Ten-year risk of cerebrovascular accidents in incident rheumatoid arthritis: a population-based study of trends over time.

Abstract

To evaluate secular trends in 10-year risk of incident cerebrovascular accidents (CVA), in incident RA relative to the general population. We conducted a retrospective study of a population-based incident cohort with RA onset from 1997 to 2004 in British Columbia, Canada, with matched general population controls (2:1), using administrative health data. RA and general population cohorts were divided according to year of RA onset, defined according to the first RA visit of the case definition. Incident CVA was defined as the first CVA occurring within 10 years from the first RA visit. Secular trend was assessed using delayed-entry Cox models with a two-way interaction term between the year of RA onset and indicator of RA vs general population. Linear, quadratic and spline functions of year of RA onset were compared with assess non-linear effects. The model with the lowest Akaike Information Criterion was selected. Overall, 23545 RA and 47090 general population experienced 658 and 1220 incident CVAs, respectively. A spline Cox model with a knot at year of onset 1999 was selected. A significant decline in risk of CVA was observed in individuals with RA onset after 1999 [0.90 (0.86, 0.95); P = 0.0001]. The change in CVA risk over time differed significantly in RA with onset from 1999 onwards compared with the general population (P-value of interaction term = 0.03), but not before 1999 (P = 0.06). Our findings suggest that people with RA onset from 1999 onwards, had a significantly greater decline in 10-year risk of CVA compared with the general population.