Technology Report. Gastro-duodenal Ulcers Associated with the Use of Non-steroidal Antiinflammatory

Transcription

1 Technology Report Issue 38 September 2003 Gastro-duodenal Ulcers Associated with the Use of Non-steroidal Antiinflammatory Drugs: A Systematic Review of Preventive Pharmacological Interventions

2 Publications can be requested from: CCOHTA Carling Avenue Ottawa, Ontario, Canada K1S 5S8 Tel. (613) Fax. (613) or download from CCOHTA s web site: Cite as: Rostom A, Dubé C, Jolicoeur E, Boucher M, Joyce J. Gastro-duodenal ulcers associated with the use of non-steroidal anti-inflammatory drugs: a systematic review of preventive pharmacological interventions. Ottawa: Canadian Coordinating Office for Health Technology Assessment; Technology report no 38. Reproduction of this document for non-commercial purposes is permitted provided appropriate credit is given to CCOHTA. CCOHTA is a non-profit organization funded by the federal, provincial and territorial governments. Legal Deposit National Library of Canada ISBN: (print) ISBN: (electronic version) Publications Mail Agreement Number:

3 Canadian Coordinating Office for Health Technology Assessment Gastro-duodenal Ulcers Associated with the Use of Non-steroidal Anti-inflammatory Drugs: A Systematic Review of Preventive Pharmacological Interventions Alaa Rostom MD MSc FRCPC 1 Catherine Dubé MD MSc FRCPC 1 Emilie Jolicoeur MD 1 Michel Boucher BPharm MSc 2 Janet Joyce MLS 2 September University of Ottawa, Ottawa, Ontario 2 Canadian Coordinating Office for Health Technology Assessment, Ottawa, Ontario

4 Reviewers These individuals kindly provided comments on this report. External Reviewers Isabelle Chabot, PhD Manager, Health Economics and Outcomes Research Merck Frosst Canada & Co. Kirkland, Quebec John M. Fardy, MD MSc FRCPC Associate Professor of Medicine (Gastroenterology) Memorial University of Newfoundland St. John's, Newfoundland and Labrador Gail Huxley, Scientific Development Manager National Medicine Boehringer Ingelheim (Canada) Ltd. Burlington, Ontario John K. Marshall, MD MSc FRCPC Assistant Professor Division of Gastroenterology McMaster University Hamilton, Ontario Betsy Miller, BScPharm MSc Executive Director Patient Access and Outcomes Research Pharmacia Canada Inc. Mississauga, Ontario Andreas Maetzel, MD MSc PhD Scientist, Assistant Professor Division of Clinical Decision Making University Health Network University of Toronto Toronto, Ontario Alan B.R. Thomson, MD PhD FRCPC FACP FACG Professor of Medicine University of Alberta Edmonton, Alberta Angie Wong, BScPharm MSc Manager, Outcomes Research Pharmacia Canada Inc. Mississauga, Ontario This report was also reviewed by Pfizer Canada Inc., Kirkland, Quebec. CCOHTA Scientific Advisory Panel Reviewers Kenneth Marshall, MD Professor of Family Medicine (retired) University of Western Ontario London, Ontario George Wells, PhD Director, Department of Epidemiology and Community Medicine University of Ottawa Ottawa, Ontario This report is a review of existing public literature, studies, materials and other information and documentation (collectively the source documentation ) which are available to CCOHTA. The accuracy of the contents of the source documentation on which this report is based is not warranted, assured or represented in any way by CCOHTA and CCOHTA does not assume responsibility for the quality, propriety, inaccuracies or reasonableness of any statements, information or conclusions contained in the source documentation. CCOHTA takes sole responsibility for the final form and content of this report. The statements and conclusions in this report are those of CCOHTA and not of its Panel members or reviewers. i

5 Authorship Alaa Rostom, the principal investigator, participated in all aspects of the project and was responsible for the primary writing of the report. Catherine Dubé participated in the design of the project, selection of relevant studies and analysis of the data. She also critically reviewed all drafts of the report. Emilie Jolicoeur participated in the selection of included studies and in data abstraction and assisted in analysis and interpretation of the results. She also critically reviewed all drafts of the report. Michel Boucher participated in the design of the project, selection of relevant studies and writing of the initial draft and all report revisions. He also coordinated the external review process. Janet Joyce was responsible for designing and conducting the literature search, writing the search methods, preparing the search strategies appendix and compiling the bibliography. Conflict of Interest Dr. John K. Marshall declared that he has received speaker s fees from Merck Frosst Canada & Co. ii

6 REPORT IN BRIEF September 2003 Drugs to Prevent Gastro-duodenal Ulcers Associated with the Use of NSAIDs Technology Name Non-steroidal anti-inflammatory drugs (NSAIDs) Gastroprotective agents: misoprostol, histamine type-2 receptor antagonists (H 2 RAs), proton pump inhibitors (PPIs) Cyclooxygenase (COX)-2 selective NSAIDs: celecoxib, rofecoxib, meloxicam. Disease/Condition Traditional NSAIDs, which are used to treat painful arthritic and inflammatory disorders, can produce gastro-intestinal (GI) adverse effects. The incidence of serious complications is low, but it is an important clinical issue because of the widespread use of these medications. Technology Description Gastroprotective agents can help protect the stomach when an NSAID is used. COX is an enzyme. Type 1 COX (COX- 1) is involved in protecting the stomach lining, while the second type (COX-2) may be important in promoting the pain of arthritic diseases. Traditional NSAIDs inhibit the effects of COX-1 and COX-2. NSAIDs that selectively inhibit COX-2 theoretically should have little effect on the GI tract. The Issue The efficacy and safety profiles of available gastroprotective agents should be compared when they are used to protect against NSAID-induced GI damage. The GI safety profiles of COX-2 selective NSAIDs should be compared with those of traditional non-selective NSAIDs. Assessment Objectives 1. To assess how well gastroprotective agents protect against the upper GI damage caused by traditional nonselective NSAIDs. 2. To compare the upper GI damage caused by COX-2 selective NSAIDs with that caused by traditional nonselective NSAIDs. 3. To compare the upper GI damage caused by COX-2 selective NSAIDs with that caused by placebo. Method For the first objective, a Cochrane Collaboration metaanalysis was updated. For the other objectives, a literature search was used to identify studies that assessed the GI safety of the newer COX-2 selective NSAIDs. Conclusions Gastroprotective agents Misoprostol, PPIs and double doses of H 2 RAs are effective at reducing the risk of endoscopically identified NSAID-induced ulcers. Standard doses of H 2 RAs are ineffective at reducing the risk of endoscopically identified NSAID-induced ulcers. Misoprostol is the only agent that has been shown to reduce the risk of NSAID-induced clinically important ulcer complications. Its use, however, is associated with significant adverse effects, particularly at higher doses. COX-2 selective NSAIDs COX-2 selective NSAIDs are associated with a lower risk of endoscopically identified ulcers and of clinically important ulcer complications when compared with traditional non-selective NSAIDs in general. COX-2 selective NSAIDs were found to be safer than naproxen and ibuprofen (high dose), but no significant difference was found between the COX-2 selective NSAIDs reviewed and diclofenac. Preliminary results indicate that the reduced GI complication rate due to celecoxib may be lost when it is administered with acetylsalicylic acid (ASA). This has not been tested for rofecoxib. Meloxicam does not seem to be safer than traditional non-selective NSAIDs. It is unclear whether the co-administration of a COX- 2 selective NSAID and a gastroprotective agent significantly improves safety over the use of a COX- 2 selective NSAID alone or the use of a traditional non-selective NSAID with gastroprotection. This summary is based on a comprehensive health technology assessment report available from CCOHTA s web site ( Rostom A, Dubé C, Jolicoeur E, Boucher M, Joyce J. Gastroduodenal ulcers associated with the use of non-steroidal anti-inflammatory drugs: a systematic review of preventive pharmacological interventions. Canadian Coordinating Office For Health Technology Assessment (CCOHTA) Carling Avenue, Ottawa, ON, Canada K1S 5S8 Tel: Fax: CCOHTA is an independent, non-profit health research agency funded by the federal, provincial and territorial governments.

7 EXECUTIVE SUMMARY Background Non-steroidal anti-inflammatory drugs (NSAIDs) are commonly used to treat arthritic and inflammatory syndromes and acute and chronic pain. These agents, however, are associated with gastro-intestinal (GI) complications such as perforation or hemorrhage, with an occurrence of about 2% per year in average risk patients and as high as 10% per year in high risk patients. Although the former rate is low, it is important because of the common use of these medications. Several pharmacological strategies have been developed to reduce the risk of NSAID-induced GI complications. One approach is the use of concurrent prophylactic gastroprotective agents such as misoprostol, histamine type-2 receptor antagonists (H 2 RAs) and proton pump inhibitors (PPIs). Alternatively, newer cyclooxygenase-isoform type 2 (COX-2) selective NSAIDs have been developed in an attempt to offer greater GI safety. Objectives Our objectives in this systematic review are to: 1. assess the effectiveness of common pharmacological interventions used for the prevention of the upper GI toxicity that is associated with non-selective NSAIDs 2. compare the upper GI toxicity of COX-2 selective NSAIDs with that of non-selective NSAIDs, with or without the concomitant use of gastroprotective agents 3. assess the upper GI toxicity of the COX-2 selective NSAIDs available in Canada by comparing them with placebo. Methods A literature search was used to identify randomized controlled trials (RCTs) of prophylactic agents used to prevent upper GI toxicity and RCTs that assessed the GI safety of the newer COX-2 selective NSAIDs [celecoxib (Celebrex ), meloxicam (Mobicox ), rofecoxib (Vioxx )]. Study selection and data extraction were performed in duplicate by independent reviewers. Data were analyzed using RevMan 4.1 in accordance with accepted meta-analysis techniques. Results The results of this meta-analysis show that misoprostol, PPIs, and double doses of H 2 RAs are effective at reducing the risk of endoscopically identified NSAID-induced gastric and duodenal ulcers. Standard doses of H 2 RAs, however, are ineffective at reducing the risk of NSAIDinduced gastric ulcers. Misoprostol is the only prophylactic agent that has been evaluated in a trial assessing clinical ulcers as opposed to endoscopic ulcers. It has been shown that misoprostol reduces the risk of NSAID-related ulcer complications, but its use is associated with significant adverse effects, particularly at higher doses. Compared with most non-selective NSAIDs, COX-2 selective NSAIDs are associated with significantly fewer endoscopically identified ulcers and clinically important ulcer complications and with fewer treatment withdrawals due to GI symptoms. An exception is the comparator NSAID diclofenac. When diclofenac is compared with the COX-2 selective NSAIDs studied, no iv

8 statistically significant differences are observed for any of these endpoints. In one study, the use of rofecoxib is associated with more cardiovascular events when compared with naproxen. Another study shows that the use of acetylsalicylic acid (ASA) may diminish the benefit of celecoxib when compared with non-selective NSAIDs. High-quality studies fail to show any significant benefit from meloxicam as compared with the non-selective NSAIDs piroxicam and diclofenac with respect to clinically important ulcer complications. Inclusion of data from six meloxicam efficacy studies, however, reveal a statistically significant relative risk reduction of 52% compared with non-selective NSAIDs, but this inclusion does not show any benefit of meloxicam over diclofenac. Last, the benefit from the growing clinical use of COX-2 selective NSAIDs with added prophylaxis remains untested. Conclusions Gastroprotective agents Misoprostol, PPIs, and double doses of H 2 RAs are effective at reducing the risk of endoscopically identified gastric and duodenal NSAID-induced ulcers. Standard doses of H 2 RAs are ineffective at reducing the risk of endoscopically identified NSAID-induced gastric ulcers. Misoprostol is the only prophylactic agent that has been evaluated in a true clinical outcome trial. It reduces the risk of NSAID-related ulcer complications, but its use is associated with significant adverse effects, particularly at higher doses. PPIs are effective at healing NSAID-induced ulcers even when patients continue the use of NSAIDs. Misoprostol, however, may be more effective than PPIs at reducing the risk of NSAID-induced gastric ulcers. COX-2 selective NSAIDs With regards to GI toxicity, COX-2 selective NSAIDs are safer than naproxen and high-dose ibuprofen and are better tolerated than non-selective NSAIDs in general. No significant difference between the COX-2 selective NSAIDs considered and diclofenac was found, suggesting that COX-2 selective NSAIDs may not be a different class of agents. Instead, they may be part of a continuum of agents that includes the standard NSAIDs but with differing GI toxicities. A superior GI safety profile over one non-selective NSAID does not imply that a given COX-2 selective NSAID is safer than all non-selective NSAIDs. Since COX-2 selective NSAIDs have little anti-platelet effect, the co-administration of ASA is inevitable in patients with an indication for low-dose ASA. This raises preliminary concerns that will need to be confirmed in ongoing studies. These concerns are based on the apparent absence of benefit of celecoxib over non-selective NSAIDs when ASA is coadministered with celecoxib in the CLASS study, and on the absence of data regarding the co-administration of rofecoxib and ASA, since patients requiring ASA are excluded from the VIGOR trial. The benefit of the growing clinical use of COX-2 selective NSAIDs with added prophylaxis remains untested. v

9 TABLE OF CONTENTS EXECUTIVE SUMMARY...iv ABBREVIATIONS...viii 1 INTRODUCTION Background Technology Overview OBJECTIVES METHODS Literature Search Data Retrieval and Quality Assessment Selection of relevant studies Quality assessment Study Inclusion Criteria Types of participants Types of interventions Types of studies Outcome measures Statistical Analysis Heterogeneity Publication Bias Subgroup Analyses Sensitivity Analyses RESULTS Study Selection Prophylactic Gastroprotective Therapy Misoprostol H 2 RAs PPIs Head-to-head comparisons COX-2 Selective NSAIDs COX-2 selective NSAIDs versus non-selective NSAIDS DISCUSSION Prophylactic Gastroprotective Therapy Misoprostol H 2 RAs PPIs COX-2 Selective NSAIDs vi

10 6 CONCLUSIONS Gastroprotective Agents COX-2 Selective NSAIDs REFERENCES Appendix 1: Literature Search Strategy Appendix 2: Data Extraction Form Appendix 3: Outcomes Appendix 4: Jadad Scale vii

11 ABBREVIATIONS AMI ARR ASA CI COX FDA GI H 2 RA NDA NNH NNT NS NSAID OA PPI POB PUB RA RCT RR RRR SAE acute myocardial infarction absolute risk reduction acetylsalicylic acid confidence interval cyclooxygenase Federal Drug Administration gastro-intestinal histamine-2 receptor antagonist new drug application (FDA) number needed to harm number needed to treat not significant non-steroidal anti-inflammatory drug osteoarthritis proton pump inhibitor perforation, obstruction, bleeding perforation, ulcer, bleeding, or symptomatic ulcer rheumatoid arthritis randomized controlled trial relative risk relative risk reduction serious adverse event viii

12 1 INTRODUCTION 1.1 Background Non-steroidal anti-inflammatory drugs (NSAIDs) are agents used in the management of patients with arthritic and inflammatory conditions. 1 Their established efficacy makes NSAIDs among the most frequently used medications. The estimated world market is over $6 billion annually. 2 In Canada, these widely prescribed agents generate several million dollars in sales every year (Table 1). Table 1: Estimated number of prescriptions and associated sales for NSAIDs dispensed in Canadian retail pharmacies Type of NSAIDs 1999 Prescriptions (sales) Non-selective 7,960,831 ($241 million) COX-2 selective 1,906,580 ($101 million) Total 9,867,411 ($342 million) Source: Canadian Compuscript, IMS Health 2000 Prescriptions (sales) 6,145,700 ($179 million) 6,282,132 ($364 million) 12,427,832 ($543 million) 2001 Prescriptions (sales) 5,404,140 ($153 million) 7,724,157 ($467 million) 13,128,297 ($620 million) Table 1 shows that the number of prescriptions for NSAIDs has been increasing over the last three years. This increase is driven by the new selective cyclooxygenase type 2 (COX-2) selective NSAIDs, suggesting prescribers and patients are not only replacing existing nonselective therapy with more selective agents, but also increasing their use of NSAIDs in general. NSAIDs are linked to gastro-intestinal (GI) toxicities. 1,3-13 Their use is commonly associated with symptoms such as nausea and dyspepsia, which correlate poorly with serious adverse GI events. 13,14 Endoscopic ulcers occur in as many as 40% of chronic NSAID users, 3 but as many as 85% of these ulcers may never become clinically important. 13,15 Serious NSAID-induced complications such as hemorrhage, perforation or death occur collectively with an incidence of about 2% per year in average risk NSAID users, and up to 10% per year in high risk patients. 13 Because of the large numbers of individuals involved, however, the use of these drugs is linked to over 70,000 hospitalizations and over 7,000 deaths annually in the United States. 5 The use of prophylactic therapy to prevent NSAID-induced GI complications is considerable, given the large number of patients taking these drugs. Routine use of prophylactic therapy is recommended for patients with risk factors for peptic ulcer disease such as previous peptic ulcers, age over 60 years, concomitant use of corticosteroids, prior discontinuation of NSAIDs for GI side effects, high daily dose of NSAIDs and significant disability. 9 1

13 1.2 Technology Overview It is hypothesized that two mechanisms may be responsible for NSAID-induced gastric damage: an irritant effect and a systemic effect mediated by COX inhibition, leading to reduced prostaglandin synthesis. At least three isoforms of COX have been identified. Two are of clinical significance for this report. 16 COX-1 is usually expressed (constituative) in gastric mucosa and in other organs. Its inhibition may impair the gastric mucosa s ability to protect itself. COX-2, an inducible isoform, may be expressed in the presence and involved in the promotion of inflammation. 17 This co-enzyme may also be involved in neoplastic conditions. 18 Inhibition of COX-2 is believed to be responsible for the anti-inflammatory and analgesic effects of NSAIDs. 17 The unsatisfactory therapeutic profile of classic NSAIDs prompted the development of three strategies in an effort to curtail their adverse effects: 1. the substitution of less toxic agents such as acetaminophen, when possible 2. the use of prophylactic gastroprotective agents, such as misoprostol, histamine 2 -receptors antagonists (H 2 RAs) or proton pump inhibitors (PPIs), with non-selective NSAIDs 3. the use of newly developed, more selective COX-2 NSAIDs. Each gastroprotective agent or class of agents has a different mechanism of action. Misoprostol (Cytotec ) is a synthetic prostaglandin E 1 analogue with antisecretory and cytoprotective properties. H 2 RAs (cimetidine, famotidine, nizatidine, ranitidine) reduce the secretion of gastric acid by competitively and selectively inhibiting the action of histamine on H 2 receptors of the parietal cells. PPIs (lansoprazole, omeprazole, pantoprazole) are specific inhibitors of gastric secretion; they act by irreversibly binding to K + -H + -ATPase (an enzyme that transports acid across the parietal cell). 17 The use of COX-2 selective NSAIDs is an alternative to the combined use of prophylactic gastroprotective therapy and non-selective NSAIDs. 17 When this review was undertaken in 2002, there were three agents marketed as COX-2 selective NSAIDs in Canada: celecoxib (Celebrex TM ), meloxicam (Mobicox TM ) and rofecoxib (Vioxx ). 2

14 2 OBJECTIVES The objectives of this systematic review are to: 1. assess the effectiveness of misoprostol, H 2 RAs and PPIs for the prevention of upper GI toxicity associated with non-selective NSAIDs 2. compare the upper GI toxicity of COX-2 selective NSAIDs with that of non-selective NSAIDs with or without the concomitant use of gastroprotective agents 3. assess the upper GI toxicity of the COX-2 selective NSAIDs available in Canada by comparing them with placebo. 3 METHODS A systematic review of randomized controlled trials (RCTs) was conducted to meet the three objectives. The first objective was addressed by updating the Cochrane Collaboration systematic review published in 2000 by Rostom et al. 19,20 For the other two objectives, the review was expanded to include trials that compare the new COX-2 selective NSAIDs to non-selective NSAIDs. 3.1 Literature Search A Dialog OneSearch on MEDLINE, ToxFile, EMBASE, BIOSIS Previews, Pharmaceutical News Index (PNI) and Current Contents Search for published and scientific meeting literature was performed. The Cochrane Library was searched separately. The search was done in accordance with published recommendations. 21,22 No language restrictions were applied. In addition to the searches that were done for the original Cochrane review, 20 the literature search was updated to cover January 2000 to September 2002 (Appendix 1). We also sought literature to assess the GI safety of the three COX-2 selective NSAIDS available in Canada (Appendix 1). We searched the web sites of the International Network of Agencies for Health Technology Assessment (INAHTA), specialized databases [e.g. University of York National Health Service (NHS) Centre for Reviews and Dissemination (CRD)] and Conference Papers Index. Internet searching (e.g. using Google TM ), was done to identify health technology assessment reports, meeting abstracts and grey literature. Registries were searched for ongoing trials. Content experts were asked to provide additional studies and information regarding ongoing trials. Since the assessment of the GI safety of the three COX-2 selective NSAIDs available in Canada was not an update but an original review, manufacturers of these drug were contacted for relevant information. 3

15 3.2 Data Retrieval and Quality Assessment Selection of relevant studies The reference lists of all potentially relevant articles, including reviews, meta-analyses, consensus statements and conference proceedings, were reviewed to identify potentially relevant trials. The output of the electronic search was screened to ensure that relevant articles were captured (MB, AR). The abstracts of the identified articles were assessed for potential inclusion in the review by two independent reviewers (AR, CD). The full articles were retrieved in duplicate. Two independent reviewers assessed the potentially relevant trials, selected the RCTs and extracted the relevant data (AR, EJ). For each included RCT, data were collected regarding the trial design, patient characteristics, intervention drugs, dosages, NSAIDs used and outcomes, using an extraction form (Appendix 2 and 3). Differences were resolved by referring to the RCTs and establishing consensus. Any remaining differences were resolved with the help of an independent gastroenterology or statistics expert. It was necessary to obtain data for two unpublished RCTs and to supplement the data for several published trials from the FDA web site and by contacting the FDA. Included RCTs were classified as being primary or secondary prophylaxis trials and by period of outcome Quality assessment All RCTs were scored for quality by two independent reviewers (AR, EJ) using a modification of the scale developed by Jadad et al 23 (Appendix 4). For each study, the adequacy of randomization and blinding and the completeness of follow-up were determined. The quality of allocation concealment was also assessed. 24 Differences were resolved by consensus. 3.3 Study Inclusion Criteria Types of participants Included participants met three conditions: they had osteoarthritis (OA), rheumatoid arthritis (RA) or other arthritic conditions; they were older than 18 years; and they required chronic (four weeks or longer) use of an NSAID. If patients younger than age 18 were enrolled, the trial was included if >75% of the participants were over age Types of interventions Objective 1: The pharmaceutical agents considered included misoprostol, H 2 RAs or PPIs used for the prevention of NSAID-induced upper GI toxicity. For this review, standard-dose H 2 RA was defined as the equivalent of ranitidine 150 mg twice daily, while double-dose H 2 RA was defined as the equivalent of ranitidine 300 mg twice daily. Misoprostol daily dosages were classified as low (400 µg), medium (600 µg) or high (800 µg). 4

16 Objectives 2 and 3: The pharmaceutical agents considered included the COX-2 selective NSAIDs available in Canada: celecoxib (Celebrex TM ), meloxicam (Mobicox TM ) and rofecoxib (Vioxx ). For this review, low-dose COX-2 selective NSAIDs was defined as celecoxib 200 mg bid or less, rofecoxib 25 mg daily or less and meloxicam 7.5 mg daily. High-dose COX-2 was defined as celecoxib 400 mg bid, rofecoxib 50 mg daily and meloxicam 15 mg daily. Definition of COX-2 selective NSAIDs: Variability exists in the literature regarding the criteria by which an NSAID is classified as COX-2 selective and the techniques used to make this determination. The most accepted technique involves determination of the COX-2 IC50 to COX-1 IC50 ratio (a ratio of the concentrations of the drug that result in 50% blockage of the COX-2 and COX-1 isoenzymes) through a whole blood assay. A value below 1 indicates greater affinity for COX-2 inhibition than COX-1 inhibition: the lower the value, the greater the COX-2 selectivity. A ratio below 1, however, does not guarantee COX-2 selectivity in clinical practice since other factors are at play. These include COX-2 selectivity at the gastric mucosa and the effect of clinical dosages of the drug on its COX-2 selectivity (i.e. an agent may be COX-2 selective only at subtherapeutic doses). Also, reported COX-2 to COX-1 IC50 ratios for the available COX-2 selective NSAIDs differ Types of studies Only RCTs were included. All trials had to meet the following criteria: the duration of NSAID exposure was four weeks or longer (chronic NSAID exposure) the proportion of patients with endoscopic ulcers or clinical GI events could be determined endoscopic ulcers were defined as being at least 3 mm in diameter or could be distinguished from erosions based on the authors descriptions (it was noted whether endoscopy was performed based on symptoms or as part of a protocol). Objective 1: RCTs of misoprostol, H 2 RAs or PPIs were considered eligible for inclusion if these drugs were assessed for their ability to prevent NSAID-induced upper GI toxicity and the intervention was compared to placebo or to another agent thought to be effective in this clinical setting. Objective 2: RCTs of COX-2 selective NSAIDs (celecoxib, rofecoxib, meloxicam) were considered eligible for inclusion if: (a) for COX-2 selective versus non-selective NSAIDs, the trial assessed the drugs upper GI toxicity, compared to non-selective NSAID therapy without concurrent use of gastroprotective agents b) for COX-2 selective versus non-selective NSAIDs with prophylaxis, the trial assessed the drugs upper GI toxicity compared to non-selective NSAID therapy with concurrent use of gastroprotective agents. Objective 3: RCTs of COX-2 selective NSAIDs were considered eligible for inclusion if the study assessed the drugs upper GI toxicity compared to placebo. 5

17 Primary prophylaxis studies included patients who did not have an ulcer as determined during screening endoscopy. Studies were considered secondary prophylaxis trials if the enrolled patients had NSAID-induced ulcers that were healed in a preceding study. Short-term studies (<4 weeks) on healthy volunteers were not considered, since the clinical importance of acute NSAID-induced erosions and submucosal hemorrhages is thought to be minimal Outcome measures There are several measurable endpoints used to assess upper GI toxicity in NSAID clinical trials. Each endpoint is analyzed separately. The primary outcomes are: endoscopic ulcer: a commonly used surrogate outcome that may lead to more serious clinically important adverse events; defined as a mucosal defect at least 3 mm in diameter or distinguished from erosions based on the authors descriptions clinical GI events: clinically important adverse events such as upper GI hemorrhage, perforation, pyloric obstruction and death GI symptoms: dyspepsia, nausea, abdominal pain or diarrhea other measures of toxicity or lack of efficacy overall treatment withdrawals treatment withdrawals due to specific symptoms symptoms not causing treatment withdrawals. 3.4 Statistical Analysis In prior work, outcome data are expressed as a proportion for all endpoints. Review Manager (RevMan) version 4.1 was used to analyze the dichotomous (binary) data obtained. Endoscopic, clinical and symptom-based outcomes were analyzed separately. The results in this review are expressed as the relative risk (RR) of the outcome obtained with the intervention as compared with outcomes obtained with placebo or comparators, using a fixed effect model. Initial explanations for any heterogeneity are based on study differences or groupings and adjusted appropriately. A random-effects model was used to combine heterogeneous trials only if it was clinically and statistically appropriate. The results were expressed as RRs rather than the standard odds ratios since the concept of risk would be simpler than odds. The absolute risk reduction (ARR) and the number needed to treat (NNT) were calculated for appropriate clinical endpoints to help frame the magnitude of the treatment effects. When clinically and statistically appropriate, differences between groups in stratified analyses were formally tested for significance using techniques described by Deeks and Altman. 31 A qualitative systematic review was conducted for endpoints and trials that could not be appropriately combined. 6

18 For objective 2b, it was decided a priori that if no trials could be found that compared the GI toxicity of COX-2 selective NSAIDs to that of non-selective NSAIDs with prophylaxis, attempts would be made at indirect comparison if the trials were similar in: design patient population (including OA, RA, age, co-morbidities); comparator drugs, i.e. non-selective NSAIDs study endpoints: endoscopic ulcers, clinical GI events (hemorrhage, perforation, pyloric obstruction or death), symptoms (nausea, vomiting, dyspepsia, abdominal pain or diarrhea). If indirect comparisons are made, they are explained in the report. 3.5 Heterogeneity Heterogeneity is tested using a chi-square test with N-1 degrees of freedom, where N equals the number of trials contributing data. A p value of less than 0.10 is considered to be evidence of statistical heterogeneity. 32 L Abbé plots and standard forest plots are used to depict the observed heterogeneity graphically. 33 Galbraith plots show the COX-2 selective NSAID study endpoints of clinical ulcer complications and endoscopic gastric ulcers. These also present the precision of each included trial. 34 Attempts are made to explain any heterogeneity through the exploration of clinical and statistical factors related to the observation. All graphs are produced in Microsoft Excel Publication Bias The existence of publication bias was explored through the use of the inverted funnel plot using data on the primary outcome of endoscopic gastric ulcers and clinical ulcer complications. 3.7 Subgroup Analyses Subgroup analyses are based on dose, duration of therapy and primary versus secondary prophylaxis trials. Analyses by dosages are important for misoprostol where higher doses are associated with more side effects, for H 2 RAs where standard doses may be ineffective at reducing NSAID ulcer risk and for newer COX-2 selective NSAIDs where selectivity for the COX-2 isoenzyme may be diminished at higher doses. The subgroup analyses are also stratified by intervention (COX-2 selective NSAIDs) and by comparator non-selective NSAIDs. 7

19 3.8 Sensitivity Analyses We did sensitivity analyses to test the robustness of the results. The variables considered were: the mean quality score as a cutoff to define high- and low-quality trials the definition of endoscopic ulcer size (>3 mm versus >5 mm) (this analysis was to be assessed at the outset but these data are only available in a subset of the COX-2 selective NSAID trials, so the usefulness is limited; overall, data for ulcers >3 mm are used) type of NSAID used, COX-2 selective NSAID individually and through cross-tabulations. 8

20 4 RESULTS 4.1 Study Selection We identified 898 abstracts in the initial search strategy. Two reviewers (AR, MB) independently assessed these abstracts and deemed 241 to be potentially relevant. For these, the full publication was obtained and independently evaluated for inclusion (AR, CD). In total, 42 RCTs fulfilled the inclusion criteria. The data of included trials were extracted in duplicate using a standard extraction sheet (AR, EJ). AR and EJ also independently rated the quality of each included RCT. Five trials considered prophylaxis against the GI toxicity of non-selective NSAIDS, whereas 37 focused on COX-2 selective NSAIDs (Figure 1). Table 2a describes the five new non-selective NSAID prophylaxis trials. Table 2b describes the 22 ulcer complication trials (including the eight meloxicam efficacy trials). These are further divided into endoscopic ulcer trials (Table 2c: seven trials) and clinical ulcer complications trials (Table 2d: seven trials and eight meloxicam efficacy trials). Side effects from efficacy trials are detailed in another table (Table 2e: 15 trials). A few trials that consider more than one intervention could be used in several analyses. 9

21 Figure 1: Selection of RCTs Initial search strategy 898 abstracts Potentially relevant trials 241 abstracts Fulfilled inclusion criteria 42 RCTs NSAID prophylaxis* 5 new trials COX-2 selective NSAIDs 37 trials PPI 3 trials Endoscopic ulcer 7 trials Ulcer complications** 7+8 meloxicam efficacy trials Misoprostol 2 trials Side effects from efficacy trials 23 (includes 8 meloxicam efficacy trials) * NSAID prophylaxis refers to trials of GI event prophylaxis in non-selective NSAID users. ** There were seven primary clinical ulcer complication trials, but eight meloxicam efficacy trials allowed the extraction of perforation, ulcer, bleeding (PUB) data. 10

22 Trials Graham Chan Table 2a: Characteristics of new GI event prophylaxis RCTs in non-selective NSAID users added since original Cochrane Collaboration meta-analysis Comparisons Intervention Comparator Misoprostol and NSAIDs Misoprostol and naproxen Lansoprazole and placebo Nabumetone Chan ,38 Celecoxib Diclofenac and omeprazole Dose Misoprostol 800 ug/day, lansoprazole 15 and 30 mg/day Misoprostol 400 ug/day, naproxen 500 to 1,000 mg/day, nabumetone 1,000 to 1,500 mg/day Celecoxib 400 mg/day, diclofenac 150 mg/day, omeprazole 20 mg/day Number of Patients Mean age (years) Arthritis Type Primary versus Secondary Prophylaxis * Follow-up Quality Score Various Secondary 12 weeks (median age reported) Various Secondary 24 weeks Various Secondary 6 months 5 Bianchi Porro Pantoprazole Placebo Pantoprazole 40 mg/day OA and Primary 12 weeks and NSAIDs ** RA * * Primary prophylaxis=trial in patients without previous ulcer, secondary prophylaxis=trial in patients who had recent ulcer that was healed, OA=osteoarthritis, RA=rheumatoid arthritis **NSAIDs in this trial = diclofenac, ketoprofen, indomathecin 11 11

23 Table 2b: COX-2 selective NSAID RCTs characteristics Endpoint Endoscopic Ulcer Trial Comparisons Number of Mean Age Arthritis Follow-up Patients (Years) Type Intervention Comparator Celecoxib Celecoxib NSAID Goldstein mg bid Naproxen 500 mg bid OA and RA 4, 8, 12 weeks 5 Emery mg bid Diclofenac 75 mg bid RA 24 weeks 5 Simon , RA 2, 6, 12 weeks 5 FDA study mg bid 200 mg bid 400 mg bid 50 mg bid 100 mg bid 200 mg bid Naproxen 500 mg bid Placebo Naproxen 500 mg bid Placebo FDA study mg bid Diclofenac 75 mg bid Ibuprofen 800 mg tid Rofecoxib Rofecoxib Laine mg/day 50 mg/day Quality Score 1,108 N/A OA 2, 6, 12 weeks 2 1,097 N/A OA and RA 4, 8, 12 weeks 2 Ibuprofen 800 mg tid OA 6 weeks, 3, 6 months 5 Hawkey mg/day 50 mg/day Ibuprofen 800 mg tid OA 6 weeks, 3, 6 months 5 12 Clinical Ulcer Complications Celecoxib Celecoxib NSAID Silverstein (CLASS) 400 mg bid Diclofenac 75 mg bid Ibuprofen 800 mg tid Goldstein (combined analysis study) * SUCCESS Abstract # 25 mg bid to 400 mg bid 7 dose steps 100 mg bid Naproxen 500 mg bid Diclofenac 75 mg bid Ibuprofen 800 mg tid 8, OA and RA 4, 13, 26 weeks (1 year FDA) 5 11, OA and RA 2 to 24 weeks 3 Naproxen 500 mg bid 13, OA 6, 12 weeks mg bid Rofecoxib Rofecoxib Bombardier mg/day Naproxen 500 mg bid 8, RA 6 weeks, (VIGOR) 54 4, 8,12 months Langman (combined analysis study) 25 mg/day 50 mg/day Ibuprofen 800 mg tid Diclofenac 50 mg tid Nabumetone 1,500 mg 5, OA 6 weeks, 4, 6, 12, 24 months

24 Clinical Ulcer Complications Table 2b (continued): COX-2 selective NSAID RCTs characteristics Meloxicam Meloxicam Dequeker 1998 (SELECT) mg/day Piroxicam 8, OA 4 weeks 4 Hawkey 1998 (MELISSA) mg/day Diclofenac 9, OA 4 4 weeks 100 mg/day Yocum & 3.75 to 15 mg Diclofenac OA 3 12 weeks 100 mg/day Goei & 15 mg Diclofenac OA 3 6 weeks 100 mg/day Hosie & 15 mg Piroxicam 20 mg OA 6 months 3 Hosie & 7.5 mg Diclofenac OA 3 6 months 100 mg/day Lindén & 15 mg Piroxicam 20 mg OA 6 weeks 3 Wojtulewski & 7.5 mg Naproxen 750 mg 379 N/A RA 6 months 3 Distel , to 15 mg Piroxicam 20 mg/day 6, OA and RA Up to 1 year 3 (combined analysis study) Naproxen 1,000 mg/day Diclofenac 100 mg/day Lemmel & 7.5 to 15 mg Placebo RA 3 weeks 3 Lund & 7.5 to 30 mg Placebo OA 3 weeks 3 * Combined analysis means inclusion of all initial pre-class FDA-New drug application (NDA) studies. # SUCCESS-1 trial includes the Canadian study (Haraoui et al) and the world study (which likely includes the Canadian study). & These meloxicam studies are efficacy trials. Because their GI data are poor quality, the score assigned was 3 even though they may be rated higher if considered for efficacy

25 4.2 Prophylactic Gastroprotective Therapy The NSAID prophylaxis trials found for this review are used to update the Cochrane Collaboration meta-analysis by Rostom et al. 19,20 This review is updated with five RCTs, which do not change the results of the original review. The characteristics of these trials are listed in Table 2a Misoprostol Misoprostol is the only prophylactic agent that has been shown to reduce the occurrence of clinically important ulcer complications. 13 a) Endoscopic ulcer trials In total, 22 trials assessed the effect of misoprostol on the prevention of NSAID-induced endoscopically confirmed ulcers. 13,35,36,68-86 Eleven trials with 3,687 patients compared the incidence of endoscopic ulcers with misoprostol to that with placebo after at least three months of NSAID exposure. 35,36,68,69,75-78,81,83,86 The cumulative incidence of endoscopic gastric and duodenal ulcers with placebo were 15% and 6% respectively. Misoprostol significantly reduced the RR of gastric and duodenal ulcers by 74% (RR=0.26; 95% confidence interval (CI): 0.17 to 0.39 random effects) and 53% (RR=0.47; 95% CI: 0.33 to 0.69 random effects). These RRs corresponded to 12% and 3% ARRs for gastric and duodenal ulcers respectively. The heterogeneity in these estimates was due to inclusion of all misoprostol doses in the analyses. Analysis of the trials stratified by dose eliminated the heterogeneity. Subgroup analyses: dose of misoprostol All the studied doses of misoprostol were associated with a significantly reduced risk of endoscopic ulcers. A dose-response relationship was demonstrated for endoscopic gastric ulcers. There were six trials where 2,461 patients used misoprostol 400 µg daily, 36,69,76,78,81,86 one where 928 patients used 600 µg daily 81 and seven where 2,423 patients used 800 µg daily. 35,68,75-77,81,83 Misoprostol 800 µg daily was associated with the lowest risk (RR=0.17; 95% CI: 0.11 to 0.24) of endoscopic gastric ulcers when compared with placebo, whereas misoprostol 400 µg daily was associated with a RR of 0.42 (95% CI: 0.28 to 0.67 random effects model for heterogeneity). The heterogeneity in the 400 µg/day group was the result of the addition of the Chan study. 36 This trial compared naproxen with low-dose misoprostol to nabumetone alone. In this trial, the risk of ulcers was inexplicably greater in the misoprostol group, but this result was based on the differences between the safety of the comparator NSAIDs rather than the prophylactic agent. In a sensitivity analysis, removal of the Chan study eliminated the heterogeneity without significantly altering the results, giving low-dose misoprostol prophylaxis a RR of 0.39 (95% CI: 0.3 to 0.51). This difference between high- and low-dose misoprostol reached statistical significance (p=0.0055). The intermediate misoprostol dose (600 µg daily) was not significantly different from either the low or high dose. Also, the pooled relative risk reduction (RRR) of 78% (ARR=4.7%, RR=0.22; 95% CI: 0.09 to 0.49) for prevention of duodenal ulcers with misoprostol 800 µg daily was not significantly superior to those of the lower daily dosages. 14

26 Subgroup analysis: short-term duration of NSAID exposure Eight trials, with 2,206 patients, assessed the rates of endoscopic ulcers with misoprostol compared to placebo at 1 to 1.5 months 70-72,74,75,79,80,84 Pooled results from these trials revealed an 83% RRR of gastric ulcers with misoprostol (RR=0.17; 95% CI: 0.09 to 0.31) and a 72% RRR of duodenal ulcers (RR=0.28; 95% CI: 0.14 to 0.56). One trial compared misoprostol to a newer cytoprotective agent, dosmafate, for NSAID prophylaxis and found no statistically significant difference in ulcer rates between the two. 73 b) Clinical ulcer complications Only one RCT, the MUCOSA trial, evaluated the efficacy of misoprostol prophylaxis against clinically important NSAID-induced ulcer complications. In this trial of 8,843 patients studied over six months, the incidence of clinical ulcer complications was about 1.5% per year. 13 Misoprostol 800 µg/day was associated with a statistically significant 40% risk reduction (odds ratio=0.598; 95% CI: to 0.982) of combined clinical ulcer complications (p=0.049), representing a risk difference of 0.38% (from 0.95% to 0.57%). In our analysis, the MUCOSA results represented a 51% RRR in clinical ulcer complications (RR=0.49; 95% CI: 0.29 to 0.88). Adverse effects Misoprostol was associated with a small but statistically significant 1.6-fold excess risk of treatment withdrawals due to drug-induced adverse events and a statistically significant excess risk of treatment withdrawals due to nausea (RR=1.30; 95% CI: 1.08 to 1.55), diarrhea (RR=2.40; 95% CI: 2.05 to 2.81) and abdominal pain (RR=1.36; 95% CI: 1.20 to 1.55). In this trial, 732 of 4,404 patients on misoprostol experienced diarrhea or abdominal pain, compared with 399 of 4,439 on placebo for a RR of 1.82 associated with misoprostol (p<0.001). Overall, 27% of patients on misoprosol experienced one or more side effects. 13 When analyzed by dose, only misoprostol 800 µg led to a statistically significant excess risk of treatment withdrawals due to diarrhea (RR=2.45; 95% CI: 2.09 to 2.88) and abdominal pain (RR=1.38; 95% CI: 1.17 to 1.63). The risk of diarrhea with misoprostol 800 µg/day (RR=3.25; 95% CI: 2.60 to 4.06) was significantly higher than that with 400 µg/day (RR=1.81; 95% CI: 1.52 to 2.16, p=0.0012). 19,20 Analyses by quality Both high- and low-quality misoprostol trials demonstrated a statistically significant reduction in endoscopic ulcers H 2 RAs a) Endoscopic ulcer trials Seven trials with 1,188 patients assessed the effect of standard-dose H 2 RAs on the prevention of endoscopic NSAID ulcers at one month Five trials with 1,005 patients assessed these outcomes at three months or longer. 87,91,94-96 Standard-dose H 2 RAs were effective at reducing the risk of duodenal ulcers (RR=0.24; 95% CI: 0.10 to 0.57 and RR=0.36; 95% CI: 0.18 to 0.74 at one and three or more months respectively), but not the risk of gastric ulcers [not significant (NS)]. 19,20 One trial without a placebo comparator was excluded in the pooled estimate. 94 The 15

27 failure to reduce the risk of gastric ulcers is problematic since up to 80% of NSAID ulcers are gastric. Three RCTs with 298 patients assessed the efficacy of double-dose H 2 RA for the prevention of NSAID-induced upper GI toxicity. 87,97,98 When compared with placebo, double-dose H 2 RAs were associated with a statistically significant reduction in the risk of duodenal (RR= 0.26; 95% CI: 0.11 to 0.65) and gastric ulcers (RR=0.44; 95% CI: 0.26 to 0.74). This 56% RRR in gastric ulcers corresponded to a 12% absolute risk difference (range 23.1% to 11.3%). Analysis of the secondary prophylaxis trials alone yielded similar results. b) Clinical ulcer complications We did not find any H 2 RA trials that assessed clinical ulcer complications. Symptoms H 2 RAs, in standard or double doses, were unassociated with an excess risk of total withdrawals, treatment withdrawals due to side effects or treatment withdrawals due to symptoms, when compared to placebo. Symptoms of abdominal pain were, however, significantly reduced with high-dose H 2 RAs when compared with placebo (RR=0.57; 95% CI: 0.33 to 0.98). Analyses by quality In contrast to high-quality trials, low-quality trials failed to demonstrate that standard-dose H 2 RAs prevented endoscopic duodenal ulcers. No significant differences were observed for treatment withdrawals and symptoms when trials of different quality were compared PPIs Eight RCTs with 2,181 patients assessed the effect of PPIs on the prevention of NSAID- induced upper GI toxicity. 35,38,78, Three trials compared omeprazole to placebo. 78,102,103 Of the two trials that compared a PPI to placebo and to misoprostol, one studied lansoprazole 35 and the other studied omeprazole in prophylaxis. 78 Chan compared the combination of omeprazole and diclofenac to celecoxib, 37,38 while Jensen compared omeprazole to misoprostol. 100 Another study compared pantoprazole to placebo, 99 while the last compared omeprazole to ranitidine 150 mg. 101 a) PPIs compared to placebo PPIs significantly reduced the risk of both endoscopic duodenal (RR=0.19; 95% CI: 0.09 to 0.37) and gastric ulcers (RR=0.40; 95% CI: 0.32 to 0.51) compared with placebo. The results were similar in both primary and secondary prophylaxis trials. b) Clinical ulcer complications No PPI trials that assessed clinical ulcer complications were found. 16

28 Symptoms The omeprazole trials used the same composite endpoints to define successful treatment. 78, In these trials, omeprazole significantly reduced dyspeptic symptoms as defined by the authors. Side effects were not different from those observed with placebo Head-to-head comparisons a) Misoprostol versus H 2 RAs Two trials, with 600 patients in total, compared misoprostol to ranitidine 150 mg twice daily. 81,85 Misoprostol seemed to be superior to standard-dose ranitidine for the prevention of NSAIDinduced gastric ulcers (RR= 0.12; 95% CI: 0.03 to 0.51), but not for duodenal ulcers (RR= 1.00; 95% CI: 0.14 to 7.14). b) PPIs versus ranitidine In a study of 425 patients, Yeomans et al. compared omeprazole 20 mg daily to ranitidine 150 mg twice daily in prophylaxis for patients on NSAIDs. 101 In this study, omeprazole was superior to standard-dose ranitidine for the prevention of gastric (RR= 0.32; 95% CI: 0.17 to 0.62) and duodenal ulcers (RR=0.11; 95% CI: 0.01 to 0.89). c) PPIs versus misoprostol Two secondary prophylaxis trials with a total of 838 patients 35,78 compared a PPI to misoprostol. Hawkey et al. 78 compared low-dose misoprostol (400 µg) daily to omeprazole 20 mg daily, while Graham 35 compared high-dose misoprostol (800 µg) to lansoprazole 15 or 30 mg daily. PPIs were found to be significantly superior to misoprostol for the prevention of duodenal (RR=0.29; 95% CI: 0.15 to 0.56), but not gastric ulcers. The Hawkey trial 78 showed a nonsignificant trend towards greater benefit with misoprostol over omeprazole for the prevention of gastric ulcers, while the Graham study 35 showed that misoprostol is superior to lansoprazole for the prevention of gastric ulcers. Our pooled results mirrored these findings, but statistical heterogeneity occurred. The use of a random-effects model failed to demonstrate a statistically significant benefit of misoprostol over PPIs for gastric-ulcer prevention (RR=0.59; 95% CI: 0.27 to 1.25). Jensen et al. 100 presented an abstract of a prospective, randomized parallel-group trial comparing omeprazole (20 mg bid) to misoprostol (200 µg qid) in high-risk patients who required the continued use of NSAIDs or ASA. In this trial, patients had a documented severe GI hemorrhage while on NSAIDs. Treatment failures were defined as ulcer bleeding, severe adverse effects or symptomatic ulcers. Although the sample size of 46 patients was small, the authors found fewer treatment failures with omeprazole than with misoprostol (4.4% versus 30.4%, p=0.02). If only ulcer bleeding or symptomatic ulcers were considered, however, a statistical difference was not seen (RR=0.25; 95% CI: 0.03 to 2.07). d) Symptoms In the two head-to-head comparisons of omeprazole and misoprostol, 35,78 omeprazole was associated with significantly fewer treatment withdrawals overall (RR=0.64; 95% CI: 0.45 to 0.91) and significantly fewer treatment withdrawals due to side effects (RR=0.48; 95% CI: 0.29 to 0.78). There were no significant differences between low-dose H 2 RAs and PPIs in treatment withdrawals due to side effects (RR= 1.90; 95% CI: 0.77 to 4.67) or symptoms of abdominal 17