Article Text
PDF
Abstract
Background Recent hospital-based studies have suggested a sixfold increased risk of pulmonary embolism (PE) in rheumatoid arthritis (RA) in the year following admission. We evaluated the risk of PE and deep vein thrombosis (DVT) and associated time trend among RA patients (84.5% without a history of hospitalisation during the past year) derived from the general population.
Methods We conducted a cohort study using an electronic medical records database representative of the UK general population, collected from 1986 to 2010. Primary definitions of the RA cohort (exposure) and PE/DVT outcomes required physician diagnoses followed by corresponding treatments. We estimated relative risks (RRs) of PE and DVT compared with a matched non-RA comparison cohort, adjusting for age, sex, smoking, body mass index, comorbidities and hospitalisations.
Results Among 9589 individuals with RA (69% female, mean age of 58 years), 82 developed PE and 110 developed DVT (incidence rates, 1.5 and 2.1 per 1000 person-years). Compared with non-RA individuals (N=95 776), the age-, sex- and entry-time-matched RRs were 2.23 (95% CI 1.75 to 2.86) for PE and 2.20 (CI 1.78 to 2.71) for DVT. Adjusting for other covariates, the corresponding RRs were 2.16 (CI 1.68 to 2.79) and 2.16 (CI 1.74 to 2.69). The time-specific RRs for PE were 3.27, 1.88 and 2.35 for follow-up times of <1 year, 1–4.9 years, and ≥5 years, and corresponding RRs for DVT were 3.16, 1.82 and 2.32.
Conclusions This population-based study indicates an increased risk of PE and DVT in RA, supporting increased monitoring of venous-thromboembolic complications and risk factors in RA, regardless of hospitalisation.
- Rheumatoid Arthritis
- Cardiovascular Disease
- Inflammation
Statistics from Altmetric.com
Introduction
Rheumatoid arthritis (RA) is a chronic progressive disease associated with systemic inflammation. In addition to experiencing considerable associated morbidity, RA patients have a substantially shortened life expectancy,1–3 which is primarily driven by premature cardiovascular (CV) mortality.4 ,5 In searching for the underlying aetiology of this vascular impact, investigations to date have focused heavily on the arterial atherosclerotic CV complications (eg, increased risk of myocardial infarction (MI) or ischaemic stroke)6–8 and their precursors (eg, increased carotid artery wall thickness9–11). By comparison, there is a scarcity of data on the risk of venous thromboembolism (VTE) (ie, pulmonary embolism (PE) and deep vein thrombosis (DVT)) in RA, despite several mechanisms by which RA may increase the risk of PE and DVT. For example, systematic inflammation associated with RA may modulate thrombotic responses by upregulating procoagulants, downregulating anticoagulants and suppressing fibrinolysis.12 Furthermore, chronic inflammatory conditions, such as RA, are associated with endothelial dysfunction,13–15 and may have additional underlying VTE pathophysiology (eg, antiphospholipid syndrome or venulitis).
Indeed, several hospital-based studies have reported a substantially increased risk of VTE among hospitalised patients with these inflammatory conditions,16–18 particularly in the first year after admission for RA.16 For example, the risk of PE during the first year after admission for RA was found to be increased six times, which declined over time to 1.18 times or less by year 5 and later.16 It is unknown whether this increased risk and time trend are generalisable to non-hospitalised RA patients, who constitute the vast majority of these cases in contemporary care settings in developed countries. As PE represents a common CV event (the third most common after MI and stroke),19 and is associated with fatality (∼15% within 3 months), an accurate understanding of these risks is crucial to improving the overall outcomes of RA. To address these issues, we evaluated the risk of PE and DVT among RA patients (84.5% with no history of hospitalisation during the past year) in a general population context. We also evaluated the risk according to RA duration (follow-up period).
Methods
Data source
The Health Improvement Network (THIN) is a computerised medical record database entered by general practitioners (GP) in the UK. Data on approximately 7.3 million patients are systematically recorded and sent anonymously to THIN. Because the National Health Service in the UK requires every individual to be registered with a GP regardless of health status, THIN is a population-based cohort representative of the UK general population. The computerised data in THIN includes demographic information, details from GPs’ visits, diagnoses from specialists’ referrals and hospital admissions, results of laboratory tests, and additional health information recorded systematically including height, weight, blood pressure, smoking and vaccinations (http://csdmruk.cegedim.com/our-data/our-data.html). The Read codes (the standard clinical terminology system used in UK general practice) are used to record specific diagnoses,20 ,21 whereas the coding system from the Multilex classification (a widely used drug terminology system in the UK) is used to code drugs (http://www.firstdatabank.co.uk/8/multilex-drug-data-file). Health information is recorded on-site at each practice using a computerised system with quality control procedures to maintain high data completion rates and accuracy. Thus, THIN data represent routine medical practices in a population-based setting.22 The validity of THIN for pharmacoepidemiological research has been demonstrated,23 and the database has been successfully used for important epidemiologic and clinical studies (http://csdmruk.cegedim.com/THINBibliography.pdf), including those for rheumatic conditions.24–26
Study design and cohort definitions
We conducted cohort analyses of incident VTE (ie, PE or DVT) among individuals with incident RA (RA cohort) as compared with age-, sex- and entry-time-matched individuals without RA (comparison cohort) using data from THIN. Our primary definition for the RA cohort required one RA diagnostic code with at least one use of a disease modifying antirheumatic drug (DMARD) (see supplementary file for lists of RA diagnostic codes and DMARDs). This approach combining both diagnostic codes and DMARD use has been recommended to accurately identify RA cases in healthcare utilisation data.27 Continuous enrolment in the database for at least 12 months prior to the first diagnosis of RA was required. Individuals with any other inflammatory rheumatic conditions (THIN Read diagnostic codes corresponding to ICD-9-CM 274.x, 287.x, 446.x–447.x, 695.4, and 710.x–713.x)28 prior to the start of follow-up were excluded (see online supplementary file for the list of Read codes). This RA definition has been found to have a specificity of 96% (against 1987 American College of Rheumatology criteria for RA) in the UK General Practice Research Database (GPRD),29 which has 60% patients overlapping with THIN. Secondary definitions of the RA cohort were also used to assess the robustness of our results, including: (1) an RA definition based on a single diagnostic record for RA, and (2) at least two visits with RA diagnoses recorded at least 7 days apart by physicians in THIN.28 Participants with PE or DVT prior to RA diagnosis were excluded.
For the comparison cohorts corresponding to each RA cohort, we matched up to 10 individuals without RA to each RA case based on age, sex and calendar year of study entry. Like RA cohort members, comparison cohort members were excluded if they had a diagnosis of PE, DVT or any other inflammatory rheumatic conditions prior to the start of follow-up.
Our study cohorts spanned the period 1 January 1986 through 31 May 2010. Participants entered the cohort after all inclusion criteria had been met (eg, the first date of DMARD prescription in the primary RA cohort) or a matched doctor's visit in the comparison cohorts. Participants were followed until they experienced an outcome, died, or the follow-up ended (31 May 2010), whichever came first.
Ascertainment of PE and DVT
A patient was considered to be a PE or DVT case when he or she had a recorded code of PE or DVT and received anticoagulant therapy (heparin, warfarin sodium or a similar agent).30 Since VTE is a potentially fatal disease, we also included patients with a fatal outcome because a patient may have died before he or she could receive anticoagulation treatment.30 Under this assumption, patients with a recorded code of DVT or PE were included in the absence of recorded anticoagulant therapy if there was a fatal outcome within 1 month of diagnosis, regardless of whether autopsy results were available.30 This VTE definition has been found to have a confirmation rate of 94% in the UK GPRD.30
Assessment of covariates
Demographics, lifestyle factors, comorbid medical conditions, medication use and hospitalisation records were collected from the database before the start of follow-up. THIN lifestyle exposure variables have been collected prospectively and successfully utilised in previous analyses by confirming anticipated associations, such as those among alcohol, body mass index (BMI), and gout24; smoking and lung cancer31; smoking, obesity, and MI23; and alcohol and paroxysmal atrial fibrillation.32 We defined presence of comorbidities using the GP's diagnostic records by Read codes and drug use by prescription data. Hospitalisations during the year prior to cohort entry were also assessed to account for their potential impact on the relation between RA and VTE.16
Statistical analysis
We compared the baseline characteristics across the RA cohorts and comparison cohorts. We identified incident cases of DVT and PE during follow-up and calculated person-years to estimate an incidence rate (IR) for PE and DVT, both individually and in combination (ie, DVT or PE). We estimated the cumulative incidence of each event accounting for the competing risk of death.33 The IRs for RA cohorts were compared with those for control cohorts to calculate IR ratios. We employed Cox proportional hazard regression models to assess the multivariate relative risks (RRs) after stratifying by matched variables (age, sex and calendar year of study entry). We imputed missing values for BMI and smoking using IVEware within SAS, V.9.2 (SAS Institute, Cary, North Carolina, USA), which is a sequential regression approach,34 assuming that BMI and smoking were missing at random conditioning on all covariates included in the analysis. We imputed five datasets and then combined estimates from these datasets.35 Our multivariate analyses were further adjusted for BMI (five categories in kg/m2), smoking (non-smoker, current smoker and past smoker), CV conditions, atrial fibrillation, chronic obstructive pulmonary disease, cancer, inflammatory bowel disease, diabetes, chronic kidney disease (≥stage 3), varicose veins, surgery, fracture and hospitalisations. To evaluate the impact of follow-up time (ie, duration of RA), we estimated RRs during the first year, 1–4.9 years, and ≥5 years, as was similarly conducted in a recent hospitalisation-based study.16 We also performed subgroup analyses according to age (age <60 years vs ≥60 years) and sex. For all RRs, we calculated 95% CI. All p values were two-sided.
We performed a sensitivity analysis in which we restricted our cohort selection to those with incident RA cases in 1995 and later, which was similar to a recent analysis.36 Additional sensitivity analyses were performed after excluding those with missing lifestyle covariate values and those with a history of hospitalisation during the past year.
Results
Our primary analysis included 105 365 individuals for a combined follow-up time of 583 058 person-years, during which 440 cases of incident PE and 622 cases of incident DVT were diagnosed. Table 1 summarises the baseline characteristics of the two cohorts at the time of meeting the primary definition of RA in the RA cohort, and a matched time in the comparison cohort. Compared with the non-RA group, the RA group tended to have higher frequencies of overweight or obesity status, past or current smoking status, congestive heart failure, chronic obstructive pulmonary disease, cancer, inflammatory bowel disease, surgery and hospital admissions over the past year.
Characteristics of RA and comparison cohorts at baseline (RA onset)
RA was associated with an increased incidence of PE or DVT (table 2 and figure 1). Compared with non-RA patients, the age-, sex- and entry-time-matched RRs were 2.23 (95% CI 1.75 to 2.86) for PE and 2.20 (95% CI 1.78 to 2.71) for DVT. After further adjusting for BMI, smoking, comorbid conditions and hospitalisations, the corresponding RRs were 2.16 (95% CI 1.68 to 2.79) and 2.16 (95% CI 1.74 to 2.69) (table 2). These RRs persisted in our age and sex subgroup analyses, although the background rates of PE and DVT were substantially higher in the older age group (see supplementary table S1).
RR of Incident PE and DVT According to RA Status
Cumulative incidence of pulmonary embolism (upper panel) and deep vein thrombosis (lower panel) in the 9589 patients with incident rheumatoid arthritis (RA) as compared with the 95 776 age-, sex- entry-time-matched, non-RA subjects. This figure is only reproduced in colour in the online version.
When we repeated our analysis using secondary definitions of RA, the full multivariate RRs for PE and DVT were 1.77 (95% CI 1.46 to 2.15) and 1.76 (95% CI 1.49 to 2.07) using a single diagnostic record for RA (see supplementary tables S2 and S4) and the corresponding RRs were 1.84 (95% CI 1.37 to 2.48) and 1.90 (CI 1.51 to 2.41) using at least two visits with RA diagnoses.
When we evaluated the impact of follow-up time, the RRs for PE and DVT in RA patients as compared with non-RA patients were the largest in the first year (table 3). From the follow-up period of <1 year to ≥5 years, the age-, sex- and entry-time-matched RRs for PE were 3.27, 1.88 and 2.35 for the follow-up times of <1 year, 1–4.9 years and ≥5 years, respectively, and the corresponding RRs for DVT were 3.16, 1.82 and 2.32, respectively. (Extended covariate adjustment was not sought because there was no material impact of such adjustment in the main analysis (table 2) and the number of end-points was relatively small considering the number of covariates). When we repeated our analysis using a single diagnostic record for RA, the corresponding RRs were 2.17, 2.11 and 1.48 for PE, and 1.92, 1.84 and 1.67 for DVT (see supplementary table S5).
RR for PE and DVT in RA according to follow-up period
We performed several sensitivity analyses to further ensure the robustness of our results (see supplementary table S2). Similar results were found in analyses that (1) restricted our cohort selection to 1995 and later; (2) excluded individuals with missing covariate values; and (3) excluded individuals with hospitalisation during the past year (see supplementary table S2).
Discussion
Our objective was to evaluate the risk of PE and DVT in primarily outpatient RA patients in a general population context. In this large general practice cohort representative of the UK population, we found that RA, regardless of hospitalisation, was associated with increased risks of PE and DVT at an early stage of the disease. These RRs were slightly attenuated with time but remained significantly elevated at 5 years disease duration and thereafter. Overall, we observed over twofold increased risks of PE and DVT in RA patients compared with those without RA, independent of other potential confounders such as age, sex, smoking, BMI, CV-pulmonary-renal comorbidities, cancer, fracture and history of hospitalisation. These associations persisted across age and sex subgroups. These findings provide general population-based evidence that RA, regardless of hospitalisation, is associated with an increased risk of developing PE and DVT, and shed light on corresponding risk trends over the duration of RA primarily among outpatients.
Recent hospital-based studies have found an increased risk of VTE among RA patients,17 ,18 ,37 although these data may have been confounded by indication for hospitalisation, such as severe disease activity, RA complications/comorbidities, or procedures. For example, a nationwide Swedish hospitalisation follow-up study showed that the risk of PE during the first year after admission for RA was 5.99.16 This risk dramatically decreased over time after the first year to 1.12 at 10 years and later.16 It was somewhat unclear whether some of these hospital-based data may have had selection bias associated with hospitalisation differentially affecting the coprevalence of exposure (ie, RA) and outcome (ie, PE) as compared with that in the general population. Furthermore, it is unknown whether the observed increased risks are generalisable to non-hospitalised outpatients with RA, who constitute the vast majority of these cases in contemporary care settings in developed countries, as also reflected in our study cohort. To this end, a recent population-based inception RA cohort study from Olmstead County, Minnesota, USA, analysed 12 PE and 11 DVT cases that accrued from 464 RA patients (not limited to hospitalised RA cases) over 5.9 years.36 The study reported an age- and sex-matched RR of 3.6 for VTE (no individual RR for PE or DVT reported) compared with a comparison cohort of 464 non-RA individuals followed over 6.8 years. With these relatively small sample sizes, this report did not include fully adjusted effect estimates or trends over the RA durations.
Our general population-derived, primarily outpatient RA data expand on these limited existing data with larger sample sizes. The persisting association after our extensive adjustment of comorbidities associated with RA and history of hospitalisation in the current study indicates that the observed risk is unlikely to be confounded by these factors. Further, our time-trend data appear consistent with the higher RR of PE observed in the first year after hospitalisation for RA compared with the later years in the Swedish inpatient-based study,37 although the effect estimates were not as dramatic in our primarily outpatient RA context. A similar trend was observed with our DVT outcome analysis in addition to our combined analysis of PE and DVT. These data suggest that the biologic impact of having RA on the risk of PE and DVT may be higher at an early stage of RA (eg, in the first year following initiation of DMARDs or RA diagnosis), regardless of hospitalisation, perhaps related to uncontrolled inflammatory activity before the full benefit of antirheumatic therapy is achieved.37 Nevertheless, it is noteworthy that a significantly elevated risk persisted after 5 years of follow-up using several definitions of RA in our outpatient RA care setting.
There are several mechanisms by which RA could increase the risk of VTE. In 1856, Virchow proposed three precipitants for venous thrombosis: venous stasis, increased coagulability of the blood and damage to the vessel wall.38 ,39 RA and other inflammatory arthritic conditions could affect venous stasis or increase coagulability of the blood through decreased mobility and inflammation-associated mechanisms. For example, inflammation modulates thrombotic responses by upregulating procoagulants, downregulating anticoagulants and suppressing fibrinolysis.12 Furthermore, inflammation is a key determinant of endothelial function in both arteries and veins,14 ,15 ,40 ,41 and chronic inflammatory disorders are associated with endothelial dysfunction,14 ,15 including RA.13 RA patients might have additional underlying VTE pathophysiology, such as inflammatory damage to the vessel wall due to venulitis or presence of antiphospholipid antibodies. As the reported prevalence of antiphospholipid antibodies among RA patients has ranged from 5% to 75%,36 ,42 their contribution to the observed VTE risk is potentially substantial.
Strengths and limitations of our study deserve comment. This study was performed using a large UK general practice database; therefore, findings are likely to be generalisable to the general population. Uncertainty surrounding diagnostic accuracy is a potential concern in studies that identify cases from administrative databases. However, our databases are actually electronic medical records used for patients’ care, like the GPRD, and thus, the overall accuracy is expected to be higher as reflected in many validation studies of important outcomes,43–48 including those on our cohort definition (RA)49 and outcomes (PE and DVT).30 Furthermore, our results persisted in our sensitivity analyses that used several different definitions of RA. It is conceivable that the time-points used in our RA definition might have missed some relevant periods such as those associated with a high level of uncontrolled inflammation before clinical diagnosis of RA (after onset). Nevertheless, the highest observed RRs in the first year after our RA cohort entry time suggest that our definitions are relevantly close to the time period of interest for early RA. As the current study focused on the link and time trends between RA and the risk of PE and DVT, these data call for future studies aimed at identifying the risk factors for PE and DVT in RA patients, including RA-specific factors and potentially relevant drugs (eg, DMARDs, glucocorticoids and non-steroidal anti-inflammatory drugs). Of note, a recent analysis of the British Society for Rheumatology Biologics Register suggested that anti-TNF (tumour necrosis factor) therapy is not associated with the risk of VTE.50
In conclusion, this general population study suggests that RA is associated with increased risks of PE and DVT. These findings support increased vigilance in monitoring VTE complications and potential risk-factor intervention among RA patients, regardless of hospitalisation.
Acknowledgments
This work was supported in part by grants from the NIAMS (P60AR047785) and Boston University School of Medicine. The funding sources had no role in the design, conduct or reporting of the study, or in the decision to submit the manuscript for publication.
References
Supplementary materials
Supplementary Data
This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.
Files in this Data Supplement:
- Data supplement 1 - Online tables
Footnotes
-
Contributors All authors participated in the conception, design and analyses of the study. HKC and YZ drafted the manuscript and are guarantors. All authors contributed to interpretation of the results.
-
Funding This work was supported in part by grants from the NIAMS (P60AR047785) and Boston University School of Medicine. The funding sources had no role in the design, conduct or reporting of the study, or in the decision to submit the manuscript for publication.
-
Competing interests None.
-
Ethics approval The current study was approved by the Boston University Institutional Review Board and Multicenter Research Ethics Committee.
-
Provenance and peer review Not commissioned; externally peer reviewed.