The National Data Bank for Rheumatic Diseases has conducted safety registries for Centocor, Sanofi-Aventis, and Bristol-Myers Squibb, and has received research grants from Abbott, Amgen, Wyeth-Australia, Merck, and Pfizer.
To determine the risk of myocardial infarction (MI) in patients with rheumatoid arthritis (RA) compared with that in patients with noninflammatory rheumatic disorders and to determine risk factors for MI in RA, the relationship between cardiovascular risk factors and corticosteroid use, and the relationship between RA treatment and MI.
We conducted a cohort study of MI in 17,738 patients with RA and 3,001 patients with noninflammatory rheumatic disorders who were assessed at 6-month intervals between 1999 and July 2006. We evaluated treatment effect in a nested case–control study of RA participants who were matched by age, sex, study duration, and date of study entry.
The covariate-adjusted risk of first MI in RA versus that in noninflammatory rheumatic disorders was 1.9 (95% confidence interval 1.2–2.9) (P = 0.005). In RA, MI was predicted by age, sex, education level, hypertension, smoking, exercise, prior MI, diabetes, a comorbidity index, use of low-dose aspirin and antilipemic agents, RA severity and treatment variables, and corticosteroid use. Except for obesity, predictors were of equal strength in RA and noninflammatory rheumatic disorders. The increased risk for MI in RA compared with that in noninflammatory rheumatic disorders lessened when corticosteroid users were excluded. Use of corticosteroids was associated with future development of diabetes and hypertension.
MI in RA is associated with demographic and cardiovascular risk factors and corticosteroid use. Study data support the hypothesis that RA activity causes MI and that corticosteroids are primarily a marker of RA activity. However, corticosteroids increase the risk of diabetes and hypertension and contribute to the overall risk of MI.
The risk of ischemic heart disease is substantially increased in rheumatoid arthritis (RA). Recent large epidemiologic studies have provided quantitative evidence of the association between RA and the risk of myocardial infarction (MI). The relative risk of MI was 2.0 (95% confidence interval [95% CI] 1.2–3.3) in the 2003 Women's Health Study (1), 1.9 (95% CI 1.7–2.1) in a 2006 report from a British Columbia population registry (2), 1.8 (1.2–2.4) in a community registry (3), and 1.5 (95% CI 1.2–1.8) in the UK-based General Practice Research Database (4). Older studies, including those using different study designs, have shown the risk to range from 1.1 to 5.2 (5). The mechanism by which RA exerts its effect is not clear, but it has been suggested to be independent of traditional cardiovascular risk factors (5, 6).
Baseline demographic and cardiovascular risk factors have been compared in patients who subsequently developed MI (7). However, no epidemiologic studies of MI risk are available that include a wide range of demographic, cardiovascular, behavioral, RA severity, and RA treatment risk factors. Corticosteroids are widely used in RA and are thought to be associated with hypertension, diabetes mellitus, and hyperlipidemia, all of which are risk factors for MI (8–10). However, no studies document the purported association between hypertension/diabetes mellitus and corticosteroids in RA. In addition, anti–tumor necrosis factor (anti-TNF) therapy might alter the risk of cardiovascular events. C-reactive protein (CRP), which is produced in response to TNF and interleukin-6, has been shown to be an independent risk factor for MI and stroke (11–13). In addition to initiation of CRP synthesis, TNF may activate endothelial cells, converting them into procoagulant and prothrombotic states (11, 14).
In the present study, we used a longitudinal, prospective cohort to investigate MI risk in a large, well-defined sample of patients diagnosed by rheumatology specialists as having RA, to 1) measure the overall risk of MI in RA, 2) determine the comparative predictive strength of cardiovascular risk factors for MI, 3) determine whether corticosteroids are associated with hypertension and diabetes mellitus, and 4) determine the association of individual RA therapies with MI risk.
PATIENTS AND METHODS
We studied participants in the National Data Bank for Rheumatic Diseases (NDB) longitudinal study of rheumatic disease outcomes. NDB participants are recruited from the practices of US rheumatologists and are followed up prospectively with semiannual, detailed, 28-page questionnaires, as previously described (15, 16). This study utilized NDB data in a longitudinal cohort analysis of 20,739 adult participants (ages 18–103 years), of whom 17,738 had RA and 3,001 had a noninflammatory rheumatic disorder. In contrast to the unselected patients from rheumatology practices, 6,035 RA patients were enrolled specifically as part of a treatment safety registry. Safety registry patients are those enrolled at the time they start a specific therapy (e.g., infliximab), and their inclusion may create a bias toward increased RA severity.
Patients were enrolled continuously beginning in 1999 and completed at least 2 semiannual questionnaires between January 1999 and July 2006. Rheumatic disease diagnoses were made by the patients' rheumatologists. Noninflammatory rheumatic disorders included diagnoses such as osteoarthritis, back pain syndromes, tendinitis, etc., excluding fibromyalgia. Study variables were assessed at entry into the NDB and at every subsequent completion of the semiannual questionnaire. In this study, we performed 2 separate analyses: cohort analyses that included patients with RA and/or noninflammatory rheumatic disorders and a nested case–control analysis that was restricted to RA patients.
Possible MIs were identified from study questionnaires, hospitalization records, physician reports, and death records. Only MIs that were confirmed by medical review or death records were included as MIs in this study. If hospital or death records were not available, we contacted the patient's physician and/or interviewed the patient or family with a structured, preplanned interview designed to address the reported condition. Comparison of patient self reports with medical records indicates agreement in >94% of cases. Review of potential cases was performed by 2 trained, experienced NDB staff reviewers. This review was followed by an independent physician review. Putative cases of MI with medical confirmation or, in the absence of physician or hospital data availability, with convincing patient or family report, were accepted as MIs in this study. Death records in which MI was recorded as the underlying cause must have referred to deaths that occurred within 6 months of the last questionnaire for the death to be regarded as resulting from an MI in this study in order to study effects of treatment. The index observation is defined as the 6-month period in which the patient had his or her first observed MI. For control subjects, the index observation was the matched observation(s) in the Cox regression and nested case–control analyses.
Demographic variables included age, sex, education level, body mass index (BMI), and smoking history. Comorbidity was measured by a patient-reported composite comorbidity score (range 0–9) comprising 11 present or past comorbid conditions including pulmonary disorders, MI, other cardiovascular disorders, stroke, hypertension, diabetes, spine/hip/leg fracture, depression, gastrointestinal (GI) ulcer, other GI disorders, and cancer (17), and by identification of prior MI, hypertension, and diabetes. BMI, hypertension, previous MI, and smoking history were ascertained from patient self- report. The presence of diabetes was determined by use of antidiabetic medication. We also identified the use of low-dose (81 mg) aspirin and antilipemic therapy and the amount of weekly exercise. Patients were classified as performing aerobic exercise if they performed any aerobic exercise in an average week, including swimming, biking, use of aerobic equipment, or other aerobic exercise.
Clinical variables specific to RA severity and outcome included duration of RA, total joint arthroplasties, the Health Assessment Questionnaire (HAQ) disability index (18, 19), and visual analog scales (VAS) for pain intensity and global severity. These 3 scales were used to calculate the Patient Activity Scale (PAS) score (20). The PAS is a 0–10 measure of RA disease activity in which higher values indicate greater RA disease activity. The PAS score is computed by multiplying the HAQ score by 3.33 and then dividing the sum of the VAS pain intensity score, VAS global severity score, and HAQ score by 3. The PAS is an effective measure of RA activity (21).
Assessment of treatment exposure.
Patients reported all medications used within the previous 6 months. We classified infliximab, etanercept, and adalimumab as anti-TNF agents. The average doses of drugs used in the study were as follows: methotrexate (MTX) 14 mg/week, hydroxychloroquine (HCQ) 349 mg/day, sulfasalazine 1,780 mg/day, leflunomide 20 mg/day, prednisone 7 mg/day, rofecoxib 27 mg/day, and celecoxib 310 mg/day. It was not possible to obtain self-reported doses of infliximab. The mean dose of adalimumab was 88 mg every 2 weeks, and that for etanercept was 50 mg/week. Drugs were characterized as “ever” used if they had been used at any time from the start of the study through the index event and as “current” if they had been used in the 6-month period before the index event. In addition, the cumulative duration of therapy was determined for each treatment at the index event. In the analyses of treatment effect, patients who were receiving anti-TNF therapy prior to entry into the study were excluded.
Nontreatment variables in these analyses represent values obtained in the questionnaire prior to the index observation, except for baseline values which are identified as such (e.g., history of MI at cohort entry). We analyzed cohort data with time-varying Cox regression. In the analysis of the risk of MI associated with RA, we excluded all patients who were members of safety registries. In the nested case–control analyses, we matched up to 20 controls with RA to each RA case (each RA patient with an MI) at the index observation by 10 categories of age, calendar time (6-month periods), calendar time of entry into the cohort (6-month periods), and sex using incidence density sampling (selecting controls without replacement from all persons at risk at the time of case occurrence, excluding the index case itself) (22). In these analyses, patients who had been receiving anti-TNF therapy prior to cohort entry were excluded from analysis in order to assess newly treated patients. All controls were alive and participants in the NDB at the time their matched case had the MI. Matching on cohort entry was performed to control for secular trends.
For the nested case–control analyses, we used conditional logistic regression for individually matched case–control groups to derive odds ratios (ORs) with 95% CIs as a measure of relative risk. The data shown in Table 2 were obtained at a randomly selected observation for each patient. We used Stata (version 10.0; StataCorp, College Station, TX) for all analyses. We selected a 2-tailed P value of 0.05 as significant, and we reported 95% CIs.
Table 2. Comparison of RA and NIRD subjects for MI predictor characteristics*
RA subjects (n = 17,738)
NIRD subjects (n = 3,001)
Except where indicated otherwise, values are the percent of subjects (95% confidence interval [95% CI]). Estimated percents and values are adjusted for age and sex, except for age and sex, which are adjusted for each other. See Table 1 for other definitions.
Not initially collected; ∼45% of observations were missing.
Table 1. Predictors of MI: univariate analyses adjusted for age and sex*
All RA subjects (n = 17,738)
RA subjects with first MI (n = 16,748)
All NIRD subjects (n = 3,001)
NIRD subjects without first MI (n = 2,822)
Except where indicated otherwise, values are hazard ratios (95% confidence intervals). Hazard ratios are adjusted for age and sex, except for age and sex, which are adjusted for each other. MI = myocardial infarction; RA = rheumatoid arthritis; NIRD = noninflammatory rheumatic disorder; BMI = body mass index; HAQ = Health Assessment Questionnaire; PAS = Patient Activity Scale.
P ≤ 0.05 for association with MI.
Not initially collected; ∼45% of observations were missing.
Cohort analysis: the risk of MI among RA patients and the effect of demographic and cardiovascular risk factors on MI risk.
There were 283 incident MIs (223 first MIs) in persons with RA and 35 (25 first MIs) in persons with noninflammatory rheumatic disorders during a mean followup period of 3.0 years (range 0.5–8.5 years). With adjustment for age and sex, a large series of RA MI predictors were identified that spanned demographic, educational, behavioral, cardiovascular, and RA severity variables. Significant predictors included age, sex, education level, hypertension, prior MI, diabetes, smoking, exercise, comorbidity, and use of low-dose aspirin and antilipemic agents (Table 1). The HAQ score and the PAS score were also associated with MI. However, among RA patients, duration of RA and the presence of a total joint replacement were not predictive of MI. Point estimates were similar for RA and noninflammatory rheumatic disorders. No interaction effects were found.
Differential characteristics of RA subjects and subjects with noninflammatory rheumatic disorders.
Subjects with RA and noninflammatory rheumatic disorders differed in many clinically important ways. With adjustment for age, RA patients were more likely to be men (22.2% versus 20.2%). In addition, with adjustment for age and sex, RA patients were more likely to have ever been smokers (52.3% versus 42.2%) and less likely to perform aerobic exercise (21.8% versus 26.8%) (Table 2). Subjects with noninflammatory rheumatic disorders were more likely to be obese (39.9% versus 27.8%), to have ever had hypertension (49.4% versus 39.0%), and to use low-dose aspirin (23.2% versus 12.3%) and antilipemic therapy (32.5% versus 20.7%), and had more comorbid conditions (comorbidity score 1.80 versus 1.63), in addition to other, less clinically significant, differences. Among those with RA, the mean ± SD age was 60.9 ± 13.3 years and the median duration of RA was 12.2 years. The use of RA therapies was as follows: MTX 53.7%, prednisone 39.1%, anti-TNF therapy 36.1%, HCQ 21.0%, leflunomide 17.6%, celecoxib 17.6%, and rofecoxib 7.9%.
The risk of MI in RA compared with that in noninflammatory rheumatic disorders.
To estimate the risk of MI-associated RA, we restricted RA analysis to 11,703 RA patients who were not part of drug treatment registries to be certain that the results were not biased by more severe RA cases. With adjustment for age and sex, RA subjects were at substantially greater risk for MI than those with noninflammatory rheumatic disorders (hazard ratio [HR] 1.7 [95% CI 1.1–2.4], P = 0.008). With adjustment for age, sex, college education, lifetime smoking, diabetes, hypertension, comorbidity index, low-dose aspirin, and baseline history of MI, the HR was 1.6 (95% CI 1.1–2.3) (P = 0.020). Excluding patients with a baseline history of MI from the analysis, and thereby analyzing first-ever MIs, the age- and sex-adjusted result was an HR of 1.9 (95% CI 1.2–2.9) (P = 0.006), and the fully adjusted result was an HR of 1.9 (95% CI 1.2–2.9) (P = 0.005). The covariate-adjusted HRs may be thought of as the risk for MI in RA compared with that in noninflammatory rheumatic disorders after adjustment for demographic and cardiovascular risk factors.
Nested case–control analysis: the effect of RA treatment and severity on the risk of MI.
To a conditional logistic regression model (Table 3) that contained college education, ethnicity, smoking status, diabetes, aerobic exercise, hypertension, comorbidity index, low-dose aspirin, BMI, and baseline MI status, we added the PAS score, total joint replacement status, and RA duration. The multivariable ORs for the 3 variables in the first MI model were as follows: for the PAS score, 1.1 (95% CI 1.0–1.2) (P = 0.036); for total joint replacement, 1.1 (95% CI 0.8–1.6) (P = 0.381); for RA duration per 10 years, 0.9 (95% CI 0.9–1.1) (P = 0.227). We then added all treatment variables to the models that already contained demographic, cardiovascular, and RA severity risk factors.
Adjusted for college education, ethnicity, smoking status, diabetes, aerobic exercise, hypertension, comorbidity index, low-dose aspirin, BMI, baseline MI status, PAS score, total joint replacement status, and RA duration. “Ever” therapy indicates that patients received that therapy at any time up to and including the index observation. “Current” therapy indicates that patients were receiving that therapy within 6 months of the index event. OR = odds ratio; 95% CI = 95% confidence interval (see Table 1 for other definitions).
All RA patients not receiving anti–tumor necrosis factor (anti-TNF) therapy at entry into the study. These included 255 patients with a history of MI prior to or subsequent to entry into the study and 4,924 controls.
All RA patients not receiving anti-TNF therapy at entry into the study and with no history of MI at entry into the study. These included 198 patients who had their first MI subsequent to entry into the study and 3,776 controls.
P values are for the association of RA treatment with MI, and were significant only for prednisone monotherapy.
Most common monotherapies used. Groups consist of patients simultaneously receiving these therapies but not any of the other therapies listed above or below. Patients may be receiving other disease-modifying antirheumatic drugs. Numbers of subjects in the RA subjects (all) group/RA subjects (first MI) group were as follows: for methotrexate (MTX), 709/527; for MTX + anti-TNF, 652/555; for MTX + anti-TNF + prednisone, 607/467; for MTX + prednisone, 480/351; for prednisone, 308/201.
In the 4 models examined in the monotherapy section of Table 3, prednisone was associated with subsequent MI, regardless of whether all MIs or first MIs were studied, or whether “ever” or “current” treatments were evaluated. The ORs in these analyses ranged from 1.4 (95% CI 1.0–1.8) to 1.7 (95% CI 1.3–2.3). The relationship of prednisone to other therapies is shown graphically in Figure 1. Neither anti-TNF therapy nor any other therapy was significantly associated with MI risk in any analysis.
In an effort to better understand these results, we examined the 5 most common treatments (used by ∼53% of patients) in combinations of 1 to 3 therapies (Table 3). The ORs for MI with the 3 treatment combinations that included prednisone were >1.0, while the ORs with the nonprednisone treatments were <1.0. The combination of the 2 groups not receiving prednisone differed significantly from the combination of the 3 groups receiving prednisone in the risk of MI (P = 0.033 and P = 0.012 for all MIs and first MIs, respectively) (see also Figure 1). These results demonstrate that the association of prednisone with MI was independent of other treatments.
We further investigated the relationship of prednisone and anti-TNF treatment to first MI as a function of exposure time. There was no significant association between MI and the duration of exposure to anti-TNF therapy (OR 1.1 [95% CI 1.0–1.2], P = 0.140) or the duration of prednisone therapy (OR 1.1 [95% CI 1.0–1.1], P = 0.208). However, the risk of MI was associated with increasing the current dose of prednisone. The respective ORs at 1–5 mg/day, 6–10 mg/day, and >10 mg/day compared with no prednisone were 1.7 (95% CI 1.2–2.3), 1.9 (95% CI 1.2–3.1), and 2.0 (95% CI 0.9–4.7). The risk was significantly greater in the group receiving >5 mg/day than in the group receiving 1–5 mg/day (P = 0.031).
The relationship between corticosteroid use and cardiovascular risk factors.
We studied the relationship between a subject's use of corticosteroids while being a member of the RA cohort and hypertension, diabetes, antilipemic therapy, and obesity, since these factors have been suggested to be increased by corticosteroid use. Table 4 displays the age- and sex-adjusted percent of patients with these conditions who were using and the percent who were not using corticosteroids at the index observation. Hypertension was significantly more common in patients receiving prednisone (49.1% versus 45.6%), but there were no differences according to prednisone status in the proportions of patients with diabetes or obesity or using antilipemic drugs. The mean and median prednisone-equivalent doses were 8.1 mg/day and 5.0 mg/day, respectively. Sixty-four percent of patients using corticosteroids used prednisone doses of ≤5 mg/day, 69% used prednisone doses of ≤7.5 mg/day, and 87% used prednisone doses of ≤10 mg/day.
Table 4. Comparison of prevalent risk factors in rheumatoid arthritis patients using prednisone and in those not using prednisone*
Did not use prednisone
Values are the age- and sex-adjusted percent of patients with these conditions using and not using prednisone at the index observation. The prevalence of a given risk factor within patients was determined at a randomly selected observation. Prednisone use is at any time during the study prior to the randomly selected observation.
Not initially collected; ∼28% of observations were missing.
To further explore the effect of corticosteroids on cardiovascular risk factors, we performed time-varying Cox regression analyses in all RA patients. In these analyses, we excluded patients who were receiving prednisone at their first observation, and for the respective analyses, we excluded those who already had diabetes, hypertension, or obesity. Antilipemic therapy wasn't included in the analyses because of missing data. As shown in Table 5, prednisone use was associated with future development of diabetes and hypertension, although no clear trend with dose could be seen. The duration of prednisone use was also associated with the development of diabetes and hypertension. For diabetes, each year of prednisone use yielded an HR of 1.18 (95% CI 1.06–1.32) (P = 0.003). For hypertension, the HR was 1.12 (95% CI 1.04–1.21) (P = 0.004). There was no association between obesity and the duration of prednisone therapy.
Table 5. Assessment of the risk of developing MI risk factors in RA patients using prednisone and in those not using prednisone*
HR = hazard ratio; 95% CI = 95% confidence interval (see Table 1 for other definitions).
Adjusted for age and sex.
Because of the findings in these analyses, we returned to the original Cox regression analyses of the risk of MI conferred by RA. As noted above, the age- and sex-adjusted HR for RA compared with noninflammatory rheumatic disorders was 1.7 (95% CI 1.1–2.4) (P = 0.008). However, when all patients using prednisone were excluded, the HR was 1.5 (95% CI 0.9–2.4) (P = 0.100), and the HR was 1.3 (95% CI 0.9–2.0) (P = 0.192) when only RA patients using prednisone were excluded. Similarly, in the analyses of first MI, the age- and sex-adjusted HR was 1.9 (95% CI 1.2–2.9) (P = 0.006). When all subjects with prednisone use were omitted, the HR was 1.7 (95% CI 1.0–3.0) (P = 0.063), and when only RA patients with prednisone use were omitted, the HR was 1.5 (95% CI 0.9–2.4) (P = 0.133).
The risk ratio of 1.9 for first MIs in RA observed in the current study confirms results from recent population and community-based studies (1–4). However, we also examined a series of risk factors that have not been generally available in previous epidemiologic studies of MI. Among RA subjects, we found that conventional cardiovascular risk factors—hypertension, preexisting cardiovascular disease, diabetes, current and past smoking, exercise, and the number of comorbid conditions, but not obesity—were also predictive of MI (Table 1). We noted that the magnitude of the predictive effect was similar in RA and noninflammatory rheumatic disorders. After adjustment for age and sex, cardiovascular risk factors were more common in the noninflammatory rheumatic disorders group, particularly with respect to obesity, hypertension, and diabetes (Table 2). However, lifetime smoking was 24% more common among RA subjects than among those with noninflammatory rheumatic disorders. From these data we conclude that, with the exception of BMI, cardiovascular risk factors are important predictors of MI in RA, but are not different in magnitude in subjects with RA and non-RA rheumatic disease. We also conducted the analyses shown in Tables 1 and 2 after excluding patients who were members of safety registries and obtained results similar to those shown in Tables 1 and 2 (data not shown).
In their 2004 report of 287 cases of RA before 1990 in the Nurses' Health Study, Solomon et al reported that cardiovascular risk factors were similar among those who had RA and those who did not (7). In a case–control study of 236 RA patients with controlling for age, sex, smoking, diabetes, hypercholesterolemia, systolic blood pressure, and BMI, Del Rincón et al concluded that the high incidence of cardiovascular events was not explained by traditional risk factors (6). In another study, those authors used high-resolution ultrasound to measure the carotid intima-media thickness (IMT) and plaque in 631 RA patients. They reported that demographic factors explained 11–16% of IMT variance, cardiovascular risk factors explained 4–12%, and RA manifestations explained 1–6% (23). The analyses in the current study yielded similar conclusions. The receiver operating curve areas under the curve for the nested models of risk of all MIs and first MIs in RA shown in Table 3 were 0.627 and 0.676 for combined RA severity and current treatment factors, respectively, and increased to 0.657 and 0.696, respectively, when demographic and cardiovascular risk factors were added.
Why should we and others (6, 23) be unable to detect a strong relationship between RA severity factors and MI when there is clear evidence of increased risk of MI in persons with RA? It is likely that RA produces cumulative damage that is not detectable by conventional cross-sectional studies. Even in our study, which had longitudinal data, patients were observed for only 19% of their RA course and during that time only at semiannual intervals. Although the use of average or cumulative values of predictors available to investigators will improve predictability, this effect is likely to be small (24).
Another possible reason is that our study did not include physician joint counts and laboratory tests. However, studies of mortality that did include these variables showed them to have minimal additional predictive ability (24). Furthermore, studies of arterial IMT demonstrated almost no improvement in statistical prediction by inclusion of CRP level, erythrocyte sedimentation rate, or Larsen radiographic score (25, 26). Kitas and Erb and Del Rincón et al have reviewed the possible mechanisms by which RA itself might lead to cardiovascular events (5, 6). Nevertheless, there is still no clearly understood or documented pathway from RA to MI.
We did find a positive association between MI and corticosteroid use (Table 3 and Figure 1). Corticosteroids might increase cardiovascular risk by increasing the prevalence of hypertension, diabetes, and hyperlipidemia (8–10). The relevance of the corticosteroid and cardiovascular risk factor link is not always clear, however, since study populations, doses, and comorbid conditions in reported studies limit generalizability of the results to persons with RA. The association between corticosteroid use and cardiovascular disease has been reviewed in detail by Davis et al (27).
There are limited data on actual corticosteroid-associated risk of MI. A large retrospective population-based study showed that doses of prednisone ≥7.5 mg/day were associated with MI (28). Other studies have suggested that the apparent corticosteroid association might reflect illness severity rather than corticosteroid use (29, 30). Corticosteroids are associated with the risk of symptomatic coronary artery disease in systemic lupus erythematosus (31, 32). In RA, corticosteroids are associated with carotid plaque and arterial incompressibility (33) as well as with almost all adverse RA outcomes, including work disability, total joint replacement, and mortality (34). Davis et al described a 3-fold risk of cardiovascular disease among seropositive (but not seronegative) RA patients, but did not separate out those with MI (35); other studies also did not distinguish MI from other cardiovascular conditions (27). Wei et al evaluated cases of patients hospitalized with inflammatory arthritis (28). However, the relevance of this sample to usual RA was uncertain, the results were probably confounded, and no specific data were given for adjusted MI risk in the arthritis group.
In multivariable analyses (Table 3), we found that current or ever users of corticosteroids were at significant risk for future MIs and that this effect was independent of other common RA therapies (Table 3 and Figure 1). We also noted that the risk of MI in RA compared with that in noninflammatory rheumatic disorders was reduced when patients using prednisone were excluded from the analyses.
It is generally established that corticosteroid use increases the risk of diabetes mellitus (36, 37). However, we found that prednisone use was not more common among patients who had diabetes, but that new cases of diabetes were significantly associated with prednisone use (HR 1.7 [95% CI 1.3–2.2]) (Table 5). This suggests that RA patients with diabetes are switched from prednisone or are not as likely to be prescribed prednisone as persons without diabetes.
Approximately 66% of RA patients use corticosteroids over their lifetime, and ∼12% use corticosteroids for >5 years (34). We found that corticosteroid use is associated with the development of diabetes and hypertension in patients with RA (Table 5), and that hypertension is increased in RA patients using corticosteroids.
Treatments that cause the risk of MI to increase can be viewed through 2 pathways. The first is a relatively quick effect by rapid alteration of cardiovascular risk factors. The model for this effect is rofecoxib (38). The rofecoxib effect is immediate, increases further with exposure, and decreases with treatment discontinuation. The second hypothesized mechanism is a gradual accumulation of risk in which some function of duration of exposure is related to the risk of the outcome, but this effect presumably does not decrease with treatment discontinuation. An example of this is the effect of corticosteroids on osteoporosis and cataracts.
The present study shares with all observational studies the limitation of potential residual confounding (39). There is strong evidence that corticosteroids are prescribed for persons with more severe RA (34), and there is evidence that more severe RA is associated with MI (40). We controlled for severity with the PAS and total joint replacement, and we included a very strong marker for severity, anti-TNF therapy, since such therapy is generally reserved for severe RA and RA treatment failure. In addition, the finding that corticosteroids increase the risk of hypertension and diabetes in RA patients further suggests a possible harmful role for corticosteroids. The results of the current study are similar to those of the well-controlled study by Davis et al and suggest general agreement when all of their patients are considered together. Even so, it is likely that some residual confounding exists in the current study and in other studies (27, 35), and that the risk associated with corticosteroids is partially a severity risk and partially a treatment risk.
In summary, we have confirmed the risk of MI conferred by RA (OR 1.9) in a large prospective database of RA patients treated by rheumatologists, and we have identified and quantified risk factors for MI. We were unable to detect a protective effect of anti-TNF therapy. Corticosteroids are associated with the development of hypertension and diabetes and are probably causally associated with MI risk, although residual confounding may explain part of the corticosteroid effect.
Dr. Wolfe had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.
Study design. Wolfe, Michaud.
Acquisition of data. Wolfe, Michaud.
Analysis and interpretation of data. Wolfe, Michaud.