Appropriateness of NSAID and Coxib Prescribing for Patients With Osteoarthritis by Primary Care Physicians in Ontario: Results From the CANOAR Study

, , ,
The American Journal of Managed Care, November 2004 - Part 1, Volume 10, Issue 11 Pt 1

Objective: To assess the appropriateness of nonsteroidal antiinflammatorydrug (NSAID) use relative to recent osteoarthritistreatment guidelines from the Second Canadian ConsensusConference.

Study Design: Observational study of self-reported practice in acohort of physicians from the Canadian Osteoarthritis Rx(CANOAR) study.

Subjects and Methods: Ontario primary care physicians wererecruited from the top 10% of NSAID prescribers based on thenumber of NSAID prescriptions filled per year. Physicians wereasked to record office visits on a 1-page data collection form fromNovember 2000 to December 2001.

Results: Of 1400 physicians invited, 185 were enrolled and 119registered office visits. Data were analyzed for the first visits of5459 patients for whom a prescribed NSAID was identified, ofwhom 60% were female and 46% were older than 65 years.Coxibs were prescribed for 56% of study patients and were morecommonly used by those with recent gastrointestinal (GI) events(85%), those receiving warfarin sodium therapy (79%), and thosewith congestive heart failure (68%). Coxib use increased withincreasing global assessment of OA severity, but not patient age.Overall, 58% of prescriptions were considered appropriate givenpatient GI risk factors.

Conclusions: Most coxib and NSAID prescriptions were consistentwith the guidelines, but there was considerable underuse ofcoxibs in at-risk patients and some overuse of coxibs and of gastroprotectiveagents with NSAIDs in patients with no identified GIrisk factors. Increased recognition of relationships between patientage and NSAID-related GI risk would likely promote more appropriateuse of traditional NSAIDs, coxibs, and gastroprotectiveagents in osteoarthritis patients.

(Am J Manag Care. 2004;10:742-750)

Osteoarthritis (OA) is a joint disease characterizedby degenerative changes in the articularcartilage and progressive joint deterioration. Itis the most common type of joint disease and is a majorcause of disability.1 In Western countries, radiographicevidence of OA is present in most individuals by 65years of age and in about 80% of those older than 75years.2 In Canada, an estimated 4 million individuals 15years and older experienced arthritic disorders in 2000,and the direct and indirect costs of these diseases in1998 were estimated to be at least Can$4.4 billion.3Osteoarthritis is expected to account for about half ofthese costs.4

Osteoarthritis typically presents as pain involving 1or several joints. The primary goal of treatment is torelieve pain, thereby improving function and quality oflife. Treatment guidelines in Canada5 and the UnitedStates6 recommend nonsteroidal anti-inflammatorydrugs (NSAIDs) as second-line treatment (after acetaminophen)for mild OA and as first-line treatment formoderate-to-severe OA.

Canadian guidelines further recommend that, inpatients at high risk for upper gastrointestinal (GI)adverse events, the use of traditional NSAIDs should beaccompanied by prophylaxis with misoprostol or a protonpump inhibitor.5,7 Risk factors include a history ofpeptic ulcer disease, GI bleeding, cardiovascular disease,and aged 65 years or older.8,9

Alternatively, the use of coxib NSAIDs (cyclooxygenase-2—specific inhibitors) as first-line treatment forhigh-risk patients is also recommended by Canadianguidelines.5,7 Large clinical trials have demonstratedthat these drugs have efficacy similar to that of traditional(non-coxib) NSAIDs10-12 and produce fewer GIadverse effects.13-16 Although coxibs are not medicallycontraindicated for low-risk patients, economic modelssuggest that they may not be cost effective in thisgroup.17,18

The Canadian Osteoarthritis Rx (CANOAR) studywas designed to characterize OA patients in Ontario forwhom primary care physicians prescribed NSAIDs. Thisarticle examines the use of NSAIDs in this study cohortin relation to recommendations for their appropriateuse as published in the Second Canadian ConsensusConference arthritis treatment guidelines.5


The CANOAR study was an observational study of acohort of primary care physicians with high-volumeNSAID prescribing practices. The objective was todetermine the clinical circumstances under whichNSAIDs were prescribed for OA patients in routine clinicalpractice. The study took place in Ontario, aprovince of 11 million people under a single healthcaresystem.

Before conducting the study, a steering committeewas formed with 2 primary care physicians, 3 rheumatologists,a biostatistician, a health economist, the chiefexecutive officer of The Arthritis Society, and a patientwith OA. The committee met to review and modify aninitial draft protocol and data collection form. A reviewof the modified protocol by 4 rheumatologists, 4 epidemiologists,and 2 health economists further focusedthe study design.

One-page clinical encounter data forms using anonymoussequential patient identifiers were designed andpilot-tested in collaboration with 7 primary care physicians.The form recorded patient demographics, locationof principal OA joint involvement, global physicianand patient assessments, concomitant medical problems,history of clinically significant GI events, currentdrugs, and previous use of and current recommendationsfor OA therapies and gastroprotective agents(GPAs). The form allowed identification of the followingGI risk factors listed in the Second Canadian ConsensusConference guidelines: history of a clinically significantGI event, advanced age (aged 65-74 years was scored as1 risk factor, while age ≥ 75 years was scored as 2 riskfactors), current and continuing use of warfarin sodium,current and continuing use of corticosteroids, hypertension,congestive heart failure, renal insufficiency,and hepatic insufficiency. Renal risk factors recordedwere renal insufficiency, hypertension, congestive heartfailure, hepatic insufficiency, and concomitant antihypertensivedrugs.

The forms were designed to be usable as part of theroutines of normal clinical practice. Seven pilot physicianstested the protocol by enrolling 10 patients andprovided final assessments and recommendations aboutoffice procedures for patient enrollment and consent,data form completion, and data transmission to thedata center by toll-free, never-busy facsimile. The pilotphysicians also provided estimates of required perpatient time, for inclusion in the physician recruitmentletter.

After receiving ethics review board approval, theCANOAR study proceeded with physician recruitment.Of the 12 000 primary care physicians in Ontario, 1400of the top NSAID prescribers, based on the number ofNSAID prescriptions filled in 2000, were identified from aphysician list sourced from the IMS Health Canada aggregatedprescriber-level database. These physicians wereinvited by letter to participate. Each participating physicianwas asked to record up to 130 office visits, includingfollow-up visits, of successive patients for whom theydecided to prescribe an NSAID or a coxib (new or renewal)for OA. Participants were not provided with theSecond Canadian Consensus Conference guidelines aspart of the study protocol and were not told that theirresults would be compared with guidelines. Physicianswere reimbursed Can$20 per completed form.

Although a diagnosis of OA is a determinant of eligibilityfor coxib coverage by the Ontario Drug BenefitsProgram, which provides basic drug coverage to allOntario citizens 65 years and older, the program doesnot specify OA diagnostic criteria.19 The phrase "workingdiagnosis of OA" was therefore used on the dataform to avoid telling participating physicians how todiagnose OA. Patients were included in the study if theyhad a working diagnosis of OA by the treating primarycare physician, received a prescription for an NSAID ora coxib as part of routine care on the first study visit,and gave consent for the use of their anonymous datafor aggregate analyses.


Completed data forms were sent by toll-free facsimileto the data center. Standardized data managementprotocols for paperless and semiautomated processesadapted to the specific needs of the study were used toaggregate and process the data into a relational database.The Goodman-Kruskal &#947; statistic was used as ameasure of ordinal association. All statistical comparisonswere 2-tailed, with < .05 considered statisticallysignificant.



Of the 1400 physicians invited, 185 agreed to participateand were enrolled. It was not possible to examinewhether characteristics of participants differed fromthose of nonparticipants, because the aggregated prescriberlist was not ranked and did not preserve prescribingdata or other individual characteristics. Datacollection continued during 14 months, from November2000 to December 2001, with 119 (64% of thoseenrolled) of the physicians registering 8846 office visitsof 5947 OA patients in the study (the number ofpatients with 2, 3, and >3 visits were 1034, 431, and280, respectively). Data were analyzed for the first visitsof the 5459 patients for whom a prescribed NSAIDwas identified.

Patient Characteristics

Of the 5459 OA patients for whom data were analyzed,60% were female and 46% were older than 65years. All residents of Ontario 65 years and older havelimited public drug coverage through the Ontario DrugBenefits Program. In this study, 57% of patients youngerthan 65 years and 19% of patients 65 years or older hadprivate drug coverage.

A large number of patients had other concomitantconditions. The most common comorbidities werehypertension (34%) and coronaryartery disease (12%). A clinicallysignificant GI event hadbeen experienced by 8% ofpatients. More than a third (34%)of patients were receiving concomitantantihypertensive therapy,and 17% were receivinglow-dose aspirin (the study protocoldid not specify a thresholddose because physicians mayhave differed in their operationaldefinitions of low-dose aspirinfor cardiovascular protection).

At the time of the first studyvisit, 20% of patients werealready receiving a GPA.Acetaminophen had previouslybeen tried by 61% of patients fortheir OA symptoms, and 49%had tried 1 or more NSAIDs(Figure 1). Five percent ofpatients had tried 3 or more NSAIDs. Of the 5459patients, 1145 (21%) were receiving NSAIDs for thefirst time at their first CANOAR study visit.

Nonsteroidal Anti-inflammatoryDrug Prescribing Patterns

Coxibs were prescribed at 56% of visits in this studycohort, evenly divided between rofecoxib and celecoxib(Figure 2). Diclofenac, meloxicam, naproxen, andibuprofen were less commonly recommended. Fortypatients (0.7%) received aspirin prescriptionsat the study visit, andbecause it was unclear whether theywere for OA therapy, these prescriptionsare not included in Figure 2.






Coxib use was higher in certainhigh-risk patient subgroups. Amongthe 26 patients who had experiencedclinically significant GI events in theprevious 60 days, 85% were prescribedcoxibs (Table 1), while 74%of the 214 patients whose GI eventshad occurred within the last 10 yearsreceived coxibs. Most (79%) of the 72patients receiving warfarin therapyand most (68%) of the 163 patientswith congestive heart failure werealso prescribed coxibs. However, thepercentage (57%) of patients withhypertension who were prescribedcoxibs was similar to that (56%)among all patients, and coxib use did not vary withpatient age (Goodman-Kruskal &#947; = 0.07, SE = 0.02, =.60; data not shown). Although it appeared that coxibswere prescribed more commonly as the number of GIand renal risk factors increased (Table 2), these trendswere not statistically significant ( = .93 and = .21,respectively). However, the results suggested a thresholdeffect, with higher coxib use for patients with 4 ormore risk factors than for those with 3 or fewer. A posthoc analysis in which risk factor categories were pooledconfirmed that this difference was significant (GI andrenal risk factors, < .001 for both; Fisher exact test).Coxib use increased significantly with increasing globalpain assessments (physician and patient assessments, < .001 for both; Table 3).

Gastroprotective Agent Prescribing Patterns

Gastroprotective agents were coprescribed with anNSAID at 31% of visits. The most common reason givenfor prescribing a GPA was NSAID GI prophylaxis (39%).Other reasons included dyspepsia (19%), gastroesophagealreflux disease (25%), and a history of ulcer(8%). Misoprostol was the most commonly prescribedGPA, usually in the form of a combination pill ofdiclofenac and misoprostol, while H2 antagonists (eg,ranitidine) and proton pump inhibitors (eg, omeprazole)were also used, although less often (Table 4). Patientsreceiving non-coxib NSAIDs were more likely toreceive GPAs than were coxib patients (45% vs 21%).


Gastroprotective agent coprescriptions were morecommonly given to patients with a higher number ofGI risk factors (Goodman-Kruskal &#947; = &#8722;0.20, SE =0.04; < .001; Table 5). The percentageof patients who werecoprescribed a GPA with theirNSAID was 25% for patients youngerthan 65 years, 34% for those aged 65to 69 years, 39% for those aged 70 to74 years, 39% for those aged 75 to 79years, and 45% for those older than79 years.

Acetaminophen Use



Acetaminophen had been triedpreviously by 61% of study patients.The prevalence of its previous useincreased with increasing patientage and the number of GI and renalrisk factors ( < .001 for all). In addition,previous acetaminophen usewas more common among patientswith higher patient and physicianglobal assessments of OA severity (< .001).

Appropriateness of Prescribing

Most coxib prescriptions were for patients with 1 ormore GI risk factors, but a large percentage (39%) werefor patients with no identified risk factors and could beconsidered to be less cost effective (Table 6). Most(56%) prescriptions for a traditional NSAID alone werefor patients with at least 1 GI risk factor, for whom aGPA coprescription or a coxib would be recommended.Of 1081 prescriptions for traditional NSAIDs plusGPAs, 31% went to patients with no identified risk factors.Physicians did not indicate a reason for prescribinga GPA in 40% of prescriptions for traditionalNSAIDs plus GPAs, so these cases were excluded fromthe analysis of appropriate prescribing. In 98% of theseexcluded cases, the prescription wasfor the NSAID and GPA combinationpill Arthrotec (diclofenac sodiumand misoprostol). Somephysicians who prescribedArthrotec may not have consideredthat they were actively prescribinga GPA, which may explain the highrate of nonresponse to this question.This problem was addressedby also excluding from the analysisof appropriate prescribing all 236Arthrotec prescriptions for which areason for prescribing a GPA wasgiven. Of the remaining 413 prescriptionsfor traditional NSAIDsplus GPAs, 73% went topatients with 1 or moreidentified risk factors andwere counted as appropriate.The other 27% went topatients with no identifiedrisk factors. Of these prescriptions,63% were prescribedfor NSAID GIprophylaxis only and werecounted as inappropriate,while the remainder wereprescribed for other reasons(30%) or for NSAID GI prophylaxisand other reasons(6%) and were counted asappropriate. Therefore, 83%of prescriptions for traditionalNSAIDs plus GPAswere considered appropriate.

For patients with no identified GI risk factors, 33% ofnonexcluded prescriptions were considered appropriate,whereas for those with 1 or more risk factors, 74%were considered appropriate. Overall, 58% of nonexcludedprescriptions were considered to be appropriategiven patient GI risk factors. If the analysis was restrictedfurther to exclude prescriptions for traditionalNSAIDs plus GPAs in which the GPA was prescribed forreasons other than GI prophylaxis alone, 32% of prescriptionsfor patients with no identified GI risk factors,73% of those for patients with 1 or more risk factors, and56% overall were considered appropriate.


Treatment guidelines in the United States,6 Europe,20and Canada5 stress the importance of tailoring OAtreatment to individual patient needs. Patient age andthe presence of comorbidities should be important factorsin treatment selection. The 2000 Second CanadianConsensus Conference guidelines5 that were availablewhen this study was conducted are a relevant benchmarkfor prescribing practice at the time. Although subsequentlypublished guidelines7,21 have refined thecriteria for appropriate coxib, NSAID, and GPA prescription, they identify the same risk factors as did theSecond Canadian Consensus Conference and agree onselection of a coxib or an NSAID plus GPA for at-riskpatients. On the basis of the Second Canadian ConsensusConference guidelines, 42% of prescriptions inthe CANOAR study were inappropriate given patient GIrisk factors. There was underuse of coxibs and ofNSAIDs plus GPAs for patients with 1 or more risk factorsand overuse of these options for patients with noidentified risk factors. The percentage of prescriptionsdeemed appropriate was higher for coxibs than for traditionalNSAIDs alone (61% vs 44%) but lower than that(83%) for traditional NSAIDs plus GPAs. Greater appropriateuse of traditional NSAIDs in combination withGPAs than of coxibs may reflect the fact that coxibswere a new prescribing option when this study wasconducted.

All guidelines recommend acetaminophen as a firstlineagent, primarily because of its GI safety. In accordwith guidelines, most patients in this Ontario cohorthad previously tried acetaminophen, particularly olderpatients, those with multiple risk factors, and thosewith severe OA.

Bleeding, upper GI ulcers, and perforation occur inapproximately 1% of patients treated with non-coxibNSAIDs for 3 to 6 months.5 Among patients treated for1 year, 2% to 4% are affected. These events are costlyand can lead to reduced compliance and decreasedquality of life.22 Recent clinical trial evidence foundthat, compared with acetaminophen alone, NSAIDs incombination with a GPA produced significantly greaterimprovements in pain scores for patients with moderateOA.23 In our study, GPAs were prescribed at 31% ofvisits, and 39% of these prescriptions were for NSAIDGI prophylaxis. In addition, GPA prescriptionsincreased with the number of GI risk factors. Forpatients with the same number of GI risk factors, thosereceiving coxibs had a somewhat lower rate of GPAcoprescription compared with those receiving traditionalNSAIDs, often for reasons other than NSAID GIprophylaxis. In general practice, GPAs are commonlyand appropriately prescribed to treat conditions suchas gastroesophageal reflux disease and dyspepsia.24,25

Comparisons of non-coxib NSAIDs and coxibs suggestsimilar efficacy,10-12 and a recent clinical trial suggeststhat coxibs are similar to or better thanacetaminophen.26 Because they produce fewer GIadverse effects than traditional NSAIDs,13-16 coxibs arenow recommended in the United States and Canadaas first-line treatment for patients at risk for GI perforation,ulcer, and bleeding;5,6 recent updates ofEuropean guidelines are in agreement.27 In this study,85% of patients who had experienced recent (within60 days) clinically significant GI events were prescribeda coxib NSAID, while 74% of those whose GIevent occurred up to 10 years earlier received a coxibNSAID. Also in keeping with Canadian guidelines, acoxib NSAID was prescribed to most (79%) OApatients receiving warfarin therapy, who are at higherrisk for GI bleeding.

Osteoarthritis is the most common chronic conditionamong older persons,28 and non-coxib NSAIDs areprescribed frequently in this group.29 A study30 of noncoxibNSAID prescribing patterns in Canada reportedthat older OA patients, who have 3 times the risk forserious GI adverse effects relative to patients youngerthan 50 years, were frequently given unnecessary prescriptionsfor non-coxib NSAIDs. In addition, GI complicationsstemming from non-coxib NSAID use are aleading cause of hospitalization for older persons.29Although Canadian guidelines recommend the use ofcoxibs in patients with GI risk factors, including olderpatients,5 coxib prescribing in our cohort was not relatedto age. As a result, some older patients who couldhave benefited from a coxib received non-coxib NSAIDprescriptions instead, whereas some younger patientsfor whom traditional NSAIDs may have been appropriate5were prescribed more costly coxibs. One possibleexplanation is that age alone may not have been considereda risk factor by the physicians participating inthis study. Alternatively, NSAID prescribing practicemay have been correlated with barriers against coxibuse, such as access. The finding that older patients hadhigher rates of GPA coprescription with NSAIDs contradictsthe suggestion that physicians failed to recognizeage as a GI risk factor, suggesting instead theexistence of fewer barriers to coxib prescribing amongyounger patients than among seniors. Future analysesof the CANOAR study data will examine whether prescribingvaried by drug coverage.

Although an association has been found betweenNSAID use and impaired renal function,31 NSAIDs arenot contraindicated for patients with renal risk factors,including older patients. Rather, renal function shouldbe monitored in all high-risk patients taking NSAIDs.There does not appear to be a difference between noncoxiband coxib NSAIDs in this regard, and in ourstudy there was no relationship between coxib prescribingand elevated creatinine levels. This practicepattern is in concert with treatment guidelines.

More than a third of patients in this study werehypertensive. Meta-analyses studying the effect of noncoxibNSAIDs on blood pressure found that they elevateblood pressure and antagonize the bloodpressure—lowering effect of antihypertensive medication.32,33 The Second Canadian Consensus Conferenceguidelines noted that coxibs can also adversely affectblood pressure control, but did not mention possibledifferences in the hypertensive or prothrombotic effectsof rofecoxib and celecoxib.5 Data collection for thisstudy was largely complete before publication of ananalysis that suggested differences may exist amongcoxibs in the risk of cardiovascular events.34 It is thereforeunlikely that participating physicians prescribedcoxibs differentially in the presence of cardiovascularrisk factors. More recent meta-analyses have not foundincreased cardiovascular thrombotic event rates relativeto placebo for either coxib.35,36

This study has limitations. Prescribing patterns weremeasured via physician questionnaire responses, withoutexternal verification. Self-reported physician practicemay not accurately reflect actual practice,introducing the potential for biasing self-reports towardhigher quality levels.37 However, the CANOAR studyparticipants were not told in advance that their responseswould be evaluated against guidelines. Furthermore,forms were to be completed during routine clinical practice,minimizing the potential for self-recall bias.

Participation in the study was voluntary, and the 185physicians who agreed to participate—and the 119 whocontributed data—may not represent a random sampleof the 1400 high NSAID prescribers who were invited.Although nonresponding physicians were not followedup to determine reasons for their nonparticipation, it ispossible that some may have thought that study participationwould be onerous, given the need to fill indetailed data forms, obtain patient consent, and performmedical chart review. Some nonparticipants mayalso have found the financial incentive to be inadequate.This could bias results if, for instance, physicianswho are more likely to enroll in such a study and whoactually do participate are also more likely to follow prescribingguidelines.

High NSAID prescribers were targeted to capture thehighest number of prescriptions with the fewest physicians.Although the prescribing choices of these physiciansmay not reflect those of all primary carephysicians, their disproportionately large share ofNSAID prescriptions is expected to be representative ofprescriptions for OA in Ontario. These results are notnecessarily generalizable to other jurisdictions, however,because prescribing may differ between Canada andthe United States or other regions. For example, whereas39% of coxib prescriptions in the CANOAR studywent to patients with no GI risk factors, a recent analysisof claims data from a large preferred provider organizationin the US Midwest found that as many as 73% ofpatients given new coxib prescriptions had no evidenceof GI risk factors.38

To avoid influencing prescribing decisions, thisobservational study did not provide physicians withguidelines for OA diagnosis or therapy. This may haveincreased variability in physician ratings of OA severity,but questions on the data form about other risk factorsrequired yes or no responses that would not be predictedto vary with exposure to guidelines.

Previous use of acetaminophen as reported in thisstudy may be somewhat underestimated, as this commonlyused drug is available by prescription and overthe counter. Its previous use in either form, perhapsmany years earlier, may have been forgotten by physiciansor their patients at the time of completing thestudy data forms.

In conclusion, large, simple observational studiessuch as the CANOAR study can contribute importantpractical assessments of the appropriateness of real-worldprescribing and guide the development of futureinterventions designed to enhance the application ofevidence-based care in actual clinical practice. Mostprescriptions observed in this cohort of OA patientstreated by primary care physicians in Ontario were consistentwith guidelines from the Second CanadianConsensus Conference,5 but some could be consideredinappropriate. Areas in which improved recognition ofrisk factors might lead to further increases in theappropriate use of NSAIDs, coxibs, and GPAs in OApatients include the relationship between patient ageand GI risk, as well as the benefit of GPAs or coxibs forpatients with multiple GI risk factors.

AcknowledgmentsWe thank the 119 Ontario primary care physicians who contributed tothe data collection effort. This article was prepared with the assistance ofBioMedCom Consultants Inc, Montreal, Quebec.

From the Centre for Evaluation of Medicines, St Joseph's Healthcare (RJS, AP, CHG),and Departments of Clinical Epidemiology and Biostatistics (RJS, CHG) and Medicine (RJS,AP), McMaster University, Hamilton, Ontario; and Merck Frosst Canada Ltd, Kirkland,Quebec (MAM).

This study was made possible in part by a grant from Merck Frosst Canada Ltd.Address correspondence to: Rolf J. Sebaldt, MD, Centre for Evaluation of Medicines, StJoseph's Healthcare, 105 Main Street E, Level P1, Hamilton, Ontario, Canada L8N 1G6.

Arthritis Res.

1. Lohmander LS. What can we do about osteoarthritis? 2000;2:95-100.


2. Lawrence RC, Helmick CG, Arnett FC, et al. Estimates of the prevalence ofarthritis and selected musculoskeletal disorders in the United States. 1998;41:778-799.

Arthritis in Canada: An Ongoing Challenge.

3. Health Canada. Health Canada:Ottawa, Ontario; 2003. Catalog No. H39-4/14-2003 E.

J Rheumatol.

4. Badley EM. The effect of osteoarthritis on disability and health care use inCanada. 1995;22(suppl 43):19-22.

Can J Clin Pharmacol.

5. Tannenbaum H, Peloso PM, Russell AS, Marlow B. An evidence-based approachto prescribing NSAIDs in the treatment of osteoarthritis and rheumatoid arthritis: theSecond Canadian Consensus Conference. 2000;7(supplA):4A-16A.

Arthritis Rheum.

6. American College of Rheumatology Subcommittee on OsteoarthritisGuidelines. Recommendations for the medical management of osteoarthritis of thehip and knee: 2000 update. 2000;43:1905-1915.

Can J Gastroenterol.

7. Hunt RH, Barkun AN, Baron D, et al. Recommendations for the appropriate useof anti-inflammatory drugs in the era of the coxibs: defining the role of gastroprotectiveagents. 2002;16:231-240.

Am J Med.

8. Griffin MR. Epidemiology of nonsteroidal anti-inflammatory drug–associatedgastrointestinal injury. 1998;104:23S-29S, 41S-42S.

Cleve ClinJ Med.

9. Peura DA. Gastrointestinal safety and tolerability of nonselective nonsteroidalanti-inflammatory agents and cyclooxygenase-2–selective inhibitors. 2002;69(suppl 1):SI31-SI39.


10. Cannon GW, Caldwell JR, Holt P, et al; Rofecoxib Phase III Protocol 035Study Group. Rofecoxib, a specific inhibitor of cyclooxygenase 2, with clinical efficacycomparable with that of diclofenac sodium: results of a one-year, randomized,clinical trial in patients with osteoarthritis of the knee and hip. 2000;43:978-987.

Arch Fam Med.

11. Saag K, van der Heijde D, Fisher C, et al; Osteoarthritis Studies Group.Rofecoxib, a new cyclooxygenase 2 inhibitor, shows sustained efficacy, comparable with other nonsteroidal anti-inflammatory drugs: a 6-week and a 1-year trial inpatients with osteoarthritis. 2000;9:1124-1134.

Arch Intern Med.

12. Day R, Morrison B, Luza A, et al; Rofecoxib/Ibuprofen Comparator StudyGroup. A randomized trial of the efficacy and tolerability of the COX-2 inhibitorrofecoxib vs ibuprofen in patients with osteoarthritis. 2000;160:1781-1787.

N Engl J Med.

13. Bombardier C, Laine L, Reicin A, et al; VIGOR Study Group. Comparison ofupper gastrointestinal toxicity of rofecoxib and naproxen in patients with rheumatoidarthritis. 2000;343:1520-1528.

Arthritis Rheum.

14. Hawkey C, Laine L, Simon T, et al; Rofecoxib Osteoarthritis EndoscopyMultinational Study Group. Comparison of the effect of rofecoxib (a cyclooxygenase2 inhibitor), ibuprofen, and placebo on the gastroduodenal mucosa ofpatients with osteoarthritis: a randomized, double-blind, placebo-controlled trial. 2000;43:370-377.


15. Silverstein FE, Faich G, Goldstein JL, et al. Gastrointestinal toxicity with celecoxibvs nonsteroidal anti-inflammatory drugs for osteoarthritis and rheumatoidarthritis: the CLASS Study: a randomized controlled trial: Celecoxib Long-termArthritis Safety Study. 2000;284:1247-1255.


16. Langman MJ, Jensen DM, Watson DJ, et al. Adverse upper gastrointestinaleffects of rofecoxib compared with NSAIDs. 1999;282:1929-1933.


17. Zabinski RA, Burke TA, Johnson J, et al. An economic model for determiningthe costs and consequences of using various treatment alternatives for the managementof arthritis in Canada. 2001;19(suppl 1):49-58.

Arthritis Rheum.

18. Maetzel A, Krahn M, Naglie G. The cost effectiveness of rofecoxib and celecoxibin patients with osteoarthritis or rheumatoid arthritis. 2003;49:283-292.

19. Ontario drug benefit formulary/comparative index. Edition 38. OntarioMinistry of Health and Long-Term Care Web site. Available at: AccessedOctober 15, 2004.

Joint Bone Spine.

20. Mazieres B, Bannwarth B, Dougados M, Lequesne M. EULAR recommendationsfor the management of knee osteoarthritis: report of a task force of theStanding Committee for International Clinical Studies Including Therapeutic Trials. 2001;68:231-240.

Aliment Pharmacol Ther.

21. Dubois RW, Melmed GY, Henning JM, Laine L. Guidelines for the appropriateuse of non-steroidal anti-inflammatory drugs, cyclo-oxygenase-2–specificinhibitors and proton pump inhibitors in patients requiring chronic anti-inflammatorytherapy. 2004;19:197-208.

Scand JRheumatol.

22. Yeomans N, Wilson I, Langstrom G, et al. Quality of life in chronic NSAIDusers: a comparison of the effect of omeprazole and misoprostol. 2002;30:328-334.

Arthritis Rheum.

23. Pincus T, Koch GG, Sokka T, et al. A randomized, double-blind, crossoverclinical trial of diclofenac plus misoprostol versus acetaminophen in patients withosteoarthritis of the hip or knee. 2001;44:1587-1598.


24. Talley NJ, Stanghellini V, Heading RC, Koch KL, Malagelada JR, Tytgat GN.Functional gastroduodenal disorders. 1999;45(suppl 2):II37-II42.


25. Dent J, Jones R, Kahrilas P, Talley NJ. Management of gastro-oesophagealreflux disease in general practice. 2001;322:344-347.


26. Geba GP, Weaver AL, Polis AB, Dixon ME, Schnitzer TJ. Efficacy of rofecoxib,celecoxib, and acetaminophen in osteoarthritis of the knee: a randomized trial. 2002;287:64-71.


27. Pencharz JN, Grigoriadis E, Jansz GF, Bombardier C. A critical appraisal ofclinical practice guidelines for the treatment of lower-limb osteoarthritis. 2002;4:36-44.

Ann Intern Med.

28. MacLean CH. Quality indicators for the management of osteoarthritis in vulnerableelders. 2001;135(pt 2):711-721.

Pain Manag Nurs.

29. Buffum M, Buffum JC. Nonsteroidal anti-inflammatory drugs in the elderly. 2000;1:40-50.

Ann Intern Med.

30. Tamblyn R, Berkson L, Dauphinee WD, et al. Unnecessary prescribing ofNSAIDs and the management of NSAID-related gastropathy in medical practice. 1997;127:429-438.

Am J Med.

31. Sturmer T, Erb A, Keller F, Gunther KP, Brenner H. Determinants of impairedrenal function with use of nonsteroidal anti-inflammatory drugs: the importance ofhalf-life and other medications. 2001;111:521-527.

Arch Intern Med.

32. Pope JE, Anderson JJ, Felson DT. A meta-analysis of the effects of nonsteroidalanti-inflammatory drugs on blood pressure. 1993;153:477-484.

Ann Intern Med.

33. Johnson AG, Nguyen TV, Day RO. Do nonsteroidal anti-inflammatory drugsaffect blood pressure? a meta-analysis. 1994;121:289-300.


34. Mukherjee D, Nissen SE, Topol EJ. Risk of cardiovascular events associatedwith selective COX-2 inhibitors. 2001;286:954-959.


35. Konstam MA, Weir MR, Reicin A, et al. Cardiovascular thrombotic events incontrolled, clinical trials of rofecoxib. 2001;104:2280-2288.

Am J Cardiol.

36. White WB, Faich G, Borer JS, Makuch RW. Cardiovascular thrombotic eventsin arthritis trials of the cyclooxygenase-2 inhibitor celecoxib. 2003;92:411-418.

Int J Qual Health Care.

37. Saturno PJ, Palmer RH, Gascon JJ. Physician attitudes, self-estimated performanceand actual compliance with locally peer-defined quality evaluation criteria. 1999;11:487-496.

Am J Manag Care.

38. Cox ER, Motheral B, Frisse M, Behm A, Mager D. Prescribing COX-2s forpatients new to cyclo-oxygenase inhibition therapy. 2003;9:735-742.