Risk of Venous Thromboembolism in Patients With Rheumatoid Arthritis

Authors

  • Seoyoung C. Kim,

    Corresponding author
    • Brigham and Women's Hospital, Boston, Massachusetts
    Search for more papers by this author
    • Dr. Kim has received research support from Takeda Pharmaceuticals North America and Pfizer and tuition support for the Pharmacoepidemiology Program at the Harvard School of Public Health funded by Pfizer and Asisa.

  • Sebastian Schneeweiss,

    1. Brigham and Women's Hospital, Boston, Massachusetts
    Search for more papers by this author
    • Dr. Schneeweiss has received consultant fees (more than $10,000 each) from WHISCON, PICORI, and Booz & Company.

  • Jun Liu,

    1. Brigham and Women's Hospital, Boston, Massachusetts
    Search for more papers by this author
  • Daniel H. Solomon

    1. Brigham and Women's Hospital, Boston, Massachusetts
    Search for more papers by this author
    • Dr. Solomon has received research support from Abbott Immunology, Amgen, and Lilly; serves in unpaid roles on trials sponsored by Pfizer and Novartis; and receives royalties (less than $5,000) for UpToDate.


1620 Tremont Street, Suite 3030, Boston, MA 02120. E-mail: skim62@partners.org

Abstract

Objective

Rheumatoid arthritis (RA) is associated with cardiovascular disease (CVD), but little is known about its association with another form of vascular disorder, venous thromboembolism (VTE).

Methods

A retrospective cohort study was conducted using US insurance claims. RA and non-RA patients were matched on age, sex, and index date. Incidence rates (IRs) and rate ratios (RRs) of VTE, defined as the composite of deep vein thrombosis (DVT) or pulmonary embolism (PE), were calculated. Cox proportional hazards models compared VTE risks between RA and non-RA patients, adjusting for VTE risk factors such as CVD, surgery, hospitalization, medications, and acute-phase reactants.

Results

Over the mean followup of 2 years, the IR for VTE among RA patients was 6.1 per 1,000 person-years, 2.4 times higher (95% confidence interval [95% CI] 2.1–2.8) than the rate of non-RA patients. The IRs for both DVT (RR 2.2, 95% CI 1.9–2.6) and PE (RR 2.7, 95% CI 2.2–3.5) were higher in RA patients compared with non-RA patients. After adjusting for risk factors of VTE, the VTE risk remained elevated in RA patients (hazard ratio 1.4, 95% CI 1.1–1.7) compared to non-RA patients. The result was similar after further adjustment for elevated acute-phase reactants (hazard ratio 1.5, 95% CI 0.3–6.5). One-third of patients who developed VTE had at least 1 major VTE risk factor 90 days before and after the VTE event.

Conclusion

Our results showed an increased risk of developing VTE for RA patients compared with non-RA patients. The risk was attenuated but remained elevated even after adjusting for various risk factors for VTE.

INTRODUCTION

Venous thromboembolism (VTE), which includes deep vein thrombosis (DVT) and pulmonary embolism (PE), is a major health problem and occurs in approximately 1 per 1,000 persons in the US ([1]). The incidence increases dramatically with age. Other known risk factors for VTE are fracture of the lower extremities, joint replacement surgery, major general surgery, major trauma, malignancy, heart or respiratory failure, pregnancy, history of VTE, hormone replacement therapy, and oral contraceptive use, but do not traditionally include inflammatory diseases such as rheumatoid arthritis (RA) ([2, 3]). The link between chronic systemic inflammatory diseases, such as RA, and cardiovascular disease (CVD), including myocardial infarction and stroke, has been well documented ([4-7]). Systemic inflammation may also play an important role in the development of VTE because inflammatory cytokines such as interleukin-6 (IL-6), IL-8, and tumor necrosis factor (TNF) could modulate thrombotic responses by activating coagulation pathways ([8, 9]). Markers of systemic inflammation, such as C-reactive protein (CRP) level, fibrinogen, and factor VIII, are also found at higher levels in patients with VTE, similar to atherothrombosis ([10, 11]).

Recent studies report that patients with RA have a 1.5- to 6-fold increased risk of VTE, such as PE and DVT, compared to non-RA patients ([12-18]). Many of these studies identified their RA cohort based on a hospital discharge diagnosis of RA, which could introduce a bias to select patients with severe RA and therefore may not be generalizable to typical RA patients seen in the outpatient setting ([13-16]). Furthermore, no prior studies have examined the VTE risk in RA adjusting for acute-phase reactant levels. The objectives of this study were 1) to examine the rate of incident VTE in a cohort of patients with RA compared with those without RA in the general population, 2) to assess the VTE risk in RA patients compared to non-RA patients adjusting for a number of known risk factors for VTE as well as baseline acute-phase reactant levels, and 3) to determine the proportion of VTE cases in the presence of major VTE risk factors such as recent hospitalization, surgery, and malignancy during the followup period.

Box 1. Significance & Innovations

  • Incident venous thromboembolism (VTE) occurred in 6 per 1,000 rheumatoid arthritis (RA) patient-years, 2.4 times higher than the rate of age-, sex-, and index date–matched patients without RA.
  • There was a 40% increased risk of developing VTE in RA patients compared with non-RA patients after adjusting for baseline VTE risk factors.
  • One-third of patients who developed VTE had at least 1 major VTE risk factor such as acute hospitalization, surgery, and malignancy diagnosis 90 days before and after the VTE event.

SUBJECTS AND METHODS

Data source

We conducted a cohort study using the claims data from a commercial US health plan that insures primarily working adults and their family members, and a small Medicare population for the period January 1, 2001 through June 30, 2008. This database contains longitudinal claims information, including medical diagnoses, procedures, hospitalizations, physician visits, and pharmacy dispensing, on more than 28 million fully-insured subscribers (with medical and pharmacy coverage) to 14 Blue Cross and Blue Shield health plans across the US. Results for outpatient laboratory tests, including CRP level and erythrocyte sedimentation rate (ESR), were available on a subset of beneficiaries. Personal identifiers were removed from the data set before the analysis to protect subject confidentiality. Patient informed consent was not required. The study protocol was approved by the Institutional Review Board of Brigham and Women's Hospital.

Study cohort

Adult patients who had at least 2 visits 7 days apart coded with the International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM) code 714.xx for RA were eligible for the RA cohort. The index date for the RA cohort was the date of the first dispensing of a disease-modifying antirheumatic drug (DMARD) after at least 12 months of continuous health plan eligibility; therefore, all persons in the RA cohort were required to have had 2 diagnoses and at least 1 filled prescription for a DMARD at the start of followup. A previous validation study showed that RA patients can be accurately identified using a combination of diagnosis codes and DMARD prescriptions in insurance claims data ([19]). Patients with 1 or more diagnostic codes of DVT or PE, solid tumors, hematologic malignancies, or myelodysplastic syndrome recorded in the 12-month period prior to the index dates were excluded. To ensure that we only included incident cases of VTE, subjects with claims for DVT or PE, or dispensings for anticoagulants in the 12-month period prior to the index dates, were also excluded from both cohorts (for the list of diagnosis codes for VTE and the names of anticoagulants that were excluded, see Supplementary Appendix A, available in the online version of this article at http://onlinelibrary.wiley.com/doi/10.1002/acr.22039/abstract).

The non-RA cohort, defined as adults who never had a diagnosis of RA during the study period, were identified and the aforementioned exclusion criteria were applied. The index date for the non-RA cohort was the date of the second physician visit after at least 12 months of continuous health plan eligibility. The non-RA cohort was then matched to the RA cohort on age, sex, and index date (±30 days) in a 4:1 ratio to mirror the health care utilization of the RA cohort. Patients in both cohorts were followed from the index date to the first of any of the following censoring events: development of DVT, PE, loss of health plan eligibility, end of study database, or death.

Outcome definition

The primary outcome was defined as a hospitalization for the composite end point of incident VTE (either PE or DVT) based on discharge diagnosis codes within the study database ([20, 21]). The secondary outcome was defined more strictly, with a combination of hospital discharge diagnosis codes for VTE and a prescription for anticoagulants within 10 days after the discharge date (see Supplementary Appendix A, available in the online version of this article at http://onlinelibrary.wiley.com/doi/10.1002/acr.22039/abstract).

Covariates

Variables potentially related to development of VTE were assessed using data from the 12 months before the index date. These variables included demographic factors (age and sex), comorbidities (hypertension, CVD, chronic obstructive pulmonary disease, diabetes mellitus, heart failure, chronic kidney disease, liver disease, malignancy, smoking, varicose veins, obesity, and history of various types of surgeries), medications (oral contraceptives and hormone replacement therapy), and health care utilization factors (the numbers of visits to any physicians and acute care hospitalizations and the number of different prescription medications). To quantify patients' comorbidities, we also calculated the Deyo-adapted Charlson Comorbidity Index based on the ICD-9-CM ([22, 23]). The comorbidity index is a summative score based on 19 major medical conditions, including myocardial infarction, congestive heart failure, peripheral vascular disease, dementia, cerebrovascular disease, pulmonary disease, renal disease, hepatic disease, diabetes mellitus, ulcer, connective tissue disease, cancer, and human immunodeficiency virus infection. A score of 0 represents absence of comorbidity and a higher score indicates a greater number of comorbid conditions. Outpatient laboratory data such as levels of acute-phase reactants (i.e., ESR or CRP level) at baseline were available in a subgroup of the study cohort.

In patients who developed VTE during the followup, major clinical risk factors or provoking factors for VTE such as malignancy 90 days before and after the VTE and hospitalization and surgeries 90 days before the date of VTE events were assessed ([24]).

Statistical analyses

We compared the baseline characteristics between RA and non-RA patients. We estimated the incidence rate (IR) of VTE events with the 95% confidence interval (95% CI), calculated as the number of subjects who had a hospitalization for either DVT or PE divided by the total person-time, in both the RA and non-RA cohorts. The rate ratio (RR) with the 95% CI was estimated by dividing the rate of the VTE events among RA patients by that of the non-RA patients ([25]). To adjust for potential confounders, separate Cox proportional hazards models were used to compare the rate of VTE events among RA patients with that in the non-RA cohort ([26]). The relationship between RA and the proportion of patients without VTE was plotted on a Kaplan-Meier curve. Additionally, we conducted subgroup analyses for the primary outcome in patients for whom we had an acute-phase reactant measurement at baseline. Among these subjects, the acute-phase reactant levels were used as a covariate in Cox proportional hazards models comparing the risk of hospitalization for VTE in RA patients with that in non-RA patients. All analyses were performed using SAS statistical software, version 9.2.

RESULTS

Cohort selection

There were initially 92,827 subjects with at least 1 RA diagnosis after a 12-month enrollment period and 920,697 subjects with no RA diagnosis at any time during the entire study period. Of the potential 44,978 RA patients with at least 2 diagnoses of RA at least 7 days apart, 25,065 patients had at least 1 prescription for DMARDs. A total of 700,647 patients without RA had at least 2 physician visits after a 12-month enrollment period. There were 6,138 patients with a history of VTE in the 12-month baseline period. After applying the exclusion criteria, we selected 4 non-RA patients matched to each RA patient on age, sex, and index date (±30 days). Our final study cohort includes 22,143 RA patients and 88,572 non-RA patients. The cohort selection process is shown in Figure 1. The mean ± SD followup time was 2.0 ± 1.6 years for RA patients and 2.0 ± 1.5 years for non-RA patients.

Figure 1.

Selection of the study cohort. The final study cohort included 22,143 rheumatoid arthritis (RA) patients and 88,572 non-RA patients matched on age, sex, and index date. DMARD = disease-modifying antirheumatic drug; VTE = venous thromboembolism.

Patient characteristics

Baseline characteristics of the age- and sex-matched cohorts were compared (Table 1). The mean age was 52 years and 75% were women in both cohorts. Substantial differences across almost all other baseline characteristics were observed between the cohorts, with the prevalence of comorbid conditions that may be related to VTE risks more common in RA patients than in non-RA subjects. A recorded diagnosis of smoking, hypertension, CVD, chronic obstructive pulmonary disease, and intraabdominal surgery; hormone replacement therapy and oral contraceptive use; use of oral steroids; and health care utilization, including physician visits, hospitalization, and number of prescription drugs, was more commonly noted in patients with RA than in non-RA patients. The mean ± SD length of hospitalization among those who had at least 1 hospitalization in the baseline 365-day period was 7.2 ± 12.0 days for the RA cohort and 5.9 ± 11.1 days for the non-RA cohort. Fifteen percent of RA patients and 1% of non-RA patients had their baseline acute-phase reactants measured (see Supplementary Appendix B, available in the online version of this article at http://onlinelibrary.wiley.com/doi/10.1002/acr.22039/abstract). Thirty-seven percent of RA patients and 18% of non-RA patients with a baseline acute-phase reactant level measurement available had elevated levels. The mean ± SD CRP level at baseline was 15.2 ± 28.1 mg/liter for RA patients and 3.4 ± 6.1 mg/liter for non-RA patients. The mean ± SD ESR at baseline was 19.3 ± 21.9 mm/hour for RA patients and 10.2 ± 14.0 mm/hour for non-RA patients.

Table 1. Baseline characteristics of the study cohort in the 12 months prior to the index date*
 RA (n = 22,143)Non-RA (n = 88,572)
  1. Values are the number (percentage) unless otherwise indicated. RA = rheumatoid arthritis; COPD = chronic obstructive pulmonary disease; APR = acute-phase reactant.
  2. aMatched.
  3. bThe proportion was calculated among the subjects with APR levels available.
Followup period, mean ± SD years2 ± 1.62 ± 1.5
Demographics  
Age, mean ± SD yearsa52.2 ± 1252.2 ± 12
Womena16,513 (75)66,052 (75)
Comorbidities  
Comorbidity index, mean ± SD1.2 ± 0.80.2 ± 0.6
Diabetes mellitus1,132 (5)3,827 (4)
Obesity531 (2)1,809 (2)
Smoking1,446 (7)3,614 (4)
Varicose veins180 (1)494 (1)
Chronic kidney disease333 (2)651 (1)
Liver disease604 (3)1,370 (2)
Hypertension6,747 (30)20,880 (24)
Cardiovascular disease1,496 (7)4,222 (5)
Stroke658 (3)1,841 (2)
COPD2,754 (12)6,467 (7)
Heart failure410 (2)915 (1)
Pregnancy253 (1)1,531 (2)
Hormone replacement therapy3,304 (15)10,234 (12)
Oral contraceptives1,805 (8)6,416 (7)
Oral steroids11,730 (53)5,948 (7)
Surgery, musculoskeletal1,095 (5)1,389 (2)
Surgery, cardiovascular501 (2)1,120 (1)
Surgery, intraabdominal1,465 (7)3,481 (4)
Health care utilization, mean ± SD  
No. of total physician visits10.4 ± 7.63.8 ± 4.4
No. of hospitalizations0.2 ± 0.70.1 ± 0.5
No. of prescription medications10.3 ± 6.44.4 ± 4.6
Laboratory data  
APR levels available3,313 (15)840 (1)
Elevated APR levelsb1,210 (37)149 (18)

Risk of VTE

During the study followup, 713 patients (0.6%) of the total study population were hospitalized for VTE. Of these patients with VTE, 42.8% received a dispensing for an anticoagulant within 10 days after hospital discharge and 47.1% received a dispensing for an anticoagulant within 30 days after hospital discharge. Among those who developed VTE events during the followup time, the mean ± SD length of hospitalization was 13.8 ± 19.1 days for the RA cohort and 12.9 ± 18.9 days for the non-RA cohort. As shown in Table 2, the IR of the primary outcome, hospitalization for VTE among RA patients, was 6.1 per 1,000 person-years and 2.4 times higher than that of non-RA patients (2.5 per 1,000 person-years). The age- and sex-adjusted RR of RA patients was 2.2 (95% CI 1.9–2.6) for experiencing DVT and 2.7 (95% CI 2.2–3.5) for PE compared to non-RA patients. Similar risks were observed for the secondary outcomes, defined by both diagnosis codes and dispensings of anticoagulants.

Table 2. Incidence rates and RRs of VTE (per 1,000 person-years) of age-, sex-, and index date–matched patients*
 RA (n = 22,143)Non-RA (n = 88,572)RR (95% CI)
Cases, no.Person-yearsRates (95% CI)Cases, no.Person-yearsRates (95% CI)
  1. RR = rate ratio; VTE = venous thromboembolism; RA = rheumatoid arthritis; 95% CI = 95% confidence interval; DVT = deep vein thrombosis; PE = pulmonary embolism.
  2. aBased on a discharge diagnosis.
  3. bBased on a discharge diagnosis and a dispensing of anticoagulants within 10 days after the discharge date.
Primary definitiona       
VTE (DVT or PE)26543,2786.1 (5.4–6.9)448176,8662.5 (2.3–2.8)2.4 (2.1–2.8)
DVT19743,3714.5 (4.0–5.2)364177,0182.1 (1.9–2.3)2.2 (1.9–2.6)
PE11143,4902.6 (2.1–3.1)164177,2760.9 (0.8–1.1)2.7 (2.2–3.5)
Secondary definitionb       
VTE (DVT or PE)11543,4622.7 (2.2–3.2)190177,2091.1 (0.9–1.2)2.5 (2.0–3.1)
DVT8043,5141.8 (1.5–2.3)156177,2840.9 (0.8–1.0)2.1 (1.6–2.7)
PE6143,5511.4 (1.1–1.8)84177,3800.5 (0.4–0.6)3.0 (2.1–4.1)

The age- and sex-adjusted hazard ratio (HR) of VTE for RA patients compared with non-RA patients was 2.5 (95% CI 2.2–2.9). After adjusting for all of the potential confounders of VTE, including demographic factors, comorbidities, medications, and health care utilization characteristics (listed in Table 1), the VTE risk remained elevated in RA patients (HR 1.4, 95% CI 1.1–1.7) compared to non-RA patients. The fully adjusted HR was 1.2 (95% CI 0.9–1.5) for DVT and 1.9 (95% CI 1.3–2.7) for PE. The HRs were very similar for the secondary definitions of VTE (Table 3). Figure 2 shows the Kaplan-Meier curves for hospitalization for VTE in RA and non-RA patients. A total of 84.2% of VTE events in RA patients and 85.5% of VTE events in non-RA patients occurred in the first 3 years of followup.

Table 3. Fully adjusted hazard ratios and 95% confidence intervals for VTE in RA patients compared to non-RA patients*
 RA vs. non-RA
  1. Adjusted for age, sex, comorbidity index, comorbid conditions, medications, and health care utilization patterns listed in Table 1. VTE = venous thromboembolism; RA = rheumatoid arthritis; DVT = deep vein thrombosis; PE = pulmonary embolism.
  2. aP < 0.05.
Primary definition 
VTE (DVT or PE)1.4 (1.1–1.7)a
DVT1.2 (0.9–1.5)
PE1.9 (1.3–2.7)a
Secondary definition 
VTE (DVT or PE)1.7 (1.2–2.3)a
DVT1.2 (0.8–1.7)
PE2.5 (1.6–4.2)a
Figure 2.

Kaplan-Meier survival curves for hospitalization for venous thromboembolism in patients with and without rheumatoid arthritis (RA). The numbers of subjects at risk and events were stratified by time since the index date.

Among 713 patients who developed incident VTE, 57% of RA patients and 60% of non-RA patients had a hospitalization, surgery, or new diagnosis of malignancy, which are considered major VTE risk factors, or provoking factors 90 days before the VTE event and a diagnosis of malignancy 90 days after the time of the VTE event (Table 4). Approximately one-third of those with incident VTE had at least 1 antecedent major risk factor in the 90 days prior to VTE events.

Table 4. Presence (%) of major risk factors for VTE among patients with incident VTE (n = 713) during the followup period*
 RA (n = 265)Non-RA (n = 448)
  1. VTE = venous thromboembolism; RA = rheumatoid arthritis.
  2. aProvoked VTE was defined as occurring in the presence of known malignancy 90 days before and after the VTE and acute hospitalization and surgery 90 days before the VTE ([24]).
90 days before the outcome  
Hospitalization2932
Any surgery3842
Malignancy3019
90 days after the outcome  
Malignancy1119
90 days before and after the outcome  
Provoked VTEa5760

Subgroup analyses on laboratory data

In a subgroup of patients (n = 4,153) with the baseline acute-phase reactant levels available, a total of 22 hospitalizations (0.5%) for VTE occurred. The multivariable HR for VTE associated with RA was 1.7 (95% CI 0.5–5.9), adjusted for age, sex, and elevated acute-phase reactants. After adjusting for age, sex, comorbidity index, and elevated acute-phase reactants, the multivariable HR was 1.3 (95% CI 0.4–4.4). The fully adjusted model that further adjusted for elevated acute-phase reactants showed an HR of 1.5 (95% CI 0.3–6.5) associated with RA.

DISCUSSION

The link between RA and VTE is not yet clearly understood. It has been thought that hypercoagulability is induced by active systemic inflammation and production of cytokines such as TNFα and IL-1. These inflammatory cytokines can lead to endothelial dysfunction, down-regulation of protein C, and inhibition of fibrinolysis ([8, 9, 27]). It is also possible that patients with RA have more risk factors of VTE such as acute hospitalization, surgical procedures, physical inactivity, and other cardiovascular comorbidities compared to non-RA patients. This study found that the occurrence of hospitalization for VTE in both patients with RA and non-RA patients is uncommon, but the risk of incident VTE is increased by 40% in patients with RA compared to non-RA patients in a fully adjusted analysis accounting for more than 20 different potential confounders. The results remain consistent even after adjusting for elevated acute-phase reactants, although they are not statistically significant due to the small number of events in this subgroup.

It is well known that VTE is associated with a number of risk factors, such as lower extremity fracture, joint replacement surgery, major general surgery, major trauma, spinal cord injury, congestive heart failure, hormone replacement therapy and oral contraceptive use, and older age ([2]). It is also known that nearly all patients who develop VTE have at least one recognized clinical factor associated with VTE, and the risk of VTE increases as the number of VTE risk factors increases ([2, 28, 29]). In the present study, 57% of RA patients and 60% of non-RA patients who developed VTE during the followup time had at least one major VTE risk factor, such as a recent hospitalization, surgery, or diagnosis of malignancy 90 days before or a diagnosis of malignancy 90 days after the date of the VTE events.

Several strengths of this study are worth noting. First, we examined a large cohort of RA and non-RA patients in a population that is representative of the US commercially- insured population. Second, although we mainly relied on diagnosis codes for RA and VTE that could potentially lead to exposure and outcome misclassification, the ICD diagnosis codes for both RA and VTE have been validated and used in a number of studies ([19, 20]). Moreover, the event rates observed in this study are consistent with prior studies: the IR of VTE in non-RA patients, 1 to 2 per 1,000 person-years, is consistent with prior studies ([1, 17, 18, 30, 31]), and the IR in RA patients was 6 per 1,000 person-years, which is similar to a previously reported IR using medical records to define the outcome ([12]). Third, we conducted a subgroup analysis of patients with baseline CRP level or ESR measured and noted a consistently elevated risk of VTE, although not statistically significant due to a smaller size of the subgroup, between the RA and non-RA cohorts.

Finally, unlike most published studies that examined the risk of VTE in RA patients based on a hospital diagnosis of RA ([13-16]), our study cohort was mainly selected based on outpatient RA diagnoses (94%) to be representative of RA patients seen in an ambulatory setting. Our results are also consistent with findings from a recently published cohort study using data from the UK outpatient medical records ([17]). Acute medical hospitalization is a known risk factor for VTE regardless of RA diagnosis ([32]). Furthermore, very active RA requiring acute hospitalization could be a potential confounder for the association between RA and VTE. A Swedish study based on patients admitted to the hospital for an autoimmune disease reported that PE is a serious problem in this population, and the standardized IR for RA was 6.0 (95% CI 5.6–6.4) in the first year after the hospital admission ([16]). Another Swedish population-based study also showed a higher rate of VTE in the first year following hospitalization in both RA patients (13.1 per 1,000 person-years) and the general population (2.4 per 1,000 person-years) ([18]). In a study from the US, the adjusted RR of VTE in hospitalized RA patients who did not have an orthopedic surgery was 2.0 (95% CI 2.0–2.0) compared to non-RA patients ([13]). Similarly, a Danish population-based study showed that the adjusted RR for VTE was 1.2 (95% CI 1.0–1.3) among patients with a history of hospitalization for RA ([15]).

There are, however, limitations to our study. First, this cohort study is likely subject to residual confounding by body mass index, immobility, severity of heart failure or chronic obstructive pulmonary disease, hereditary hypercoagulable conditions, and other unmeasured risk factors, including RA disease severity as well as surveillance bias ([2, 33]). A previous US population-based study of 464 RA patients and 464 non-RA patients, however, reported that there was no significant association between VTE risk and RA disease activity, such as the presence of rheumatoid factor, the presence of joint erosions or destructive changes, or rheumatoid nodules ([12]). We assessed a prior diagnosis of VTE, use of anticoagulants, and a number of variables potentially related to a future VTE event using the data from the 12 months prior to the index date, but this time period might not be long enough to capture all of the information on a preexisting diagnosis of VTE and/or potential confounders. We also used multivariable Cox models that were simultaneously adjusted for more than 20 risk factors of VTE, the comorbidity index, and health care utilization patterns to minimize the effect of such confounders.

Second, we relied on prescription dispensing records in the database to determine patients' medication exposures, including anticoagulants. This may not be the most accurate way to verify individuals' daily medication exposures, but it is considered one of the best ways to ascertain medication exposure status in nonexperimental settings ([34]). Lastly, this study was not designed to address a potential role of DMARDs, including TNFα inhibitors, in the risk of incident VTE among patients with RA. There are limited data suggesting an increased risk of VTE associated with the use of DMARDs ([12, 35, 36]). The exact effects of DMARDs, either traditional or biologic DMARDs, on the risk of VTE in RA patients should be further studied, given the large degree of potential for confounding by indication.

Third, because our primary outcomes were based on a hospital discharge diagnosis of VTE, not an admission diagnosis, some of the VTE cases might have occurred during the hospitalization for a different reason. In this study, mild cases of VTE treated in the outpatient setting were not included.

In conclusion, our results showed an increased risk of incident VTE, both DVT and PE, for RA patients compared with non-RA patients. The risk was attenuated but remained elevated after adjusting for known risk factors for VTE such as CVD, surgery, hospitalization, and medications and elevated acute-phase reactant levels. Future research is needed as to whether treatment with DMARDs could modify the risk of VTE in patients with RA.

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. Kim 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. Kim, Schneeweiss, Liu, Solomon.

Acquisition of data. Kim, Schneeweiss, Liu, Solomon.

Analysis and interpretation of data. Kim, Schneeweiss, Liu, Solomon.

Ancillary