Dr. Suissa has received honoraria (less than $10,000 per year) from Bristol-Myers Squibb and Sanofi-Aventis.
Antirheumatic drug use and the risk of acute myocardial infarction
Article first published online: 27 JUL 2006
Copyright © 2006 by the American College of Rheumatology
Arthritis Care & Research
Volume 55, Issue 4, pages 531–536, 15 August 2006
How to Cite
Suissa, S., Bernatsky, S. and Hudson, M. (2006), Antirheumatic drug use and the risk of acute myocardial infarction. Arthritis & Rheumatism, 55: 531–536. doi: 10.1002/art.22094
- Issue published online: 27 JUL 2006
- Article first published online: 27 JUL 2006
- Manuscript Accepted: 23 NOV 2005
- Manuscript Received: 10 MAY 2005
- Canadian Institutes of Health Research
- Fonds de la recherche en santé du Québec
- Distinguished Investigator award from the Canadian Institutes of Health Research
- Observational studies;
- COX-2 inhibitors;
- Rheumatoid arthritis
To assess the risk of acute myocardial infarction (AMI) associated with the use of disease-modifying antirheumatic drugs (DMARDs) and other medications commonly used in rheumatoid arthritis (RA).
We conducted a nested case-control analysis within a cohort of subjects with RA, observed between 1999 and 2003, identified from the PharMetrics claims database. For each first AMI hospitalization identified during followup, 10 controls matched on sex, age, and time of study entry were randomly selected from the cohort. Conditional logistic regression was used to estimate the rate ratio (RR) of AMI associated with the current use of anti-RA therapy, as measured from dispensed prescriptions, after adjustment for AMI risk factors.
The cohort included 107,908 subjects (average age 54 years at cohort entry). During followup, 558 AMI cases occurred (3.4 per 1,000 per year). AMI rate was significantly decreased with the current use of any DMARD (adjusted RR 0.80, 95% confidence interval [95% CI] 0.65–0.98). This effect was consistent across all DMARDs, including methotrexate (RR 0.81, 95% CI 0.60–1.08), leflunomide (RR 0.28, 95% CI 0.12–0.65), and other traditional DMARDs (RR 0.67, 95% CI 0.46–0.97), but not biologic agents (RR 1.30, 95% CI 0.92–1.83). AMI rate increased with the use of glucocorticoids (RR 1.32, 95% CI 1.02–1.72) but not with nonselective nonsteroidal antiinflammatory drugs (RR 1.05, 95% CI 0.81–1.36) or cyclooxygenase 2 (COX-2) inhibitors (RR 1.11, 95% CI 0.87–1.43).
DMARD use is associated with a reduction in AMI risk in patients with RA. No risk increase was found with the COX-2 inhibitors in this population.
Rheumatoid arthritis (RA) is a chronic inflammatory arthropathy associated with significant morbidity and reduced survival (1). The most common cause of death in RA is cardiovascular disease (CVD), with much higher incidence and mortality rates among patients with RA than the general population (2). These increases do not appear to be fully explained by traditional risk factors such as smoking, diabetes mellitus, hypertension, hypercholesterolemia, and obesity (3). The severity of RA, however, has been associated with a higher risk of death due to CVD (4). These observations have contributed to the hypothesis that chronic inflammation may mediate the increased risk of CVD in persons with RA (5). This hypothesis has gained support from recent evidence pointing to the role of inflammation in the development of CVD in the general population (6, 7). If the chronic inflammation present in persons with RA does increase coronary risk, it can be hypothesized that treatments aimed at controlling RA activity may affect the rate of cardiovascular events.
Several of the medications used to treat inflammation in individuals with RA have been associated with CVD. Glucocorticoids promote hypertension, dyslipidemia, and glucose intolerance and have been associated with the development of atherosclerosis in patients with RA independent of their cardiovascular risk factors and clinical status (8). However, whether glucocorticoid treatment as such is atherogenic or whether it is a marker of systemic inflammation, which itself has been proposed as a cause of accelerated atherosclerosis, or severity of disease remains the subject of controversy. Some researchers have even proposed that glucocorticoids, by alleviating systemic inflammation, may actually decrease the risk of atherosclerosis and CVD in patients with RA. A recent review of the effects of glucocorticoids on cardiovascular mortality in patients with RA concluded that the true effect, and even the direction of the effect, of glucocorticoids on cardiovascular mortality in these patients remains unclear (9). Traditional nonsteroidal antiinflammatory drugs (NSAIDs), particularly naproxen, have been shown to protect against CVD (10–12), but not all studies have found this effect (13, 14). The selective cyclooxygenase 2 (COX-2) enzyme inhibitors have been associated with an elevated risk of CVD in randomized trials (15–17) as well as observational studies and meta-analyses (18–20).
However, there have been few studies of the effects of disease-modifying antirheumatic drugs (DMARDs) on the occurrence of CVD in patients with RA. For traditional DMARDs, hydroxychloroquine has been found to have a beneficial effect on lipid profiles (21), and methotrexate was recently reported to be protective against death from CVD in patients with RA (22). With the newer DMARDs, however, the data are contradictory. Some studies suggest that leflunomide and tumor necrosis factor α (TNFα) antagonists may adversely affect a patient's cardiovascular risk profile due to increases in blood pressure and lipids (23, 24), and recent studies suggest that some of these agents may in fact decrease the risk of CVD (25).
Therefore, to date, the question of whether anti-RA medications, particularly the newer DMARDs, have an effect on CVD outcomes remains largely unanswered. We conducted a cohort study to examine the effect of medications commonly used in the treatment of RA on the risk of acute myocardial infarction (AMI) requiring hospitalization.
SUBJECTS AND METHODS
Role of the funding source.
Sanofi-Aventis made all arrangements to acquire the databases. The authors independently designed the study, analyzed the data, interpreted the results, and wrote the manuscript.
The study cohort was formed using the PharMetrics Patient-Centric Outcomes Database, a North American insurance claims database (26). This database consists of standardized information on claims data from 75 different health insurance plans and currently encompasses more than 55 million unique patients. The database includes data on all physician visits, hospitalizations, and prescribed medications dispensed in these populations. Data from January 1995 to December 2003 were used for this study. To protect confidentiality, this database does not permit access to patients' medical records. The database has been used previously to assess the hepatic effects of DMARDs (27).
We identified all subjects with a diagnosis of RA based on data (International Classification of Diseases, Ninth Revision [ICD-9] code 714) from physician visits between January 1, 1999 and December 31, 2003. Although data were available starting in January 1995, we chose January 1999 as the start date because several of the drugs of interest in this study, particularly leflunomide, biologic agents, and COX-2 inhibitors, were only widely available on the market after that time. Cohort entry was defined as the date of the first prescription for an anti-RA medication after January 1, 1999. The anti-RA drugs included DMARDs (namely methotrexate, antimalarials [hydroxychloroquine and chloroquine], leflunomide, biologic agents, sulfasalazine, gold compounds, minocycline, penicillamine, and immunosuppressants) as well as glucocorticoids, NSAIDs, and COX-2 inhibitors. The cohort entry date of subjects with <12 months of eligibility in the health insurance plan prior to cohort entry was moved forward in time and redefined as the first prescription for an anti-RA drug after a year of eligibility in the database. All subjects were followed from the date of cohort entry until one of the following occurred: termination of health plan enrollment, death, the end of the study period (December 31, 2003), or the outcome of interest, namely AMI (definition below). Subjects had to be free of the outcome of interest (AMI), including old myocardial infarction (ICD-9 code 412), during this 1-year period and had to be ≥18 years at cohort entry.
The outcome was the first occurrence of AMI requiring hospitalization during followup. These events were identified from all inpatient physician encounters with an ICD-9 code of 410.
Nested case-control design.
Because of the complexity and time-varying nature of drug exposures, we used, as an alternative to the cohort design, a nested case-control approach within the cohort, which increases the ease of analysis with inconsequential loss of power (28). For each case of AMI identified in the cohort, we randomly selected 10 controls after matching on sex, age (within 2 years), and month and year of cohort entry, and after ensuring that each control was still at risk for a first-time AMI on the day the case occurred. The date of AMI occurrence in the case was designated as the index date for each case-control set.
For all subjects, all drugs received during the year prior to the index date, including DMARDs and other non-DMARD RA drugs, were identified from dispensed prescription data. The DMARDs were considered together as a class and were also divided into 4 groups: methotrexate, leflunomide, biologic agents available at the time (etanercept, infliximab, and anakinra), and all other DMARDs (hydroxychloroquine, chloroquine, sulfasalazine, gold, azathioprine, minocycline, penicillamine, chlorambucil, cyclophosphamide, and cyclosporine). The drugs, commonly used as concomitant treatment for RA, were grouped into glucocorticoids, NSAIDs, and COX-2 inhibitors. In view of previous studies on NSAIDs and COX-2 inhibitors, which found that among NSAIDs naproxen may exhibit a cardioprotective effect (10–12), the NSAIDs were separated into naproxen and all others, and COX-2 inhibitors were also separated into rofecoxib and celecoxib (the 2 COX-2 inhibitors available during this period).
Age was used as a basic covariate defining the study population. The identification of comorbid conditions was based on diagnoses made during the year prior to cohort entry based on ICD-9 codes for outpatient and inpatient physician encounters. Prior CVD was defined as the diagnosis (based on any physician encounter) of ischemic heart disease, heart failure, stroke, peripheral arterial disease, and other CVDs. Other comorbidities that we considered as cardiovascular risk factors included hypertension, diabetes mellitus, and hypercholesterolemia. We also identified other comorbid conditions including respiratory disease and cancer.
Total person-time of followup was cumulated to estimate the rate of occurrence of AMI in the cohort. Conditional logistic regression was used with the nested case-control sample to estimate the rate ratio (RR) of AMI for each of the anti-RA medication classes. Current exposure to these medications was defined as a prescription dispensed during the 30-day period prior to the index date. The DMARDs were analyzed as a class and also as 4 groups (methotrexate, leflunomide, biologic agents, and all other DMARDs). To create mutually exclusive DMARD exposures, we defined these groups hierarchically as methotrexate monotherapy, leflunomide (with or without other DMARDs), biologic agents (with or without other DMARDs but not leflunomide), and all other DMARDs as the last group. For the drugs commonly used as concomitant therapy, namely glucocorticoids, NSAIDs, and COX-2 inhibitors, current exposure was considered a binary factor indicating whether a prescription for the given agent was dispensed during the 30-day period prior to the index date. These factors were included in conditional regression models, with additional adjustment for age and comorbidity, the latter using stepwise regression.
The cohort included 107,908 subjects who had a diagnosis of RA and a dispensed anti-RA prescription with at least 1 year of information prior to cohort entry. The average age of the subjects at cohort entry was 54 years, and 74% were women. The entire cohort generated 166,194 person-years of followup. During followup, 558 cases of AMI requiring hospitalization occurred, with an overall rate of 3.4 per 1,000 per year.
The AMI cases and their matched controls were 65 years of age at the index date, and the cases occurred on average 14 months after cohort entry (Table 1). Fifty-five percent of the cases and matched controls were women. The cases had appreciably more comorbidity, including CVD and risk factors, as well as noncardiovascular conditions during the year prior to cohort entry, thus justifying the inclusion of comorbidity in adjusted analyses.
|Cases (n = 558)||Controls (n = 5,580)|
|Age, mean ± SD years†||65 ± 12||65 ± 12|
|Followup, mean ± SD months†||14 ± 11||14 ± 11|
|Comorbidity (1 year prior to cohort entry)|
|Ischemic heart disease||19||8|
|Peripheral arterial disease||6||2|
|Other cardiovascular disease||8||7|
|Cardiovascular risk factors|
|DMARD use 1 year prior to cohort entry||37||39|
The adjusted RR of an AMI for the current use of any DMARD was 0.80 (95% confidence interval [95% CI] 0.65–0.98) relative to no current use (Table 2). This effect remained relatively consistent for the current use of each DMARD category, including methotrexate monotherapy (RR 0.81, 95% CI 0.60–1.08), leflunomide (RR 0.28, 95% CI 0.12–0.65), and other traditional DMARDs (RR 0.67, 95% CI 0.46–0.97), but not biologic agents (RR 1.30, 95% CI 0.92–1.83). With respect to the concomitant medications commonly used in RA, the rate of AMI was not significantly increased with the current use of traditional NSAIDs (RR 1.05, 95% CI 0.81–1.36) and COX-2 inhibitors (RR 1.11, 95% CI 0.87–1.43), but was increased with glucocorticoids (RR 1.32, 95% CI 1.02–1.72) relative to noncurrent use. The adjusted RR of AMI with etanercept was 0.63 (95% CI 0.34–1.17) and with infliximab was 1.58 (95% CI 0.82–3.05) (data not shown).
|AMI cases (n = 558)||Controls (n = 5,580)||Crude RR||Adjusted†|
|DMARDs (current use)|
|No current use (reference)||416||3,945||1.00||1.00||Reference|
|Current DMARD use||142||1,695||0.78||0.80||0.65–0.98|
|All other DMARDs§||34||479||0.69||0.67||0.46–0.97|
|Other anti-RA drugs (current use)|
The effects of nonselective NSAIDs and COX-2 inhibitors are shown in Table 3. In particular, the RR for current naproxen use was 0.98 (95% CI 0.59–1.64). Among the 2 COX-2 inhibitors, the RR of AMI with celecoxib was 1.03 (95% CI 0.75–1.40), whereas for rofecoxib it was 1.26 (95% CI 0.89–1.80).
|AMI cases (n = 558)||Controls (n = 5,580)||Crude RR||Adjusted†|
|Nonselective NSAIDs (current use)|
|COX-2 inhibitors (current use)|
In a large cohort of subjects with RA, we found that current exposure to disease-modifying drugs appears to have a protective effect against AMI. In contrast, NSAIDs and COX-2 inhibitors were not found to alter this risk.
The finding of a decreased AMI risk with DMARD therapy may have several possible explanations. A likely explanation may be an indirect result of the beneficial effect of DMARDs on cardiac risk factors in persons with RA. For instance, the effectiveness of DMARDs in improving physical activity may contribute to a decrease in cardiovascular risk factors. In addition, glucocorticoids have been hypothesized to increase atherosclerotic disease (8, 29), and DMARDs, through their steroid-sparing effect, may reduce cardiovascular risk. Finally, the ability of DMARDs to decrease systemic inflammation has also been hypothesized to mediate a reduction in cardiac risk (22, 25).
Although the absence of an increased risk with the use of both COX-2 inhibitors celecoxib and rofecoxib can be considered surprising, previous studies are not consistent in this respect. Increased rates of cardiovascular events were reported in several randomized controlled trials (15–17), observational studies, and meta-analyses (18–20), but other trials and studies did not find such an effect in either RA or non-RA populations (19, 30–32). Moreover, several studies that did find an elevated risk found that the risk varied according to the COX-2 inhibitor, the dose, or the study population (19, 20, 33–35), variations that may explain our findings.
We did not find a protective effect of nonselective NSAIDs on AMI risk. Earlier studies have suggested that naproxen in particular, possibly due to its antiplatelet effect, might reduce cardiothrombotic events in non-RA and RA populations (10–12). However, these findings were not replicated by more recent studies (13, 14, 32).
The current study has several possible limitations. This study was exclusively based on computerized claims data, making it impossible to validate the outcome of AMI. However, limiting the cases to those hospitalized for AMI likely increases the validity of the diagnoses. It was also not possible to verify whether the subjects actually used their dispensed anti-RA medications, but the severity of this disease and the nature and effectiveness of the treatments in decreasing symptoms make compliance more likely. The database did not permit confirmation of the RA diagnosis, although the physician encounter data combined with a dispensed prescription for an anti-RA medication also likely increase the validity of the RA diagnosis. Our study was possibly subject to residual confounding. The RRs that we found for DMARDs could be due to physicians prescribing these drugs to patients according to the presence of CVD risk factors. However, we reduced this form of confounding by restricting the analyses to subjects with a first AMI, and by adjusting for various comorbid conditions. Nevertheless, data on unmeasured potential confounders such as weight, physical activity, and smoking were not available in this study and could have affected the results if physicians prescribed DMARDs selectively to healthier patients. Finally, it is conceivable that a protective effect of current DMARD use is due to the better health of current users if DMARD users who become at risk for an AMI stop their treatment before the event occurs. This was not the case in our study, as the noncurrent users who had stopped within the previous year did not have an increased AMI risk.
The increasing use of DMARDs in the last few years has helped improve the treatment of patients with RA, and newer agents have been an important development for patients who do not respond to or have contraindications to methotrexate or other traditional DMARDs. The cardiovascular effects of newer DMARDs, specifically leflunomide and TNFα antagonists, have been controversial, with some data suggesting that DMARDs may increase blood pressure and lipids, and other data showing a beneficial effect on cardiovascular risk. More data will be necessary to ascertain the cardiovascular effects of these drugs. In conclusion, our finding that the use of DMARDs is associated with a reduction in the risk of AMI in patients with RA suggests that the benefits of DMARDs in RA may extend beyond their arthritis-remitting effects.
We thank Dr. Juhaeri Juhaeri of Sanofi-Aventis for his assistance with the database acquisition.
- 26Pharmetrics Inc. Pharmetrics patient-centric database. Watertown (MA): Pharmetrics Inc; 2005.
- 28Novel approaches to pharmacoepidemiological study design and statistical analysis. In: StromB, editor. 4th ed. Pharmacoepidemiology. New York: John Wiley & Sons; 2005. p. 812–29..