Multivariate models* assessing the link between exposure to corticosteroids during the various time-windows and occurrence of severe infection.

*Adjusted for mucosal or internal bleeding at diagnosis, lung disease, kidney disease, cardiac disease, diabetes mellitus, exposure to rituximab in the six months before index date, to azathioprine, mycophenolate, cyclosporine and polyvalent immunoglobulin in the month before index date as well as splenectomy before index date. Age and disease duration until index date were neutralized by matching.†Index date: date of infection for cases.Abbreviations: 95% CI: 95% confidence interval; OR: odds ratio.Each line shows the adjusted result of exposure to corticosteroids with a given time-window period.