Cardiovascular disease risk in rheumatoid arthritis: Progress, debate, and opportunity

Authors


Introduction

Rheumatoid arthritis (RA) is a systemic inflammatory disease associated with a substantially increased risk of cardiovascular disease (CVD), leading to reduced lifespan, diminished quality of life, and increased health care costs. Much progress has been made in understanding the mechanisms and impact of this public health problem over the past decade, and strong evidence exists in key areas, including the estimation of the contribution of RA to cardiovascular mortality and identifying a link between methotrexate use and lower CVD mortality. However, there remain key areas of debate.

Recently, there has been a veritable explosion of research into the link between RA and CVD. A Medline search of the combined terms “CVD” and “RA” returns 147 articles and 38 reviews published in just the past year, and a staggering 868 articles and 204 reviews over the past 5 years. However, only within the last 2 decades has the recognition of CVD as a major contributor to mortality in RA been appreciated, leading to a reduction in lifespan of a decade or more in some cases (1, 2). Beyond mortality, the impact of CVD on quality of life (3), combined with the associated increase in health care costs (4), render CVD in RA a major public health problem.

Addressing any public health problem involves several key steps, including measuring the problem's magnitude, identifying susceptible subgroups, isolating mechanistic determinants, and testing the efficacy and effectiveness of interventions. The body of literature addressing these steps varies in size and quality, resulting in an incomplete assessment of the breadth of the problem. For example, while multiple studies across diverse populations have established a strong base of support for RA as a risk factor for higher rates of hospitalized myocardial infarction (MI) and CVD mortality, the volume and quality of evidence is lower for end points that are either difficult to define clinically or lack reliability when ascertained from administrative data sources (e.g., heart failure) and for estimates of risk among subgroups of patients with RA (i.e., those characterized by disease features, disease activity, or severity). There is even less evidence regarding the isolation of mechanistic determinants, and importantly, no trials have been published to demonstrate that any particular intervention can reduce CVD risk in RA. The latter is urgently needed, as clinicians caring for RA patients are currently required to guess at optimal management by applying untested results from observational studies or extrapolating from non-RA populations. While these deficiencies in progress are considerable, their recognition allows for the development of a focused research agenda.

Well-written reviews (5, 6) on the topic of CVD risk in RA have been published. Rather than reiterating the content of these, we sought to frame CVD in RA in the context of its public health impact. First, while it is not possible within the confines of this review to provide a systematic synthesis of all of the relevant literature, we have summarized several key studies for the aspects of CVD in RA in which relatively abundant or high-quality evidence is available. Second, we explore key areas of debate of CVD epidemiology, pathogenesis, and treatment/prevention for which evidence is lacking, with a special emphasis on the research agenda required to address these controversies.

Progress

Truly high-grade evidence (such as well-designed meta-analyses, systematic reviews, randomized controlled trials, or prospective cohort studies) is available for very few aspects of CVD in RA. However, there are several areas where strong evidence allows conclusions to be made with reasonable assurance. These include higher CVD mortality and morbidity in RA compared to the general population, and lower CVD risk in methotrexate-treated patients.

Progress: CVD events and mortality are more frequent in RA, the increased risk is shared by men and women, and CVD events lead to reduced years of life for patients with RA.

Numerous longitudinal studies have demonstrated a greater risk of CVD events (7–15) and CVD mortality in RA (1, 7, 9, 11, 16–25). Relative to the general population, patients with RA are at an approximately 1.5–2-fold higher risk of MI (9, 11, 21, 26–28), a 1.4–2.7-fold higher risk of stroke (9–11, 15, 29), and a 1.3–1.7-fold higher risk of heart failure (10, 30, 31).

In a recent meta-analysis, Avina-Zubieta et al (32) compiled the results of 24 observational studies (comprising 111,758 patients). Compared to non-RA controls, there was a 50% increase in all-cause CVD mortality (standardized mortality ratio [SMR] 1.5, 95% confidence interval [95% CI] 1.39–1.61), with a 59% increase in death from ischemic heart disease and a 52% increase in death from cerebrovascular disease. Risk was even higher in studies enrolling patients after implementation of the 1987 American College of Rheumatology (ACR) criteria for the classification of RA (SMR 1.67, 95% CI 1.55–1.81); therefore, the risk of CVD death may be even higher in RA patients, assuming that the criteria prompted inclusion of only well-defined RA cases after 1987 (33). Inception cohort studies were the only group that did not show a significantly increased CVD mortality risk (SMR 1.19, 95% CI 0.86–1.68), but the pooled sample size was small compared to noninception cohorts, and the duration of followup was shorter.

In the same meta-analysis (32), the relative increase in CVD mortality and events was equal between men and women. Accordingly, RA patients experience a reduction in lifespan by an average of 5 to 15 years in both men and women with RA (1, 34). These sex equalities in outcomes are interesting, considering that some studies have identified a higher relative burden of subclinical atherosclerosis in men with RA compared to controls than in women compared to controls (35–38), suggesting that an increase in atherosclerotic burden per se may not be the sole contributor to the increased event rates observed across studies.

Coincidentally, with implementation of therapeutic and lifestyle interventions for CVD risk reduction, overall rates of mortality in the US population have improved dramatically over the past 4 decades (39). However, several recent reports suggest that this decline in mortality has not occurred in patients with RA (Table 1). This is surprising, considering data suggest that there may be a temporal trend toward milder disease in RA (as captured by disease activity scores, functional disability, measures of inflammation, and hospitalizations) (44–46) and the recent emphasis on tight control of disease activity for all RA patients. The widening mortality gap is concerning, but there are at least 2 possible explanations. First, the lead time from exposure to the most effective treatments (methotrexate, tumor necrosis factor [TNF] inhibitors) and the ascertainment of CVD events may require longer followup to observe a reduction in risk. Second, a reduction in CVD mortality may have been equaled by increases in mortality from other causes, such as infection or malignancy. These competing risks could serve to stabilize mortality rates, even in the setting of temporal improvements in CVD mortality. This is supported by a study by Krishnan et al (22) demonstrating a temporal decline in mortality from acute MI in 3,862 RA patients followed up between the years 1980 and 1997. In that study, patients with incident RA after 1990 did not have excess mortality from acute MI compared to the general population.

Table 1. Studies of mortality trends comparing patients with rheumatoid arthritis to the general population*
Author, year (ref.)Study typePopulation studiedNo. of person-yearsMortality outcomeMortality trend to general
  • *

    ↓ = decrease over time (i.e., approach normal); MI = myocardial infarction; ↑ = mortality relative to the general population was found to increase over time in the population studied; ↔ = no change over time.

  • This study did not provide person-years or average followup time; an estimate was calculated from followup times for each study included in the meta-analysis.

Bjornadal et al, 2002 (20)Population-based cohortSweden489,048All cause
Krishnan et al, 2004 (22)Incidence cohortNorth America22,209Acute MI
Gonzalez et al, 2007 (40)Incidence cohortUS11,705All cause
Bergstrom et al, 2009 (41)Prevalence cohortSweden2,472Cardiovascular
Meune et al, 2009 (42)Meta-analysis of cohortsMultiple91,916 patients, mean ± SD 14 ± 6 years of followupCardiovascular
Radovits et al, 2010 (43)Inception cohortThe Netherlands1,049 patients, mean ± SD 9 ± 6 years of followupAll cause

Progress: CVD event rates are lower in RA patients treated with methotrexate.

Methotrexate, the most frequently used disease-modifying antirheumatic drug (DMARD) for the treatment of RA, has been associated with a lower risk of cardiovascular death (47), cardiovascular morbidity (48–51), MI (49), and heart failure (52) compared to other treatments. The authors of a recent systematic review concluded that methotrexate use was associated with a 15–89% lower risk of CVD events in patients with RA (53). Of note, all published studies reviewed by the authors maintained a trend toward reduction in CVD mortality and CVD morbidity with methotrexate treatment (regardless of significance).

Despite concordance among observational studies, the lack of confirmation from randomized controlled trial(s) raises the possibility of bias (e.g., confounding by indication) to account for the protective effect of methotrexate. Indeed, 1 study (published as a letter) reported increased all-cause mortality in RA patients with preexistent CVD who were treated with methotrexate (54). While also not interventional, and perhaps explained by the lack of routine folate supplementation in the cohort, this report raises the possibility of a lack of benefit with methotrexate in certain subpopulations of patients.

The mechanism behind cardiovascular risk reduction with methotrexate is unknown. While it could be mediated by a methotrexate-related reduction in inflammatory cytokines, there are other possibilities. For example, Reiss et al (55) recently demonstrated the in vitro ability of methotrexate to prevent interferon-γ–induced transformation of lipid-laden macrophages into foam cells, an effect potentially mediated by promotion of reverse cholesterol transport by methotrexate. Such a mechanism may explain why lower CVD event rates are not seen with other nonbiologic DMARDs (e.g., sulfasalazine and leflunomide); however, data for other nonbiologic DMARDs are not as extensive, and their study is complicated by concomitant use with methotrexate, confounding ascertainment of independent drug effects.

Debate: key issues of controversy

Despite the recent advances in the understanding of CVD risk in RA described in prior sections, there remain several notable issues of debate, largely due to a lack of sufficient evidence to support sound conclusions. We have grouped these debates into 3 main themes: those of natural history, pathophysiology, and treatment.

Natural history

Debate: is CVD risk increased prior to RA disease onset?

There is debate as to when increased CVD risk manifests during the course of RA. Among studies of patients with established disease, RA duration has been a fairly consistent RA-specific factor independently associated with cardiovascular events (21) and subclinical atherosclerosis (35, 56–66), although this association has not been uniform across studies (28–30, 36–38, 67–77). Radovits et al (43), comparing mortality rates in an inception cohort of 1,049 RA patients versus the general population, found that a difference in observed and expected mortality emerged after 10 years of followup. Presumably, this association was mediated by the cumulative effect of the RA disease process, assumed to be cumulative systemic inflammation. However, there are other possible contributors to the observed increased risk with RA duration, such as the effects of accumulated deformity and disability on physical fitness and overall comorbidity (78–80). This is supported by their finding that sex, age, disability, comorbidity, and disease severity were all predictors of mortality. Other investigations have mirrored these findings. Studying an incidence cohort, Kroot et al (81) found no increase in mortality in patients with RA after up to 10 years of followup from disease onset. More recently, Holmqvist et al (82) found no increased occurrence of ischemic heart disease, MI, or angina before the onset of RA symptoms.

However, there is evidence supporting an increase in CVD event rates prior to RA diagnosis. From an incidence cohort of 603 RA patients in Rochester, Minnesota (27), RA patients were significantly more likely to have been hospitalized for acute MI or to have experienced an unrecognized MI during the 2-year period prior to fulfillment of ACR criteria. In a separate, cross-sectional study of 538 RA patients (83), RA patients with extraarticular manifestations had greater rates of hypertension and thrombosis than RA patients without extraarticular manifestations (83) prior to the diagnosis of RA. If accurate, these findings are potentially in conflict with the construct of accumulated risk over the course of RA duration, and could suggest that shared environmental or genetic risk factors for both the development of RA and CVD may contribute to this early risk.

There are a number of ways to reconcile these findings. For one, both autoimmunity and elevated systemic inflammation have been shown to precede clinical RA (84), potentially contributing to CVD risk. Second, most RA patients experience symptoms of articular inflammation for months to years prior to RA diagnosis (85), a lead time in which CVD events could manifest. Third, it is possible that there is a different set of risk factors contributing to early-onset CVD in RA compared to late-onset CVD.

The ideal research study to address this question would be prospective and longitudinal, following large numbers of patients (i.e., those at risk for the development of RA, such as those with a confirmed family history) over many years for predefined hard CVD end points. While this would require significant time commitment and financial support, it could provide a definitive answer to the question of whether patients are at an increased risk of CVD before the onset of clinically apparent RA.

Pathophysiology

Debate: is atherosclerotic burden higher in RA patients?

The evidence for increased subclinical atherosclerosis in RA is not definitive, and most is derived from cross-sectional studies. Using sensitive noninvasive techniques, the majority of studies demonstrate that subclinical atherosclerosis, vascular stiffness, and endothelial dysfunction are more prevalent in RA compared to control groups (Table 2).

Table 2. Comparative studies of carotid ultrasound, coronary artery calcium, endothelial function, and pulse wave analysis*
Author, year (ref.)No. of RA patientsFemales, %Excludes CVD or CVD risk factorsAge, mean ± SD or median yearsRA duration, mean ± SD or median yearsCVD outcomeOutcome associated with inflammation (ESR/CRP level)
  • *

    RA = rheumatoid arthritis; CVD = cardiovascular disease; ESR = erythrocyte sedimentation rate; CRP = C-reactive protein; NR = not reported; ↑ = marker was higher in RA than in controls; cc = common carotid; IMT = intima-media thickness; ↔ = marker was not significantly different between RA and controls; ↓ = marker was significantly lower in RA than in controls; FMD = flow-mediated endothelial-dependent vasodilatation.

  • Excluding carotid ultrasound studies with <40 patients and coronary artery calcium studies, endothelial function studies, and pulse wave analysis studies with <25 patients.

  • If the study excluded patients with a history of CVD or more than 1 traditional CVD risk factor, it was marked “yes.”

  • §

    Study divided RA patients into early and late groups by RA duration.

Carotid ultrasound  studies       
 Kumeda et al, 2002 (57)13888Yes55 ± 11NR↑ max cc IMT, ↑ max femoral IMTNo
 Park et al, 2002 (56)53100Yes55 ± 36 ± 5↑ mean cc IMTYes
 Alkaabi et al, 2003 (86)4050No5715↑ mean cc IMTDid not measure
 Del Rincon et al, 2003 (87)20489No6060↔ max IMT (internal and common), ↔ plaqueYes
 Gonzalez-Juanatey et al, 2003 (67)4777Yes59 ± 1316 ± 9↑ mean cc IMT, ↑ plaqueNo
 Abu-Shakra et al, 2005 (59)5786No52 ± 1513 ± 9↔ “ultrasound score”Did not measure
 Gerli et al, 2005 (68)10173Yes63 ± 1111 ± 9↔ mean bifurcation IMTNR
 Grover et al, 2006 (88)5791Yes42 ± 88 ± 5↑ mean cc IMTDid not measure
 Pahor et al, 2006 (89)70100No42 ± 610 ± 6↑ mean cc IMT, ↑ plaqueYes
 Pamuk et al, 2006 (36)6389Yes506↑ mean cc IMT, ↑ plaqueDid not measure
 Roman et al, 2006 (70)9898No48 ± 1312↓ mean cc IMT, ↑ plaqueNo
 Daza et al, 2007 (72)6585Yes44 ± 812↑ mean cc IMTNo
 Hannawi et al, 2007 (71)4068No537↑ mean cc IMT, ↑ plaqueYes
 Georgiadis et al, 2008 (90)4075Yes53 ± 13<1↑ cc IMT (single measurement)NR
 Kerekes et al, 2008 (64)5277Yes51 ± 1211 ± 9↑ mean cc IMTYes
 Mahajan et al, 2008 (65)10094Yes44 ± 1113 ± 4↑ mean cc IMT, ↑ plaqueDid not measure
 Pereira et al, 2009 (75)7190Yes49 ± 12NR↔ mean cc IMT, ↔ plaqueNo
 Schott et al, 2009 (76)93100Yes5314↔ mean cc IMT, ↔ plaqueNo
 Kobayashi et al, 2010 (38)19561Yes59 ± 99↑ max bulb IMT, ↑ plaqueNo
 Ristic et al, 2010 (77)42100Yes45 ± 107 ± 5↑ cc, bulb, and internal IMT (mean and max)Yes
Coronary artery  calcium  studies       
 Chung et al, 2005 (35)14169NoEarly RA: 51, late RA: 58§Early RA: 2, late RA: 20§↑ Agatston score in established RAYes
 Giles et al, 2009 (37)19561Yes60 ± 99↑ Agatston score in men with RAYes
 Kao et al, 2008 (91)105100Yes52 ± 816 ± 10↔ Agatston scoreYes
 Rho et al, 2009 (92)16969No54 ± 12NR↑ Agatston scoreYes
 Wang et al, 2009 (66)8586Yes54 ± 1213 ± 11↑ Agatston scoreNo
Endothelial function  studies       
 Gonzalez-Juanatey et al, 2003 (93)5576Yes59 ± 12NR↓ FMD (↑ endothelial dysfunction)NR
 Van Doornum et al, 2003 (94)2576Yes46 ± 814 ± 9↔ FMDNo
 Vaudo et al, 2004 (58)3288Yes50 ± 711 ± 8↓ FMD (↑ endothelial dysfunction)Yes
 Arosio et al, 2007 (62)65100Yes47 ± 66 ± 3↓ FMD (↑ endothelial dysfunction)Yes
 Kerekes et al, 2008 (64)5277Yes51 ± 1212 ± 9↓ FMD (↑ endothelial dysfunction)Yes
Pulse wave analysis  studies       
 Van Doornum et al, 2003 (94)2576Yes46 ± 813.7 ± 9↑ pulse wave velocityNR
 Maki-Petaja et al, 2006 (95)7781Yes57 ± 1313 ± 10↑ pulse wave velocityYes
 Arosio et al, 2007 (62)65100Yes47 ± 66 ± 3↑ pulse wave velocityNo
 Avalos et al, 2007 (61)11765NoEarly RA: 51, late RA: 60§Early RA: 2, late RA: 19§↔ pulse wave velocity, ↑ Augmentation indexNo
 Inaba et al, 2007 (73)30100Yes61 ± 100.9 ± 1↑ brachial–ankle pulse wave velocityNo
 Wallberg-Jonsson et al, 2008 (96)3077No54 ± NR24↑ stiffness index at baselineNo
 Pieringer et al, 2009 (97)3686Yes46 ± 816 ± 7↑ Augmentation indexNo

However, there are important discordances between aspects of these studies. For example, with carotid ultrasound, some groups found increased plaque prevalence but no difference (or even a decrease) in intima-media thickness (IMT) (70, 75), and others found no difference in IMT or carotid plaque (76, 87). The sources of these discrepancies may be related to differences in the size and composition of the populations studied, differences in imaging and interpretive methods between studies, and unaccounted confounders. Nevertheless, these inconsistencies between studies give a mixed picture of truth, and the clinical significance of the findings (i.e., whether the greater burden of atherosclerosis translates to hard outcomes) is unclear.

Many studies employ measures of vascular function, such as pulse wave analysis, acetylcholine-induced endothelium-dependent vasodilatation, and flow-mediated endothelium-dependent vasodilatation, assuming that these measures correlate with atherosclerosis and subsequent CVD risk (98). However, these associations have only been validated in non-RA patients (99), and their interpretation may be confounded by the marked inflammatory milieu of RA. Even the more direct, structural markers, such as carotid IMT and coronary artery calcification (CAC), are subject to measurement and prognostic uncertainty (100) and, in the case of CAC, may miss clinically relevant uncalcified coronary plaque (101).

Perhaps more importantly, there are few studies addressing upstream predictors of subclinical atherosclerosis (particularly those unique to RA), and fewer have addressed the downstream impact of subclinical atherosclerosis on CVD events and mortality in RA. Answers to these questions are prerequisite to designing interventional trials (both to identify targets for therapy and to legitimize the use of subclinical markers as outcomes), and studies addressing downstream impacts are required before subclinical disease findings can be used prognostically. While these data could potentially help clinicians in the field, there has been only 1 small study of 47 RA patients that explored a subclinical disease marker (in this case, common carotid IMT and plaque) as a predictor of CVD events in RA patients (102). In the study, both higher levels of IMT and the presence of plaque were associated with subsequent CVD events over 5 years of followup.

Debate: does systemic inflammation accelerate atherogenesis and promote plaque rupture in RA?

Elevated levels of nonspecific systemic inflammatory markers (e.g., erythrocyte sedimentation rate [ESR] and C-reactive protein [CRP]) predict future cardiovascular events and mortality in both the general population (103) and in RA patients (24, 25, 69, 104). For example, 1 study found a significant correlation between the last recorded ESR and cardiovascular events in patients with RA (104). Another study, following an incidence cohort of 603 RA patients for an average of 15 years, found that ≥3 recorded ESR values of ≥60 mm/hour were associated with a 2-fold higher hazard of cardiovascular death compared to those with lower values (69). Providing biologic support, studies of atherosclerosis-prone mice demonstrate lower endothelial activation (105) and reduction in the extent of atherosclerosis (106) from gene knockout or neutralization of TNFα. These findings suggest that chronically elevated levels of circulating inflammatory cytokines accelerate the process of atherogenesis in RA.

While compelling, support for this hypothesis has not been consistent across studies. As shown in Table 2, among the 29 investigations of subclinical atherosclerosis that included an assessment of associations with systemic inflammatory markers, only 14 showed a significant association. In 1 study (38), higher interleukin-6 levels were strongly and significantly associated with carotid atherosclerosis in matched controls, but demonstrated no association in RA patients. This inconsistency invokes the possibility that chronic exposure to circulating cytokines may have a less potent effect on atherogenesis in RA than observed in animal studies or the general population, particularly since inflammatory markers in RA patients are increased many-fold relative to non-RA patients, yet CVD risk is increased by only approximately 50%. However, most of the RA studies were cross-sectional, powered only to observe large effects, and enrolled patients with established disease. Since inflammatory markers tend to fluctuate with time and treatment, a single cross-sectional assessment later in the RA disease course may not accurately reflect cumulative exposure. Stronger associations of indirect measures of cumulative inflammation, such as RA duration (57, 59, 62, 65, 67, 88) or disease severity (37), with measures of atherosclerosis provide circumstantial evidence of causality, but these measures are indirect and could reflect pathogenic processes other than systemic inflammation.

Ideally, these questions would be addressed with an observational study including longitudinal assessment of systemic inflammation in RA cohorts followed from disease onset to provide an accurate measure of cumulative exposure. This study type, combined with an intervention demonstrating a reduction in atherogenesis with targeted suppression of systemic inflammation, would provide additional assurance that inflammation is playing a causal role. If inflammation does indeed exert a relatively less potent effect on atherogenesis in RA, the mechanisms mediating this reduction deserve investigation, both for understanding disease pathogenesis in general and for identifying the most effective interventional targets.

Recent studies (37, 38) exploring sex differences in subclinical carotid and coronary atherosclerosis have demonstrated that relative differences in atherosclerotic burden between RA patients and sex-respective controls were lower for women compared to men. Is there a mechanism that could account for increased events in subgroups of RA patients who do not have extensive atherosclerosis? Aubry et al (107), exploring autopsied coronary arteries of 41 RA patients and 82 age-, sex-, and autopsy date–matched controls, found no difference between groups in the severity or extent of atherosclerosis. However, RA patients were less likely to have multivessel disease and were more likely to have vulnerable and inflamed plaques than the controls. In support, studies of the general population have shown that in those with higher levels of systemic inflammation, the fibrous caps of complicated fibroatheromas contain more inflammatory cells, particularly at the rupture-prone shoulder regions (108). Combined with the general increase in thrombotic potential in RA (109), isolated, inflamed vulnerable coronary plaques could lead to higher event rates in the context of lower overall atherosclerotic burden. This distinction is important for RA patients, since imaging modalities increasingly used for cardiovascular risk stratification, such as the Agatston calcium score derived from cardiac computed tomography, may be less predictive of events than in the general population, potentially giving RA patients with lower Agatston scores a false sense of security.

Atherothrombosis has received less study in RA compared to atherogenesis, perhaps because techniques to assess plaque vulnerability and atherothrombotic potential are less widely available. Stamatelopoulos et al (110) reported that the carotid plaques in RA patients with active disease had significantly lower gray-scale median on ultrasound, a marker of plaque vulnerability, than RA patients with inactive disease, providing support that active disease promotes plaque vulnerability. Newer techniques in development, such as enhanced magnetic resonance imaging, could help distinguish inflamed rupture-prone plaques from stable plaques, independent of atherosclerotic burden. These studies are needed for understanding cardiovascular pathogenesis in RA and for validating RA-specific cardiovascular risk–stratification tools. Importantly, the use of nonsteroidal antiinflammatory drugs (NSAIDs) is common in RA patients. The potential for NSAIDs to contribute to atherothrombosis in RA is a rich area of controversy that has been thoroughly reviewed elsewhere (111).

Debate: what role do traditional CVD risk factors play in RA atherogenesis?

Across a number of studies, traditional cardiovascular risk factors, such as those considered in the Framingham risk equation (i.e., smoking, diabetes mellitus, hypertension, hyperlipidemia), are not substantially different in RA compared to control populations (8, 112, 113). This finding has led to an assertion that traditional cardiovascular risk factors may not account for the difference in CVD risk between RA patients and controls and that RA-associated factors (such as systemic inflammation) must be the sole contributors to the observed increase in cardiovascular risk. If this was the case, then interventions targeting cardiovascular risk factors may not have any ability to narrow the risk differential between RA patients and the general population. However, this assumes that the impact of cardiovascular risk factors on CVD is of equal magnitude in both RA and non-RA patients. While a number of studies have shown that traditional cardiovascular risk factors are indeed independently associated with cardiovascular events (114, 115) and subclinical atherosclerosis (35, 116, 117) in RA, few have explored the potential synergistic interaction between RA disease factors and traditional cardiovascular risk factors. In 1 study, del Rincon et al (118) demonstrated a synergism between systemic inflammation and CVD risk factors on measures of atherosclerosis, such that the magnitude of the association of ESR levels with IMT and carotid plaque was dependent on the number of traditional cardiovascular risk factors. There were no associations of ESR with carotid measures noted in RA patients without traditional risk factors.

Interestingly, some studies have suggested that traditional cardiovascular risk factors may even have smaller effects in RA (i.e., antagonistic interaction). For example, Gonzalez et al (114) demonstrated that male sex, smoking, and prior cardiovascular events imparted significantly less CVD event risk for RA patients compared to non-RA controls. Other CVD risk factors, such as insulin resistance and metabolic syndrome, may contribute to CVD in RA independent of traditional risk factors. For instance, RA patients were more likely to be insulin resistant than controls from the general population (117, 119, 120) and even patients with other autoimmune diseases (121), and RA patients with high-grade inflammation are more likely to be insulin resistant than those with low-grade inflammation (117). Insulin resistance and the metabolic syndrome have also been found to correlate with subclinical atherosclerosis in RA patients (36, 117, 122).

Altered adipocytokine profiles, such as higher leptin (123, 124), resistin (125), and adiponectin levels (123, 126), have been reported in RA. While there has been some inconsistency, particularly with regard to leptin (127–129), these findings are intriguing because adipocytokines appear to interact with inflammation, insulin resistance, atherosclerosis, and CVD risk in the general population (130, 131). However, the contribution of adipocytokines to atherosclerotic risk in RA needs to be elucidated. For example, Rho et al (132) found that higher concentrations of leptin were associated with an attenuation of the proatherosclerotic effects of insulin resistance on coronary calcification.

Therefore, although the prevalences of most traditional risk factors appear to be similar between RA and non-RA populations, the impact of each of these risk factors in RA is decidedly less certain and deserves further study. For nontraditional risk factors, both the prevalence and the effects are quite uncertain in RA, and most studies to date have employed cross-sectional designs, making it impossible to infer any cause-and-effect relationships. There is a clear need for longitudinal and interventional studies to elucidate the effects of these putative risk factors and determine the mechanisms by which they might increase risk.

Treatment

Debate: do biologic or glucocorticoid therapies modify CVD risk or atherogenesis in RA?

There has been great interest in the ability of the biologic DMARDs to ameliorate CVD risk, presumably via specific targeting of many of the cytokines implicated in atherogenesis and atherothrombosis. However, there has been substantial debate over the effect of this class of drugs (particularly TNF inhibitors) on CVD end points, subclinical markers, and risk factors.

There have been observational reports of decreased incidence of combined CVD events (133) and congestive heart failure (30, 52) in TNF inhibitor–treated RA patients compared to other therapies. In 1 study (134), the independent association of biologics on the risk of hospitalized cardiovascular events was similar to that of methotrexate. As interesting as these findings may be, other studies have yielded less compelling results. For instance, 1 study (135) found a lower risk of MI in responders to anti-TNF therapy compared to nonresponders but no actual difference between anti-TNF–treated and –untreated patients, and another study (52) found no difference in the risk of acute MI with anti-TNF therapy. Other studies have also yielded inconclusive results (136–138). Moreover, there have been studies documenting increased heart failure risk with TNF inhibitor treatment (139), and there have been sporadic cases of incident heart failure after commencement of TNF inhibitor therapy (140). These conflicting reports highlight the need for interventional studies in which the effects of exposures independent of the indication for treatment can be better elucidated.

Furthermore, studies of the association of TNF inhibition with measures of vascular structure and/or function (Table 3) and intermediate CVD end points have also yielded mixed results. There are reports documenting improved endothelial function (95, 141, 147) and decreased vascular stiffness (95, 146, 150), but also those reporting no difference or only temporary improvements in these measures (152). One study showed no difference in the progression of carotid IMT over 3 years among patients treated with TNF inhibitors versus those treated with conventional therapy (153), and other studies have confirmed this finding in the short term (147, 149). In contrast, Del Porto et al (144) showed a significant reduction in common carotid IMT at 12 months among responders to TNF inhibitors compared to nonresponders. Although a small study (n = 39), these prospective findings provide some suggestion that effective inhibition of TNFα in particular or inflammation in general could affect regression of atherosclerosis in some vascular beds.

Table 3. Interventional studies of TNF inhibitors and subclinical markers of cardiovascular disease in RA patients*
Author, year (ref.)AgentType of studyNo. of subjects (females)Study length, weeksAge, mean ± SD yearsRA duration, mean ± SD yearsMarkerImprovementComments
  1. * TNF = tumor necrosis factor; RA = rheumatoid arthritis; IFX = infliximab; FMD = flow-mediated endothelium-dependent vasodilatation; ETA = etanercept; ADA = adalimumab; AI = augmentation index; Ach-EDV = acetylcholine-induced endothelium-dependent vasodilatation; cc = common carotid; IMT = intima-media thickness; DAS44 = Disease Activity Score in 44 joints; ACR = American College of Rheumatology; MTX = methotrexate; HCQ = hydroxychloroquine; SSZ = sulfasalazine; DMARD = disease-modifying antirheumatic drug; skin Ach-EDV = iontophoretic Ach-EDV (microvascular function); RCT = randomized controlled trial; NR = not reported; PWV = pulse wave velocity; AS = ankylosing spondylitis; PsA = psoriatic arthritis; DAS28 = Disease Activity Score in 28 joints.

Hurlimann et al, 2002 (141)IFXOpen label11 (9)1246 ± 59 ± 2FMDYes 
Van Doornum et al, 2005 (142)ETA, ADA, IFXOpen label14 (8)655 ± 48 ± 1AINoThe study was powered to 0.8 for an average reduction in AI of 3.5%
Cardillo et al, 2006 (143)IFXOpen label10 (4)048 ± 150.6 (range 0.2–1.0)Ach-EDVYesThere was a control group of 10 healthy volunteers for comparing baseline Ach-EDV only
Del Porto et al, 2007 (144)ETA, IFXOpen label, controlled30 (25)5255 ± 97 ± 5cc IMTYesThere was a significant correlation between cc IMT and the DAS44. Control group consisted of RA patients who declined TNF inhibitors. The study only evaluated patients who responded to TNF inhibitor therapy by ACR criteria
Bosello et al, 2008 (145)IFXOpen label10 (9)1453 ± 813 ± 10FMDTransient onlyFMD improved after each infusion but returned to baseline by the next assessment time. There was a control group of 20 healthy volunteers for comparing baseline FMD only
Galarraga et al, 2009 (146)ETAOpen label, controlled26 (22)1657 ± 1114 ± 10AIYesControl group consisted of RA patients about to be prescribed MTX as part of clnical care, and had significantly shorter disease duration and disease severity
Sidiropoulos et al, 2009 (147)ADA, IFXOpen label, controlled12 (7)7254.8 ± 1513 ± 8FMD cc IMTYes (FMD only)cc IMT did not change. Control group consisted of patients starting MTX, HCQ, and SSZ as first DMARD, and had significantly shorter disease duration
Van Eijk et al, 2009 (148)ETAOpen label15 (5)440 ± 1011 ± 8Skin Ach-EDVYesThere was a control group of 12 healthy volunteers for comparing baseline EDV only
Wong et al, 2009 (149)IFXPost hoc analysis of RCT26 (22)5648 ± 12NRPWV cc IMTYes (PWV only)PWV was significantly lower after 56 weeks of treatment with IFX; cc IMT and plaque did not change. Duration of IFX was significantly associated with lower PWV on multivariate analysis
Angel et al, 2010 (150)ETA, ADA, IFXOpen label, controlled35 (18)1247 ± 1252 ± 15AI and PWVYes (PWV only)PWV was improved, but AI did not change. The study also included non-RA patients (AS and PsA). Control group consisted of RA, AS, or PsA patients meeting indication for TNF therapy but postponing their medication start because of a positive Mantoux test
Galarraga et al, 2010 (151)ETA, IFXOpen label, controlled30 (26)1657 ± 1154 ± 11Skin Ach-EDVNoControl group consisted of patients beginning MTX, and had significantly shorter disease duration and less severe disease at baseline. Treatment responders by the DAS28 (TNF + MTX groups) showed improved responses to Ach. Comparison between responders in TNF and MTX groups was impossible due to the small number of responders to MTX

Studies of the effects of TNF inhibitors on traditional CVD risk factors have focused primarily on lipid levels. A systematic review including 18 publications found that, on balance, lipid levels tended to increase, but the atherogenic index (total cholesterol/high-density lipoprotein cholesterol) tended to remain the same (154). Although there are fewer studies, RA patients treated with TNF inhibitors have shown at least temporary improvements in nontraditional risk factors, such as insulin resistance (155), number of circulating endothelial progenitor cells (156), levels of prothrombotic markers (157), adiponectin (158, 159), and levels of vascular adhesion molecules (160).

It is difficult to draw firm conclusions from these studies. First, there are no blinded, placebo-controlled trials assessing either hard or soft CVD end points. Many of the studies are underpowered and have very limited followup (weeks only), perhaps too short of a period to reach steady-state concentrations of the agents or expect the establishment of lasting biologic effects. Third, most studies enrolled patients according to standard of care. Biologic therapy is typically considered for more severe disease, increasing the risk of confounding by indication (and indeed, many studies of subclinical disease used control groups with significantly lower disease activity and/or duration). Moreover, it is also possible that the CVD end points may be more immutable in later stages of disease. Ideally, longer-term controlled trials with randomized treatment allocation are needed in order to accurately assess the effects of TNF inhibition on cardiovascular risk factors, atherogenesis, and hard CVD outcomes.

Fewer data are available for the effect of the non–TNF inhibitor biologics (i.e., anakinra, rituximab, abatacept, and tocilizumab) on cardiovascular risk factors and atherogenesis. Because the different biologics may have agent-specific effects on cardiovascular risk, separate research of each agent is warranted.

Data for glucocorticoids are sparser. In the general population, glucocorticoid excess is associated with obesity, insulin resistance, dyslipidemia, hypertension, atherosclerosis, and vascular remodeling (161). However, glucocorticoids also have antiinflammatory and antiproliferative properties, and administration has been associated with reduced atherogenesis and restenosis in animal models (162). Moreover, the inflammatory milieu of RA adds further complexity, and the problem of confounding by indication comes to the forefront in most studies. Some studies have found links between glucocorticoid use and increased plaque presence and arterial incompressibility in RA patients (163, 164). However, another study reported by Hafstrom et al (165) showed no significant difference in IMT, plaque prevalence, or flow-mediated dilatation between prednisolone-treated and -untreated patients after 5 years, although treated patients did have higher systolic blood pressures. At this point, it seems reasonable to use glucocorticoids only when necessary and at the lowest possible dose. For firmer conclusions to be made, a large, prospective, randomized placebo-controlled study will be needed.

Debate: should RA patients receive treatment with statins and aspirin to reduce CVD risk?

Recent reports have identified similarities between RA and diabetes mellitus with regard to CVD event rates (166) and the prevalence of subclinical atherosclerosis (110), prompting the consideration of RA as a diabetes mellitus risk equivalent. There is additional concern that RA patients may be undertreated with respect to traditional risk factors and that this may be a potential mechanism for their increased CVD risk. Toms et al (167) reported that approximately 50% of RA patients without clinical CVD or diabetes mellitus were dyslipidemic by National Cholesterol Education Program criteria, and depending on the risk stratification algorithm, between 2% and 26% were considered eligible for statin therapy. However, only 7% of patients identified as being high risk by 1 or more guidelines were receiving statin therapy, and interestingly, the majority of at-risk patients were only identified by 1 or 2 of the 4 methods used for stratification. Accordingly, the current European League Against Rheumatism (EULAR) guidelines for CVD risk management in RA call for annual cardiovascular risk assessment using locally adopted risk-stratification tools (e.g., Framingham) and intervention based on national guidelines for treatment (168). Acknowledging higher CVD event rates in RA, current EULAR guidelines recommend using a multiplier of 1.5 for RA patients meeting 2 or more of the following criteria: disease duration greater than 10 years, seropositivity, or extraarticular manifestations. However apt, the use of this multiplier has not been prospectively evaluated in RA patients, and validation studies are needed.

Pharmacotherapy for the primary prevention of CVD in other high-risk groups has focused on the hydroxymethylglutaryl-coenzyme A reductase inhibitors (statins) and low-dose aspirin. The antiinflammatory and cardioprotective effects of statins have been well documented in animal (169) and human (170) studies. In a landmark trial (Justification for the Use of Statins in Prevention: an Interventional Trial Evaluating Rosuvastatin; JUPITER) studying individuals without hyperlipidemia but with a CRP level of ≥2.0 mg/liter, Ridker et al (170) demonstrated a reduction in low-density lipoprotein cholesterol and CRP level by 50% and 37%, respectively, and a striking 56% reduction in CVD events with statin therapy compared to placebo.

To date, however, statins have only been studied in RA for their DMARD potential. In 2004, McCarey et al (171) reported the results of a double-blind, randomized, placebo-controlled trial of atorvastatin in RA. When compared to placebo after 6 months, atorvastatin-treated patients showed modest but significant reductions in the Disease Activity Score, ESR, and CRP level, along with beneficial effects on lipids. However, since there have been no published investigations into the effect of statins on CVD risk reduction in RA patients, it is unclear whether RA patients (especially those without hyperlipidemia) would incur the same reduction in CVD events observed in the JUPITER trial. In the absence of RA-specific data, the favorable safety profile of statins in RA patients (171) could support their use as a primary CVD prevention. However, the long-term cost-effectiveness of statins in RA patients without hyperlipidemia has been questioned (172). An interventional trial of atorvastatin addressing CVD end points in RA (Trial of Atorvastatin for the Primary Prevention of Cardiovascular Events in RA) is currently underway, and should provide some valuable clinical insight. Until these results are available, it would appear reasonable and safe to recommend statin therapy for all RA patients with hyperlipidemia, and possibly in those without hyperlipidemia who have persistently elevated inflammatory markers or other risk factors (i.e., long-term corticosteroid use). However, it is difficult to recommend a universal prescription of statins to all RA patients, particularly for those without hyperlipidemia and low levels of systemic inflammation, without specific investigation.

Aspirin, the prototypical antiplatelet agent, has consistently demonstrated a benefit in secondary CVD prevention (173, 174), and to a lesser extent for primary CVD prevention (175), in the general population. The benefit for primary prevention is greater in high-risk populations, such as individuals with diabetes mellitus (176). However, aspirin for primary or secondary prevention of CVD in RA has received little research. In the only study to explore aspirin use and CVD outcomes in RA, Linos et al (177) retrospectively explored CVD event rates in a cohort of RA patients according to exposure to aspirin. There was a possible benefit in men; however, small numbers and lack of random allocation to aspirin render these findings inconclusive. While it may seem logical to proceed with treatment, prolonged low-dose aspirin therapy comes with a risk of major bleeding. Across studies of the general population, the number needed to treat with aspirin to prevent a CVD event is approximately equivalent to the number needed to harm with a major bleeding episode (175). Equating CVD risk in RA with that of diabetes mellitus might suggest a favorable risk/benefit ratio in RA. However, bleeding risk is also higher in RA patients, even independent of the use of glucocorticoids and NSAIDs (178). Therefore, without specific demonstration of a favorable benefit to harm profile, aspirin for primary CVD prevention in RA patients is highly questionable, especially for those at low risk based on traditional risk factors.

Discussion

Elevated CVD risk is a significant public health problem that contributes greatly to the increased morbidity and shortened lifespan of individuals with RA. Over the past decades, there has been great progress into the understanding of the extent and severity of atherosclerosis and CVD risk in RA, but much work remains. It is particularly important to continue the search for mechanistic determinants and to delineate the apparent links between increased inflammation, traditional and nontraditional risk factors, atherosclerotic burden, and actual CVD events. There is a clear need for long-term interventional trials with CVD end points, especially for statins and biologic therapy.

Although there is immediate need for randomized efficacy studies of CVD prevention measures in RA, numerous challenges must be overcome in order to design and execute the longitudinal studies and interventional trials needed to address these questions. The high cost and the large numbers of patients required for studies of cardiovascular end points are substantial barriers. For example, the JUPITER trial enrolled 17,802 patients studied for a maximum of 5 years (170). Beyond cost, time, and number of patients, accounting for variations in disease characteristics and deciding when in the disease course to intervene adds further complexity. However arduous, the ability to ensure every RA patient the full extent of their productive lifespan is the potential reward for these endeavors.

AUTHOR CONTRIBUTIONS

All authors were involved in drafting the article or revising it critically for important intellectual content, and all authors approved the final version to be published. Dr. Giles 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 conception and design. Kramer, Giles.

Acquisition of data. Kramer, Giles.

Analysis and interpretation of data. Kramer, Giles.

Acknowledgements

We are indebted to Dr. Joan Bathon for her advice and guidance during the preparation of this manuscript.

Ancillary