Background Infection is a major cause of morbidity and mortality in SLE patients. It would be helpful to have a tool to predict the risk that an individual patient with SLE will develop serious infection.
Objectives To develop a predictive risk calculator algorithm (SCORE) that assesses the probability of serious infection (i.e. leading to hospitalization) in SLE patients and to test it in an independent cohort.
Methods The SCORE was developed using data from the RELESSER (Spanish Society of Rheumatology Lupus Registry) cohort of 3658 SLE patients. A Cox regression model for repeated events (Andersen-Gill) was applied to assess which demographic and clinical factors were independently associated with increased risk of developing serious infection (Table 1). The SCORE was then validated using retrospective data from the UCLH (University College London Hospital) cohort including 699 SLE patients. Median SCORE values were compared between sub-groups of patients using the U-Mann-Whitney test.
Results Among 699 SLE UCLH patients, 98 (14%) developed serious infection. We compared these patients with 111 SLE controls who have never suffered serious infection. The characteristics of both groups are summarized in table 2. The infection group were more likely to have suffered previous infection (P=0.001) and/or hospitalized for SLE (P<0.001) and had renal and joint disease (P=0.005). Over a quarter of the infection group died from their infection. Median (Md) SCORE at diagnosis in SLE patients with infection was 4.27 (IQR 3.18) which was significantly higher than in the control group (Md 2.55, IQR 3.79) (z=3.341; P=0.0008). Md SCORE before infection in patients was 5.3 (IQR 3.68) which was significantly higher than SCORE at diagnosis (z=-5.733; P≤0.001) in those patients. By ROC analysis, we defined three possible cut-offs to distinguish patients with and without infection. The area under the ROC curve was 0.66 (CI 95% 0.56 to 0.71). A cut-off for SCORE at diagnosis >3.18 identified patients who would develop serious infection with sensitivity (S)76.5% and specificity (SPC) 50.5%. For SCORE >3.79, S was 64.3% and SPC 57.7%. For SCORE >4.24, S was 64.3% and SPC 60.4%.
Conclusions We have developed a SCORE for predicting risk of serious infection in SLE and validated it in an independent cohort. Given the potential mortality from such infections, this SCORE could be clinically useful though the moderate sensitivity and specificity necessitate caution and further prospective studies.
Disclosure of Interest None declared