Relationship Between Selective Cyclooxygenase-2 Inhibitors
and Acute Myocardial Infarction in Older Adults
Daniel H. Solomon, MD, MPH; Sebastian Schneeweiss, MD, ScD; Robert J. Glynn, PhD, ScD; Yuka Kiyota, MD, MPH; Raisa Levin, MSc; Helen Mogun, MSc; Jerry Avorn, MD Background—Although cyclooxygenase-2 inhibitors (coxibs) were developed to cause less gastrointestinal hemorrhage
than nonselective nonsteroidal antiinflammatory drugs (NSAIDs), there has been concern about their cardiovascularsafety. We studied the relative risk of acute myocardial infarction (AMI) among users of celecoxib, rofecoxib, andNSAIDs in Medicare beneficiaries with a comprehensive drug benefit.
Methods and Results—We conducted a matched case-control study of 54 475 patients 65 years of age or older who
received their medications through 2 state-sponsored pharmaceutical benefits programs in the United States. Allhealthcare use encounters were examined to identify hospitalizations for AMI. Each of the 10 895 cases of AMI wasmatched to 4 controls on the basis of age, gender, and the month of index date. We constructed matched logisticregression models including indicators for patient demographics, healthcare use, medication use, and cardiovascular riskfactors to assess the relative risk of AMI in patients who used rofecoxib compared with persons taking no NSAID, takingcelecoxib, or taking NSAIDs. Current use of rofecoxib was associated with an elevated relative risk of AMI comparedwith celecoxib (odds ratio [OR], 1.24; 95% CI, 1.05 to 1.46; P⫽0.011) and with no NSAID (OR, 1.14; 95% CI, 1.00to 1.31; P⫽0.054). The adjusted relative risk of AMI was also elevated in dose-specific comparisons: rofecoxib ⱕ25mg versus celecoxib ⱕ200 mg (OR, 1.21; 95% CI, 1.01 to 1.44; P⫽0.036) and rofecoxib ⬎25 mg versus celecoxib ⬎200 mg (OR, 1.70; 95% CI, 1.07 to 2.71; P⫽0.026). The adjusted relative risks of AMI associated with rofecoxib useof 1 to 30 days (OR, 1.40; 95% CI, 1.12 to 1.75; P⫽0.005) and 31 to 90 days (OR, 1.38; 95% CI, 1.11 to 1.72; P⫽0.003)were higher than ⬎90 days (OR, 0.96; 95% CI, 0.72 to 1.25; P⫽0.8) compared with celecoxib use of similar duration.
Celecoxib was not associated with an increased relative risk of AMI in these comparisons.
Conclusions—In this study, current rofecoxib use was associated with an elevated relative risk of AMI compared with
celecoxib use and no NSAID use. Dosages of rofecoxib ⬎25 mg were associated with a higher risk than dosages ⱕ25
mg. The risk was elevated in the first 90 days of use but not thereafter. (Circulation. 2004;109:2068-2073.)
Key Words: cyclooxygenase inhibitors 䡲 myocardial infarction 䡲 aging
The Vioxx and Gastrointestinal Outcomes (VIGOR) trial celecoxib with the rates in the placebo arms of several trials compared the gastrointestinal safety of rofecoxib 50 of aspirin, the coxibs were associated with an elevated risk.5 mg/d with naproxen 1000 mg/d in patients with rheumatoid Pooled analyses of rofecoxib randomized clinical trials, arthritis who did not take aspirin regularly.1 Although the trial including VIGOR, suggested that there may be a statistically found that patients taking rofecoxib had fewer serious gas- significantly increased risk of cardiovascular events in pa- trointestinal outcomes, there were more acute myocardial tients taking rofecoxib compared with naproxen, but this risk infarctions (AMIs) with rofecoxib than naproxen. This study was not seen when rofecoxib was compared with other could not discern the extent to which the difference in AMI nonsteroidal antiinflammatory drugs (NSAIDs) or with pla- could be explained by a protective effect of naproxen2–4 cebo.6,7 A reanalysis of the Celecoxib Long-term Arthritis and/or an increased risk associated with the selective cyclo- Safety Study (CLASS), which compared celecoxib with oxygenase (COX)-2 inhibitor (coxib).
ibuprofen or diclofenac, found no increase in the risk of AMI Previous studies on the association between coxibs and associated with celecoxib.8 A large observational study sug- AMI have provided conflicting results. In an analysis com- gested that rofecoxib at dosages ⬎25 mg was associated with paring the rates of AMI in phase III trials of rofecoxib and an approximately 2-fold increased risk of AMI compared Received October 7, 2003; revision received January 22, 2004; accepted February 5, 2004.
From the Division of Pharmacoepidemiology and Pharmacoeconomics (D.H.S., S.S., R.J.G., Y.K., R.L., H.M., J.A.) and Division of Rheumatology, Immunology, and Allergy (D.H.S.), Brigham and Women's Hospital, Harvard Medical School, Boston, Mass.
Drs Solomon, Schneeweiss, and Avorn have received salary support from an unrestricted research grant from Pfizer. Other than Dr Cannuscio, an employee of Merck, no authors have direct personal financial relationships with any pharmaceutical company.
Correspondence to Daniel H. Solomon, MD, MPH, Division of Pharmacoepidemiology and Pharmacoeconomics, Brigham and Women's Hospital, Harvard Medical School, 1620 Tremont St, Suite 3030, Boston, MA 02120. E-mail 2004 American Heart Association, Inc.
Circulation is available at
Solomon et al
COX-2 Inhibitors and Acute Myocardial Infarction
with celecoxib or no NSAID, whereas rofecoxib ⱕ25 mg was Coxib and Nonselective NSAID Use
not associated with an elevated risk.9 A smaller observational The study database contained information on all prescription drugs study found no increased risk of AMI with either coxib, but filled by eligible beneficiaries, including drug name, dosage, fre- dosage was not addressed.
quency, and days of supply. The exposures of interest were the use of celecoxib or rofecoxib on the index date. During the study period, In 2002, more than 41 million prescriptions were filled in both drugs were covered by the prescription benefit programs the United States for coxibs,11 making any potential relation- without restriction, and copayments were less than $10. The risk of ship between coxibs and AMI a substantial clinical and public AMI associated with these agents was compared with several health issue. We undertook an observational study examining reference groups: use of the other coxib, no NSAID or coxib,ibuprofen, naproxen, or other NSAIDs. Prescriptions filled on the the association between rofecoxib, celecoxib, NSAIDs, and index date were excluded in the primary analyses. Persons with AMI in a large population of older adults for whom complete prescriptions for more than one of the drug categories on the index information was available on prescription medication use and date were included in both categories.
Two dosage and 3 duration categories were defined a priori for all relevant exposures. Dosage categories for the coxibs and NSAIDs were split at the modal daily dosage. For example, the modal dosageof celecoxib was 200 mg, so current use was categorized as ⱕ200 mg or ⬎200 mg. For rofecoxib, the modal dosage was 25 mg; current All persons studied were Medicare beneficiaries who received use was dichotomized as ⱕ25 mg or ⬎25 mg. Dosage categories prescription medications through the Pennsylvania Pharmaceutical were created for the NSAIDs on the basis of the same methodology.
Assistance Contract for the Elderly or the New Jersey Pharmaceu- For each individual study drug, 3 duration categories were created: 1 tical Assistance Program for the Aged and Disabled during 1998, to 30 days, 31 to 90 days, and ⬎90 days.
1999, and 2000. These 2 programs cover medication expenses forlow-income elderly with annual household incomes between $10 000 and $17 000. To be included, participants had to be enrolled Covariates were defined on the basis of data from the year before the and active users of Medicare and the respective prescription drug study period. Although information for most of these patients and benefit program from 1998 through their index date (defined below), covariates was available for longer than 12 months, we restricted the as demonstrated by presence in the program eligibility files and ascertainment to this period to reduce potential bias that might arise filling at least 1 prescription as well as having at least 1 healthcare because of varying lengths of covariate assessment. The covariates encounter in each 6-month period.
assessed include age, gender, race, previous MI, angina, coronary From this pool of eligible persons (n⫽310 229), we excluded artery revascularization, congestive heart failure, ischemic cerebro- patients who had illnesses that might have obscured any potential vascular accident, diabetes, hypertension, use of a lipid-lowering relationship between coxibs and AMI. These included persons with drug (statin), use of hormone replacement therapy, use of an a serious life-threatening illness, including HIV/AIDS (n⫽114) or anticoagulant (clopidogrel, dipyridamole, ticlopidine, and warfarin), malignancy (n⫽50 973), and persons with a coagulopathy use of an NSAID in 1998, rheumatoid arthritis, osteoarthritis, (n⫽5403). We also excluded persons with a hospitalization during presence of a hospitalization, number of visits for ambulatory care, 1998 who received a diagnosis of AMI that was not the principal number of comorbid medical conditions,15 and number of different discharge diagnosis (n⫽2441).
medications used.
All patient identifiers and all traceable information were deleted Several variables of interest were not available within the study from the case-control study database to protect patients' privacy. The database, including body mass index, tobacco use, aspirin use, and Human Subjects Committee of Brigham and Women's Hospital and socioeconomic status. In theory, these variables could be differen- the Centers for Medicare and Medicaid Services approved this study.
tially related to use of a coxib, use of an NSAID, and AMI.16–18 Wetherefore analyzed data from the Medicare Current Beneficiary Acute Myocardial Infarction
Survey,19 a nationwide in-home survey conducted among 8785 The case-defining event was a hospitalization in 1999 or 2000 with beneficiaries ⱖ65 years old in 1999 with a 97% response rate. We a discharge diagnosis code of AMI (ICD-9-CM 410) in the first or compared patients' body mass index, tobacco use, aspirin use, annual second position. The length of hospitalization must have been at least household income, and educational attainment between patients 3 days and no more than 180 days, unless the patient died. This was reporting use of celecoxib (n⫽562), rofecoxib (n⫽244), and an found to be an accurate algorithm for defining AMI in another study NSAID (n⫽1302). In these analyses, body mass index was compa- population.12 To assess the accuracy of this algorithm in our study rable in both groups of coxib users (celecoxib, 27.5 kg/m2 versus population, we identified a subset of patients with Medicare diag- rofecoxib, 27.2 kg/m2, P⫽0.2) and similar to that of NSAID users nosis codes for AMI and had their primary hospital records re- (27.7 kg/m2, P⫽0.5 versus coxib users). Current tobacco use was viewed. We chose all patients from Pennsylvania taking a coxib or equally common in both groups of coxib users (celecoxib, 8.7% an NSAID who had a Medicare diagnosis code for AMI in 1998 versus rofecoxib, 7.0%, P⫽0.5) and was more common among (n⫽1525), as well as a random subset of those not taking these NSAID users (9.8%, P⫽0.005). Aspirin use was similar in both agents (n⫽675). Trained chart abstractors blinded to the study coxib groups (celecoxib, 8.2% versus rofecoxib, 11.5%, P⫽0.2) and question reviewed the charts using a review form developed as part among NSAID users (10.2%, P⫽0.4). The proportion of persons of the Cardiovascular Coordinating Project.13 On the basis of the with an educational level of college or higher was not statistically primary medical records, we determined whether each admission different between coxibs (celecoxib, 29.6% versus rofecoxib, 31.8%, met criteria for an AMI established by the World Health Organiza- P⫽0.11) or between coxibs and NSAIDs (26.5%, P⫽0.2). The mean tion.14 The Medicare ICD-9-CM diagnosis plus the length-of-stay annual household income of both coxib groups was similar (P⫽0.7) requirements had a positive predictive value of 93% (95% CI, 92% and higher than that for NSAID users (P⫽0.0001).
to 94%). We identified 10 895 hospitalizations for AMI in theeligible study population on the basis of this algorithm.
Four control subjects (controls) who did not sustain an AMI The distribution of covariates was assessed in each exposure cate- during the study period were identified for each case. The date of gory. The unadjusted odds ratio (OR) between each covariate and hospitalization for AMI was the index date for cases. A randomly AMI was then examined separately for New Jersey and Pennsylva- selected date was the index date for controls. Controls were matched nia. The CIs for the crude ORs for each state overlapped for every to cases on the basis of age (⫾1 year), gender, and the month of covariate; data from both states were combined for the multivariable analyses. All covariates were tested in multivariable conditional May 4, 2004
Baseline Characteristics of Study Population by Exposure Category
Age, y, mean⫾SD Nursing home resident in previous year No. of physician visits, mean⫾SD Hospitalized in previous year Comorbid conditions, mean⫾SD No. of different prescription drugs, mean⫾SD History of previous myocardial infarction History of angina History of coronary revascularization History of congestive heart failure History of a cerebrovascular accident Use of hormone replacement therapy Use of any anticoagulant* Rheumatoid arthritis Previous nonselective NSAID use Values are n (%) unless noted. Current use refers to use on the index date. Persons who were current users of multiple agents (n⫽117) are counted in the appropriate column for each drug used.
*Anticoagulants include clopidogrel, dipyridamole, ticlopidine, and warfarin.
logistic regression models conditioning on all matching factors. On were not considered exposed. Another set of sensitivity analyses the basis of a backward selection routine with a threshold of P⬍0.2, redefined the no NSAID use reference group to only persons who anticoagulant use, previous hospitalization, osteoarthritis, and nurs- had never been exposed to an NSAID during the study period.
ing home residence were dropped from all versions of the adjusted Finally, we assessed the relationship between coxibs and AMI in model. The remaining covariates were included in all multivariable subgroups of persons with rheumatoid arthritis, a history of MI, or conditional logistic regression models. The model was rerun for each NSAID use during the baseline period.
reference group (the alternative coxib, no NSAID, ibuprofen, The data and all analyses were under control of the authors. An naproxen, and other NSAIDs). A secondary analysis excluded independent review of the study protocol and statistical program- persons exposed to multiple agents on the index date (n⫽117). The ming was performed by an epidemiologist external to the study results were virtually identical to the main analyses and are not sponsor and project team. All analyses were conducted using SAS statistical software (version 8.2).
We assessed the relationship between dosage categories and AMI for the coxibs and NSAIDs using similar multivariable regression models in which the dosage was classified as less than or equal to the The baseline characteristics of patients are shown in Table 1.
modal dosage or greater than the modal dosage. For example, inanalyses comparing the coxibs, the most commonly used dosages of The study population was primarily elderly women with a rofecoxib (ⱕ25 mg) were compared with the most commonly used mean age ⬎80 years in all drug use groups. More than 85% dosages of celecoxib (ⱕ200 mg). Users of rofecoxib ⬎25 mg were of patients were white. The study population used substantial then compared with users of celecoxib ⬎200 mg.
healthcare resources. Risk factors for AMI, such as diabetes Several sensitivity analyses were undertaken. On the basis of previous findings that first-time users may be at the highest risk for and hypertension, were common, and previous cardiovascular cardiovascular events associated with coxibs,9 we constructed con- disease was frequent. In the baseline period, more than 5% of ditional regression models that considered persons exposed only if the population had sustained a previous MI, more than 13% their use on the index date was their first time using a coxib. We had angina, more than 12% had congestive heart failure, and examined the relationship between AMI and the duration of exposure more than 9% had a previous ischemic stroke. Statins were to coxibs in this group of users. Assessing the effect of durationamong first-time users provides a more precise estimate of the actual used by more than 15% of all patients. As seen in Table 1, period of exposure, because persons with intermittent prescriptions patients using celecoxib or rofecoxib were similar with regard Solomon et al
COX-2 Inhibitors and Acute Myocardial Infarction
Adjusted Association Between Coxibs and Acute
but did not reach statistical significance compared with no current NSAID (OR, 1.14; 95% CI, 1.00 to 1.31), naproxen(OR, 1.17; 95% CI, 0.90 to 1.52), and ibuprofen (OR, 1.21; Adjusted Odds Ratio 95% CI, 0.92 to 1.58). Relatively few patients were currentusers of naproxen (n⫽331) or ibuprofen (n⫽263), contribut- Exposure (reference group) ing to the wide CIs. Celecoxib was not associated with an Rofecoxib (celecoxib) 1.24 (1.05–1.46) elevated risk of AMI in these analyses.
Celecoxib (no current use) 0.93 (0.84–1.02) In all comparisons related to dose, use of rofecoxib ⬎25 Rofecoxib (no current use) 1.14 (1.00–1.31) mg/d was associated with a higher adjusted relative risk of Celecoxib (naproxen) 0.95 (0.74–1.21) AMI than rofecoxib ⱕ25 mg. The adjusted relative risk ofrofecoxib Rofecoxib (naproxen) 1.17 (0.90–1.52) ⬎25 mg (OR, 1.70; 95% CI, 1.07 to 2.71) was higher than that seen for ⱕ25 mg (OR, 1.21; 95% CI, 1.01 to Celecoxib (ibuprofen) 0.98 (0.76–1.26) 1.44) compared with celecoxib ⬎200 mg or ⱕ200 mg. The Rofecoxib (ibuprofen) 1.21 (0.92–1.58) magnitude in elevation of relative risk was similar when Celecoxib (other NSAID) 0.95 (0.82–1.10) rofecoxib was compared with no current NSAID, naproxen, Rofecoxib (other NSAID) 1.17 (0.99–1.38) ibuprofen, and other NSAIDs. Neither celecoxib dosage was associated with an elevated risk of AMI in any comparison.
1.20 (1.12–1.29) Sensitivity analyses that considered only the first-time use of a coxib or NSAID during the study period provided No. of physician visits findings very similar to those of the primary analysis. A 1.11 (1.05–1.18) sensitivity analysis comparing rofecoxib users with patients 1.11 (1.05–1.17) who had no use of either a coxib or NSAID since January 1, 1.09 (1.02–1.16) 1999, produced results nearly identical to those of the primary Comorbid conditions analysis (OR, 1.14; 95% CI, 0.99 to 1.31; P⫽0.062).
1.25 (1.20–1.31) We also examined the relationships between duration of 1.42 (1.33–1.52) coxib exposure and AMI in first-time users. Compared with No. of different drugs celecoxib use of similar duration, rofecoxib use for 1 to 30days was associated with an elevated risk of AMI (OR, 1.43; 1.14 (1.08–1.19) 95% CI, 1.12 to 1.83; P⫽0.005). A similar elevation was 1.18 (1.12–1.25) associated with 31 to 90 days of rofecoxib use (OR, 1.46; 1.48 (1.42–1.54) 95% CI, 1.14 to 1.86; P⫽0.003), but no elevation in AMI risk 1.15 (1.11–1.20) was observed with ⬎90 days of rofecoxib use (OR, 1.04; Previous myocardial infarction 1.56 (1.48–1.66) 95% CI, 0.77 to 1.38; P⫽0.8). The elevated relative risk of 1.31 (1.25–1.38) AMI seen in patients taking rofecoxib for 90 days or less was Previous coronary revascularization 0.78 (0.69–0.89) not restricted to those taking ⬎25 mg. Compared with Congestive heart failure 1.37 (1.31–1.44) patients taking celecoxib ⱕ200 mg for 1 to 90 days, theadjusted relative risk of AMI associated with rofecoxib Cerebrovascular accident 1.07 (1.01–1.27) mg (OR, 1.37; 95% CI, 1.15 to 1.63; P⫽0.0004) was similar 1.00 (0.94–1.04) to the adjusted relative risk for rofecoxib ⬎25 mg (OR, 1.38; Use of hormone replacement therapy 0.88 (0.79–0.98) 95% CI, 0.80 to 2.37; P⫽0.3). No duration category for Rheumatoid arthritis 1.16 (1.02–1.31) celecoxib use was associated with an elevated risk.
Previous nonselective NSAID use 0.97 (0.90–1.04) Subgroup analyses that focused on patients with previous Conditional logistic model matched on age, gender, and month of index date.
AMI (n⫽4698) and compared persons taking rofecoxib with All other variables listed were adjusted for in the multivariable models. The those taking celecoxib found no elevation in relative risk number of AMIs in each exposure group was as follows: celecoxib 425, associated with rofecoxib (OR, 0.91; 95% CI, 0.60 to 1.38; rofecoxib 225, ibuprofen 49, naproxen 63, other NSAID 371, and no current use P⫽0.6). Analyses restricted to patients with rheumatoid arthritis (n⫽1088) also found no elevation in AMI risk witheither coxib. These subgroup analyses were limited by small to baseline characteristics. Compared with NSAID users, numbers of patients.
coxib users were somewhat less healthy during the baselineperiod, with more health service use, hypertension, previous MIs, cerebrovascular accidents, angina, and cardiovascular We studied the relationship between coxibs, NSAIDs, and medication use (statins and anticoagulants).
hospitalization for AMI in a large population of older The results of the multivariable conditional logistic regres- patients. The study database contained information on more sion models are shown in Table 2. After control for all than 50 000 older adults in 2 US states with complete available confounders, rofecoxib was associated with an prescription drug coverage. The main analyses, as well as elevated risk of AMI compared with persons who were taking dose- and duration-specific analyses, found an elevated risk celecoxib (OR, 1.24; 95% CI, 1.05 to 1.46). The adjusted of AMI associated with rofecoxib but not with celecoxib. The relative risk of AMI associated with rofecoxib was elevated risk was higher in persons taking ⬎25 mg of rofecoxib and May 4, 2004
during the first 90 days of use and was observed consistently index, aspirin use, tobacco use, income status, and educa- in relation to several reference groups.
tional attainment. A comparison of people taking coxibs with It is important that these findings be considered in light of those taking NSAIDs suggests that unmeasured confounding previous research. In the VIGOR trial, which compared 50 by each of these factors may result in a small degree of bias mg of rofecoxib with 1000 mg of naproxen in patients with toward the null. In addition to the potential for bias by rheumatoid arthritis, the risk of AMI was elevated in patients unmeasured confounders, these results may be influenced by treated with rofecoxib.1 Patients were not allowed to take residual confounding by factors that were incompletely as- aspirin during the trial. In an analysis that compared data sessed in this administrative database, such as severity of from phase III randomized clinical trials of celecoxib and cardiovascular risk factors. However, the relationship be- rofecoxib with data from the placebo groups of 4 aspirin tween available covariates and AMI is consistent with results primary prevention trials, the annualized MI rates for patients from previous observational studies. Third, it is possible that randomized to either celecoxib or rofecoxib were higher than some patients prescribed coxibs and/or NSAIDs used them on rates for the meta-analysis of the placebo groups.5 This an as-needed basis. Thus, patients may not have been exposed analysis has been criticized because the coxib trials included to the drug on all days of the calculated prescription period, osteoarthritis and rheumatoid arthritis patients, and the latter leading to potential misclassification of exposure status. If the group has been observed to have an elevated baseline risk of pattern of misclassification was similar across drugs, the bias AMI.20 The control population was characterized by rela- would be toward the null value. Alternatively, if it varied by tively low rates of AMI. A reanalysis of data from the CLASS drug or dose, as a function of the indication for the medica- trial, in which patients were allowed to take aspirin, found noelevation in risk of AMI associated with celecoxib.
tion (such as acute versus chronic pain) or the efficacy of the observational study conducted in the Tennessee Medicaid treatment, the magnitude and direction of bias could be population found that rofecoxib at dosages ⬎25 mg/d was toward or away from the null value. We have no compelling associated with a nearly 2-fold increased risk of AMI com- reason to believe that this misclassification of exposure pared with nonuse of any NSAID.9 Our findings differ from would have differed by drug. Finally, one must consider the the pooled analyses of rofecoxib randomized controlled trials, generalizability of findings on the basis of data from an older, which showed no significant increase in cardiovascular low-income population in 2 states, whose prescription drug events compared with non-naproxen NSAIDs.6,7 In addition, use was slightly higher than the national average. Because the a recently published observational analysis from Ontario also elderly are among the most frequent users of coxibs, the study found no increased risk of AMI associated with any dosage of population examined is relevant.
rofecoxib.10 This analysis excluded persons who were pre- Several biological pathways could underlie a potential scribed a coxib for ⬍30 days. The findings of our study association between selective COX-2 inhibition and coronary suggest that the first 30 days of use may include a period of events. Although NSAIDs inhibit both COX isoforms, selec- elevated risk. Finally, rofecoxib dosages ⬎25 mg, which tive inhibition of COX-2 results in decreased prostacyclin, a were associated with the highest relative risk of AMI in this vasodilator and moderator of platelet activation, without study and the study by Ray and colleagues,9 were not reported reducing COX-1– dependent thromboxanes, contributors to separately in the Ontario study.
platelet aggregation and vasoconstriction.21,22 Emerging data There are important potential limitations to the present support a varied role for COX-2 in the vascular bed, with study. One is the concern about possible misclassification of important functions in vascular resistance,23 late precondi- end points using Medicare use data. We studied the accuracy tioning,24 endothelial function,25,26 and atherogenesis.27,28 of the AMI diagnosis codes and found that they had a positive Data from rat models of hypertension suggest that celecoxib predictive value of 93% compared with primary hospital may be associated with improvements in endothelial function records. However, patients who suffered an AMI and were and reductions in oxidative stress29; neutral findings have not hospitalized because of sudden death or a silent event been reported for rofecoxib and diclofenac.30 Although both would not be counted in these analyses for any exposure rofecoxib and celecoxib, like most NSAIDs, have been group. In addition, it is possible that some cases sustained associated with hypertension, several large head-to-head their AMI during the hospitalization. If so, these patients may randomized controlled trials have reported higher rates not have been exposed to the medications of interest for a among patients treated with rofecoxib31; other smaller studies period of time before their event. This may have influencedthe results if patients taking one particular medication before in healthy adults suggest similarity between coxibs.
admission were more likely to suffer an AMI during the In conclusion, we observed an elevated risk of hospitaliza- course of a hospitalization. However, we have no reason to tion for AMI among elderly Medicare enrollees treated with believe that this was the case. Second, similar to all retro- rofecoxib. This risk was higher in persons taking ⬎25 mg of spective observational studies, these results may have been rofecoxib than in patients taking the most common dosages biased because of confounding by factors not observable in used of ⱕ25 mg. The risk was elevated during the first 90 Medicare use data. We examined this possibility using data days of exposure but not thereafter. We did not find an from the in-home Medicare Current Beneficiary Survey and elevated risk of AMI for persons taking celecoxib. Because of found that people taking rofecoxib or celecoxib were similar the important potential public health implications, our find- with respect to 5 variables known to be independent risk ings should be followed up by additional clinical and mech- factors for cardiovascular end points, including body mass anistic studies, several of which are ongoing.
Solomon et al
COX-2 Inhibitors and Acute Myocardial Infarction
fatality rates in 38 populations from 21 countries in four continents.
This work was supported by Merck & Co through an unrestricted WHO/MONICA Project. Circulation. 1994;90:583– 612.
15. Romano R, Roos LL, Jollis JG. Adapting a clinical comorbidity index for research grant to Brigham and Women's Hospital. The authors had use with ICD-9-CM administrative data: differing perspectives. J Clin sole responsibility for study design, data interpretation, and publica- tion of findings. Dr Solomon was also supported by grants from the 16. Palmieri V, de Simone G, Arnett DK, et al. Relation of various degrees Arthritis Foundation and the National Institutes of Health (AR- of body mass index in patients with systemic hypertension to left ven- 48616 and AR-48264). We want to thank an epidemiologist who tricular mass, cardiac output, and peripheral resistance. Am J Cardiol.
participated actively in the study design, statistical analysis and interpretation of the data, and preparation of the manuscript. We are 17. Weir MR, Maibach EW, Bakris GL, et al. Implications of a health also grateful to Rhonda Bohn, ScD, who performed an independent lifestyle and medication analysis for improving hypertension control.
review of the study protocol and statistical programming.
Arch Intern Med. 2000;160:481– 490.
18. Kington RS, Smith JP. Socioeconomic status and racial and ethnic dif- ferences in functional status associated with chronic diseases. Am J Pub 1. Bombardier C, Laine L, Reicin A, et al. Comparison of upper gastroin- Health. 1997;87:805– 810.
testinal toxicity of rofecoxib and naproxen in patients with rheumatoid 19. Adler GS. A profile of the Medicare Current Beneficiary Survey. Health arthritis. N Engl J Med. 2000;343:1520 –1528.
Care Financing Review. 1994;15:153–163.
2. Solomon DH, Glynn RJ, Levin R, et al. Nonsteroidal anti-inflammatory 20. Solomon DH, Karlson EW, Rimm EB, et al. Cardiovascular morbidity drug use and acute myocardial infarction. Arch Intern Med. 2002;162: and mortality in women diagnosed with rheumatoid arthritis. Circulation.
1099 –1104.
3. Rahme E, Pilote L, LeLorier J. Association between naproxen use and 21. Lipsky PE, Brooks P, Crofford LJ, et al. Unresolved issues in the role of protection against acute myocardial infarction. Arch Intern Med. 2002; cyclooxygenase-2 in normal physiologic processes and disease. Arch Intern Med. 2000;160:913–920.
4. Watson DJ, Rhodes T, Cai B, et al. Lower risk of thromboembolic 22. FitzGerald GA, Patrono C. The coxibs, selective inhibitors of cyclooxy- cardiovascular events with naproxen among patients with rheumatoid genase-2. N Engl J Med. 2001;345:433– 442.
arthritis. Arch Intern Med. 2002;162:1105–1110.
23. Topper JN, Cai J, Falb D, et al. Identification of vascular endothelial 5. Mukherjee D, Nissen SE, Topol EJ. Risk of cardiovascular events asso- genes differentially responsive to fluid mechanical stimuli: cyclooxygen- ciated with selective COX-2 inhibitors. JAMA. 2001;286:954 –959.
ase-2, manganese superoxide dismutase, and endothelial cell nitric oxide 6. Konstam MA, Weir MR, Reicin A, et al. Cardiovascular thrombotic synthase are selectively up-regulated by steady laminar shear stress. Proc events in controlled, clinical trials of rofecoxib. Circulation. 2001;104: Nat Acad Sci U S A. 1996;93:10417–10422.
2280 –2288.
24. Bolli R, Shinmura K, Tang XL, et al. Discovery of a new function of 7. Reicin A, Shapiro D, Sperling RS, et al. Comparison of cardiovascular cyclooxygenase (COX)-2: COX-2 is a cardioprotective protein that alle- thrombotic events in patients with osteoarthritis treated with rofecoxib viates ischemia/reperfusion injury and mediates the late phase of precon- versus nonselective nonsteroidal anti-inflammatory drugs (ibuprofen, ditioning. Cardiovasc Res. 2002;55:506 –519.
diclofenac, and nabumetone). Am J Cardiol. 2002;89:204 –209.
25. Chenevard R, Hurlimann D, Bechir M, et al. Selective COX-2 inhibition 8. White WB, Faich G, Whelton A, et al. Comparison of thromboembolic improves endothelial function in coronary artery disease. Circulation.
events in patients treated with celecoxib, a cyclooxygenase-2 specific inhibitor, versus ibuprofen or diclofenac. Am J Cardiol. 2002;89: 26. Cheng Y, Austin SC, Rocca B, et al. Role of prostacyclin in the cardio- vascular response to thromboxane A2. Science. 2002;296:539 –541.
9. Ray WA, Stein CM, Daugherty JR, et al. COX-2 selective non-steroidal 27. Burleigh ME, Babaev VR, Oates JA, et al. Cyclooxygenase-2 promotes anti-inflammatory drugs and risk of serious coronary heart disease.
early atherosclerotic lesion formation in LDL receptor-deficient mice.
Circulation. 2002;105:1816 –1823.
10. Mamdani M, Rochon P, Juurlink DN, et al. Effect of selective cycloox- 28. Cipollone F, Prontera C, Pini B, et al. Overexpression of functionally ygenase 2 inhibitors and naproxen on short-term risk of acute myocardial coupled cyclooxygenase-2 and prostaglandin E synthase in symptomatic infarction in the elderly. Arch Intern Med. 2003;163:481– 486.
atherosclerotic plaques as a basis of prostaglandin E2-dependent plaque 11. The Online Magazine for Pharmacists. March 17, 2003. Available at: 29. Hermann M, Camici G, Fratton A, et al. Differential effects of selective d/data/2003/0317/past_issues_show. Accessed April 17, 2003.
cyclooxygenase-2 inhibitors on endothelial dysfunction in salt-induced 12. Petersen LA, Wright S, Normand SLT, et al. Positive predictive value of hypertension. Circulation. 2003;108:2308 –2311.
the diagnosis of acute myocardial infarction in an administrative database.
30. Title LM, Giddens K, McInerney MM, et al. Effect of cyclooxygenase-2 J Gen Intern Med. 1999;14:555–558.
inhibition with rofecoxib on endothelial dysfunction and inflammatory 13. Ellerbeck EF, Jencks SR, Randford MJ, et al. Quality of care for markers in patients with coronary artery disease. J Am Coll Cardiol.
Medicare patients with acute myocardial infarction: report on a four state pilot of the Cooperative Cardiovascular Project. JAMA. 1995;273: 31. Whelton A, White WB, Bello AE, et al, for the SUCCESS-VII Investi- 1509 –1514.
gators. Effects of celecoxib and rofecoxib on blood pressure and edema 14. Myocardial infarction and coronary deaths in the World Health Organi- in patients ⬎ or ⫽65 years of age with systemic hypertension and zation MONICA project: registration procedures, event rates, and case- osteoarthritis. Am J Cardiol. 2002;90:959 –963.


importance, sufficient time must elapse before initiating TCA treatment in a 1 Includes reports of "dry lips", "dry throat", and "thirst" patient being withdrawn from fluoxetine, given the long halflife of the parent 2 Includes reports of "pruritus exacerbated" and active metabolite (at least 5 weeks may be necessary). 3 Includes report of "increased irritation at application site"

No job name

How Does Psychotherapy Influence Personality?A Theoretical Integration John D. MayerUniversity of New Hampshire A given type of psychotherapy (e.g., psychodynamic) is associated with aset of specific change techniques (e.g., interpreting defenses, identifyingrelationship themes). Different change techniques can be conceived of asinfluencing different parts of personality (e.g., interpreting defense increasesconscious awareness). An integrated model of personality is presented.Then, change techniques from different theoretical perspectives are assignedby judges to areas of personality the techniques are believed to influence.The results suggest that specific change techniques can be reliably sortedinto the areas of personality. Thinking across theoretical perspectives leadsto important new opportunities for assessment, therapy outcome research,and communication with patients concerning personality change. ©2004Wiley Periodicals, Inc. J Clin Psychol 60: 1291–1315, 2004.