A Cohort Study to Assess Risk of Cutaneous Small Vessel Vasculitis Among Users of Different Oral Anticoagulants

Abstract

Purpose:

Cutaneous small vessel vasculitis (CSVV) was identified as a safety signal among patients treated with Direct Oral Anticoagulants (DOAC). This study aimed to determine if CSVV risk differed among patients with atrial fibrillation (Afib) who newly initiated warfarin or a DOAC.

Methods:

We identified enrollees aged ≥21 years diagnosed with Afib who newly initiated rivaroxaban, dabigatran, apixaban, and warfarin in the Sentinel Distributed Database from October 19, 2010, to February 29, 2020. We selected and followed patients who did not have evidence of the following in the 183 days prior to initiating treatment: CSVV diagnosis, dispensing of other study drugs, select autoimmune diseases or autoimmune medications, cancer diagnoses or chemotherapeutic treatment, kidney dialysis or transplant, alternative anticoagulation indications, or an institutional (nursing home, hospice, hospital) stay on the treatment initiation date (index date) until CSVV outcome or pre-specified censoring. We conducted 1:1 propensity score matching in six comparisons.

Results:

CSVV incidence rates for DOACs and warfarin ranged from 3.3 to 5.6 per 10,000 person years in our matched Afib population. The adjusted CSVV hazard ratio (HR) and 95% confidence interval (CI) was 0.94 (0.64, 1.39) for rivaroxaban vs. warfarin; 1.17 (0.67, 2.06) for dabigatran vs. warfarin; 0.85 (0.62, 1.16) for apixaban vs. warfarin; 0.86 (0.49, 1.50) for rivaroxaban vs. dabigatran; 0.99 (0.68, 1.45) for rivaroxaban vs. apixaban; and 1.70 (0.90, 3.21) for dabigatran vs. apixaban.

Conclusion:

We did not find significant evidence of differential CSVV risk in pair-wise comparisons of DOACs and warfarin.

Plain Language Summary

Cutaneous small vessel vasculitis (CSVV) has been reported among patients treated with oral anticoagulants. It is unknown if one oral anticoagulant carries a higher CSVV risk than others. The objective of this study is to investigate if CSVV risk differs among new users of different oral anticoagulants who have atrial fibrillation (Afib). We identified patients with Afib 21years and older in a distributed claims database who started using one of four oral anticoagulants during our study period. In addition, we required the patients to have no CSVV, autoimmune disease, cancer, kidney disease, or other reasons for anticoagulation besides Afib. We matched new users of each oral anticoagulant on their patient characteristics, followed them up, and compared their risk of CSVV to each other. Among new users of the four oral anticoagulants, incidence rate of CSVV ranged from 3.3 to 5.6 per 10,000 person years. The adjusted hazard ratios ranged from 0.86 to 1.70 with none being statistically significant. Although the incidence of CSVV was higher in our study population than previously reported, we did not find the risk of CSVV to be significantly different among matched new users of the four different oral anticoagulants.

Introduction

Vasculitis is an inflammatory process primarily affecting blood vessels resulting in the destruction of the vessel wall that could lead to hemorrhage, ischemia, and/or infarction.¹ Vasculitis is a heterogeneous group of rare diseases with many etiologies, and classification is typically based on the size of the affected blood vessels.¹ Cutaneous small vessel vasculitis (CSVV) is a form of vasculitis that is typically limited to the skin but may also be associated with extracutaneous disease (e.g., joint, kidney, gastrointestinal involvement) or systemic vasculitis. Morphologically, CSVV presents as a symmetric palpable purpura of the lower extremities but can sometimes involve skin on the trunk and upper extremities. Lesions typically present in crops and may be associated with itching, pain, a burning sensation, or may be surrounded by hemorrhagic bullae.

Clinical presentation such as urticarial lesions and ulcerative vesicles or nodules may be indicative of deeper or medium vessel involvement.² Patients presenting with the classical morphology of CSVV must be evaluated for systemic involvement.³ A delay or misdiagnosis can lead to systemic effects resulting from restricted blood flow to vital organs.⁴ CSVV diagnosis can be confirmed by skin biopsy ideally taken from a lesion present for 18 to 48 hours to avoid non-specific results. ^1,3

In the medical literature, CSVV has been described using multiple terms including leukocytoclastic vasculitis (LCV), hypersensitivity vasculitis, allergic vasculitis, drug-induced vasculitis, hypersensitivity angiitis, and cutaneous leukocytoclastic angiitis.^1,5 CSVV is rare with incidence estimated to be between 15 and 45 cases per million adults per year.^6,7 CSVV can be triggered by medications, infections, auto-immune disease, or malignancy. Therapeutic agents most often implicated in CSVV development are penicillins, cephalosporins, sulfonamides, phenytoin, and allopurinol.^8,9 CSVV can also be idiopathic when there is an absence of any causative factor.³

For drug-induced CSVV, symptoms typically occur within 7 to 10 days of drug exposure; however, cases occurring after longer durations of drug exposure have also been documented with various agents including warfarin and heparin.^10,11 Discontinuation of the implicated drug agent or treatment with corticosteroid and immunosuppressive agents have been effective in resolving CSVV in most cases.^4,12

Direct oral anticoagulants (DOACs) such as dabigatran, rivaroxaban, apixaban, edoxaban, and betrixaban are the preferred treatment for reducing the risk of stroke in patients with atrial fibrillation (Afib) and in the prevention and treatment of venous thromboembolism (VTE).^13,14 Vasculitis adverse events, including CSVV, were reported in pre-marketing trials for the approval of some DOACs, but incidence was low. For example, in the ROCKET-AF trial (7,111 patients treated with rivaroxaban) the incidence of cutaneous vasculitis was 0.01% with no cases in the warfarin arm (7,125 patients treated with warfarin).¹⁵ Both pre- and post-marketing data also documented a low prevalence (<1%) of mild hypersensitivity reactions such as rash and severe hypersensitivity reactions such as anaphylaxis and angioedema in patients treated with dabigatran, rivaroxaban, and apixaban.^16–18 Warfarin is labeled for vasculitis in the adverse reaction section of the prescribing information, but DOACs are not, despite labeling for other hypersensitivity reactions. This suggests that these drugs may induce an immune response in some patients.

In post-marketing data, continued case reports submitted to the U.S. Food and Drug Administration (FDA) Adverse Events Reporting System (FAERS) describing CSVV after DOAC exposure prompted a review of all CSVV cases reported to FAERS and published in the medical literature. Additional work was also undertaken to characterize CSVV among patients exposed to DOACs in the Sentinel Distributed Database (SDD). This descriptive work showed similar temporal relationships and clinical patient characteristics among CSVV cases reported to FAERS and in the SDD, suggesting a possible association between DOACs and CSVV.¹⁹

Although case reports have implicated oral anticoagulants generally as risk factors for CSVV, it is unknown if any individual anticoagulant confers a higher CSVV risk than the others. The objective of this study was to assess the adjusted comparative risk of CSVV among patients with atrial fibrillation (Afib) who newly initiated a DOAC (dabigatran, rivaroxaban, or apixaban) or warfarin in the SDD to determine if CSVV risk is differential across oral anticoagulants using a cohort design.

Methods

Data Source

This retrospective new-user cohort study was performed in the SDD. The SDD is a nationally representative distributed data network containing curated electronic health data. At the time of this study, the SDD comprised 16 participating health plans (i.e., data partners) and approximately 354 million unique patient identifiers in the United States.²⁰ FDA uses the SDD for safety signal refinement and regulatory decision-making. The FDA launched the Sentinel System to streamline epidemiologic research by combining analytic datasets that continuously meet rigorous quality assurance checks²¹ _and standard SAS programs that may be customized to meet the needs of each unique analysis. This infrastructure allows the FDA to rapidly conduct safety analyses and maximally leverage the Sentinel infrastructure to address emergent drug safety questions, while minimizing complexity and cost.

Study Design

We identified patients who were new users of rivaroxaban, dabigatran, apixaban, or warfarin aged 21 or older with at least 6 months of continuous medical and drug coverage prior to the first index dispensing between October 19, 2010, and February 29, 2020. We allowed gaps in coverage of up to 45 days. We did not include edoxaban and betrixaban since both were approved more recently and utilization during the study period was low. Eligible patients had a diagnosis of Afib in the 183 days prior to and including the index dispensing date. We excluded patients with evidence of an institutional stay or of any other study drug of interest dispensed on the index date or with evidence of any the following in the 183 days prior to and including the index dispensing date: 1) CSVV diagnosis of hypersensitivity angiitis, vasculitis of the skin, or allergic purpura in the ambulatory visit (AV), other ambulatory visit (OA), or emergency department (ED) care setting or steroid treatment for CSVV; 2) diagnosis of select autoimmune diseases including rheumatoid arthritis, lupus, Crohn’s disease, Sjogren’s syndrome, dermatomyositis, polymyositis, and cryoglobulinemia or treatment for these autoimmune disorders; 3) cancer diagnoses or chemotherapeutic treatment; 4) kidney dialysis or transplant; or 5) alternative anticoagulation indications, including deep vein thrombosis, pulmonary embolism, or joint replacement surgery prophylaxis. We followed patients from the index date until censoring which occurred when: 1) a gap in anticoagulant days of supply exceeded three days; 2) patients initiated any other DOAC or warfarin (for DOAC cohorts) or any DOAC (for warfarin cohorts); 3) study outcome occurred; or 4) patients were disenrolled, died, or data were no longer available.

Outcome Definition

We defined the outcome of CSVV as a diagnosis of hypersensitivity angiitis, vasculitis of the skin, or allergic purpura in the AV, OA, or ED care settings and a dispensing of an oral or topical steroid within 90 days of the diagnosis date. Prior to conducting this study, we used the Sentinel Patient Episode Profile Retrieval (PEPR) analytic tool (version 3.3.0) to randomly generate an individual-level line listing of the claims profile of 90 patients with a CSVV diagnosis in the Truven MarketScan database, a nationally representative commercial claims database that is approximately half the size of the SDD. Patients may appear in both databases, but the exact amount of overlap is unknown. For each potential CSVV case identified, two FDA epidemiologists and study co-authors independently reviewed a chronologic transcript of all diagnosis, procedure, and dispensing claims generated between the 90 days before and after the CSVV date to determine cases that were most likely to represent true DOAC-induced CSVV. The results of this manual review were used to define the study outcome of CSVV. The outcome date was defined as the date of CSVV diagnosis.

Statistical Method

We compared the risk of CSVV for the following six comparisons: 1) rivaroxaban vs. warfarin; 2) dabigatran vs. warfarin; 3) apixaban vs. warfarin; 4) rivaroxaban vs. dabigatran; 5) rivaroxaban vs. apixaban; and 6) dabigatran vs. apixaban. For each of the six comparisons, we fitted a logistic regression model within each Data Partner to estimate the probability of exposure, or propensity score, based on the following potential confounders: age (continuous variable), sex, autoimmune diseases, hematological blood disorders, viral infections, bacterial infections, anti-infectives, nonsteroidal anti-inflammatory drugs, psychoactive drugs, cardiovascular and diuretic drugs, beta-adrenergic receptor agonists, and anticonvulsants. A full list of factors included in the propensity score model has been published on the sentinel website.²² For each comparison, we 1:1 matched exposed patients and their comparators within each Data Partner based on their propensity score using a nearest neighbor matching algorithm with a caliper of 0.05. Cox proportional hazard regression model was used to estimate the hazard ratio and 95% confidence interval in the matched cohort after aggregation across Data Partners. The Cox model was stratified by Data Partner, which allowed the baseline hazard function to be different across Data Partners; however, the hazard ratios were assumed to be the same across Data Partners.^23–25

Ethical Approval

This FDA Sentinel System project is a public health surveillance activity conducted under the authority of the US Food and Drug Administration and Centers for Disease Control and Prevention and is accordingly not subject to Institutional Review Board oversight.²⁶

Results

In our study population of Afib patients, warfarin was the most used anticoagulant, followed by apixaban, rivaroxaban, and dabigatran. We identified 328,249 rivaroxaban new users (mean age ± standard deviation (SD): 73.3 ± 9.6 years), 142,328 dabigatran new users (72.7 ± 9.9 years), 532,973 apixaban new users (75.5 ± 9.6 years) and over 617,000 warfarin new users (75.4 ± 9.8 years) across all three warfarin pairwise comparisons. Prior to matching, new users of warfarin tended to be more Caucasian, older, more likely to have comorbidities such as hematological blood disorders, especially anemia and hemophilia, more likely to take cardiovascular and diuretic drugs, especially amiodarone, atenolol, and furosemide, and more likely to utilize healthcare compared to new users of rivaroxaban, dabigatran, and apixaban. New users of rivaroxaban, and dabigatran were generally similar with regards to their baseline demographic and clinical characteristics. Apixaban users were slightly older, more likely to have comorbidities, more likely to utilize healthcare services, especially inpatient and ambulatory encounters, than new users of rivaroxaban and dabigatran. After propensity score matching, the matched cohorts were balanced on all measured covariates. (Tables 1 and 2) Standardized mean differences for all variables before and after propensity score matching are provided in the supplemental tables.

Table 1.

Baseline Characteristics of New Initiators of Direct Oral Anticoagulants and Warfarin After Propensity Score Matching

	Rivaroxaban	Warfarin	SMD	Dabigatran	Warfarin	SMD	Apixaban	Warfarin	SMD
Number of Patients	320,363	320,363	-	142,197	142,197	-	503,885	503,885	-
Demographics, %
Age (years), mean*	73.8	73.8	0.001	72.8	72.8	−0.004	75.7	75.7	0.002
Age (years)
21–64	15.0	16.1	−0.035	19.2	19.5	−0.010	11.0	11.8	−0.028
65 – 74	33.8	36.2	0.053	37.2	35.6	0.035	35.3	32.8	0.052
75 – 84	32.6	33.6	−0.020	31.2	32.3	−0.024	34.8	36.7	−0.040
85 – 99	13.6	14.2	−0.015	12.3	12.6	−0.008	18.9	18.6	0.007
Sex
Female*	46.1	46.1	0.001	45.0	44.6	0.007	49.8	49.6	0.003
Male*	53.9	53.9	−0.001	55.0	55.4	−0.007	50.2	50.4	−0.003
Race
White	75.0	76.1	−0.035	73.5	73.9	−0.013	77.7	78.5	−0.027
Black	4.9	4.8	0.002	4.4	4.5	−0.007	5.2	5.1	0.004
Asian	1.4	1.1	0.025	2.5	2.3	0.018	1.3	1.1	0.018
American Indian	0.3	0.3	−0.007	0.3	0.4	−0.011	0.3	0.4	−0.013
Native Hawaiian	0.1	0.1	0.005	0.2	0.3	−0.012	0.1	0.1	0.007
Unknown	8.3	17.5	0.038	19.0	18.6	0.020	15.5	14.9	0.031
Hispanic Origin	1.4	1.2	0.015	2.1	2.1	−0.005	1.2	1.3	−0.006
Year
2010	-	4.3	-	1.2	4.3	−0.196	-	4.1	-
2011	0.1	16.9	−0.635	24.4	17.0	0.186	-	16.6	-
2012	5.3	15.9	−0.349	18.1	15.8	0.083	-	15.7	-
2013	12.1	14.4	−0.068	11.1	14.3	−0.086	1.9	14.3	−0.466
2014	15.3	12.2	0.089	8.3	12.3	−0.134	6.6	12.3	−0.195
2015	13.3	10.5	0.085	6.7	10.8	−0.148	11.8	10.6	0.036
2016	13.1	8.7	0.139	9.3	8.9	0.016	16.0	9.0	0.212
2017	15.4	7.1	0.266	8.8	7.1	0.062	16.7	7.3	0.292
2018	13.8	5.5	0.285	7.2	5.4	0.078	22.8	5.6	0.506
2019	11.7	4.4	0.270	3.9	4.0	−0.006	24.1	4.5	0.586
2020	1.1	0.5	0.070	0.1	0.4	−0.053	3.0	0.5	0.194
Baseline Comorbidities, %
Charlson Comorbidity Score, mean*	2.6	2.6	0.005	2.4	2.4	0.011	3.1	3.1	0.006
Autoimmune diseases*	5.1	5.2	0.001	5.3	4.8	0.025	5.6	5.6	0.002
Blood disorders*	19.1	19.1	0.000	17.2	16.7	0.014	23.8	23.6	0.005
Viral infections*	1.3	1.3	0.003	1.3	1.2	0.013	1.3	1.3	0.001
Chronic bacteremia*	0.8	0.8	−0.003	0.6	0.6	−0.004	1.1	1.1	0.001
Anti-infectives*	30.1	29.6	0.012	28.6	27.5	0.024	30.9	30.7	0.004
Non-steroidal anti-inflammatory drugs (NSAIDS)*	6.4	6.2	0.006	7.2	6.6	0.026	5.5	5.5	−0.000
Psychoactive drugs*	4.3	4.2	0.006	3.9	3.5	0.020	4.5	4.5	0.001
Cardiovascular and diuretic drugs*	61.6	61.6	0.001	62.6	62.7	−0.003	64.5	64.8	−0.007
Beta-adrenergic receptor agonists*	10.7	10.5	0.007	10.3	9.5	0.029	11.2	11.2	0.001
Anticonvulsants*	0.6	0.7	−0.004	0.8	0.7	0.010	0.6	0.6	−0.003
Baseline Health Utilization, mean
Number of ambulatory encounters	10.0	10.9	−0.102	9.4	9.9	−0.060	10.6	11.3	−0.084
Number of emergency room encounters	0.4	0.4	0.074	0.4	0.4	0.026	0.5	0.4	0.084
Number of inpatient hospital encounters	0.5	0.5	−0.062	0.5	0.5	−0.019	0.6	0.6	−0.045
Number of non-acute institutional encounters	0.1	0.2	−0.037	0.1	0.1	−0.044	0.2	0.2	−0.035
Number of other ambulatory encounters	4.8	5.6	−0.082	4.5	5.2	−0.074	5.9	6.8	−0.079
Number of unique drug classes	8.1	8.1	−0.005	7.9	7.8	0.026	8.5	8.5	−0.001
Number of generics	8.6	8.6	−0.002	8.4	8.3	0.032	9.0	9.0	0.001
Number of filled prescriptions	19.2	19.0	−0.003	19.0	17.8	0.083	19.7	20.1	−0.029

^*Included in the propensity score model

SMD = Standardized Mean Difference

Table 2:

Baseline Characteristics of New Initiators of Direct Oral Anticoagulants After Propensity Score Matching

	Rivaroxaban	Dabigatran	SMD	Rivaroxaban	Apixaban	SMD	Dabigatran	Apixaban	SMD
Number of Patients	125,338	125,338	-	331,796	331,796	-	125,718	125,718	-
Demographics, %
Age (years), mean*	73.1	73.1	−0.002	73.5	73.5	−0.006	73.2	73.3	−0.008
Age (years)
21–64	17.4	17.9	−0.013	16.5	16.4	0.002	17.8	17.2	0.019
65 – 74	38.2	37.1	0.023	38.0	38.2	−0.003	37.1	38.1	−0.022
75 – 84	31.5	32.2	−0.014	32.1	31.5	0.012	32.2	31.3	0.021
85 – 99	12.8	12.9	−0.002	13.4	13.9	−0.014	13.0	13.5	−0.015
Sex
Female*	45.5	45.7	−0.004	45.7	45.8	−0.003	45.7	45.9	−0.004
Male*	54.5	54.3	0.004	54.3	54.2	0.003	54.3	54.1	0.004
Race
White	72.5	73.1	−0.018	73.7	74.1	−0.013	73.1	72.9	0.006
Black	4.6	4.3	0.010	4.8	4.8	0.003	4.4	4.5	−0.008
Asian	1.4	1.5	−0.009	1.3	1.2	0.015	1.5	1.2	0.024
American Indian	0.3	0.3	0.004	0.3	0.3	0.007	0.3	0.3	0.000
Native Hawaiian	0.1	0.1	0.010	0.1	0.1	0.001	0.1	0.1	−0.011
Unknown	21.2	20.8	0.018	19.7	19.6	0.007	20.7	21.0	−0.011
Hispanic Origin	1.4	1.3	0.005	1.4	1.1	0.027	1.4	1.1	0.026
Year
2010	-	1.3	-	-	-	-	1.3	-	-
2011	0.1	27.6	−0.874	0.1	-	-	27.5	-	-
2012	5.5	21.6	−0.485	5.3	-	-	21.5	-	-
2013	12.2	13.1	−0.026	12.1	2.1	0.400	13.1	2.1	0.422
2014	15.4	9.0	0.195	15.4	6.8	0.275	9.1	6.9	0.080
2015	13.5	6.6	0.232	13.4	12.0	0.041	6.6	12.0	−0.188
2016	13.0	8.3	0.151	13.1	15.9	−0.079	8.4	15.9	−0.230
2017	15.2	6.5	0.284	15.3	16.6	−0.037	6.5	16.5	−0.317
2018	13.7	3.9	0.351	13.7	22.5	−0.231	3.9	22.6	−0.574
2019	11.4	2.1	0.379	11.5	23.8	−0.327	2.1	23.8	−0.685
2020	1.2	0.1	0.130	1.1	2.9	−0.132	0.1	2.9	−0.229
Baseline Comorbidities, %
Charlson Comorbidity Score, mean*	2.4	2.4	−0.011	2.6	2.6	0.007	2.4	2.4	0.013
Autoimmune diseases*	5.5	5.7	−0.009	5.1	5.0	0.005	5.7	5.3	0.015
Blood disorders*	18.1	18.6	−0.012	18.8	18.3	0.012	18.7	17.8	0.024
Viral infections*	1.3	1.3	−0.001	1.3	1.3	0.001	1.3	1.2	0.005
Chronic bacteremia*	0.5	0.6	−0.005	0.8	0.7	0.002	0.6	0.6	0.004
Anti-infectives*	28.9	29.4	−0.013	30.0	29.7	0.007	29.4	28.3	0.025
Non-steroidal anti-inflammatory drugs (NSAIDS)*	5.9	6.2	−0.010	6.6	6.5	0.004	6.2	5.8	0.016
Psychoactive drugs*	3.6	3.8	−0.008	4.3	4.2	0.004	3.8	3.5	0.011
Cardiovascular and diuretic drugs*	62.5	62.9	−0.007	61.0	61.0	0.000	63.0	62.7	0.006
Beta-adrenergic receptor agonists*	9.7	10.1	−0.012	10.6	10.5	0.005	10.1	9.8	0.011
Anticonvulsants*	0.8	0.8	0.000	0.6	0.6	0.003	0.8	0.8	0.000
Baseline Health Utilization, mean
Number of ambulatory encounters	9.8	10.0	−0.034	10.0	10.2	−0.023	10.0	9.9	0.012
Number of emergency room encounters	0.4	0.4	0.057	0.4	0.5	−0.018	0.4	0.4	−0.078
Number of inpatient hospital encounters	0.4	0.5	−0.062	0.5	0.5	−0.014	0.5	0.4	0.045
Number of non-acute institutional encounters	0.1	0.1	−0.002	0.1	0.1	−0.002	0.1	0.1	−0.003
Number of other ambulatory encounters	4.3	4.4	−0.016	4.7	4.7	0.002	4.4	4.3	0.014
Number of unique drug classes	7.9	8.1	−0.058	8.1	8.1	−0.000	8.1	7.9	0.057
Number of generics	8.3	8.6	−0.059	8.5	8.5	0.001	8.6	8.4	0.060
Number of filled prescriptions	18.6	20.2	−0.104	19.1	18.6	0.031	20.2	18.2	0.133

^*Included in the propensity score model

SMD = Standardized Mean Difference

CSVV crude incidence rates for the DOACs and warfarin ranged from 3.3 to 5.6 per 10,000-person years in our matched Afib population. The adjusted CSVV hazard ratio (HR) and 95% confidence interval (CI) was 0.94 (0.64, 1.39) for rivaroxaban vs. warfarin; 1.17 (0.67, 2.06) for dabigatran vs. warfarin; 0.85 (0.62, 1.16) for apixaban vs. warfarin; 0.86 (0.49, 1.50) for rivaroxaban vs. dabigatran; 0.99 (0.68, 1.45) for rivaroxaban vs. apixaban; and 1.70 (0.90, 3.21) for dabigatran vs. apixaban. (Table 3)

Table 3.

Propensity Score Matched Unconditional Analysis of Incidence of CSVV by DOACs and Warfarin

	Number of New Users	Number of Events	Incidence Rate per 10,000 Person Years	Hazard Ratio (95% CI)	P-Value
Rivaroxaban	320,363	53	4.1	0.94 (0.64, 1.39)	0.765
Warfarin	320,363	51	4.5
Dabigatran	142,197	26	5.1	1.17 (0.67, 2.06)	0.576
Warfarin	142,197	23	4.6
Apixaban	503,885	75	3.8	0.85 (0.62, 1.16)	0.298
Warfarin	503,885	82	4.6
Rivaroxaban	125,338	25	4.9	0.86 (0.49, 1.50)	0.586
Dabigatran	125,338	24	5.6
Rivaroxaban	331,796	55	4.1	0.99 (0.68, 1.45)	0.977
Apixaban	331,796	54	4.2
Dabigatran	125,718	24	5.6	1.70 (0.90, 3.21)	0.099
Apixaban	125,718	16	3.3

Discussion

To our knowledge, this is the first published study to comparatively assess CSVV risk among patients who newly initiated an oral anticoagulant. In our study, new users of rivaroxaban and dabigatran were similar with respect to demographics and clinical characteristics collected in the six-months baseline period prior to initiating a DOAC. While new users of warfarin and apixaban were slightly older and sicker, patient characteristics were balanced after propensity score matching. We did not find a statistically significant increase in CSVV risk for any propensity score matched comparisons between DOACs and warfarin, or between DOACs. However, we did observe a non-significant increased risk of CSVV for dabigatran compared to warfarin, apixaban, and rivaroxaban that may warrant further investigation.

In the premarketing phase 3 trials, the incidence of vasculitis adverse events was low (0.1– 0.2%) among DOAC users and similar in users of comparator drugs (i.e., warfarin or placebo). However, the incidence rates of CSVV among DOACs and warfarin initiators was lower and ranged from 3.3 to 5.6 per 10,000-person years in our matched Afib population. Few data are available on the incidence of vasculitis or drug-induced vasculitis in the literature. A retrospective population-based study reported the incidence of biopsy-proven LCV (age and sex-adjusted) to be 45 per million person-years (95% CI: 35 – 54).²⁷ Another study reported an annual incidence rate of biopsy proved drug associated LCV of 17.5 cases per million patients.²⁸. Compared to reported incidence rates in published observational studies, the estimated incidence rates in our study seem particularly high. Higher incidence of CSVV in our study population may be due to outcome misclassification, given that the positive predictive value of our outcome algorithm is unknown, and we counted all CSVV cases that occurred during the exposure of interest and did not limit patient follow-up to a short period immediately following the oral anticoagulants initiation. In addition, other observational studies used biopsy to confirm CSVV, which may underestimate the true incidence. Nevertheless, it is still reassuring to see that we did not see a differential CSVV risk between DOACs and warfarin in our propensity matched cohort.

AFib is the most common cardiac arrhythmia of clinical significance, occurring in 1.5 – 2% of the general population with prevalence increasing with age.^29–31 In most cases, AFib is a chronic condition, and most patients require life-long anticoagulant therapy.^14,32 DOACs are an improvement over warfarin for treatment of AFib as they can be given at fixed doses, have a more predictable anticoagulation profile, and do not require frequent laboratory monitoring. Moreover, they have a clinically acceptable risk of bleeding and an overall good safety profile. While mild and severe hypersensitivity reactions including CSVV have been reported in patients receiving DOACs, reports of life-threatening hypersensitivity reactions are still rare. Therefore, the impact of hypersensitivity reactions such as CSVV specifically on the risk-benefit profile of DOACs is minimal.

Despite the lack of differential CSVV risk between DOACs and warfarin or among DOACs in our study, a potential association between oral anticoagulants and CSVV cannot be completely ruled out. Therefore, early recognition of hypersensitivity reactions including CSVV is imperative because prompt discontinuation of the offending DOAC is typically followed by rapid recovery without sequelae. While, bleeding is still the most concerning adverse event associated with oral anticoagulants, rare severe hypersensitivity reactions to anticoagulants can also cause significant morbidity. Therefore, it’s important to raise awareness about non-bleeding related adverse events associated with DOACs that receive less attention and may go unrecognized.

Our findings should be interpreted in the context of certain limitations. First, CSVV is a rare event; therefore, it is possible that our study was underpowered to identify a significant differential association for specific DOACs, like dabigatran, despite our large cohort sizes. Second, our definition of CSVV was not validated, so the positive predictive value of the procedure and outcome codes are unknown. Based on a manual review of 90 CSVV cases in the Truven MarketScan dataset, we required that CSVV diagnosis be followed by an oral or topical dispensing of steroid treatment within 90 days of the qualifying CSVV diagnosis date. However, given unknown PPV, there is potential for non-differential outcome misclassification that could bias our estimate towards the null. Third, because drug related CSVV is an acute outcome which is expected to occur immediately after drug initiation, we might have included CSVV cases unrelated to oral anticoagulants initiation, since we did not limit follow-up to the immediate post-initiation period. This might bias our estimate towards the null. We chose not to limit patient follow-up due to the possibility of delayed hypersensitivity reaction evident from FAERS data that showed that CSVV can occur up to 547 days after oral anti-coagulant use, although median time to CSVV onset was 10 days.¹⁹ Fourth, given the regulatory nature of this work, limitations of the current sentinel system, and our objective to rapidly but rigorously assess a new and rare safety signal, some sensitivity analyses like assessing shorter time windows after oral anticoagulant use and limiting study period to when drugs being compared were both available were not conducted. These sensitivity analyses could have tested the robustness of our study findings, but we were concerned about potential loss of power with shorter time periods due to the rarity of CSVV. For this study, the regulatory question was to quantify the risk of CSVV after starting warfarin or a DOAC in a large, robust longitudinal database with denominator information to complement data available through FAERS and to rule out more than a 2-fold differential risk across the oral anticoagulants. Finally, given the observational nature of the analysis, residual confounding by unmeasured variables or variables not included in the propensity score (e.g., calendar year) may have biased the study results.

The objective of our study was to follow-up on an emerging and rare safety signal. Our study provides some assurance that risk of CSVV is likely low with real world use of DOACs. However, there is a possible elevated risk of CSVV with dabigatran compared to warfarin and other DOACs among Afib patients newly initiating oral anti-coagulants. Overall, the benefit-risk and safety profile of DOACs remain favorable for its intended indication.

Key Points.

• Cases of CSVV have been reported among patients treated with oral anticoagulants.

• This study aimed to determine if a differential CSVV risk exist for warfarin and the DOACs.

• We conducted 1:1 propensity score matching in six warfarin and DOAC comparisons among enrollees aged 21 years and older diagnosed with Afib who newly initiated warfarin or a DOAC with no prior CSVV diagnosis or treatment at baseline.

• Incidence rate of CSVV ranged from 3.3 to 5.6 per 10,000 person years in the matched population.

• We found no significant evidence of differential CSVV risk in the pair-wise comparisons of warfarin and DOACs.