Celecoxib versus omeprazole and diclofenac in patients with osteoarthritis and rheumatoid arthritis (CONDOR): a randomised trial
Summary
Background Cyclo-oxygenase (COX)-2-selective non-steroidal anti-infl ammatory drugs (NSAIDs) and non-selectiveNSAIDs plus a proton-pump inhibitor (PPI) have similar upper gastrointestinal outcomes, but risk of clinicaloutcomes across the entire gastrointestinal tract might be lower with selective drugs than with non-selective drugs.We aimed to compare risk of gastrointestinal events associated with celecoxib versus diclofenac slow releaseplus omeprazole.
Methods We undertook a 6-month, double-blind, randomised trial in patients with osteoarthritis or rheumatoidarthritis at increased gastrointestinal risk at 196 centres in 32 countries or territories.Patients tested negative forHelicobacter pyloriand were aged 60 years and older or 18 years and older with previous gastroduodenal ulceration. Weused a computer-generated randomisation schedule to assign patients in a 1:1 ratio to receive celecoxib 200 mg twicea day or diclofenac slow release 75 mg twice a day plus omeprazole 20 mg once a day. Patients and investigators weremasked to treatment allocation. The primary endpoint was a composite of clinically signifi cant upper or lowergastrointestinal events adjudicated by an independent committee. Analysis was by intention to treat. This trial isregistered with ClinicalTrials.gov, number NCT00141102.
Findings 4484 patients were randomly allocated to treatment (2238 celecoxib; 2246 diclofenac plus omeprazole) andwere included in intention-to-treat analyses. 20 (0•9%) patients receiving celecoxib and 81 (3•8%) receiving diclofenacplus omeprazole met criteria for the primary endpoint (hazard ratio 4•3, 95% CI 2•6–7•0; p<0•0001). 114 (6%)patients taking celecoxib versus 167 (8%) taking diclofenac plus omeprazole withdrew early because of gastrointestinaladverse events (p=0•0006).
Interpretation Risk of clinical outcomes throughout the gastrointestinal tract was lower in patients treated with aCOX-2-selective NSAID than in those receiving a non-selective NSAID plus a PPI. These fi ndings should encouragereview of approaches to reduce risk of NSAID treatment.
FundingPfizer Inc.
Introduction
Present approaches to reduce risk of upper gastro-intestinal adverse events associated with non-steroidalanti-inflammatory drugs (NSAIDs) recommend use of anon-selective NSAID plus a proton-pump inhibitor (PPI)or a cyclo-oxygenase (COX)-2 selective NSAID alone.1–3Although these guidelines, which are based on previouswork by ourselves and others,4–7suggest that bothstrategies reduce risk to the upper gastrointestinal tract,they do not address gastrointestinal adverse eventsoriginating beyond the duodenum. Although NSAID useis associated with injury to the small bowel and colon,4, 8–10leading to overt bleeding, ulceration, occult blood loss, ordevelopment of anaemia, evidence suggests that COX-2selective NSAIDs are associated with fewer mucosallesions of the small bowel than are non-selective NSAIDsplus a PPI.11,12Prespecifi ed prospective trials directlycomparing clinically relevant sequelae of these strategiesthroughout the gastrointestinal tract are not available.
To address the absence of a comprehensive yetclinically practical endpoint, we proposed a compositeendpoint of clinically significant events throughout thegastro intestinal tract (referred to in the protocol asclinically significant upper or lower gastrointestinalevents) as a common platform for comparison of futureclinical trials of gastrointestinal safety.13This endpointwas designed to assess several potential outcomesrelevant to clinical practice, ranging from discontinuationof treatment due to presumed significant occult bloodloss to admission to hospital for life-threateningcomplications. We also included presumed significantoccult gastrointestinal blood loss in the definition of theendpoint because this outcome is relevant in view of thenumber of patients at risk and the potential downstreamclinical implications and economic effect.
Since damage to the small bowel and colon is not acid-dependent,14,15we postulated that the risk of clinicaloutcomes across the entire gastrointestinal tractassociated with celecoxib would be lower than thatassociated with diclofenac plus omeprazole. To addressthis hypothesis, we aimed to compare treatment withcelecoxib and diclofenac plus omeprazole in patients with osteoarthritis and rheumatoid arthritis at increasedgastrointestinal risk.
Methods
Study design and patients
We undertook a double-blind, triple-dummy, parallel-group randomised trial at 196 active (having recruitedmore than one patient) centres in 32 countries orterritories. One site was excluded because it did notcomply with International Conference on HarmonisationGood Clinical Practice guidelines. The protocol wasapproved by local institutional review boards; allparticipants provided written informed consent.
Patients with a clinical diagnosis of osteoarthritis orrheumatoid arthritis were eligible if they were expectedto need regular NSAID treatment for at least 6 months.Enrolment included those aged 60 years or older with orwithout a history of gastroduodenal ulceration orgastrointestinal haemorrhage. Patients aged between18 years and 59 years were enrolled if they had adocumented history of gastroduodenal ulceration orgastrointestinal haemorrhage more than 90 days beforescreening. Patients also had to test negative forHelicobacter pyloriat the screening visit or have confirmederadication of the infection at a rescreening visit.
Patients were excluded if they had concomitant use ofantiplatelet or anticoagulant drugs, ischaemic heartdisease, heart failure, peripheral arterial disease,cerebrovascular disease, gastrointestinal haemorrhageor active gastroduodenal ulceration less than 90 daysbefore screening, infl ammatory bowel disease, gastricsurgery besides a patch repair, erosive oesophagitis,gastric-outlet obstruction, or active malignant disease.Other exclusion criteria were alcohol and substancemisuse, allergy to diclofenac, celecoxib, omeprazole, orsulphonamides, serum alanine transaminase oraspartate transaminase concentrations more than1•5 times and serum creatinine concentration morethan 1•2 times the upper limit of normal (according tothe central laboratory defi nition), and a haemoglobinconcentration lower than 115 g/L.
Randomisation and masking
Patients were randomly assigned in a 1:1 ratio to receiveeither celecoxib 200 mg twice a day (Pfi zer Inc, NewYork, NY, USA) or diclofenac slow release 75 mg twice aday (Novartis Pharmaceuticals UK Ltd, Camberley, UK)plus omeprazole 20 mg once a day (AstraZeneca LP,Westborough, MA, USA) for 6 months. Patients werestratified first by individual investigator sites and thenaccording to history of gastroduodenal ulceration (yes orno). A computer-generated randomisation schedule(block size four; generated by the Pfi zer Global Researchand Development randomisation group) was used, withconcealed allocation of study drugs in consecutivelynumbered, sealed bottles. Patients and investigatorswere masked to treatment allocation.
Procedures
After providing informed consent, patients underwentphysical examination and laboratory testing at screening(visit 1). Eligible patients returned at visit 2 and wererandomly assigned to treatment groups. Patients couldtake antacids and non-NSAID analgesic drugs, includingparacetamol up to 4 g per day and histamine-2-receptorantagonists no more than 3 days per week. Corticosteroids(prednisolone ≤10 mg daily), disease-modifyingantirheumatic drugs, or biological treatments were onlyallowed if patients had been taking a stable dose for12 or more weeks at randomisation. After randomisation,doses could be adjusted if clinically indicated for diseasemanagement; however, patients were not allowed tostart treatment with any of these agents during thestudy. NSAIDs other than study drugs, other antiulcerdrugs, cytotoxic agents, lithium, and iron supplementswere prohibited.
Patients returned to the clinic at months 1, 2, 3, and 6.At every visit, drug compliance, use of concomitantdrugs, safety, Patient’s Global Assessment of Arthritis,16haemoglobin concentrations, and biochemical markerswere assessed. Any patient taking less than 80% or morethan 120% of the allocated dose of study drug wasregarded as non-compliant. Safety was assessed on thebasis of physical examination, laboratory tests, andadverse events and serious adverse events. Patient’sGlobal Assessment of Arthritis16was measured on aLikert scale of 1–5, where 1 was very good and 5 very poor.A fi nal visit occurred at month 6, unless the patientdiscontinued treatment early.
The primary endpoint was a composite of clinicallysignificant events occurring throughout the gastrointestinaltract. Defi nitions of the endpoint criteria were described byChan and colleagues13and are shown in the webappendix.Components of the primary endpoint were gastroduodenal,small-bowel, or large-bowel haemorrhage; gastric-outletobstruction; gastroduodenal, small-bowel, or large-bowelperforation; clinically signifi cant anaemia of definedgastrointestinal or presumed occult gastrointestinal origin(including possible blood loss from the small bowel); andacute gastrointestinal haemorrhage of unknown origin(including presumed small-bowel haemorrhage). Clinicallysignificant anaemia was defi ned in the protocol as adecrease in haemoglobin of 20 g/L or more, or a decreasein haematocrit of at least 10 percentage points.17
Patients with suspected gastrointestinal events wereasked to return for an event visit, at which the protocolrecommended that three faecal occult blood tests andgastroscopy be done. If the investigator could not identifya source of upper gastrointestinal blood loss, colonoscopywas recommended. Patients without an identified sourcewould undergo investigations per protocol to excludenon-gastrointestinal origin and further endoscopic orradiological investigations, as deemed appropriate by thelocal investigator. Members of an independent maskedadjudication committee decided whether the primary endpoint was attained in accordance with the predefinedcriteria (webappendix). If the source of bleeding wasidentifi ed, the event was adjudicated as clinicallysignificant anaemia of defi ned gastrointestinal origin.Without a source, if no clinical or laboratory evidence of anon-gastrointestinal source of anaemia was identified,the event was adjudicated as clinically significant anaemiaof presumed occult gastrointestinal origin, includingpossible small-bowel blood loss.
Key secondary endpoints were Patients’ GlobalAssessment of Arthritis (obtained at every visit),16clinically signifi cant events throughout thegastrointestinal tract plus symptomatic ulcers (definedas ulcer on endoscopy in a patient with dyspepsia),moderate-to-severe abdominal symptoms, and withdrawaldue to gastrointestinal adverse events.
An independent data safety and monitoring committeeoversaw the overall safety of the trial. From October, 2005,serious cardiovascular adverse events were adjudicated byan independent masked cardiovascular event committee.Potential cardiovascular events were classified as primary,according to the Anti-Platelet Trialists Collaborationcriteria (acute myocardial infarction, stroke, andcardiovascular deaths)18or secondary, predefined asunstable angina, coronary revascularisation, transientischaemic attack, venous and peripheral arterial vascularthrombotic events, and congestive heart failure.
Statistical analysis
SAS (version 8.02) was used for all analyses. With anassumed rate of the primary endpoint at month 6 of 1•1%for celecoxib and 2•3% for diclofenac plus omeprazole,19and a drop-out rate of roughly 20%, a sample size of4402 would achieve 80% power to detect the treatmentdiff erence at a 5% level of signifi cance using the χ2 test.
Only events confi rmed by the adjudication committeewere included in the analysis. The primary and secondaryanalyses were by intention to treat, including all patientsrandomly allocated to treatment. We assessed the primaryendpoint using a life-table (actuarial) extension of theMantel-Haenszel method,20stratified by region and historyof gastroduodenal ulceration (yes or no). Patients whodiscontinued treatment during the study or who completedthe study with no events were censored. Predefi ned regionswere western Europe, South America, Asia, and easternEurope. We computed rates of the primary endpoint atmonth 6 in each group and the rate diff erence between thetwo treatment groups using Zhang and Klein’s method.21We analysed changes from baseline in the Patients’ GlobalAssessment of Arthritis16using a generalised linear model,with last measurements carried forward.
A planned interim analysis was undertaken by the datasafety and monitoring committee after 50% of patientscompleted the trial. Using a group sequential approach,we intended to stop the trial early for superiority if therewas an overwhelming body of evidence. On reviewingthe interim data, the data safety and monitoringcommittee did not recommend changes in study conduct;hence, the adjusted nominal level of significance for thefi nal analysis was set to 0•049.This study is registered at ClinicalTrials.gov, numberNCT00141102.
Role of the funding source
Authors employed by Pfi zer (MB and HN) participated inmonitoring of study progress and collection and analysisof data. The sponsor did not participate in adjudication ofpotential gastrointestinal and cardiovascular events. FC,AL, JS, and JG had full access to all data and had finalresponsibility for the decision to submit for publication.
Results
Between Oct 31, 2005, and May 11, 2009, 8098 potentiallyeligible patients were screened; 4484 patients wereenrolled and were included in intention-to-treat analyses(2238 celecoxib, 2246 diclofenac plus omeprazole;figure 1, table 1). Predominant reasons for screeningfailure were H pyloriinfection (2476 patients, 75%) andlow baseline haemoglobin (217, 6%). 876 patientsinitially infected with H pyloriwere successfully treatedand then randomly allocated to celecoxib (444,51%) and diclofenac plus omeprazole (432, 49%).Median duration of study treatment was 176 days (range1–318) in the celecoxib group and 175 days (1–225) in the diclofenac plus omeprazole group. 493 (22%) patients in the celecoxib group discontinued treatment early
(42 discontinuations related to study drug), as did 616 (27%) patients in the diclofenac plus omeprazole group (83 related to study drug). 2193 (98%) patients in the celecoxib group and 2179 (97%) in the diclofenac plus omeprazole group had drug adherence between80% and 120%.
Site investigators reported 253 potential endpoint casesto the gastrointestinal adjudication committee forassessment (71 in the celecoxib group and 182 in thediclofenac plus omeprazole group). The committeeidentifi ed 20 primary endpoints in patients receivingcelecoxib and 81 in patients taking diclofenac plusomeprazole. With a Cox proportional hazard model, theproportion of patients reaching the primary endpointduring the 6-month study period was 0•9% (95% CI0•5–1•3) in the celecoxib group and 3•8% (2•9–4•3) inthe diclofenac plus omeprazole group (diff erence 2•9%,2•0–3•8%; p<0•0001) (table 2, fi gure 2). The hazard ratiowas 4•3 (2•6–7•0) in favour of celecoxib. The maindriving force behind the primary endpoint was ahaemoglobin decrease of 20 g/L or more (table 2). Fewerpatients in the celecoxib group than in the diclofenacplus omeprazole group had a significant decrease inhaemoglobin (15 vs77 patients). Of the 92 patients whohad a decrease of 20 g/L or more with or without definedlesions, 50 had haemoglobin concentrations lower than115 g/L (which was the central laboratory definition ofanaemia for both sexes); five (10%) were in the celecoxibgroup and 45 (90%) were in the diclofenac plusomeprazole group.
63 patients had events that were judged by thecommittee to be clinically significant anaemia ofpresumed occult gastrointestinal origin, includingpossible blood loss from the small bowel. Of thesepatients, 56 (89%) had a gastroscopy, 27 (43%) had acolonoscopy, and seven (11%) had neither; an alternativeexplanation was not identifi ed for any of these patients,thus the possibility remains that the small bowel was thesource of bleeding. 152 potential gastrointestinal events were adjudicated as not reaching the primary endpoint.51 of these events were in the celecoxib group (18 [35%]were anaemia due to non-gastrointestinal causes and33 [65%] did not meet prespecifi ed criteria) and 101 werein the diclofenac plus omeprazole group (26 [26%] wereanaemia due to non-gastrointestinal causes and 75 [74%]did not meet prespecified criteria).
Least-squares mean change from baseline to visit 6 inPatient’s Global Assessment of Arthritis16showed animprovement of 0•75 (0•02) in the celecoxib group and0•77 (0•02) in the diclofenac plus omeprazole group(p=0•41). For the secondary composite endpoint ofclinically signifi cant events throughout thegastrointestinal tract plus symptomatic ulcers, fewerevents were reported for patients who received celecoxib(25 patients, 1%) than for patients who received diclofenacplus omeprazole (92, 5%; p<0•0001). The number ofpatients with moderate-to-severe abdominal symptomsat month 6 was 336 (16%) for the celecoxib group and384 (19%) for the diclofenac plus omeprazole group(p=0•03). 114 (6%) patients in the celecoxib group and167 (8%) in the diclofenac plus omeprazole groupwithdrew early because of gastrointestinal adverse events(p=0•0006).
Table 3 shows a summary of adverse events. Two patientsin each group died. Reported causes of death werepulmonary embolism and bronchopneumonia in thecelecoxib group and two cases of cardiac arrest in thediclofenac plus omeprazole group. 28 potentialcardiovascular events were assessed by the adjudicationcommittee; 21 treatment-related events were confirmed in20 patients. 11 were primary Anti-Platelet TrialistsCollaboration events; six of these were in the celecoxib groupand fi ve were in the diclofenac plus omeprazole group (twomyocardial infarctions and three strokes in each group, plusone pulmonary embolism [the cardiovascular death] in thecelecoxib group). Nine confi rmed events were secondary:one unstable angina, two transient ischaemic attacks, oneperipheral arterial event, and four cases of venous thrombosisin the celecoxib group and one transient ischaemic attack inthe diclofenac plus omeprazole group.
Discussion
In this population of patients with osteoarthritis orrheumatoid arthritis who were not using antiplatelet andanticoagulant drugs, the rate of clinically significantgastrointestinal events was four times higher in thosereceiving diclofenac plus omeprazole than in thosereceiving celecoxib. As in our previous trial, rates ofupper gastrointestinal bleeding did not differ betweentreatment groups.4However, we noted large differencesfor the likelihood of clinically significant blood loss fromthe gastrointestinal tract. For patients with substantialdecreases in haemoglobin and defi ned lesions, our datasupport the contribution of the upper gastrointestinaltract as a potential site of blood loss. Notably, thefrequency of upper gastrointestinal ulcers or erosions associated with haemoglobin decreases was significantlyhigher in the diclofenac plus omeprazole group than inthe celecoxib group. These results confi rm our previousfi ndings in patients with high gastrointestinal risk.22
Reductions in haemoglobin in the absence of a definedlesion were more than fi ve times more likely for patientsreceiving diclofenac plus omeprazole than for thosereceiving celecoxib. We recognise that this componentof the primary endpoint was aff ected by the clinical judgment of the masked adjudication committee. Theiradjudication was on the basis of endoscopic andradiological assessment and consideration of non-gastrointestinal sources of haemoglobin reduction. Adecrease in haemoglobin is an important yet poorlyrecognised fi nding because occult blood loss was not apredefined event in previous trials. Unlike overtbleeding, occult gastrointestinal blood loss does notnecessarily lead to hospital admission, as in our study.However, one should not underestimate the relevance ofa signifi cant decrease in haemoglobin because in clinicalpractice this fi nding often leads to further investigationand premature discontinuation of treatment. Results ofepidemiological studies have consistently shown thateven mildly low or low-normal haemoglobinconcentrations were independently associated withincreased risk of frailty, poor functional outcomes,admission to hospital, and mortality.23–25
In support of the primary analysis, treatment withcelecoxib was associated with a lower rate of moderate-to-severe abdominal symptoms and withdrawal because ofgastrointestinal adverse events than was treatment withdiclofenac plus omeprazole. These small but significantdifferences contradict the results of a recent meta-analysis26suggesting that co-therapy with non-selective NSAIDs plusa PPI was better tolerated than was a COX-2 selectiveNSAID alone. This meta-analysis, however, was hamperedby scarcity of head-to-head data for comparison of COX-2selective NSAIDs versus non-selective NSAIDs plus a PPI.
This trial has important strengths. First, we directlyand rigorously compared two widely advocated strategiesto reduce NSAID-associated gastrointestinal risk. Second,unlike previous trials of gastrointestinal outcomes thatfocused only on upper gastrointestinal events, we used acomprehensive composite endpoint to cover many facetsof gastrointestinal outcomes that are relevant to clinicalpractice, ranging from discontinuation of treatment dueto decreases in haemoglobin to admission to hospitalbecause of complications. Third, our fi ndings will affectclinical practice because they show the importance ofclinical awareness of decreasing haemoglobin associatedwith NSAID use, even if the patient is without apparentclinical disease.
Our study had limitations. First, we excluded patientstaking aspirin. Although this criterion reduced theconfounding eff ect of aspirin in the gastrointestinalmucosa, our results cannot be extrapolated to patients atcardiovascular risk using aspirin, for whom we advocatealternative treatment approaches.1Second, theadjudication of presumed occult gastrointestinal bleedingwas a diagnosis by exclusion rather than by directconfi rmation of the source of blood loss. Thus, we cannotsay with certainty that the bleeding site was thegastrointestinal tract, let alone the small bowel. However,because we used masked adjudication, we believe thatany misclassifi cation of presumed occult gastrointestinalbleeding should have been equally distributed betweenthe two groups. When we excluded this limitation bycomparing only (post hoc) the other components of theprimary endpoint, the overall significant differencebetween treatments was maintained, favouring celecoxib(p=0•0035). Third, our study was not powered nordesigned to assess the diff erence in cardiovascularoutcomes between treatment groups. Furthermore,censoring of data because of potential gastrointestinalevents reduced duration of treatment exposure and thusthe likelihood of development of cardiovascular events,especially in the diclofenac plus omeprazole group. As aconsequence, the cardiovascular data should beinterpreted with caution because they might not showthe true cardiovascular hazard of the two treatments.
COX-2 selective NSAIDs were developed to provideanti-inflammatory therapy; avoiding COX-1 inhibitionand sparing the enzyme in the gut supports mucosalintegrity. Several large outcome studies19,27,28have shownreductions in upper gastrointestinal tract ulceration andcomplications for patients using these drugs comparedwith non-selective NSAIDs. Although PPIs effectivelyreduce upper gastrointestinal ulceration,4,5the CONDORtrial, by directly comparing these strategies forgastrointestinal risk reduction in a large randomisedtrial, has provided a new understanding of thegastrointestinal effect of these two seemingly similartreatment approaches.
Since guidelines recommend that selection of NSAIDtherapy be driven by consideration of both cardiovascularand gastrointestinal eff ects of treatment,29–31CONDOR hasprovided new data relevant to patients requiring anti-infl ammatory therapy who are at increased gastrointestinalbut not increased cardiovascular risk. In this population,the gastrointestinal outcomes of a COX-2 selective NSAIDwere quite diff erent to those of a non-selective NSAID plusa PPI. Further understanding of the cardiovascularoutcomes of these two strategies of gastrointestinal riskreduction requires the results of ongoing trials that havebeen designed directly to address that important clinicalquestion. The fi ndings of the CONDOR trial shouldencourage guideline committees to review their treatmentrecommendations for arthritis patients.
Contributors
FKLC proposed the research idea to the sponsor, and the protocol wasdeveloped by FKLC, AL, JS, and JLG. An independent executivecommittee provided input and oversight throughout the study. Asteering committee undertook governance of the study, and the safety ofthe trial, including the recommendation of the results of the interimanalysis, were assessed by an independent data safety and monitoringcommittee. Two independent masked committees adjudicated potentialgastrointestinal and cardiovascular events (see webappendix). MFBmonitored and oversaw the study progress and coordinated workbetween investigators and various committees. Data analysis was doneby HN, and MFB and HN were responsible for data interpretation. Thereport was prepared by the authors, with editorial support. All authorsread, revised, and approved the final report.
Conflicts of interest
FKLC, JLG, AL, and JS are all consultants to Pfi zer. FKLC is theInternational Associate Editor of the American Journal ofGastroenterology and has also served as a consultant for Eisai, Takeda, and Otsuka. He has received grants from Pfi zer and has been paid lecture fees(including service on speakers’ bureaus) by Pfi zer, AstraZeneca, andTakeda. JLG reports having served as a consultant to AstraZeneca, TAP,Takeda, Novartis, Pozen, Logical Therapeutics, Proctor and Gamble, PLX,Wyeth, Astellas, Amgen, Given, GlaxoSmithKline, and Merck, and hasreceived grant support and honoraria from Pfi zer, AstraZeneca, TAP,Takeda, Novartis, Pozen, Logical Therapeutics, Amgen, and Given. He hasalso been paid lecture fees (including service on speakers’ bureaus) byAstraZeneca, TAP, Takeda, Novartis, Pozen, Logical Therapeutics, andGiven. AL reports having received grants and lecture fees from Pfizer andAstraZeneca, and having been involved in the steering committee ofstudies conducted by Pfi zer (CONDOR) and AstraZeneca (Energib andGades Studies). JS reports serving as a consultant to AstraZeneca, Takeda,Pozen, Bayer, Novartis, and Nicox and has received a grant fromAstraZeneca. He has also been paid lecture fees (including service onspeakers’ bureaus) by AstraZeneca and Takeda. MFB and HN are bothcurrently employees of Pfi zer Inc and own Pfizer stock.
Acknowledgments
The study was sponsored by Pfi zer Inc. We thank all the study siteinvestigators (listed in webappendix), and the members of the data safetymonitoring, executive, cardiovascular adjudication, and steeringcommittees, and George Sands for critical review and comments,Mark Li for statistical advice, and Simon Lowry, Liviu Niculescu,Gail Cawkwell, and Mitchell Gandelman for their invaluable support.Editing and formatting of the report was provided by L Prevost ofPAREXEL, funded by Pfizer Inc.
References
1 Scheiman JM, Fendrick AM. Summing the risk of NSAID therapy.Lancet2007; 369:1580–81.
2 Chan FK, Abraham NS, Scheiman JM, Laine L. Managementof patients on nonsteroidal anti-infl ammatory drugs: a clinicalpractice recommendation from the First International WorkingParty on Gastrointestinal and Cardiovascular Eff ects of NonsteroidalAnti-infl ammatory Drugs and Anti-platelet Agents.Am J Gastroenterol 2008; 103:2908–18.
3 Zhang W, Moskowitz RW, Nuki G, Abramson S, Altman RD,Arden N. OARSI recommendations for the management of hip andknee osteoarthritis, Part II: OARSI evidence-based, expertconsensus guidelines. Osteoarthritis Cartilage 2008; 16:137–62.
4 Chan FK, Hung LC, Suen BY, et al. Celecoxib versus diclofenacand omeprazole in reducing the risk of recurrent ulcer bleedingin patients with arthritis. N Engl J Med2002; 347:2104–10.
5 Lai KC, Chu KM, Hui WM, et al. Celecoxib compared withlansoprazole and naproxen to prevent gastrointestinal ulcercomplications. Am J Med2005; 118:1271–78.
6 Lanas A, Bajador E, Serrano P, et al. Nitrovasodilators, low-doseaspirin, other nonsteroidal antiinfl ammatory drugs, and the risk ofupper gastrointestinal bleeding. N Engl J Med2000; 343:834–39.
7 Ray WA, Chung CP, Stein CM, et al. Risk of peptic ulcerhospitalizations in users of NSAIDs with gastroprotective cotherapyversus coxibs. Gastroenterology 2007; 133:790–98.
8 Allison MC, Howatson AG, Torrance CJ, Lee FD, Russell RI.Gastrointestinal damage associated with the use of nonsteroidalantiinflammatory drugs. N Engl J Med 1992; 327:749–54.
9 Bjarnason I, O’Morain C, Levi AJ, Peters TJ. Absorption of51chromium-labeled ethylenediaminetetraacetate in inflammatorybowel disease. Gastroenterology1983; 85:318–22.
10 Graham DY, Opekun AR, Willingham FF, Qureshi WA. Visiblesmall-intestinal mucosal injury in chronic NSAID users.Clin Gastroenterol Hepatol2005; 3:55–59.
11 Goldstein JL, Eisen GM, Lewis B, Gralnek IM, Zlotnick S, Fort JG.Video capsule endoscopy to prospectively assess small bowel injurywith celecoxib, naproxen plus omeprazole, and placebo.Clin Gastroenterol Hepatol2005; 3:133–41.
12 Goldstein JL, Eisen GM, Lewis B, et al. Small bowel mucosal injuryis reduced in healthy subjects treated with celecoxib compared withibuprofen plus omeprazole, as assessed by video capsule endoscopy.Aliment Pharmacol Ther2007; 25:1211–22.
13 Chan FKL, Cryer B, Goldstein JL, et al. A novel composite endpointto evaluate the gastrointestinal (GI) eff ects of nonsteroidalantiinfl ammatory drugs through the entire GI tract. J Rheumatol2010; 37:167–74.
14 Bjarnason I, Takeuchi K. Intestinal permeability in the pathogenesisof NSAID-induced enteropathy. J Gastroenterol2009; 44:23–29.
15 Lanas A, Panés J, Piqué JM. Clinical implications of COX-1and/or COX-2 inhibition for the distal gastrointestinal tract.Curr Pharm Des2003; 9:2253–66.
16 Arnett FC, Edworthy SM, Bloch DA, et al. The AmericanRheumatism Association 1987 revised criteria for the classificationof rheumatoid arthritis. Arthritis Rheum1988; 31:315–24.
17 FDA Advisory Committee Briefi ng Document. Review of celecoxiblong-term arthritis safety study (CLASS). http://www.fda.gov/ohrms/dockets/ac/01/briefi ng/3677b1_05_gi.pdf (accessed Jan 4,2010).
18 Antiplatelet Trialists’ Collaboration. Collaborative overview ofrandomised trials of antiplatelet therapy-I: prevention of death,myocardial infarction, and stroke by prolonged antiplatelet therapyin various categories of patients. BMJ1994; 308:81–106.
19 Silverstein FE, Faich G, Goldstein JL, et al. Gastrointestinal toxicitywith celecoxib vs nonsteroidal anti-infl ammatory drugs forosteoarthritis and rheumatoid arthritis: the CLASS study:a randomised controlled trial. JAMA2000; 284:1247–55.
20 Koch GG, McCanless I, Ward JF Jr. Interpretation of statisticalmethodology associated with maintenance trials. Am J Med1984;77:43–50.
21 Zhang MJ, Klein J. Confi dence bands for the diff erence of twosurvival curves under proportional hazard models.Lifetime Data Anal 2001; 7: 243–54.
22 Chan FK, Hung LC, Suen BY, et al. Celecoxib versus diclofenac plusomeprazole in high-risk arthritis patients: results of a randomiseddouble-blind trial. Gastroenterology2004; 127:1038–43.
23 Zakai NA, Katz R, Hirsch C, et al. A prospective study of anemiastatus, hemoglobin concentration, and mortality in an elderlycohort: the Cardiovascular Health Study. Arch Intern Med2005;165:2214–20.
24 Chaves PH, Semba RD, Leng SX, et al. Impact of anaemia andcardiovascular disease on frailty status of community-dwelling olderwomen: the Women’s Health and Aging Studies I and II.J Gerontol A Biol Sci Med Sci2005; 60:729–35.
25 Penninx BW, Pahor M, Woodman RC, Guralnik JM. Anemia in oldage is associated with increased mortality and hospitalization.J Gerontol A Biol Sci Med Sci2006; 61:474–79.
26 Spiegel BM, Farid M, Dulai GS, Gralnek IM, Kanwal F. Comparingrates of dyspepsia with coxibs vs NSAID+PPI: a meta-analysis.Am J Med2006; 119:448.e27–36.
27 Singh G, Fort JG, Goldstein JL, et al, for the SUCCESS-IInvestigators. Celecoxib versus naproxen and diclofenac inosteoarthritis patients: SUCCESS-I Study. Am J Med2006;119: 255–66.
28 Bombardier C, Laine L, Reicin A, et al, for the VIGOR Study Group.Comparison of upper gastrointestinal toxicity of rofecoxib andnaproxen in patients with rheumatoid arthritis. N Engl J Med2000;343: 1520–28.
29 Peters MJ, Symmons DP, McCarey D, et al. EULAR evidence-basedrecommendations for cardiovascular risk management in patientswith rheumatoid arthritis and other forms of inflammatoryarthritis. Ann Rheum Dis2010; 69: 325–31.
30 Zhang W, Doherty M, Leeb BF, et al. EULAR evidence basedrecommendations for the management of hand osteoarthritis:report of a Task Force of the EULAR Standing Committee forInternational Clinical Studies Including Therapeutics (ESCISIT).Ann Rheum Dis2007;66:377–88.
31 Zhang W, Nuki G, Moskowitz RW, et al. OARSI recommendationsfor the management of hip and knee osteoarthritis: part III:changes in evidence following systematic cumulative update ofresearch published through January 2009. Osteoarthritis Cartilage2010; 18:476–99.
