Professional Documents
Culture Documents
1093/rheumatology/kel428
Advance Access publication 25 January 2007
KEYWORDS: Non-steroidal antiinflammatory drugs, COX-2 inhibitors, Acetaminophen, Acute myocardial infarction, Gastrointestinal bleeding, Elderly
patients, Administrative database, Retrospective cohort.
Introduction COX-2 inhibitors for pain relief and may be inadequate for
many patients [30–32]. It is therefore important to assess the GI
The risks and benefits of COX-2 inhibitors vs non-selective and CV risks and benefits of COX-2 inhibitors vs NS-NSAIDs and vs
non-steroidal anti-inflammatory drugs (NS-NSAIDs) are not acetaminophen, so that the appropriate therapy choices can be made.
clear. COX-2 inhibitors have a lower potential for causing This study evaluated the associations between rofecoxib, celecoxib,
gastrointestinal (GI) bleeding compared with NS-NSAIDs [1, 2], ibuprofen, diclofenac, naproxen and acetaminophen, and the risk
but may have a higher potential for causing acute myocardial of hospitalization for AMI, hospitalization for GI bleeding and
infarction (AMI) compared with naproxen [2–4]. Clinical practice the combined outcome AMI/GI bleeding, whichever occurred first
guidelines recommend the concurrent use of low-dose aspirin among elderly patients, both users and non-users of aspirin.
with COX-2 inhibitors for patients in need of cardiovascular
(CV) protection [5–7]. However, aspirin use with COX-2
inhibitors may offset their GI benefit [1] and, the use of some Methods
NS-NSAIDs with aspirin may have a deleterious effect on
We conducted a population-based retrospective cohort study
aspirin’s anti-platelet activity [8–10].
using health-care records of patients aged 65 years who filled
The CV safety of NS-NSAIDs has not been adequately studied
at least one prescription for rofecoxib, celecoxib, NS-NSAIDs
[6, 11, 12]. While one clinical trial found higher rates of AMI
or acetaminophen between April 1999 and December 2002
in patients randomized to rofecoxib as opposed to those
in Quebec, Canada. Hospitalization records for AMI and for
randomized to naproxen [2], other published clinical trials and
GI bleeding between January 1997 and December 2002 were also
a meta-analysis of published and unpublished trials did not
obtained for these patients from the hospital discharge summary
find any difference in CV risks between rofecoxib or celecoxib and
database. AMI and upper GI bleeding were identified using
either ibuprofen or diclofenac and found an increased risk
the international classification of disease 9th review (ICD-9)
associated with ibuprofen and diclofenac vs placebo [2–4, 13–
diagnosis codes: AMI: 410.x; GI bleeding: 531.x–534.x, 578.0,
15]. Epidemiological studies assessing the CV risk of NS-NSAIDs
578.1 and 578.9.
have had conflicting results [16–27]. A recent meta-analysis of
Permission from the Government of Quebec ethics committee, the
observational studies found an increased CV risk with diclofenac
Commission d’accès à l’information, was obtained to use the data.
but not with any other NS-NSAID [28].
The date of the first filled prescription for rofecoxib, celecoxib,
Acetaminophen use at more than 500 mg per day has been
any NS-NSAIDs or acetaminophen was considered as the
found to increase the risk of hypertension in older women [29]
patient’s index date. Patients with any hospitalization for AMI
and acetaminophen alone is less effective than NS-NSAIDs and
or GI bleeding in the 27 months prior to the index date
1
were excluded. Patients who filled prescriptions for two different
Department of Medicine and 2Research Institute, McGill University Health Centre,
study drugs (e.g. celecoxib and ibuprofen) on the index date were
Montreal, Canada.
also excluded.
Submitted 16 August 2006; revised version accepted 30 November 2006.
Correspondence to: E. Rahme, Division of Clinical Epidemiology, Outcome measure
McGill University Health Centre, Division of Clinical Epidemiology (V), 687
Pine Avenue West, V Building Montreal, Quebec Canada, H3A 1A1. The primary outcome of interest was the first hospitalization
E-mail: elham.rahme@mcgill.ca for AMI or GI bleeding. Only hospitalization for patients
435
ß The Author 2007. Published by Oxford University Press on behalf of the British Society for Rheumatology. All rights reserved. For Permissions, please email: journals.permissions@oxfordjournals.org
436 E. Rahme and H. Nedjar
discharged from acute care hospitals with AMI or GI bleeding were not included in the model to facilitate the analysis.
recorded as the most responsible diagnosis (primary discharge Discontinuous time intervals removes the subjects from the risk
diagnosis) were considered. Hospitalizations with AMI or sets during the time of no-exposure [35]. HRs were adjusted for
GI bleeding recorded as in-hospital complications were not baseline patient characteristics. All statistical analyses were
counted, and AMI hospitalizations where patients were performed using SAS for Unix, version 8.2 (SAS Institute Inc.,
discharged alive and the total length of stay was <3 days Cary, NC, USA).
were also not counted [33].
Subgroup analyses
Drug exposures Two subgroup analyses were conducted. First, the main analysis
Exposure to a study drug was defined as the number of days was repeated including only the subgroup of patients who did
of medication supplied, as recorded in the database, in addition not have any prescription for a COX-2 inhibitor, an NS-NSAID
to a grace period of 25% of this number [34]. A hospitalization or acetaminophen during the year preceding the index date
TABLE 1. Patient characteristics at the index date (%) and exposure and outcome occurrence during follow-up
No aspirin Aspirin
OA, osteoarthritis; RA, rheumatoid arthritis; CHF, congestive heart failure; IHD, ischaemic heart disease; COPD, chronic obstructive pulmonary disease; CVD, cerebrovascular disease; GPA,
gastroprotective agents; AMI, acute myocardial infarction.
Risks and benefits of COX-2 inhibitors vs non-selective NSAIDS 437
TABLE 2. Results of Cox regression model with time-dependent exposure to determine the association between drug exposure and AMI, GI and AMI/GI
hospitalization among all patients and among patients with osteoarthritis
HR (95% CI)
All patients OA patients
Adjusted for age; sex; diagnosis in the prior year of ischaemic heart disease, heart failure, renal failure, chronic obstructive pulmonary disease, rheumatoid arthritis, osteoarthritis,
anaemia or blood disease, alcohol or drug abuse, gastric ulcers; prescriptions in the prior year for antihypertensive agents, lipid-lowering agents, antidiabetic agents, vasodilators,
gastroprotective agents; prescriptions in the prior 90 days for anticoagulants or corticosteroids; prescriptions for gastroprotective agents at the index date; chest pain in the prior 30
days.
were receiving aspirin at the index date. Among the 3 98 730 patients with osteoarthritis exhibited higher HR of AMI/GI
who were not receiving aspirin at the index date, 55 867 received hospitalization in all exposure categories vs acetaminophen.
rofecoxib, 81 932 celecoxib, 102 021 NS-NSAIDs and 158 910 While celecoxib remained the least toxic NSAID among both
acetaminophen. Among the 112 141 patients who were receiving aspirin and non-aspirin users, rofecoxib also seemed advanta-
aspirin at the index date, 14 843 received rofecoxib and aspirin, geous among non-users of aspirin and naproxen lost its advantage
20 421 celecoxib and aspirin, 22 374 NS-NSAIDs and aspirin among aspirin users (Table 2).
and 54 503 acetaminophen and aspirin. Patients receiving
acetaminophen on the index date, whether users or non-users Sensitivity analyses
of aspirin, were older, more likely to be women, more likely
to have a risk factor for AMI and more likely to have a risk Sensitivity analyses results were similar to those of the
factor for GI bleeding compared with patients receiving either main analysis when the grace period was reduced from 25% of
a COX-2 inhibitor or an NS-NSAID (Table 1). the days of medication supplied to 0% or increased to 100%.
The results of the time-dependent Cox regression models Results were also similar to those of the main analysis when
are shown in Table 2. Compared with acetaminophen use outcomes that occurred when two different study drugs had
(with no aspirin), the adjusted HR (95% CI) for AMI been supplied during overlapping days were attributed to the
hospitalization among non-users of aspirin were as follows: medication dispensed first. The HR for the combined outcome
rofecoxib 1.14 (1.00, 1.31), celecoxib 0.97 (0.86, 1.10), ibuprofen of GI bleeding (hospitalization and emergency room), or for
1.04 (0.68, 1.59), diclofenac 1.17 (0.96, 1.43) and naproxen AMI/GI, were not different from those of the main analysis (data
1.16 (0.89, 1.51); and among users of aspirin, they were: rofecoxib not shown).
1.28 (1.08, 1.51), celecoxib 1.17 (1.01, 1.35), ibuprofen 1.39 (0.80,
2.41), diclofenac 1.25 (0.94, 1.67), naproxen 1.03 (0.67, 1.58), and Discussion
acetaminophen 1.18 (1.05, 1.32); while the adjusted HR (95% CI)
of hospitalization for GI bleeding among non-users of Our results show that in general, celecoxib was the least AMI/GI
aspirin were: rofecoxib 1.60 (1.31, 1.95), celecoxib 0.82 (0.66, toxic among both users and non-users of aspirin. In contrast to
1.01), ibuprofen 1.11 (0.56, 2.16), diclofenac 1.18 (0.86, 1.62), the Adenoma Prevention with Celecoxib (APC) study, the risk of
naproxen 2.75 (2.05, 3.69); and among users of aspirin they were: AMI with celecoxib was similar to that of acetaminophen in
rofecoxib 3.22 (2.59, 4.00), celecoxib 1.85 (1.48, 2.31), ibuprofen general but seemed higher among patients with osteoarthritis [3].
1.81 (0.75, 4.40), diclofenac 3.06 (2.16, 4.35), naproxen 2.37 The increase in AMI hospitalizations observed with rofecoxib vs
(1.40, 3.99), and acetaminophen 1.56 (1.31, 1.87). acetaminophen was similar to that observed with ibuprofen and
Among non-users of aspirin, naproxen seemed the least diclofenac but higher than that observed with naproxen corrobor-
advantageous, increasing the risk of AMI/GI hospitalization ating the results of the rofecoxib clinical trials [13, 14].
by 59% compared with acetaminophen, while celecoxib was the Surprisingly, the rates of GI bleeding observed in those
most advantageous carrying a similar AMI/GI risk compared receiving diclofenac and ibuprofen were lower than expected
with acetaminophen. Among users of aspirin, the overall AMI/GI and were not consistent with the results of the rofecoxib and
risk of naproxen was not statistically different from that celecoxib trials [1, 2]. It was difficult to determine from the data if
of acetaminophen. Both naproxen and celecoxib were the least these differences in results were due to chance, given the multiple
toxic in this patient group. modelling conducted in this study, or to differences in patient
characteristics that were not fully adjusted for in the models.
These analyses were conducted using a large, population-based,
Subgroup analyses
well-validated medical database.
Lower HR of hospitalization for AMI or GI bleeding were Nevertheless, this study has limitations. First, differences may
observed in general among all NS-NSAIDs or COX-2 inhibitors exist between patients prescribed acetaminophen and those
vs acetaminophen groups when we considered new users, with prescribed COX-2 inhibitors or NS-NSAIDs in variables that
the exception of the naproxen group who exhibited a higher HR. were not available in the database such as smoking, obesity and
In the comparison of HR between exposure categories, similar alcohol consumption. As with the measured concomitant diseases,
trends to those observed in the main analysis were found (data not the prevalence of these conditions is expected to be higher in the
shown). Compared with patients included in the main analysis, acetaminophen group and in COX-2 inhibitors groups relative to
438 E. Rahme and H. Nedjar
the NS-NSAID group. Therefore, the overall risk of AMI/GI with 11 Fries JF, Murtagh KN, Bennett M, Zatarain E, Lingala B, Bruce B. The rise and
decline of nonsteroidal anti-inflammatory drug-associated gastropathy in rheumatoid
NS-NSAIDs may have been underestimated. A second limitation arthritis. Arthritis Rheum 2004;50:2433–40.
is that the information on compliance with the prescription 12 Silverstein FE, Graham DY, Senior JR et al. Misoprostol reduces serious
regimen is not available from the database and could not be gastrointestinal complications in patients with rheumatoid arthritis receiving
accounted for in the analyses. A third limitation is that the data on nonsteroidal anti-inflammatory drugs. A randomized double-blind placebo-controlled
trial. Ann Intern Med 1995;123:241–9.
over-the-counter medication purchases are not available from the 13 Konstam MA, Weir MR, Reicin A et al. Cardiovascular thrombotic events in
database, so non-prescription use of ibuprofen, acetaminophen or controlled, clinical trials of rofecoxib. Circulation 2001;104:2280–8.
aspirin could not be assessed. However, patients had strong 14 Weir MR, Sperling RS, Reicin A, Gertz BJ. Selective COX-2 inhibition and
financial incentives to obtain these drugs via prescription rather cardiovascular effects: a review of the rofecoxib development program. Am Heart J
2003;146:591–604.
than by over-the-counter purchases, which are not reimbursed by
15 White WB, Faich G, Whelton A et al. Comparison of thromboembolic events in
the provincial drug plan. In 1998, according to Santé Québec, the patients treated with celecoxib, a cyclooxygenase-2 specific inhibitor, versus
government health agency, 17.0%, 2.2% and 46.3% of the elderly ibuprofen or diclofenac. Am J Cardiol 2002;89:425–30.
who consumed NS-NSAIDs, aspirin and acetaminophen, respec- 16 Rahme E, Pilote L, LeLorier J. Association between naproxen use and protection