Author + information
- Received April 16, 2003
- Revision received August 11, 2003
- Accepted August 19, 2003
- Published online March 17, 2004.
- Stephen E Kimmel, MD, MS, FACC*,†,* (, )
- Jesse A Berlin, ScD*,
- Muredach Reilly, MB*,†,
- Jane Jaskowiak, BSN, RN*,
- Lori Kishel, MS*,
- Jesse Chittams, MS* and
- Brian L Strom, MD, MPH*
- ↵*Reprint requests and correspondence:
Dr. Stephen E. Kimmel, University of Pennsylvania School of Medicine, 717 Blockley Hall, 423 Guardian Drive, Philadelphia, Pennsylvania 19104-6021, USA.
Objectives This study was designed to determine if non-aspirin non-steroidal anti-inflammatory drugs (NANSAIDs) are associated with lower odds of myocardial infarction (MI) and if NANSAIDs, particularly ibuprofen, interfere with aspirin's cardioprotective effect.
Background The NANSAIDs may reduce the risk of MI but may also interfere with aspirin's cardioprotective effect.
Methods A case-control study was conducted, with cases of first, nonfatal MI identified prospectively and controls identified randomly from the community.
Results The use of NANSAIDs was associated with a significant reduction in MI among those not using aspirin (adjusted odds ratio [OR] 0.53; 95% confidence interval [CI]: 0.42 to 0.67). This was true for both ibuprofen (adjusted OR 0.52; 95% CI: 0.39 to 0.69) and naproxen (adjusted OR 0.48; 95% CI: 0.28 to 0.82). Although aspirin itself was associated with decreased odds of MI in those not also using NANSAIDs (adjusted OR relative to no aspirin use 0.79; 95% CI: 0.63 to 0.98), it was not associated with decreased odds of MI among those who were using NANSAIDs (OR 1.28; 95% CI: 0.85 to 1.94; p value for interaction = 0.026). The association of aspirin and reduced odds of MI diminished with increasing frequency of NANSAID use (test for interaction p = 0.006), particularly for ibuprofen (p = 0.018). Among frequent (4 times/week) NANSAID users, the OR for aspirin versus no aspirin was 2.04 (95% CI: 1.06 to 3.94). Users of prophylactic aspirin plus frequent ibuprofen had an OR relative to aspirin-only users of 2.03 (95% CI: 0.60 to 6.84).
Conclusions In the absence of aspirin use, NANSAIDs are associated with reduced odds of MI. In those using aspirin, NANSAIDs do not provide additional protection. Additional study is needed to determine the clinical impact of using NANSAIDs along with aspirin for cardioprotection.
It has been estimated that more than 70 million prescriptions and 30 billion over-the-counter tablets of non-aspirin non-steroidal anti-inflammatory drugs (NANSAIDs) are sold annually in the U.S. alone and that 13 million Americans use NANSAIDs on a regular basis (1). Although they are effective analgesics and anti-inflammatories, they may also have an effect on myocardial infarction (MI) risk.
Nonselective NANSAIDs (those that inhibit both the platelet cyclooxygenase enzyme responsible for thromboxane A2formation and subsequent platelet aggregation) cyclooxygenase (COX)-1, and the COX-2 enzyme (2)may reduce the risk of MI in the absence of aspirin use. Decisions regarding the choice of drugs for analgesia and inflammation currently are influenced by the need to minimize bleeding risk without consideration of possible benefits of MI protection (1). Although several studies have suggested a protective effect of NANSAIDs, particularly naproxen (3–5), on MI, others have not (6,7). However, these studies used electronic prescription records, so they could not measure the fairly common use of over-the-counter NANSAIDs or aspirin (8)and, therefore, could not provide definitive conclusions. Relatedly, a trial comparing the COX-2 inhibitor rofecoxib with the NANSAID naproxen found a higher rate of MI among the COX-2 users (9). This could be due to an increased risk from the COX-2 inhibitor, decreased risk from the NANSAID, or both.
The NANSAIDs, particularly ibuprofen, could also interfere with the clinical benefit of aspirin in reducing MI risk. Pharmacologic data have demonstrated this interaction (10), but the clinical consequences are unknown.
We conducted a case-control study specifically to determine the effects of nonselective NANSAIDs on MI risk and the interaction between NANSAIDs and aspirin. We hypothesized that nonselective NANSAIDs would reduce the risk of first, nonfatal MI in the absence of aspirin use, but that NANSAIDs, particularly ibuprofen, could also interfere with the cardioprotective effect of aspirin.
Study site and participants
The study was conducted among 36 hospitals in a five-county region surrounding Philadelphia. Cases were subjects between the ages of 40 and 75 years with a first, nonfatal MI, hospitalized from May 1998 through April 2001 and identified prospectively. Myocardial infarction was validated using criteria from the Minnesota Heart Survey (11). Of the 1,151 cases, 677 had sufficient data to validate MI. Of these, 581 (86%) met the criteria for MI; the remaining 96 were excluded. Excluding the 191 cases that did not have complete information to verify MI and the 283 for whom medical records were not available did not change the study results, so they are included. The participation rate among eligible cases was 60%. Approximately four community controls with no history of MI were selected for each case using random digit dialing. The participation rate among known eligible controls was 54%.
Exposure and covariate data were collected for all cases and controls using the same structured telephone interview in identical settings (the participants' homes). Subjects were not told of the study hypothesis at any point during the study.
In order to maximize the validity of NANSAID exposure information, cases were interviewed only if they could be reached within four months of their MI. To further maximize recall (12,13), all cases and controls were prompted for exposure information with indication-specific questions about NANSAID use, examination of photographs with pictures of NANSAID products (including pictures of their containers when available) mailed to them before the interview, and reading of all other available NANSAID names for which pictures were not available. All participants also were asked to have all of the containers for the medications that they took during the index week (defined in the following text) available during the interview.
Definition of exposure
The a priori definition of exposure was any NANSAID or aspirin use within one week before the index date (the date of first onset of symptoms of MI for cases and the date of the telephone interview for controls).
The primary analysis of the study focused, a priori, on the association between nonselective NANSAIDs and the odds of MI in the absence of aspirin use. We also compared the effects of individual NANSAIDs among single NANSAID users relative to non-users. Based on recent pharmacological data (10), we also planned, prior to examining our data, to analyze the effects of aspirin on the odds of MI, stratifying by NANSAID use and focusing on ibuprofen, the only NANSAID to date known to antagonize the antiplatelet effects of aspirin (10). Because of concerns of uncontrollable confounding among those with a history of coronary artery disease, we adjusted for coronary disease by exclusion in the aspirin analyses. In these analyses we considered “frequent users” of NANSAIDs to be those using NANSAIDs at least 4 times/week and “infrequent users” to be those who used NANSAIDs <4 times/week. Finally, in order to estimate the potential effect of adding an NANSAID to a regimen of prophylactic aspirin, we compared the effects of NANSAIDs plus aspirin to aspirin use alone among those who reported using aspirin to prevent cardiovascular disease.
In order to adjust for possible confounding, we performed multiple logistic regression analyses that included, a priori, all variables in Table 1. We also tested for confounding (defined as a change in the odds ratio [OR] of ≥10% with adjustment) for the following: calendar year; alcohol consumption; history of renal dysfunction, heart failure, or arthritis; and use of either beta-blockers, calcium channel blockers, angiotensin-converting enzyme inhibitors, nitrates, cholesterol lowering medications, vitamins, or hormone replacement therapy.
In order to estimate the possible effects of recall bias, we compared those reporting NANSAID use (without acetaminophen use) with those reporting use of acetaminophen only. We also analyzed the effect of NANSAIDs among two groups of subjects most likely to have the most accurate recall: cases and controls who had all of their medication bottles from the index week available during the interview and cases who were interviewed within two months of their index date (12). Finally, we reviewed admission records of cases to determine if NANSAID use documented in the medical record was recalled in the telephone interview.
In order to estimate the possible effect of nonparticipation bias, we used the results from two nonparticipation substudies. In the first, we compared the chart-documented use of NANSAIDs among 129 medical records of non-participant cases with the chart documentation of NANSAID use among participant cases. In the second, we interviewed 182 consecutive non-participant controls and asked about analgesic and anti-inflammatory drug use in the prior week. This was compared with the results from identical questions asked of participant controls. We then derived an OR for NANSAIDs and MI, adjusted for nonparticipation (14).
The study was approved by the institutional review boards of the University of Pennsylvania and all participating hospitals.
The distribution of risk factors by NANSAID use is shown separately among cases and controls in Table 1.
Association between NANSAIDs and MI
Aspirin and NANSAIDs were used by 27% and 30% of the study population, respectively. Non-prescription NANSAID use made up 78% of all NANSAID use.
The use of NANSAIDs in the index week was associated with a significant reduction in MI odds among those who did not use aspirin (Table 2) (adjusted OR 0.53; 95% confidence interval [CI]: 0.42 to 0.67). Both ibuprofen (adjusted OR 0.52; 95% CI: 0.39 to 0.69) and naproxen (adjusted OR 0.48; 95% CI: 0.28 to 0.82) were associated with lower MI odds among non-users of aspirin. Utilization rates were too low to analyze other NANSAIDs. Among those using aspirin, NANSAIDs were not associated with lower odds of MI (adjusted OR 0.83; 95% CI: 0.58 to 1.17) (Table 2).
Effect of NANSAIDs on the aspirin-MI association
Among those not using NANSAIDs, aspirin use was associated with a statistically significant reduction in odds of MI, compared with no aspirin use (adjusted OR 0.79; 95% CI: 0.63 to 0.98) (Fig. 1). Among those using NANSAIDs in the same week as aspirin, there was no protection from aspirin (OR 1.28; 95% CI: 0.85 to 1.94; p for interaction by NANSAID use = 0.026). This interaction was suggested for ibuprofen (interaction p = 0.057) but not naproxen use (interaction p = 0.69) (Fig. 1), suggesting a potential difference between the two medications.
As the frequency of NANSAID use increased from “none” to “infrequent” (1 to 3 times/week) to “frequent” (≥4 times/week), the ORs for aspirin versus non-aspirin users increased from 0.78 to 0.97 to 2.04 (p = 0.006 for linear trend) (Fig. 2). This interaction by NANSAID frequency was present for ibuprofen (p = 0.018) but not naproxen (p = 0.86).
The OR for MI among those using NANSAIDs plus prophylactic aspirin versus prophylactic aspirin alone was 0.92 (95% CI: 0.46 to 1.81). The corresponding OR among those using ibuprofen plus prophylactic aspirin versus prophylactic aspirin alone was 1.01 (95% CI: 0.47 to 2.20), and the OR for other NANSAIDs plus aspirin versus aspirin alone was 0.61 (95% CI: 0.17 to 2.21). However, those using frequent ibuprofen plus aspirin had an OR relative to aspirin-only users of 2.03 (95% CI: 0.60 to 6.84), while the OR for those with infrequent ibuprofen plus aspirin use was 0.60 (95% CI: 0.21 to 1.66; p value comparing the two ORs = 0.12). The ORs for other NANSAIDs plus aspirin versus aspirin alone were 0.84 (95% CI: 0.18 to 3.93) when the NANSAIDs were used frequently and 0.35 (95% CI: 0.04 to 3.46) when the NANSAIDs were used infrequently.
Assessment of possible recall and nonparticipation bias
Among the 182 non-participant controls, 89 (48.9%) reported use of an analgesic. Among 210 participant controls interviewed during the same months, 119 (56.6%) reported use of an analgesic for the same indications. Among 129 non-participant cases with medications recorded in the medical records, 9 (7.0%) had a NANSAID documented compared with 66 (9.1%) of 724 participant cases.
In the second substudy, 47 cases had NANSAID use at some time before admission (not necessarily in the index week) documented in their medical record. Of these 47 cases, 38 (81%) recalled their NANSAID use during their telephone interview. Of the 10 cases whose records explicitly documented NANSAID use in the index week, all 10 (100%) recalled their NANSAID use. A total of 37% of the NANSAIDs recalled were obtained without prescription. Among 481 cases that denied NANSAID use on telephone interview, 472 (98%) cases did not have any NANSAIDs documented in the medical record. In the nine cases that did have NANSAIDs documented, the records did not indicate if the case was actively using the NANSAID at the time of MI.
None of the analyses that adjusted for recall or nonparticipation bias substantially altered the results (Table 3).
This study demonstrates a significant association between nonselective NANSAID use and lower odds of first, nonfatal MI in the absence of concurrent aspirin use. The lower OR for NANSAIDs than for aspirin was likely due to residual confounding by indication among aspirin users, and should not be interpreted as demonstrating that NANSAIDs have a greater effect on MI risk than aspirin.
The lower risk of MI among aspirin users, relative to non-users, was not observed among concomitant NANSAID users, particularly those using NANSAIDs with greater frequency. This interaction was suggested for ibuprofen (overall p value for interaction of 0.057), particularly with increasing frequency of ibuprofen (p = 0.018), but not naproxen (all p > 0.6). The lack of a significant interaction with naproxen could be due to limited power and, therefore, does not establish a lack of interaction.
It is interesting to note that, in the setting of frequent NANSAID use, aspirin users had a statistically significantly increased risk of MI relative to non-users of aspirin (Fig 2). This suggests that perhaps subjects taking frequent NANSAIDs and using aspirin because of increased risk of MI were not obtaining similar cardioprotection from their aspirin compared with aspirin users who were not using NANSAIDs or using them infrequently (Fig 2). To attempt to evaluate this possibility, we directly compared users of NANSAIDs plus prophylactic aspirin to users of aspirin alone. We did not find an increased risk overall, but the OR among frequent and infrequent users of ibuprofen were 2.03 and 0.60, respectively, suggesting that there might be an effect of regular ibuprofen use. However, the CIs were very wide. Therefore, our study cannot definitively determine the cardiac risk from the use of NANSAIDs, and ibuprofen in particular, among patients using aspirin. However, a recent study of patients with known cardiovascular disease, a population different from ours, has demonstrated a higher rate of overall mortality and cardiovascular death (not MI per se) among users of ibuprofen plus aspirin relative to aspirin alone (15).
Prior studies of nonselective NANSAIDs and MI
Several prior studies, including a randomized trial (16), have suggested that NANSAIDs themselves might protect against MI (16–18), but five nonexperimental studies have produced inconsistent results (3–7). Our study provides a possible explanation for these discrepant findings; the results of these prior studies may be explained by their use of electronic prescription records to estimate NANSAID use. As a result, they could not identify non-prescription NANSAID use nor estimate the effect of NANSAIDs separately in non-aspirin users. With respect to the former, a large proportion of NANSAID use is likely to be non-prescription (19)(80% of all NANSAID use among our controls, and even 47% of NANSAID among the Medicaid population). These non-prescription NANSAID users would have been characterized as “unexposed” in these prior database studies. The lack of data on non-prescription aspirin use (20)also prohibited the analysis of non-aspirin users separately from aspirin users. When, in our study, we compare prescription NANSAID users to all other participants (i.e., including non-prescription NANSAID users along with non-users of NANSAIDs) and do not exclude aspirin users, our OR for NANSAIDs and MI is 1.0. This bias would be more pronounced for ibuprofen than naproxen, because ibuprofen is a more commonly used over-the-counter drug (8)and because of the potential interaction of ibuprofen with aspirin. Thus, the same analysis above yields an OR of 1.1 for ibuprofen and 0.6 for naproxen. Therefore, it is possible that the results of the prior studies, including the differences observed between naproxen and ibuprofen, are due to these biases. Inability to control for important confounders in administrative databases (e.g., body mass index, physical activity) may also account for some of the differences.
Our results also suggest that at least some of the difference in MI rates between naproxen and rofecoxib in the absence of aspirin use noted in the Vioxx Gastrointestinal Outcomes Research study (9)could be due to a protective effect of naproxen. However, different results comparing celecoxib with other NANSAIDs (ibuprofen and diclofenac) (21)and recent data on rofecoxib (22)raise the possibility that there could be differences in cardiovascular risk among COX-2 inhibitors (23).
Several potential limitations should be considered in interpreting our results. Although our analysis comparing NANSAIDs with acetaminophen suggests that there was some recall bias, this bias was not sufficient to create the study findings. Because acetaminophen is an over-the-counter analgesic, like many of the NANSAIDs in this study, a comparison of NANSAIDs with acetaminophen should not be biased by differential recall. In fact, this comparison should overestimate the effect of recall bias because we did not use memory aids to improve recall for acetaminophen (12,13). In addition, other adjustment for recall did not alter the results (Table 3). Finally, adjusting for the 81% recall of NANSAID use among cases in our chart substudy, our OR for NANSAIDs among non-aspirin users would be 0.56 and among aspirin users would be 0.89. Second, selection bias was minimized in this study by randomly selecting controls from the same source population as the cases. Nonparticipation bias, a type of selection bias, is possible, but adjustment for estimates of nonparticipation did not alter the results. Third, uncontrolled confounding is always possible, even with our extensive adjustment for almost all known cardiac risk factors. Fourth, because we included only nonfatal MIs, a false association could occur if NANSAIDs had no effect on, or increased the risk of, fatal MI, particularly sudden death. However, clinical trial data suggest that this is very unlikely (24), consistent with the effects of NANSAIDs after excluding fatal coronary heart disease events in a prior study (7). Even if NANSAIDs had no effect on the estimated 30% of MIs that are fatal (25), our results would still be significant (OR of 0.68). Of course, further study is needed to determine the effects of NANSAIDs specifically on fatal MI.
Our results suggest that nonselective NANSAIDs may, themselves, reduce the risk of MI. If this unintended benefit of MI protection is confirmed with additional studies (and, optimally, randomized trials), the risk-benefit ratio of nonselective NANSAIDs will have to be reassessed. However, the cardioprotective effects of aspirin are well established, and nonselective NANSAIDs should not be considered alternatives to aspirin for prevention. Although our study cannot determine definitively if adding ibuprofen to aspirin use increases the risk of MI relative to aspirin use alone, recent data suggest that this could be the case (15). Further study is needed to assess the clinical relevance of the pharmacological interaction of ibuprofen with aspirin.
The authors acknowledge Sandy Barile for her editorial assistance, Max Herlim for his programming assistance, and William H. Sauer, MD, for his assistance with data coding.
☆ Supported by a grant from the National Institutes of Health. A supplement for research still ongoing was provided by Searle Pharmaceuticals Inc. (now Pharmacia). Dr. Strom has served as a consultant for Pfizer (associated with Pharmacia) and for most other pharmaceutical manufacturers who make either nonselective NANSAIDs or COX-2 inhibitors. Dr. Kimmel has also served as a consultant to the manufacturers of nonselective NANSAIDs, all unrelated to NANSAIDs. Dr. Kimmel currently receives grant support from Merck for research on COX-2 inhibitors. Dr. Reilly has grant support from Merck, including for one study of NANSAIDs and COX-2 inhibitors.
- confidence interval
- myocardial infarction
- non-aspirin non-steroidal anti-inflammatory drug
- odds ratio
- Received April 16, 2003.
- Revision received August 11, 2003.
- Accepted August 19, 2003.
- American College of Cardiology Foundation
- Bombardier C,
- Laine L,
- Reicin A,
- et al.
- Mascioli S.R,
- Jacobs D.R.J,
- Kottke T.E
- West S.L,
- Strom B.L,
- Poole C
- Brochier M.L
- Inman W.H.W,
- Rawson N.S.B,
- Wilton L.V
- Griffin M.R,
- Yared A,
- Ray W.A
- Silverstein F.E,
- Faich G,
- Goldstein J.L,
- et al.
- Antithrombotic Trialist's Collaboration
- American Heart Association