Predictors of infection requiring hospitalization in patients with systemic lupus erythematosus: a time-to-event analysis

Objectives: To evaluate the predictors of serious infection in patients with systemic lupus erythematosus (SLE). Methods: Serious infections were identified in SLE patients in a prospectively-followed single centre cohort. Associations of serious infection with disease-related variables and medication use were analysed using Cox and related regression models. Results: 346 patients were followed for a mean (SD) of 6.6 (3.7) years. 86 episodes of serious infection were observed, with an incidence rate of 3.8 episodes per 100 person-years. Patients who had serious infection had higher baseline SLE Damage Index (SDI) and Charlston Comorbidity Index (CCI); they were also more likely to have high disease activity status (HDAS), and higher disease activity in multiple clinical domains, higher flare rates, higher time-adjusted prednisolone dose exposure, and less time in lupus low disease activity state (LLDAS). Patients who have received cyclophosphamide, rituximab and mycophenolate were more likely to have experienced serious infection. After multivariable adjustment in Cox regression analysis, cyclophosphamide, higher SDI score, and higher disease activity were associated with an increased hazard of first serious infection. History of previous serious infection conferred the highest risk. Lymphopenia was also a modest but statistically significant predictor of serious infection. Conclusion: History of previous serious infection was the strongest predictor of serious infection in our SLE cohort. This study also suggests that clinical factors such as damage accrual, disease activity, and choice of immunosuppressant, can each have an independent risk in predicting serious infection particularly the first episode.