinically relevant to both doctors and patients. 12 However, large RCTs are difficult to perform in this area, and individual studies have not been powered to evaluate potentially small differences in dyspepsia cure between endoscopy and test and treat. Furthermore, because the raison d etre of test and treat is to reduce costs, it is surprising that trials have either not reported cost data or have done so incompletely. With the current data, it is therefore not possible to evaluate the incremental cost-effectiveness between initial endoscopy and H pylori test and treat. In an attempt to address these uncertainties, the Dyspepsia Trials Collaborators Group has been prospectively registering and incorporating trials into a review to Abbreviations used in this paper: ICER, incremental cost-effectiveness ratio; INB, incremental net benefit; PPI, proton pump inhibitor; RCTs, randomized controlled trials; SMD, standardized mean difference; WMD, weighted mean difference by the American Gastroenterological Association /05/$30.00 doi: /j.gastro

2 June 2005 HELICOBACTER PYLORI TEST AND TREAT 1839 inform the development of dyspepsia management guidelines. 13 We have now conducted an individual patient data meta-analysis of these RCTs to determine accurately which of the 2 strategies is the more costeffective for the initial management of dyspepsia. We have also compared the efficacy of both strategies in terms of cure of dyspepsia in all trial patients, as well as in important prespecified subgroups of patients. Materials and Methods Meta-analysis at trial level is restricted to pooling the data in an aggregated format reported by the trials. Although contact with authors may provide additional information not reported in the publication, this method has a number of limitations that arise because of the heterogeneity between different trials. This heterogeneity has several sources, which can occur as a result of differences between subject populations, application of interventions, effects measures, coding of resource data, and unit costs. Access to the individual patient trial data may resolve these issues, allowing predefined and standardized subgroup analyses, recoding of effect measures, and recoding and application of standard unit cost data, as well as utilizing a more complete dataset, which may also be augmented by imputation of any missing data. Search Strategy The study followed an analysis plan peer-reviewed and published in the Cochrane library. 13 The prospective trials register was supplemented by searches of the Cochrane database of RCTs and Medline until December Eligibility Criteria RCTs were eligible if they compared a prompt endoscopy strategy with a test and treat approach for the initial management of dyspepsia in adults in primary or, on first referral to, secondary care and reported outcome data including symptom resolution, resource use, and cost-effectiveness. We defined dyspepsia as a cluster of symptoms attributable to the upper gastrointestinal tract, including both epigastric pain and heartburn. This is a broader definition than the Rome working party, 14 but studies have shown that in uninvestigated patients in primary care there is considerable overlap in terms of symptoms, and symptom patterns are not predictive of underlying pathology. 15 As a result, many studies include patients with heartburn. Instead, we used the ability of the individual patient data analysis to create subgroups to explore the effect of initial symptom on outcome. Data Cleaning and Validation Researchers were contacted and asked to provide us with their original data sets, as well as their study protocols, validated questionnaires used in the trials, data dictionaries, and analysis plans. These were translated into English, when required, with the help of the researchers. The data sets were then transformed, and the authors original analyses were replicated to ensure the integrity of the data obtained. When discrepancies arose, these were clarified, and the data set amended if necessary. Following this, the data sets for each trial were standardized with respect to variable labels, and reporting of symptom and cost data, to allow direct comparisons to be made among the trials whenever possible. The symptom data were rationalized, with the exclusion of questionnaire items such as diarrhea or flatulence and to include only the components of dyspepsia as defined above (epigastric/upper abdominal pain, heartburn, regurgitation, and nausea). Data Synthesis Symptom data were analyzed at 12 months because all the trials collected data at this stage of follow-up. This was done in 2 ways: first, as a total score derived from the 4 key symptoms mentioned above. A mean and standard deviation were calculated for each trial and the results pooled. Second, they were dichotomized into the presence or absence of those 4 symptoms. This was performed using the symptom frequency or severity response, depending on the questionnaire used in each trial, with absence of symptoms defined as a negative response to that symptom. 16 Patients were defined as asymptomatic at trial completion if they reported none of these symptoms. Data were expressed as intention-to-treat analyses. Prespecified subgroup analyses examining symptom status at 12 months were conducted for patients according to sex, age (less than 50 years or 50 years and over), predominant symptom at trial entry (epigastric pain or heartburn), and initial H pylori status. Cost data were derived from the total dyspepsia-related resource use per patient at trial completion. These were separated into primary and secondary care costs (including primary care and outpatient consultations with dyspepsia and inpatient admissions as a consequence of dyspepsia), costs of prescribed drugs for dyspepsia (using total defined daily doses of acid suppression drugs and number of courses of eradication therapy), and investigation rates (number of barium meals, upper GI endoscopies, abdominal ultrasound scans, and breath tests). Costs were obtained for the US setting for each of these to obtain a total cost per patient. Drug costs were obtained from the average retail prices for pharmaceuticals. 17 Physician costs, including procedures, were obtained from the American Medical Association procedural code book and the 2003 Medicare fee schedule. These are outlined in Table 1. Traditionally, in health-economic studies, the outcome of interest is the incremental cost-effectiveness ratio (ICER). 18 This is a comparative measure of the difference in cost ( C) divided by the difference in benefit ( E) between 1 strategy and another (ICER C/ E). However, it is incorrect to pool costs separately from effects because they are correlated at the individual patient level. It is self-evident that symptomatic patients will tend to use more resources (drugs and others) than asymptomatic patients. In addition, the ICER is a ratio,

3 1840 FORD ET AL GASTROENTEROLOGY Vol. 128, No. 7 Table 1. Costs Used in Obtaining a Total Cost per Patient Cost ($ US) General practitioner visit 170 Outpatient visit 232 Inpatient day 550 PPI (1-month single dose) H 2 RA (1 month) Prokinetic (1 month) 70 Antacid (1 month) 8.49 Eradication therapy 152 Urea breath test 80 Endoscopy 450 Barium meal Abdominal ultrasound scan 118 and ratios cannot be pooled in conventional meta-analysis. 19 For this reason, we use the incremental net benefit (INB). This is derived by rearranging the previous formula such that INB in monetary terms E C, in which is the maximum amount a third-party payer or patient is willing to pay per dyspepsia case symptom-free per year, E is the benefit (assigned a value of 1 if the patient is asymptomatic and 0 if they are symptomatic), and C is the difference in cost. The INB can be calculated at the individual patient data level and is normally distributed. 20 This allowed us to calculate a mean net benefit and standard deviation for each trial arm and pool the results. We were then able to perform sensitivity analyses to different levels of willingness to pay (ie, different values of ). Statistical Analysis The impact of the 2 management strategies on total dyspepsia symptom score and presence of dyspepsia at 12 months were expressed as a standardized mean difference (SMD) and a relative risk (RR), respectively, along with their 95% confidence intervals (CI). Cost and cost-effectiveness data were reported as weighted mean differences (WMD), with a 95% CI. Where significant heterogeneity between trial results was detected, a random effects model was used. All statistical analyses were performed using SPSS for Windows version 11.5 (SPSS Inc, Chicago, IL) and Stata version 8.0 (Stata Corporation, College Station, TX). Results Five trials were identified comparing management based on a prompt upper GI endoscopy strategy with an H pylori test and treat approach for the initial management of dyspepsia. Three trials 9,10 (Dr. Paul Myres, personal communication, May 2004) were registered with the collaborator s group, of which one 10 was published as a full paper. One trial (Myres) had closed early because of poor recruitment, but full follow-up data had been collected on most of the patients recruited, expressly for the purpose of this analysis. Two trials 8,11 were found in published form, and the authors provided full data sets for analysis. Individual trial characteristics and demographics are given in Table 2. Four of the trials included H pylori testing as part of the treatment strategy in both arms of the trial, and 1 of these only included H pylori positive patients. Three of the trials set an upper age limit for patient eligibility. Only 1 trial excluded patients with symptoms suggestive of gastroesophageal reflux disease, 8 and 2 of the studies 10,11 recorded the predominant symptom at trial entry. All trials excluded patients with symptoms suggestive of upper gastrointestinal malignancy. The trials included a total of 1924 patients, 946 of whom underwent prompt endoscopy and 978 of whom were assigned to test and treat. Patient demographics according to assigned strategy are given in Table 3. The SMD in symptom scores at 12 months for patients undergoing endoscopy compared with test and treat was 0.11 (95% CI: ) (Figure 1). However, when the data were dichotomized, the combined relative risk of remaining symptomatic at 12 months with endoscopy compared with test and treat was 0.95 (95% CI: ) (Figure 2). There was therefore a small, but statistically significant, improvement in dyspepsia symptom status with endoscopy-based management. With respect to the predefined subgroup analyses examining symptom status at 12 months, there were no differences between the 2 strategies for females (RR 0.95; 95% CI: ) or for males (RR 0.97; 95% CI: ). In terms of patient age, there was a small, but statistically significant, effect in favor of prompt endoscopy in patients aged 50 years and above (RR 0.90; 95% CI: , P.05), compared with no difference in effect in the under 50 years of age group (RR 0.97; 95% CI: ). Predominant symptom at trial entry was reported in a total of 929 patients. There appeared to be no overall difference in effect between the 2 strategies for either patients with predominant epigastric pain (RR 0.96; 95% CI: ) or those with predominant heartburn (RR 0.99; 95% CI: ), although the power of the analysis is lower on account of predominant symptom at trial entry only being reported in 2 trials. Finally, initial H pylori status was reported in a total of 1539 patients. There was no overall difference in effect in either H pylori positive patients (RR 0.93; 95% CI: ) or H pylori negative patients (RR 0.97; 95% CI: ). Cost data were available for 1771 patients. The WMD in total cost for all patients was $389 (95% CI: $276 $502) (Figure 3), indicating that prompt endoscopy cost more per patient than test and treat. When examining

4 June 2005 HELICOBACTER PYLORI TEST AND TREAT 1841 Table 2. Trial Characteristics and Patient Demographics Arents Duggan Lassen McColl Myres Setting Primary care Primary care Primary and Secondary care Primary care secondary care Country Netherlands England Denmark Scotland Wales Upper age limit (y) None 70 None Duration of followup (mo) Method of noninvasive H Laboratory-based serology Near-patient serology 13 C-urea breath test 14 C-urea breath test Laboratory-based serology pylori testing for test and treat arm Patients undergoing Yes No Yes Yes Yes endoscopy tested for H pylori as well Patients undergoing endoscopy given H pylori eradication per protocol if positive Yes No No Yes No Eradication regimen Lansoprazole, amoxicillin, and metronidazole/ clarithromycin PPI, metronidazole and clarithromycin Lansoprazole, amoxicillin, and metronidazole Omeprazole, clarithromycin, and amoxicillin/ metronidazole At physicians discretion Duration of treatment 1 week 1 week 2 weeks 1 week At physicians discretion No. of patients 270 (141 test and treat ) 385 (198 test and treat ) 500 (250 test and treat ) 708 (356 test and treat ) 61 (33 test and treat ) Mean age (SD) 44 (14) 41 (12) 46 (16) 36 (9) 34 (6) Sex (%) 129 male (48) 208 male (54) 230 male (46) 377 male (53) 28 male (46) H pylori positive (%) 102 (38) 46 (23)* 141 (28) 352 (50) 61 (100)** a Patients assigned to prompt endoscopy were not tested for H pylori. b Only H pylori positive patients were eligible for trial inclusion. the cost data in the 4 separate areas, it became apparent that most of this increased burden was accounted for by the cost of investigations in the prompt endoscopy patients (WMD $318; 95% CI: $285 $350) (Table 4). Despite the small effect difference in favor of endoscopy, when the mean net benefit for each trial arm was pooled, increasing the value for willingness to pay per patient symptom-free ( ) from $0 to $1000 did not make prompt endoscopy a cost-effective approach Table 3. Patient Demographics According to Strategy Assignment Prompt UGIE (n 946) Test and treat (n 978) Mean age (SD) 40 (13) 41 (13) Males (%) 487 (51) 483 (49) H pylori positive (%) a 334 (44) 322 (41) Epigastric pain (%) b 297 (49) 266 (44) Heartburn (%) b 162 (27) 204 (34) Smoker (%) 369 (39) 405 (41) a Four trials reported initial H pylori status in a total of 1539 patients (759 prompt UGIE, 780 test and treat ). b Two trials reported predominant symptom at trial entry in a total of 1208 patients (602 prompt UGIE, 606 test and treat ). Figure 1. Standardized mean difference in symptom scores at 12 months with prompt endoscopy compared with test and treat., Individual RCT effect (area proportional to study size);, pooled effect.

5 1842 FORD ET AL GASTROENTEROLOGY Vol. 128, No. 7 Table 4. Weighted Mean Difference in Costs for Prompt Endoscopy vs Test and Treat Cost (US $) Weighted mean difference 95% Confidence interval Total cost Primary care costs Secondary care costs Investigation costs Drug costs Figure 2. Relative risk of remaining symptomatic at 12 months with prompt endoscopy compared with test and treat., Individual RCT effect (area proportional to study size);, pooled effect. (WMD $330; 95% CI: $236 to $423) (Figure 4). Prompt endoscopy eventually became cost-effective if the willingness to pay per patient symptom-free of dyspepsia was $180,000. To examine the influence that cost of endoscopy had on the cost-effectiveness of the 2 strategies, a sensitivity analysis was performed, with the cost of endoscopy reduced to $80, which is in line with charges in some European countries but probably underestimates the true economic cost of endoscopy. Even in this situation, prompt endoscopy only became cost-effective when the willingness to pay per patient symptom-free of dyspepsia reached $40,000. Critics of the adoption of a test and treat approach for the initial management of dyspepsia are concerned about the possibility of a missed diagnosis of upper gastrointestinal malignancy. Of the 1924 patients who participated in these trials, 5 (0.26%) were found to have such a diagnosis during follow-up. Among the prompt endoscopy patients, 1 patient was found to have a gastric adenocarcinoma, 1 a low-grade gastric MALToma that responded to eradication therapy, and 1 a gastric lymphoma. Of the individuals assigned to test and treat, diagnoses included 1 advanced gastric carcinoma with liver metastases, which was diagnosed 3 weeks postrandomization, and 1 esophageal carcinoma. Of these 5 patients, 1 died as a direct result of their diagnosis during follow-up. There were no adverse events reported as a direct result of endoscopy. Discussion This study, the first of its kind in this field, has used rigorous methodology to confirm that a prompt endoscopy strategy for the initial management of dyspepsia confers a clinically small, but statistically significant, benefit in terms of patients symptom resolution compared with a test and Figure 3. Weighted mean difference in total cost at 12 months for prompt endoscopy compared with test and treat., Individual RCT effect (area proportional to study size);, pooled effect. Figure 4. Weighted mean difference in net benefit for prompt endoscopy vs test and treat at a willingness to pay of $1000., Individual RCT effect (area proportional to study size);, pooled effect.

6 June 2005 HELICOBACTER PYLORI TEST AND TREAT 1843 treat approach. The explanation for this difference is not clear. We noted an increase in proton pump inhibitor (PPI) prescribing in those patients assigned to prompt endoscopy. However, when a subgroup analysis was performed to examine the relative risk of remaining symptomatic in patients according to PPI usage, there were no significant differences detected. The power of this post hoc analysis is limited, and it may be that the small difference observed is due to greater use of PPIs postendoscopy. PPIs are the most effective treatment for acid-related dyspeptic symptoms, particularly heartburn, and statistically small differences in use might be amplified by their effect on symptoms. An alternative explanation for this benefit may be the reassurance effect of a normal endoscopy, which was alluded to elsewhere, 21 but is thought to be quite short-lived. In addition, the trials were unblinded, and the possibility that this might have led to a small bias in favor of endoscopy cannot be excluded. Blinding in pragmatic trials is a controversial issue, some authors favoring a real world situation without blinding, others not. 22 Blinding between endoscopy or testing alone would not have been possible in this case. It is worth noting that a lack of blinding of outcome assessors would be a serious problem. However, in these trials, outcomes were measured by the completion of validated questionnaires by patients. The correlation between patients and clinicians assessments of gastroesophageal reflux symptoms has been found to be poor, and it has been recommended that clinical trials should use patient self-completed outcome measures for dyspepsia and reflux symptoms. 23,24 The point remains that, despite the small difference in effect, the prompt endoscopy strategy is not cost-effective at a realistic level of willingness to pay for cure of dyspepsia. Large meta-analyses have the power to detect small differences in effect that, although statistically significant, may have little clinical relevance. Of the 3 trials in published form, 8,10,11 all reported on symptom outcome in terms of mean change in symptom score, whereas one 10 also recorded proportion of days symptomatic in the last year and another 11 dichotomized patients into symptomatic or asymptomatic. None of these trials found any significant differences in symptom status, using any of the above measures, between the 2 strategies. All 3 trials found that resource use was comparable for both approaches, with the caveat that those assigned to test and treat underwent fewer endoscopies. In addition, 2 of the studies 8,10 reported an increased level of PPI consumption in the prompt endoscopy group, as we noted. This individual patient data meta-analysis has clarified the issues left unresolved by the original Cochrane review. 13 The principal differences between the trial-based metaanalysis and the individual patient data meta-analysis were the incorporation of cost data and a reduction in heterogeneity between the trials, with the emergence of a small difference in effect. The trial-based meta-analysis found no significant difference in either total symptom score or dichotomous symptom outcome between the 2 strategies at trial completion, but the confidence intervals were wide, and there was significant heterogeneity between trial results when pooled. It was postulated that this heterogeneity may have arisen because of differences in the efficacy of test and treat according to trial setting, with only 1 of the trials being exclusively based in primary care. The lack of heterogeneity observed in our study may be due to the exclusion of nondyspeptic symptoms, such as flatus and tenesmus, from the analyses. Alternatively, our data may support the review s hypothesis because we now have more primary care-based trials included in the meta-analysis. With respect to resource use, the review confirmed the significant reduction in endoscopy utilization with a test and treat approach reported by the original studies. This is supported by our finding that the significant difference in costs between the 2 strategies was largely attributable to an increase in investigative costs in the prompt endoscopy patients compared with their test and treat counterparts. None of the trials incorporated in the original review reported their full health economic data; therefore, a cost-effectiveness analysis based on trial data rather than modeling has not previously been available. There are some potential limitations of the present study. Unpacking validated questionnaires, such as the Gastrointestinal Symptom Rating Scale 25 and the Glasgow dyspepsia severity score, 26 could be criticized on the basis that the performance of the modified score is unknown. In addition, the dichotomization of symptom status could be considered an oversimplification of a complex issue. However, we believe that the exclusion of unrelated gastrointestinal symptoms and the dichotomization process are necessary and justified to allow pooling at the individual patient data level and are more clinically meaningful. Another difficulty we encountered was that reference costs for drugs, consultations, and investigations for Denmark and The Netherlands did not exist, and we were therefore obliged to apply either United Kingdom or United States reference costs to resource use reported in these trials. This method does not allow for substitution effects, whereby the difference in cost of investigations within a country (eg, noninvasive H pylori testing and endoscopy) lead to greater use of the cheaper resource. Although this concept has some theoretic validity, it is unlikely that resource use at individual patient level is significantly directed by purely economic considerations. In addition, these are RCTs, in which the principal determinants of cost were not directed by choice but by the randomization process. These factors may affect the generalizability of our findings, particularly in countries in which

7 1844 FORD ET AL GASTROENTEROLOGY Vol. 128, No. 7 the cost of endoscopy is lower than in the United States. In that the absolute difference in effect between the 2 strategies is so small, the cost of endoscopy would have to be very low indeed (lower than the lowest feasible estimate considered in our sensitivity analysis), and the cost of H pylori testing very high, for the prompt endoscopy strategy to become cost-effective at a more realistic level of willingness to pay. Finally, it should be remembered that all studies included in this meta-analysis reported on symptom status of included patients at 12 months. We have little data regarding the effects of these 2 approaches on the long-term natural history of dyspepsia, although 1 researcher has now completed a 6-year follow-up study on the original cohort of included patients, and findings show that the difference in resource use at 12 months appears to continue thereafter, with no difference in the symptom status between the 2 strategies and a continued increase in resource use in the prompt endoscopy patients. 27 This study has confirmed that the cost of prompt endoscopy as a first-line approach for the management of dyspepsia without alarm symptoms is prohibitive in everyday clinical practice, and a test and treat strategy should be preferred. References 1. Howden CW, Hunt RH. Guidelines for the management of Helicobacter pylori infection. Am J Gastroenterol 1998;93: Hunt RH, Thomson ABR. Canadian Helicobacter pylori Consensus Conference. Can J Gastroenterol 1998;12: Malfertheiner P, Megraud F, O Morain C, Hungin APS, Jones R, Axon A, Graham DY, Tytgat G. Current concepts in the management of Helicobacter pylori infection The Maastricht Consensus Report. Aliment Pharmacol Ther 2002;16: Briggs AH, Sculpher MJ, Logan RPH, Aldous J, Ramsay ME, Baron JH. Cost effectiveness of screening for and eradication of Helicobacter pylori in management of dyspeptic patients under 45 years of age. BMJ 1996;312: Fendrick AM, Chernew ME, Hirth RA, Bloom BS. Alternative management strategies for patients with suspected peptic ulcer disease. Ann Intern Med 1995;123: Ofman JJ, Etchason J, Fullerton S, Kahn KL, Soll AH. Management strategies for Helicobacter pylori-seropositive patients with dyspepsia: clinical and economic consequences. Ann Intern Med 1997;126: Silverstein M, Petterson T, Talley N. Initial endoscopy or empirical therapy with or without testing for Helicobacter pylori for dyspepsia: a decision analysis. Gastroenterology 1996;110: Arents NLA, Thijs JC, van Zwet AA, Oudkerk Pool M, Gotz JM, van de Werf GT, Reenders K, Sluiter WJ, Kleibeuker JH. Approach to treatment of dyspepsia in primary care: a randomized trial comparing test-and-treat with prompt endoscopy. Arch Intern Med 2003;163: Duggan AE, Elliott CA, Hawkey CJ, Logan RFA. Does initial management of patients with dyspepsia alter symptom response and patient satisfaction? Results from a randomised trial. Gastroenterology 1999;116(Suppl 4):G Lassen AT, Pedersen FM, Bytzer P, Schaffalitzky de Muckadell OB. Helicobacter pylori test-and-eradicate versus prompt endoscopy for management of dyspeptic patients: a randomised trial. Lancet 2000;356: McColl KEL, Murray LS, Gillen D, Walker A, Wirz A, Fletcher J, Mowat C, Henry E, Kelman A, Dickson A. Randomised trial of endoscopy with testing for Helicobacter pylori compared with non-invasive H pylori testing alone in the management of dyspepsia. BMJ 2002;324: Roland M, Torgerson DJ. Understanding controlled trials: what are pragmatic trials? BMJ 1998;316: Delaney BC, Forman D, Moayyedi P. Initial management strategies for dyspepsia (Cochrane Review). Cochrane Library, Issue 2. Chichester, UK: John Wiley & Sons, Drossman DA, Thompson WG, Talley NJ, Funch-Jensen P, Janssens J, Whitehead WE. Identification of subgroups of functional bowel disorders. Gastroenterol Int 1990;3: Chiba N, Thompson ABR, Barkun AN, Armstrong D, Veldhuyzen Van Zanten SJO, White RJ, Escobedo S, Sinclair P. The Rome II definition of dyspepsia does not exclude patients with GERD in primary care. Gastroenterology 2003;124:A Fraser A, Delaney BC, Moayyedi P. Symptom-based outcome measures for dyspepsia and GERD trials: a systematic review. Am J Gastroenterol 2005;100: Prescription drug prices and information. Available at: www. drugstore.com. Accessed August Mandelblatt JS, Fryback DG, Weinstein MC, Russell LB, Gold MR. Assessing the effectiveness of health interventions for costeffectiveness analysis. Panel on cost-effectiveness in health and medicine. J Gen Intern Med 1997;12: Stinnett AA, Paltiel AD. Estimating cost-effectiveness ratios under second-order uncertainty: the mean ratio versus the ratio of means. Med Decis Making 1997;17: Briggs AH. Statistical approaches to handling uncertainty in health economic evaluation. Eur J Gastroenterol Hepatol 2004;16: Bytzer P, Hansen JM, Schaffalitzky de Muckadell OB. Empirical H2-blocker therapy or prompt endoscopy in management of dyspepsia. Lancet 1994;343: Freemantle N, Drummond MF. Should clinical trials with concurrent economic analyses be blinded? JAMA 1997;277: McColl E, Junghard O, Wiklund I, Revicki DA. Assessing symptoms in gastroesophageal reflux disease. Am J Gastroenterol 2005;100: Moayyedi P, Delaney BC. Measuring gastroesophageal reflux symptoms: musings from Marrakech. Am J Gastroenterol 2005; 100: Svedlund J, Sjodin I, Dotevall G. GSRS a clinical rating scale for gastrointestinal symptoms in patients with irritable bowel syndrome and peptic ulcer disease. Dig Dis Sci 1988;33: El-Omar EM, Banerjee S, Wirz A, McColl KEL. The Glasgow Dyspepsia Severity Score a tool for the global measurement of dyspepsia. Eur J Gastroenterol Hepatol 1996;8: Lassen AT, Hallas J, Schaffalitzky de Muckadell OB. Helicobacter pylori test and eradicate versus prompt endoscopy for management of dyspeptic patients: 6.7-year follow-up of a randomised controlled trial. Gut 2004;53: Received November 23, Accepted February 23, Address requests for reprints to: Brendan C. Delaney, Professor of Primary Care, Department of Primary Care and General Practice, University of Birmingham, Edgbaston, Birmingham, B15 2TT, United Kingdom. b.c.delaney@bham.ac.uk; fax: (44) Paul Moayyedi has received speaker s fees and research funds from AstraZeneca, Wyeth Laboratories, and Abbott Laboratories. Brendan C. Delaney has received speaker s fees from AstraZeneca, Eisai, Wyeth, and AxCan Pharma; holds grants from the MRC and NHS R&D program; and is supported by an NHS R&D Primary Care Career Scientist Award (No. CSA99/008).