Thrombosis causes substantial morbidity and mortality in systemic lupus erythematosus (SLE) and occurs with greater frequency and at a younger age than in the general population.1 Although several risk factors for thrombosis in SLE have been identified,2 these do not completely explain the thrombosis burden. Antiphospholipid antibody (aPL) positivity has been one of the most frequently studied risk factors for thrombosis in SLE.3^–5 These autoantibodies have been associated with an increased risk of venous and arterial thrombosis but the mechanism by which these antibodies increase risk is not well understood.3 Interestingly, in the general population, ethnicity also appears to influence thrombosis risk.2 Such ethnic differences have not been well-studied in SLE, although patients with SLE of certain ethnic subgroups experience worse outcomes.6

Our goal was to identify risk factors for thrombosis in SLE using a large and ethnically diverse study sample that has also been characterised for a wide range of potential risk factors.

PATIENTS AND METHODS

We studied 1930 patients from the UCSF Lupus Genetics Project, a multiethnic cohort of patients with SLE.7 The protocol was approved by the Institutional Review Board at the University of California, San Francisco (California, USA). Subjects were recruited from hospitals, community rheumatology clinics, lupus support groups, referral centres in northern California, and nationwide outreach efforts. SLE diagnosis was rigorously confirmed by medical record review according to the American College of Rheumatology (ACR) criteria.8 9 Baseline questionnaire data included demographic information, clinical characteristics and behavioural factors.

Covariates examined as potential predictor variables included ethnicity, smoking, age at SLE diagnosis, duration of SLE and immunomodulating medications (see table 1 for specific medications examined and other variable definitions). Medication history was collected by questionnaire, supplemented by medical record information and analysed as “ever” versus “never use”.

Medical records obtained from patients’ rheumatologists or other treating physicians were reviewed to document thromboses reported in patient questionnaires and identify any additional thromboses. Thrombosis events included deep vein thrombosis, pulmonary embolism, cerebral vascular accident, myocardial infarction, retinal vein thrombosis and miscarriage (three consecutive events in the first trimester or ⩾1 in the second or third trimesters). Medical records were also reviewed for results of aPL testing, including lupus anticoagulant (LAC) measured by Russell Viper Venom Time (RVVT) and anticardiolipin (ACL) IgG or IgM autoantibodies. Subjects were considered aPL positive if at least one of these autoantibodies were documented.

Explanatory variables potentially influencing thrombosis risk were examined for association in bivariate analysis using the χ² test (for categorical variables) and Student’s t-test (for continuous variables). These variables included aPL status, ethnicity, smoking, duration of disease, age at SLE diagnosis, medication use history and history of nephritis. In multivariate analyses, we used logistic regression to analyse associations between significant predictor variables identified from the preliminary bivariate analyses and risk of thrombosis. Odds ratios and 95% confidence intervals were calculated.

Subgroup analyses were then performed in which the explanatory variable aPL was considered separately as LAC and ACL (IgG or IgM) and immunomodulating medications were considered individually. Sensitivity analyses were performed to analyse venous and arterial thromboses separately. We also performed subgroup analyses to compare results when socio-economic status indicators (education and income) were included as covariates.

To further evaluate the protective effect of hydroxychloroquine and thrombosis risk observed in the current study, we used propensity scores to adjust for confounding by indication.10 11 All statistical analyses were performed using the STATA SE software, version 9.0 (StataCorp, College Station, Texas, USA).

RESULTS

Characteristics for the 1930 SLE study participants are shown in table 1. Four hundred and twenty-six subjects (22%) had at least one documented thrombosis, 119 (6%) subjects had two or more documented thromboses, and 516 subjects (27%) were aPL positive.

Explanatory variables associated with thrombosis in bivariate analyses are shown in table 2.

Explanatory variables found to increase the risk for thrombosis in the multivariate model (table 3) included smoking (OR 1.26, p = 0.011), nephritis (OR 1.35, p = 0.036), aPL positivity (OR 3.22, p<10⁻⁹), disease duration (OR 1.26 per 5 years with SLE, p = 0.027×10⁻⁷) and immunomodulating medications (OR 1.40, p = 0.011). We found that younger age at SLE diagnosis was protective for thrombosis in our multivariate model (OR 0.52, p = 0.001 for age ⩽20 years at diagnosis). Treatment with hydroxychloroquine was protective for thrombosis when adjusting for other explanatory variables (OR 0.67, p = 0.008). When the propensity scores were added as a covariate to the original model using logistic regression, the association between hydroxychloroquine use and thrombosis remained significantly protective (OR 0.62, p = 4.91×10⁻⁴).

Results of subgroup analyses indicated that history of treatment with azathioprine (OR 1.36, p = 0.023) and cyclophosphamide (OR 1.42, p = 0.025) were significant risk factors for thrombosis as was ACL positivity (OR 2.02, p = 0.037×10⁻⁴). Analysis of a subgroup of 222 patients with ACL titres available indicated that a high ACL titre was more strongly associated with thrombosis than low or intermediate titre (OR 2.08, p = 0.058). Further, results of sensitivity analyses indicated that African-American ethnicity showed a protective trend for venous thromboses (OR 0.49, p = 0.066), Asian-American ethnicity was specifically protective for deep vein thrombosis (OR 0.41, p = 0.047), and female gender was protective for pulmonary embolism (0.40, p = 0.022) and myocardial infarction (OR 0.49, p = 0.049). Finally, when subjects with multiple thromboses were compared with subjects who had no thrombosis, results were similar but with more pronounced ORs (table 3).

DISCUSSION

These findings confirm in a large, multi-ethnic SLE cohort that patients with aPL positivity, a history of smoking, immunomodulating medication use, longer disease duration and older age at SLE onset are more likely to have had at least one thrombosis. Importantly, these data are the first to confirm in a large and ethnically diverse study sample that hydroxychloroquine use is protective. Furthermore, these data suggest that Asian-American and African-American ethnicity as well as female gender may be protective for certain thrombosis subtypes, though these findings require confirmation in other study populations.

aPL positivity is a well-established risk factor for thrombosis. Love and Santoro found that 42% of LAC positive and 40% of patients with SLE who were ACL positive had a history of thrombosis.12 We found that, when controlling for other explanatory variables, subjects who were aPL positive had more than three times the odds of having an event compared with aPL-negative subjects, confirming previous research findings.

The first study to show that smoking was a significant risk factor for vascular events in SLE was by Toloza et al.4 Other studies that focused on cardiovascular thrombosis events in SLE demonstrated that older age at diagnosis and longer duration of disease also increase risk for thrombosis.4 13 Our study reinforces these findings.

That thrombosis risk may differ for African-American and Asian-American patients with SLE is a new finding. Because this study controlled for important risk factors and potential sources of confounding, it appears that Asian-Americans and African-American patients with SLE may have different, yet unidentified, risk factors for thrombosis compared with other ethnic groups. Explanations offered in previous investigations include the lower prevalence of known genetic risk factors (eg, factor V Leiden polymorphism) for ethnic groups other than Caucasians and the lower prevalence of obesity in Asian-Americans.2 Other risk factors that might differ between ethnic groups, such as smoking and disease severity, were controlled for in our study and, therefore, are unlikely to explain these differences in thrombosis rates. In addition, results of an analysis performed for a subgroup of 800 patients for which levels of income and education were accounted for were consistent with our main results (data not shown). However, potential false positive findings resulting from multiple testing and the smaller sample sizes in our sensitivity analyses necessitate confirmation of these results.

Our finding that hydroxychloroquine use is protective for thrombosis is intriguing. Hydroxychloroquine use has been suggested to offer protection from thrombosis in a few small previous studies.14^–16 Hydroxychloroquine has antithrombotic potential through multiple mechanisms, including inhibition of platelet aggregation and adhesion,17 cholesterol-lowering mechanisms17 and blockade of aPL production.18

Limitations of our study relate primarily to its retrospective nature. Because patients were recruited from a variety of settings, aPL were measured in different labs using a variety of assays. In addition, aPL were measured in the context of clinical care and not necessarily at the time of thrombosis. As a result, these autoantibodies may have been checked more frequently in patients who experienced a thrombosis (resulting in ascertainment bias). Although we did not have standardised measures of SLE severity or damage, we did have extensive information about disease severity that allowed us to account for this important factor in our analyses. We found that immunomodulating medications and hydroxychloroquine use were significantly associated with (or protective for) thrombosis risk. Further, our use of propensity scores supports a protective effect of hydroxychloroquine treatment as opposed to an association with mild disease.

In summary, this study confirms that certain risk factors suggested in previous, smaller, studies are significantly associated with thrombosis in patients with SLE and identifies additional risk factors that may influence this important SLE outcome. Additional work is needed to clarify ethnic differences in thrombosis in SLE, particularly explanatory variables for these differences. Importantly, this study also provides rigorous evidence that patients with known risk factors for thrombosis, such as aPL, might decrease their thrombosis risk by taking hydroxychloroquine; however, this association needs to be tested in prospective randomised controlled trials.