Abstract
Objective. Cardiovascular (CV) disease (CVD) risk is increased in rheumatoid arthritis (RA). However, longterm followup studies investigating this risk are scarce.
Methods. The CARRÉ (CARdiovascular research and RhEumatoid arthritis) study is a prospective cohort study investigating CVD and its risk factors in 353 patients with longstanding RA. CV endpoints were assessed at baseline and 3, 10, and 15 years after the start of the study and are compared to a reference cohort (n = 2540), including a large number of patients with type 2 diabetes (DM).
Results. Ninety-five patients with RA developed a CV event over 2973 person-years, resulting in an incidence rate of 3.20 per 100 person-years. Two hundred fifty-seven CV events were reported in the reference cohort during 18,874 person-years, resulting in an incidence rate of 1.36 per 100 person-years. Age- and sex-adjusted HR for CV events were increased for RA (HR 2.07, 95% CI 1.57–2.72, p < 0.01) and DM (HR 1.51, 95% CI 1.02–2.22, p = 0.04) compared to the nondiabetic participants. HR was still increased in RA (HR 1.82, 95% CI 1.32–2.50, p < 0.01) after additional adjustment for CV risk factors. Patients with both RA and DM or insulin resistance had the highest HR for developing CVD (2.21, 95% CI 1.01–4.80, p = 0.046 and 2.67, 95% CI 1.30–5.46, p < 0.01, respectively).
Conclusion. The incidence rate of CV events in established RA was more than double that of the general population. Patients with RA have an even higher risk of CVD than patients with DM. This risk remained after adjustment for traditional CV risk factors, suggesting that systemic inflammation is an independent contributor to CV risk.
Rheumatoid arthritis (RA) is associated with increased cardiovascular (CV) morbidity and premature death of CV origin, when compared to the general population1. Several underlying mechanisms are suggested. Traditional CV risk factors are associated with the development of CVD in RA, similar to the general population1, including hypertension (HTN), dyslipidemia, smoking, type 2 diabetes, a sedentary lifestyle, and obesity. RA is characterized by chronic systemic inflammation (i.e., not limited to the joints), which is thought to be another major contributor to CVD development in these patients2,3,4. Systemic inflammation appears to increase CV risk independently and beyond traditional CV risk factors, while it also potentially alters existing CV risk factors in these patients3,5,6. A metaanalysis of 14 observational cohort studies showed an overall increased CV risk of about 50% in patients with RA compared to non-RA participants7. Other studies report standardized mortality ratios (SMR) as high as 2.7 compared with the general population8,9; these SMR appear to be constant over 50 years10, although this has been challenged by a study reporting decreasing mortality in the last decade11. However, study results are heterogeneous, and most studies have a followup shorter than 10 years. In 2009, our group also reported an increased CV risk in a cohort of patients with RA compared with a population-based reference cohort over 3 years of followup4. We now report on the incidence and risk of fatal and non-fatal CV events in this cohort over a maximum followup period of 15 years, compared with a representative sample cohort of the general population focusing on prevalence and incidence of type 2 diabetes mellitus (DM) and its complications.
MATERIALS AND METHODS
To investigate the risk of incident CVD in participants with RA versus DM and the general population, the CARRÉ (CARdiovascular research and RhEumatoid arthritis) study cohort and the Hoorn study cohort were compared. In both populations, data regarding demographics, RA-related variables, risk factors for CVD, and new or incident CV events after the start of the study were collected. A brief description of both study cohorts follows.
The CARRÉ study.
The CARRÉ study was initiated in 2000 with the purpose of investigating CVD and its risk factors in patients with longstanding RA. As previously described by van Halm, et al12, patients were eligible for participation if they were registered at the Jan van Breemen Institute (Reade since 2009) in Amsterdam, the Netherlands, fulfilled the 1987 American College of Rheumatology classification criteria, were diagnosed with RA between 1989 and 2001, and were aged between 50 and 75 years. Patient enrollment was done between January 2001 and January 2002. In total, 353 patients with RA were followed prospectively for 15 years. Study assessments were performed at baseline, 3 years (2004–2005), and 10 years of follow up (2010–2011). Participants were called for a CVD questionnaire in 2015 (15 yrs of followup) to assess the occurrence of CV events during the study period. All reported events since the start of the study were confirmed with medical records.
The Hoorn study
The Hoorn study is a Dutch cohort study of glucose metabolism and other CV risk factors that began in 198913. The cohort and its baseline measurements have been described in detail previously14. Briefly, a random selection of 3553 men and women, 50–75 years old, was taken from the population register. A total of 2540 (71.5%) agreed to participate, and after the exclusion of 56 non-white participants, the Hoorn Study population consisted of 2484 men and women. All Hoorn participants were subject to an extensive and repeated CV screening program similar to that used in our CARRÉ study. The local Medical Ethics Committee approved both studies (METc VUmc, Amsterdam, The Netherlands, 2001.198) and all participants gave their written informed consent at the baseline visit. An overview of the 2 cohorts is shown in Figure 1.
Study design of the CARRÉ and Hoorn study cohorts. RA: rheumatoid arthritis; IFG: impaired fasting glucose; DM: type 2 diabetes mellitus.
Assessment of RA-related values
At baseline, the demographic data, medical history, family history, and medication use of all patients with RA were recorded. Additionally, the Health Assessment Questionnaire was performed and a Disease Activity Score of 28 joints (DAS28) was calculated. IgM-rheumatoid factor (IgM-RF) and anticitrullinated protein antibody (ACPA) were assessed only at baseline. Assessed at all visits were C-reactive protein, erythrocyte sedimentation rate, and the presence of erosions in hands and feet with radiographs.
Assessment of traditional CV risk factors
CV risk factors were assessed for all subjects according to identical protocols in both studies, as described previously4. The assessments included smoking status, systolic and diastolic blood pressure (BP), body mass index (BMI; weight/height2 in kg/m2), waist to hip ratio and in fasting blood samples total cholesterol (TC), high-density lipoprotein cholesterol (HDL-C), low-density lipoprotein cholesterol (LDL-C), triglycerides (TG), TC/HDL-C ratio, glucose, HbA1c, and creatinine. HTN was defined as a systolic BP above 140 mmHg and/or a diastolic BP above 90 mmHg and/or current use of antihypertensive agents. DM was assessed in both study cohorts according to the 1999 World Health Organization criteria15. Patients were grouped according to fasting glucose levels into normal fasting glucose < 6.1 mmol/l (< 110 mg/dl), impaired fasting glucose (IFG), or insulin resistance (IR) ≥ 6.1 to < 7.0 mmol/l (110–125 mg/dl), and DM if glucose ≥ 7.0 mmol/l (≥ 126 mg/dl) or treated with glucose-lowering agents.
Assessment of CVD
Prevalent CVD at baseline and incident fatal and nonfatal CV events at followup were registered separately according to the International Classification of Diseases, 9th ed (ICD-9 codes 410.0–410.9, 435.9, 436, 443.9, and 798). CVD was defined as a verified medical history of coronary heart disease (i.e., myocardial infarction, percutaneous coronary intervention, coronary angiography with significant stenosis, stent placement, or coronary artery bypass graft), cerebral arterial disease (e.g., cerebrovascular accident, transient ischemic attack, or carotid endarterectomy), or peripheral arterial disease (e.g., ankle brachial pressure index < 0.50, peripheral arterial reconstructive surgery, or limb amputation). Unexpected deaths were considered to be CV mortality only when this was confirmed by autopsy. After study entry, participants were censored after the first new fatal or nonfatal CV event or death due to other reasons. CV events before baseline (prevalent CVD) in medical records of study participants were registered separately for additional correction in statistical analyses but were not used as new incident CV event cases for the primary outcome. The remaining patients were censored at study cessation time: March 1, 2015. Only the first CV event during followup was recorded. Patients lost to followup were censored at date of their last (event-free) followup visit. To extract data on the occurrence of CV events of the patients lost to followup, medical records were searched for the most recent medical status.
Statistical analysis
The baseline characteristics of both study cohorts were compared with parametric and nonparametric tests as appropriate. Incidence rates for fatal and nonfatal CV events were calculated per 100 person-years. The risk of developing CV events was compared between RA patients without DM, the nondiabetic general population, and patients with DM but without RA by calculating HR with Cox proportional hazards models. These hazard models were adjusted for potential confounders identified at baseline and in earlier studies4. Prevalent CVD was not excluded from the analyses, except where stated otherwise. Figure 2 was created using the corrected group prognosis method as described by Ghali, et al16. Several hazard models were made: the first model corrected adjusted for age and sex; the second model additionally adjusted corrected for systolic BP, use of anti-hypertensive agents, TC, HDL-C, statin use, smoking in pack-years, BMI, and aspirin use. The HR were first calculated for all patients. In the secondary analysis, patients with prevalent CVD were excluded. Additionally, the risk of developing CV events in patients with both RA and type 2 DM was compared to the other groups of patients. All analyses were performed with the software package IBM SPSS statistics (version 19). A p value < 0.05 was considered statistically significant.
Survival curves of participants with type 2 diabetes mellitus (black) versus the nondiabetic reference population (green), adjusted for age, sex (A), and traditional cardiovascular risk factors (B); nondiabetic participants with rheumatoid arthritis (black) versus the nondiabetic reference population (green), adjusted for age, sex (C), and traditional cardiovascular risk factors (D).
RESULTS
Baseline characteristics of both study populations
In the CARRÉ study, of the 353 patients with RA at baseline, 326 patients were included in the primary analyses (Figure 1). Their baseline characteristics as well as those from the Hoorn study reference cohort are shown in Table 1. Of these, 27 were lost to followup later on in the study because of migration, not wishing to participate because of a high burden, or death. The median followup duration was 11 years, with a minimum of 2 months and a maximum of 15 years. The majority of the patients with RA was female (65%) with a mean age of 63 ± 7 years. The median disease duration was 7 (4–10) years with a mean DAS28 score of 3.9 ± 1.3. Of the patients, 236 (72%) were IgM-RF–positive, 168 (52%) were ACPA-positive, and 263 (81%) had erosions in the hands or feet. Patients with RA were slightly older than the reference cohort and had more often prevalent CVD (15% vs 7%), HTN (61% vs 32%), a longer cumulative exposure to smoking, and more often used antihypertensive drugs, statins, or aspirin.
Patient characteristics of the RA population and the reference cohort at baseline.
New CV events in RA versus the general population
In the CARRÉ study, 95 patients developed a CV event during a median followup period of 11 years (range 2 mos to 15 yrs) and a total followup time of 2973 patient-years, resulting in a CVD incidence rate of 3.20 per 100 person-years (Table 2). In the Hoorn study, 257 participants developed a CV event during 12 years of median followup (range 1 mo to 12 yrs) and a total followup time of 18,874 patient-years, resulting in a CVD incidence rate of 1.36 per 100 person-years (Table 2). The occurring CV event subtypes are described in Appendix 1. Age- and sex-adjusted Cox regression analyses showed an HR of 1.93 (95% CI 1.51–2.45, p < 0.01) for CV events in RA (Table 2). Additional adjustment for CV risk factors resulted in an HR of 1.89 (95% CI 1.40–2.46, p < 0.01). Adjustment for prednisone use did not affect the HR significantly (data not shown). Exclusion of patients with prevalent CVD at baseline resulted in comparable HR (Table 2).
HR for new CV events in patients with RA versus the reference population.
New CV event in RA and DM versus the reference population and in RA versus DM
Before conducting these analyses, RA patients with an IFG (n = 26) or already diagnosed with DM (n = 22) were excluded from the CARRÉ study, and patients with an IFG (n = 206) were excluded from the Hoorn study. Cox regression analyses comparing incident CV events were performed with the remaining participants grouped into nondiabetic RA (n = 278), DM (n = 162), and the nondiabetic reference population (n = 1501; Table 3). The HR was increased for RA (HR 2.07, 95% CI 1.57–2.72, p < 0.01) and DM (HR 1.51, 95% CI 1.02–2.22, p = 0.04) in the age- and sex-adjusted model (Table 3, Figure 2). After adjustment for CV risk factors, the HR (1.82, 95% CI 1.32–2.50, p < 0.01) remained significantly increased in RA while it was not significant in DM (HR 1.28, 95% CI 0.85–1.92, p = 0.25; Table 3, Figure 2), with similar results after exclusion of patients with prevalent CVD (Table 3). The adjusted survival curves of the nondiabetic reference population, participants with DM, and nondiabetic participants with RA are shown in Figure 2. In addition, a direct comparison between participants with RA (n = 278) and participants with DM (n = 162) showed a higher HR for CVD in RA when compared to DM (age- and sex-adjusted HR 1.64, 95% CI 1.07–2.53, p = 0.02).
New CV events in RA and DM versus the reference population.
New CV events in subgroups of RA, insulin resistance, and DM versus the general population
The risk of developing CV events was compared between the nondiabetic general population (n = 1501), individuals with IR (n = 206), individuals with DM (n = 162), participants with RA but without IR or DM (n = 278), study participants with RA and IR (n = 26), and participants with RA and DM (n = 22). Compared to the nondiabetic general population, patients with RA and IR had the highest risk of developing a CV event when the hazard model was corrected for age and sex, followed by the other patients with RA (Table 4). Results were similar after additional correction for systolic BP, antihypertensive agents, TC, HDL-C, statin use, smoking, BMI, aspirin use, and prevalent CVD.
New CV events in RA with IR and DM versus the reference population.
DISCUSSION
In our second report on this cohort, now with a median followup of 11 years, the increased risk of CVD in RA already seen after 3 years of followup4 was confirmed, now at a level more than double that of the nondiabetic general population. Traditional CV risk factors, such as HTN, smoking, older age, and previous CVD partially explained this increased risk. However, the Cox proportional hazard models adjusted for these risk factors still showed an almost 2-fold increased CV risk in participants with RA. In participants with DM, the increased CV risk was almost fully explained by traditional CV risk factors. In RA, high-grade systemic inflammation most likely amplifies CV risk, as suggested previously3,4,5. Recently, Curtis, et al reported the highest incidence rate of acute myocardial infarction in patients with RA and DM, followed by patients with DM only, RA only, and neither RA nor DM17. We found similar results (highest HR) for participants with RA and IR or DM, but in our study, participants with RA only had a higher CVD risk than participants with DM only. This might be explained by differences in study design and population and definitions of CVD outcome. More importantly, both studies underline and establish that RA is a risk factor for the development of CVD.
Interestingly, although we had a low number of observations, the risk of IR or DM appeared additive to that of RA, suggesting that in such patients, risk factors should be approached aggressively. In participants with RA, levels of TC, LDL-C, and TG were lower than in the nondiabetic general population, while HDL-C was higher. In the past, several publications have described this phenomenon as the “lipid paradox,” in which the above-mentioned observation is not translated into a lower but paradoxically into a higher CV risk in patients with RA18,19,20. However, we identified an increased CVD risk in RA, regardless of the effect of these lipid changes on this CV risk, because we adjusted for TC, HDL-C, and statin use in our analyses. Optimal treatment of RA results in a normalization of lipid levels in these patients21,22,23,24. Unfortunately, our study also demonstrates that, although patients with RA have an increased presence of certain known risk factors for CVD, only a small proportion of patients with RA is receiving appropriate treatment with antihypertensive agents and statins. This finding is also in line with other studies25,26,27,28, suggesting that appropriate management of lifestyle factors and CV risk factors is lacking in these patients. Possible explanations could be that local CV risk management programs are ineffective or that patients and physicians are unaware of the magnitude of this risk. In addition, current CV risk assessment tools are inaccurate and risk assessment is performed using general population algorithms (e.g., SCORE and Framingham Risk Score). Some risk factors, such as lipids, are influenced by inflammation and are not an appropriate indicator for CV risk during periods of active disease20. Therefore, CV risk algorithms that accurately predict CV risk in RA as well as multidisciplinary CV risk assessment and management are certainly of additional value. Above all, creating awareness of this increased CV risk among patients and clinicians is of great importance.
The major strength of our study is its long followup duration in which the occurrence of both fatal and nonfatal CV events was recorded in patients with RA and in the general population. Our previous study had a followup duration of only 3 years, with few events, resulting in limited power, but in our current study the longterm observation and the number of events was sufficient to overcome this. However, our present study also has some limitations. The Hoorn study was conducted about 10 years before the CARRÉ study. The definition of certain diseases, their assessment, and management may have changed during this period. This may have influenced our results. Additionally, CVD incidence has declined over the last decades in the Netherlands, which could translate into less incident CVD in patients with RA29. However, comparing both groups would in this case lead to an underestimation of the CV risk in patients with RA. Additionally, the Hoorn study may have included some patients with RA that we could not identify because of insufficient data, but this would also only result in an underestimation of the CV risk in RA. Another important matter is that the prevalence of MTX and biologic treatment at baseline is not representative of the current clinical practice. Treatment guidelines have changed over the last decade and this could have influenced the CVD incidence in the RA population. However, the effect of antirheumatic treatment on CVD risk in patients with RA was not the research question of our study.
Our study demonstrates a more than 2-fold higher CV risk in patients with RA when compared with the nondiabetic general population. In our study, this risk is even higher than the CV risk of patients with DM. In accordance with our previous study from 20094, adjustment for CV risk factors still results in a significant residual CVD risk for patients with RA, indicating that systemic inflammation is likely an independent contributor to CV risk in RA. This underscores that both optimal antiinflammatory treatment of RA as well as effective CV risk management are likely of major importance to reduce CV risk in these patients.
APPENDIX 1. Cardiovascular event subtypes in both populations.
- Accepted for publication May 1, 2019.
