Venous thrombotic events are not increased in patients with rheumatoid arthritis treated with anti-TNF therapy: results from the British Society for Rheumatology Biologics Register.

OBJECTIVES: Past studies have reported conflicting rates of venous thrombotic events (VTEs) in rheumatoid arthritis (RA). The current study aimed to compare (1) the rates of VTEs in patients with RA treated with anti-tumour necrosis factor (anti-TNF) therapy versus those treated with non-biological disease-modifying antirheumatic drugs (nbDMARDs) alone and (2) the rates between each individual anti-TNF agent and nbDMARDs.
METHODS: Using data from the British Society for Rheumatology Biologics Register, a national prospective observational cohort study of biological safety in patients with RA, the authors compared the incidence of VTEs between 11 881 anti-TNF- and 3673 nbDMARD-treated patients. Analysis was limited to the first VTE per person. HRs were calculated using Cox modelling. Adjustment was made for potential confounders including surgery performed during follow-up.
RESULTS: A total of 196 first VTEs were reported (151 anti-TNF, 45 nbDMARD). Overall there was no difference in the rates of VTEs between anti-TNF- and nbDMARD-treated patients (adjusted HR 0.8 (95% CI 0.5 to 1.5)). The risk was similar across all anti-TNF agents. Rates of postoperative VTEs did not significantly differ between groups.
CONCLUSIONS: These data suggest that anti-TNF therapy is not associated with an increased risk of VTEs in RA patients.

Introduction

Rheumatoid arthritis (RA) is associated with increased mortality and co-morbidity compared with the general population.1 2 RA patients may have an increased risk of developing venous thrombotic events (VTEs), based on the high prevalence of many known VTE risk factors.3 4 Studies examining the incidence of VTEs in patients with RA have been discordant.5 6

The introduction of anti-tumour necrosis factor (anti-TNF) drugs for the treatment of RA has improved the outcomes of RA dramatically.7–9 However, there remain concerns about their long-term safety. Case reports10 11 and retrospective studies12–15 looking at VTEs in anti-TNF-treated RA patients have produced conflicting results.

To study this further, the current analysis aimed to compare (1) the rates of VTEs in RA patients treated with anti-TNF and non-biological disease-modifying antirheumatic drugs (nbDMARDs) and (2) the rates between the individual anti-TNF agents and nbDMARDs.

Methods

A full description of the methods relating to this analysis is available in our publication examining the risk of septic arthritis in this same cohort.16

In brief, the British Society for Rheumatology Biologics Register (BSRBR) is a national prospective cohort study that was established in 2001. Patients with active RA who were starting treatment with anti-TNF therapy were enrolled for observational follow-up. Three anti-TNF agents were currently in use during the study period analyses: etanercept (ETN), infliximab (INF) and adalimumab (ADA). A comparison cohort of RA patients with active disease currently receiving an nbDMARD was recruited and followed up in parallel.

Baseline information was collected regarding demographics, disease severity and co-morbidity. There were three sources of data collection during follow-up: consultant questionnaires, patient questionnaires and diaries, and the UK national cancer and death register (National Health Service Information Centre).

Adverse events from these sources were coded using MedDRA (the Medical Dictionary for Regulatory Activities).

Definition of outcome

This analysis limited outcomes to the first VTE per person. All events were verified by a BSRBR physician (JG) according to prespecified criteria (online supplementary table 1). VTEs were classified as postoperative if they occurred within 90 days of any surgery.

Statistical methods

Patients in the anti-TNF cohort were considered to be exposed from the date first treated with an anti-TNF drug until the first VTE, most recent follow-up, first missed dose of anti-TNF therapy or death. We did not include a 90-day lag window after the first missed dose in our primary analysis model as we felt that the effect of anti-TNF therapy upon thrombosis risk would cease after the drug had been eliminated from the body. For patients in the comparison cohort, follow-up started at the time of registration until first VTE, most recent follow-up form or death. Incidence rates of VTEs are presented as events per 1000 person years (pyrs) with 95% CIs. Survival analyses, performed using a Cox proportional hazards model, were used to compare the rates of VTEs between cohorts. Inverse probability of treatment weighting16 was used to adjust for confounding between the groups, including age and gender, disease severity (baseline 28-joint Disease Activity Score and Health Assessment Questionnaire), disease duration, year of entry into the study, use of baseline steroids, smoking status and co-morbidity (hypertension, chronic lung disease and diabetes). Surgery was entered into the model as a time-varying covariate, with patients considered to be at risk for 90 days postprocedure. An additional analysis evaluated the risk of postoperative VTEs using logistic regression to compare the risk of VTEs following surgery. Patients could contribute multiple surgeries to this analysis. Adjustment was made for the same confounders used in the main analysis. For all analyses, missing baseline data were replaced using multiple imputations.16 All analyses were performed using Stata V.10 software (StataCorp, College Station, Texas, USA).

Results

This analysis included 15 554 patients (11 881 anti-TNF, 3673 nbDMARD cohort). The anti-TNF cohort comprised 4139 patients starting ETN, 3475 patients starting INF and 4267 patients starting ADA. The baseline characteristics are displayed in table 1, which shows that the anti-TNF cohort was younger and comprised proportionally more women. The anti-TNF cohort also had more severe disease of a longer duration and had greater exposure to steroids at baseline.

Table 1

Baseline characteristics of the nbDMARD- and anti-TNF-treated patients

Characteristic	nbDMARD n=3673	Anti-TNF n=11 881	p Value	ETN n=4139	INF n=3475	ADA n=4267	p Value
Age, mean (SD)	60 (12)	56 (12)	0.0001	56 (12)	56 (12)	57 (12)	0.0184
Gender, % female	72	76	<0.0001	77	76	76	0.203
Disease duration, median (IQR) years	6 (1–15)	11 (6–19)	0.0001	12 (6–19)	12 (6–19)	10 (5–18)	0.0001
DAS28 score, mean (SD)	5.1 (1.3)	6.6 (1.0)	0.0001	6.6 (1.0)	6.6 (1.0)	6.5 (1.0)	0.0001
HAQ score, mean (SD)	1.5 (0.8)	2.0 (0.6)	0.0001	2.1 (0.6)	2.1 (0.5)	2.0 (0.6)	0.0001
Corticosteroids, no. (%)	837 (23)	5, 252 (44)	<0.0001	1979 (48)	1609 (46)	1664 (39)	<0.0001
Diabetes, no. (%)	247 (7)	685 (6)	0.033	255 (6)	169 (5)	261 (6)	0.026
Hypertension, no. (%)	1171 (32)	3563 (30)	0.041	1283 (31)	966 (28)	1314 (31)	0.002
Smoking history, no. (%)			0.001				0.056
Current smoker	868 (24)	2580 (22)		846 (21)	757 (22)	977 (24)
Former smoker	1454 (40)	4510 (38)		1576 (38)	1314 (38)	1620 (38)
Never smoked	1333 (36)	4714 (40)		1691 (41)	1386 (40)	1637 (39)

ADA, adalimumab; Anti-TNF, anti-tumour necrosis factor; DAS28, 28-joint Disease Activity Score; ETN, etanercept; HAQ, Health Assessment Questionnaire; INF, infliximab; nbDMARD, non-biological disease-modifying antirheumatic drug.

There were 196 verified VTEs (anti-TNF: 151, nbDMARD: 45) with an overall crude incident rate of 3.7 (95% CI 3.1 to 4.3)/1000 pyrs and 3.9 (95% CI 2.9 to 5.3)/1000 pyrs, respectively. The unadjusted HR was 1.1 (95% CI 0.8 to 1.6) and fully adjusted HR was 0.8 (95% CI 0.5 to 1.5) (table 2). We conducted sensitivity analyses to calculate HRs in those on drugs with a 90-day lag window, which did not alter the results (online supplementary table 2). Supplementary data are available detailing univariate predictors of VTEs and HRs for unimputed data (online supplementary tables 3–5).

Table 2

Crude incidence rates and hazard rates of verified first VTE in nbDMARD and anti-TNF-treated patients

	nbDMARD n=3673	Anti-TNF n=11 881	ETN n=4139	INF n=3475	ADA n=4267
Exposure (person years)	11 424	41 235	17 977	10 484	12 773
Total VTEs, n	45	151	60	49	42
VTE incidence rate/1000 person years (95% CI)	3.9 (2.9 to 5.3)	3.7 (3.1 to 4.3)	3.3 (2.5 to 4.3)	4.7 (3.5 6.2)	3.3 (2.4 to 4.4)
Unadjusted HR (95% CI)	Ref	1.1 (0.8 to 1.6)	1.0 (0.7 to 1.6)	1.4 (0.9 to 2.1)	1.0 (0.6 to 1.5)
Adjusted HR (95% CI)*	Ref	0.8 (0.5 to 1.5)	0.8† (0.4 to 1.4)	1.1† (0.6 to 1.9)	0.8† (0.4 to 1.4)

ADA, adalimumab; Anti-TNF, anti-tumour necrosis factor; ETN, etanercept; INF, infliximab; nbDMARD, non-biological disease-modifying antirheumatic drug; VTE, venous thrombotic event.

Inverse probability of treatment weighting model adjusted for age, gender, diabetes, baseline steroid exposure, smoking, hypertension, disease duration, disease severity, year of first anti-TNF drug and year of entry into study, with surgery as a time-varying covariate.

Differences between INF and ETN rates were non-significant (p=0.186). Differences between INF and ADA rates were non-significant (p=0.105).

When comparing the anti-TNF agents, the crude incidence of VTEs was highest in the INF group at 4.7 events per 1000 pyrs, compared with 3.3 for ADA- and ETN-treated patients. After adjusting for confounders, none of the hazard estimates for the individual anti-TNF agents differed significantly from the nbDMARD cohort (table 2). Analyses comparing individual anti-TNF agents with each other also showed no significant differences.

In total (allowing multiple surgeries per patient), 5584 surgical operations were reported during follow-up (table 3). Orthopaedic procedures accounted for 3948 (71%) of the surgeries performed. Twenty-one surgeries were complicated by a VTE (anti-TNF: 18/4572, nbDMARD: 3/1012). The fully adjusted OR for postoperative VTEs in the anti-TNF cohort compared with the nbDMARD cohort was 1.9 (95% CI 0.5 to 7.4) (table 3). A sensitivity analysis limited to orthopaedic procedures showed similar findings (results not shown).

Table 3

Surgery table

	nbDMARD	Anti-TNF	ETN	INF	ADA
Total surgeries (n)	1012	4572	2074	1170	1328
Total orthopaedic (n)	599	3349	1535	859	955
No. of patients having surgery (n)	699	2816	1184	743	889
Orthopaedic (n)	424	2104	908	546	650
Patients with more than one surgery (n)	227	1051	486	266	299
Orthopaedic (n)	129	726	334	178	214
No. postoperative VTEs (n)	3	18	6	6	6
Postop VTE crude rates (95% CI)	3.0 (0 to 6.3)	4.0 (2.1 to 5.8)	2.9 (0.6 to 5.2)	5.1 (1.0 to 9.3)	4.5 (0.9 to 8.2)
Unadjusted OR (95% CI)	Referent	1.3 (0.4 to 4.5)	–	–	–
Fully adjusted OR (95% CI)*	Referent	1.9 (0.5 to 7.4)	–	–	–

Adjusted for age, gender, diabetes, baseline steroid exposure, smoking, hypertension, disease duration, disease severity, year of first anti-TNF drug and year of entry into study.

ADA, adalimumab; Anti-TNF, anti-tumour necrosis factor; ETN, etanercept; INF, infliximab; nbDMARD, non-biological disease-modifying antirheumatic drug; VTE, venous thrombotic event.

Discussion

The data from this BSRBR cohort have shown that treatment with ETN, INF or ADA was not associated with a change in the risk of first VTE, either de novo or in the postoperative period.

This cohort had over 90% power to detect a doubling in the risk of VTEs but only 60% power to detect a 50% increase in the anti-TNF cohort compared with the DMARD cohort. Therefore, although we are confident that we can exclude a doubling in risk of VTEs with the TNF inhibitors studied here, differences of a smaller magnitude cannot be excluded. However, given the incidence of VTEs was 3.7/1000 pyrs in the nbDMARD cohort, smaller differences would be of questionable clinical relevance. For example, a 50% increase in relative risk would represent a small risk in absolute terms, equivalent to a number needed to treat one additional event of 540.

Comparing rates of events between the three anti-TNF cohorts must be done with caution. For example, it is important to acknowledge that differences exist relating to recruitment of patients to the anti-TNF cohorts by calendar year.17 This could be of particular relevance if we were to compare differences in the rates of postoperative VTEs when variations in rates could be partly explained by changes in the prescribing of anticoagulants following high-risk procedures. However, accepting the limitations of between-drug comparisons, our data show that the risk between ETN, INF and ADA for VTEs is broadly comparable.

The rates of postoperative VTEs in this cohort were reassuring, with approximately 4 out of every 1000 procedures being complicated by VTEs in all cohorts. Although the OR for postoperative VTEs for the anti-TNF cohort was 1.9, the 95% CI was wide and this result should not be interpreted as a significant finding. Research into this possible association could be explored further in countries with access to national registers of surgical procedures.

The main strengths of the BSRBR data have been detailed previously18 and relate to the prospective nature of the study, the large size of the cohort and the detailed data collection during follow-up. Several important limitations must also be considered when interpreting these results. The observational nature of the BSRBR predisposes to channelling bias, whereby patients with more severe disease are more likely to be enrolled into the anti-TNF cohort. Although we have made extensive efforts to adjust for this, we must acknowledge that it is likely some unmeasured confounding persists. In particular, we have only made adjustment for confounders that we record at baseline. Disease severity and steroid exposure are examples of variables that we would predict to change significantly during follow-up. Higher disease activity may well represent a prothrombotic state and would be likely to correlate to immobility; therefore, correct adjustment for this factor is clearly important. Further research is needed to establish how best we should take these factors into account.

In conclusion, this study found no evidence that anti-TNF therapy is associated with an increased risk of VTEs in RA patients, with no difference in risk observed between the anti-TNF drugs. Although this is a negative study, it has important implications. These results strengthen the evidence base and will help enable patients and healthcare professionals to make informed choices regarding therapy.