Background Patients with rheumatoid arthritis (RA) are at increased risk for stroke. The particular impact of RA and its treatment in addition to known risk factors is not clear.
Objectives To examine the impact of RA on the risk of stroke taking risk factors of cardiovascular disease (CVD) and age into account.
Methods We performed a nested case-control study within the prospectively observed cohort of the German biologics register RABBIT. Since May 2001, RA patients are enrolled at start of a conventional synthetic (cs)DMARD or biologic (b)DMARD after ≥1 csDMARD failure. Rheumatologists report clinical characteristics, RA treatment, comorbidities and the respective therapy in a standardized manner. We considered all non-haemorrhagic strokes reported until April 2015 as cases. Two control patients, also observed in RABBIT, were matched to each case by an extensive algorithm  using sex, age, enrolment period, smoking status, body mass index and comorbidities at baseline (hypertension, heart failure, coronary disease or prior cerebrovascular event, diabetes). 16 cases could not be matched to a control. A multivariable Cox proportional hazard model with a random component (shared frailty model) was applied to address the matched design. Missing data were considered by multiple imputations.
Results The matching algorithm generated utmost agreement of cases and controls in baseline characteristics (Table 1). Compared to the remaining cohort, cases and controls were older and had more CVD. Similarity between cases and controls was also found regarding the treatment of comorbid osteoporosis and diabetes whereas the treatment of CVD was significantly different. This significant difference was confirmed in multivariable analysis. Compared to patients without prevalent CV comorbidities, patients with treated CVD had a moderately elevated risk for stroke (hazard ratio (HR): 1.44 [95% CI: 0.72, 2.89]), whereas patients with untreated CVD had a highly increased risk (HR: 4.27 [2.04, 8.94]. Hospitalization for a serious adverse event within six months before stroke (correspondent index date in controls) and time-varying log-transformed C-reactive protein were further predictors for stroke (HR: 2.42 [1.46, 4.02] and 1.25 [1.06, 1.47]). Treatment with TNF inhibitors or other bDMARDs was not associated with the development of stroke (HR: 0.86 [0.56, 1.31] and 1.04 [0.61, 1.75], Reference: csDMARDs).
Conclusions Within a strictly matched design providing the same baseline risk for stroke in cases and controls, we found that higher inflammation levels during follow-up and hospitalization were associated with a higher risk for the development of stroke. However, underlying CVD which is not medically treated was found to be the strongest risk factor for a future stroke in the RABBIT cohort. These results strengthen the need for a better CVD management in RA patients.
Disclosure RABBIT is supported by a joint, unconditional grant from AbbVie, Bristol-Myers Squibb, Celltrion, Hospira, MSD Sharp&Dohme, Pfizer, Roche, and UCB.
Hansen, Ben B., and Stephanie Olsen Klopfer. Optimal full matching and related designs via network flows. Journal of Computational and Graphical Statistics (2012).
Disclosure of Interest None declared