Rheumatoid Arthritis and Incidence of Twelve Initial Presentations of Cardiovascular Disease: A Population Record-Linkage Cohort Study in England

Introduction

Studies have reported a 1.5 to 2-fold increase in risk of acute myocardial infarction (AMI), cerebrovascular disease[1–5] or cardiovascular mortality[6–8] in people with rheumatoid arthritis (RA). The absolute cardiovascular risk in individuals with this chronic inflammatory condition has been estimated to be similar to that in non-RA individuals who are approximately ten years older[9], and research suggests that the risk is similar to that seen in people with type 2 diabetes[10–12]. Both traditional risk factors and underlying disease may contribute to the increased cardiovascular risk[13]. Possible mechanisms include chronic immune dysregulation and systemic inflammation[14,15], induction of cardiovascular risk factors by medication (e.g. some disease-modifying antirheumatic drugs [DMARDs], oral corticosteroids or non-steroidal anti-inflammatory drugs), or common risk factors for RA and cardiovascular disease (CVD) such as genes[16] or smoking[17,18].

Previous studies have included only a limited range of cardiovascular endpoints, generally AMI and all-cause stroke, and most have been small or were undertaken prior to the widespread use of modern therapies for RA and to the introduction of an earlier and more aggressive pharmacological management[19].

The availability of linked electronic health records from primary care, hospital, disease registries and mortality through CALIBER (CArdiovascular research using LInked Bespoke studies and Electronic health Records)[20–23], offers the opportunity to investigate the association between RA and the initial presentation of specific types of CVDs in a contemporary population in England. The validity of both risk factor and disease measurements in CALIBER is supported by findings replicating and extending anticipated associations between phenotypic definitions for the 12 most common symptomatic CVDs and smoking[21], systolic and diastolic blood pressure[23], diabetes[24], and socio-economic deprivation[22].

The objectives of this analysis were to compare the rates of specific types of CVDs in individuals with and without RA and no prior history of CVD; and to investigate whether associations differ according to disease duration, sex, age, or presence of other autoimmune diseases in a contemporary population.

Methods

Results

Discussion

In this comprehensive and contemporary evaluation of the burden of CVDs associated with RA, we reported between thirty-six percent and two-fold higher rates of AMI, unheralded coronary death, heart failure, cardiac arrest and PAD in individuals diagnosed with the disease; but lack of association with unstable angina, cerebrovascular diseases or abdominal aortic aneurysm. Increased rates were found in both individuals with incident and prevalent RA, estimates’ size generally being similar to those observed in diabetic patients. Among individuals with incident RA, the highest CVD rates during follow-up were observed during the first year and after ten years of RA diagnosis. Rates were similar in men and women and findings were robust to multiple sensitivity analyses.

The strengths of this analysis are the population-based design of broad age span (18–99 years), large sample size (N = 133,311 in the main analysis), a cohort including all ranges of RA severity and information from individuals treated in recent years (1997–2010), and the study of the twelve most common clinical presentations of CVDs. Previous studies assessed only a limited range of cardiovascular endpoints, generally myocardial infarction and all-cause stroke, and most were small or were undertaken prior to the widespread use of modern therapies for rheumatoid arthritis and to the introduction of an earlier (i.e. within 3 months of onset) and more aggressive pharmacological management of patients with a combination of DMARDs and with short-courses of glucocorticoids for the management of flares. The cohort design and the unique data source linking EHRs covering primary care, hospital and mortality data allowed identification of a non-RA group selected from the same population and improved ascertainment of cardiovascular outcomes, which have been found to be valid[31,32] and were determined independently of RA status. Unlike previous studies, we were able to study not only the effect of adjusting for the presence of established cardiovascular risk factors but also for prescribed anti-inflammatory and cardiovascular medication currently used in clinical practice and for the presence of other diagnosed autoimmune diseases. We were also able to investigate in the same study differences in associations between patients with prevalent and incident RA, and according to RA disease duration, providing the most comprehensive picture of the burden of CVDs related to RA published to date.

One of the limitations of the study, inherent to the use of EHRs, was the possibility of RA diagnosis misclassification (e.g. mild or atypical RA patients might be less likely to be diagnosed and some diagnoses made in primary care could be inaccurate). However, we used the same diagnosis codes previously validated within the same or analogous medical databases and for which positive predictive values between 74% and 81% have been reported[27,28], and the primary analysis included only RA patients with supporting information for diagnosis. Typical DMARD medication or record of multiple diagnosis codes in primary care have been shown to increase the validity of RA diagnosis by general practitioners (≥80% sensitivity and specificity)[33]. Furthermore, although changes in clinical diagnosis and management over time could affect the estimates, ensuring that non-disease matched individuals were registered in the same practices at study inclusion minimised the introduction of this type of bias. Model adjustment for calendar-year had also negligible effects on the estimates (Figure C in S1 File).

In the cohort analysed the distribution of established cardiovascular risk factors was similar in individuals with or without RA, except for the higher prevalence of current or former smoking among RA patients. Smoking has been identified as a risk factor for the development of RA in previous studies[17] and for high antibody titres of rheumatoid factor and anti-citrullinated protein[18]. In our analysis, adjustment for the effect of smoking and other cardiovascular risk factors measured at baseline decreased the size of the estimates but did not completely explain the observed increased risk of CVDs in RA patients. However, residual confounding resulting from lack of information on known (e.g. rheumatoid factor status or lipids) or unknown risk factors cannot be completely excluded. Furthermore, precise information about time of diagnosis and the duration of and compliance with medication was not available and changes in medication use over time associated with RA flares were not investigated. The effect of anti-inflammatory drugs on CVD is complex. While reductions of cardiovascular risk could be mediated through their anti-inflammatory properties, they could also facilitate progression of atherosclerosis (e.g. corticosteroid could increase cardiovascular risk through effects on blood pressure, lipid profile and insulin resistance[34]). Indeed a meta-analysis of 280 trials of nonsteroidal anti-inflammatory drugs showed increased risk of coronary heart disease and death and heart failure associated with coxibs, diclofenac or ibuprofen use[35]. Alternatively, because oral corticosteroids and nonsteroidal anti-inflammatory drugs are generally prescribed when substantial joint inflammation is present, adjusting for anti-inflammatory medication could result in estimate underestimation. In our study, additionally adjusting for baseline prescription of anti-inflammatory medication further decreased the magnitude of the estimates but associations with myocardial infarction, heart failure and cardiac arrest remained statistically significant.

The duration of follow-up was limited, with a median time of four years per patient, but 2,125 patients with incident RA contributed with ten years or more of study follow-up to the analysis. The increased risk of CVD was observed at both, early stages of the disease and ten years or more after disease diagnosis, and in both patients with incident and prevalent disease. Results were robust to exclusion of patients who received oral corticosteroids and DMARDs or biologic therapy in the year before study entry. Although it is debated whether the timing of increase in risk of CVD is before or soon after the clinical onset of RA[18], our findings are consistent with smaller longitudinal studies reporting higher risk of coronary heart disease[36,37] and congestive heart failure[38] early after disease diagnosis. Our results suggest that the mechanism by which RA increases the cardiovascular risk is not exclusively related to a cumulative inflammatory burden but to other shared pathogenic mechanisms between the two types of disease[38].

The observed associations are consistent with results from meta-analyses that reported increased risk of fatal[6] and non-fatal[39] AMI, absence of relationship with stroke[3] and of sex-differences in these associations[39]. The lower risk of stable angina and the increased risk of initial presentation of CVD with unheralded coronary death in patients with RA are of clinical relevance and consistent with reports of a higher risk of undiagnosed mild vascular disease and unrecognised AMI and sudden cardiac death[40] in RA patients who suffer from chronic pain and frequently use analgesic drugs[18,40]. Although other less common mechanisms cannot be excluded (e.g. sensory disorders associated with neurological complications), these findings suggest that some deaths could be averted if patients are adequately informed of the high risk of unrecognised coronary disease and educated on how to recognise the symptoms. In our study, we did not find evidence of association between RA status and different types of cerebrovascular diseases, including ischaemic and haemorrhagic stroke. Evidence of the association between stroke and RA is conflicting and while a review reported increased risk of AMI but not stroke³, several meta-analyses found increased risk of both fatal and non-fatal endpoints[4,6,39]. Because higher risk of stroke was often reported in large studies and inception cohorts usually failed to find evidence of association, suggested reasons for discrepancies have been a shorter duration of RA disease and/or follow-up, less disease severity, differences in clinical management of patients (i.e. more aggressive in recent years) or age distributions in RA populations, lack of power, or absence of adjustment for the effect of cardiovascular risk factors. The numbers of events for some of the subtypes of stroke limited the power of some of the analyses but, in contrast to estimations for diabetes, we did not find evidence of association not only between RA (all RA patients, or patients with prevalent or incident RA) and specific types of stroke but also with a composite endpoint of all types of stroke investigated (data not shown) in a population with no prior history of CVD. Future analyses in larger populations and/or when longer follow-up data will be available will help to understand better the reasons for the observed lack of association with cerebrovascular endpoints.

The magnitude of cardiovascular risk estimates (similar to diabetes for several CVDs) has implications for clinical practice and supports the incorporation of regular determination and monitoring of cardiovascular risk early after RA diagnosis. While current ERC[41] and EULAR[42] guidelines recognise the importance of adequate clinical management of patients with RA to prevent CVDs, the equivalent USA guidelines, including new ACC/AHA recommendations[43], do not mention the need for cardiovascular assessment and management. Furthermore, although no intervention studies have been conducted to evaluate the effectiveness of implementing primary prevention therapy and monitoring of targets for RA patients, the study findings support the need for this type of research. Likewise, evidence of the association between RA and subsequent development of specific cardiovascular conditions other than AMI (e.g. heart failure and cardiac arrest) but a lack of relationship with others (e.g. cerebrovascular disease) in the general population has important implications for risk prediction, given that current recommended risk scores for patients with RA have been developed using angina, AMI and cerebrovascular endpoints[44].

In conclusion, regardless of the mechanisms for the observed associations with some but not all types of CVDs, our findings are important to inform both clinical practice and the choice of cardiovascular endpoints for risk prediction model development and trials targeted at patients with RA.

Supporting Information

Data Availability

Due to ethical and legal restrictions, some access restrictions apply to the data underlying the findings. The phenotype algorithms described in this paper are freely available via the CALIBER website at www.caliberresearch.org.

Funding Statement

This work was supported by grants from the Welcome Trust [WT 086091/Z/08/Z]; the UK National Institute for Health Research (RP-PG-0407-10314); and by awards establishing the Farr Institute of Health Informatics Research from the Medical Research Council (MR/K006584/1), in partnership with Arthritis Research UK, the British Heart Foundation, Cancer Research UK, the Economic and Social Research Council, the Engineering and Physical Sciences Research Council, the National Institute of Health Research, the National Institute for Social Care and Health Research (Welsh Assembly Government), the Chief Scientific Office (Scottish Government Health Directorates), and the Wellcome Trust. LS is supported by a Wellcome Trust Senior Research Fellowship in Clinical Science [098504/Z/12/Z]. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. The views and opinions expressed therein are those of the authors and do not necessarily reflect those of the National Institute for Health Research Public Health Research Programme or the Department of Health. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.