Development and validation of SLESIS-R for prediction of serious infection in patients with SLE

Patients with systemic lupus erythematosus (SLE) have an elevated susceptibility to severe infections.¹ While several risk factors have been identified, a clinically useful risk scoring instrument is unavailable.¹

Tejera Segura et al. developed and validated an SLE Severe Infection Score (SLESIS) that incorporated seven predictors, including the Katz severity index.² However, the performance of SLESIS showed moderate results, achieving an area under the receiver operating characteristic curve (AUROC) of 0.63 at diagnosis and 0.79 at the time of infection.^1,2

Rua-Figueroa, et al. recently published an article in Lupus Science & Medicine on the development and validation of an improved revised SLESIS (SLESIS-R) score.¹ Here, we summarize the key findings.

Methods¹

• Data were collected from registro de lupus eritematoso sistémico de la Sociedad Española de Reumatología (RELESSER) prospective phase (RELESSER-PRO).
• A multivariable logistic model was developed, incorporating the variables already comprising the SLESIS score, along with all other potential predictors identified through a literature review.
• Each identified predictor was translated into a score item using logistic regression coefficients. Odds ratios were rounded to the nearest integer for simplification. The sum of these values formed the SLESIS-R, with performance assessed using AUROC.
• Cut-off point was chosen based on best validity parameters (sensitivity, specificity, and likelihood ratio).
• Internal validation was conducted using a 100-sample bootstrapping process.

Key findings¹

• The development cohort included 1,459 patients who had completed 1-year follow-up or experienced infections or mortality during the study period.
  • Mean patient age was 49 years, 90% were women, and 94% were Caucasian.
  • 1.7% of patients experienced ≥1 severe infection.
• According to the adjusted multivariate model, the predictors for severe infection in the following year of SLE were age ≥60 years, previous hospitalization related to SLE, previous serious infection, and glucocorticoid dose (Figure 1).
  • The Katz severity index was eventually excluded from this revised version.
• The model exhibited adequate performance, with 97.8% correct classification. The discrimination parameters showed an AUROC of 0.874 and the Hosmer-Lemeshow test showed adequate calibration (p = 0.932).

Figure 1. Multivariate predictive model for the development of SLESIS-R* 

CI, confidence interval; d, days; GC, glucocorticoid; SLE, systemic lupus erythematosus; SLESIS-R, SLE Severe Infection Score-Revised.
*Data from Rua-Figueroa, et al.¹

• During interval validation, the model indicated appropriate discrimination parameters, with a C statistic of 0.810.
• Figure 2 shows the final SLESIS-R with score ranging from 0 to 17. The AUROC was 0.861 (0.777–0.946).
  • The cut-off chosen was ≥6, which exhibited an accuracy of 85.9% and a positive likelihood ratio of 5.48.

Figure 2. SLESIS-R index calculator* 

SLE, systemic lupus erythematous.
*Adapted from Rua-Figueroa, et al.¹