Background Accurate diagnosis of systemic lupus erythematosus (SLE) remains challenging. Despite many autoantibodies being identified, none of the currently available molecular tests including anti-dsDNA and anti-nuclear antibody (ANA), provides sufficiently high sensitivity and specificity to support the diagnosis or classification of SLE in the absence of additional clinical data

Objectives The aim of this study was to identify novel autoantibodies associated with SLE and combine these with known biomarkers to develop an improved molecular test to support the diagnosis of SLE and assist in the exclusion of confounding diseases such as rheumatoid arthritis.

Methods Serum samples (case: n=276; healthy control: n=275; confounding diseases: n=92) were age/sex matched and analysed using Discovery Array v3.0 protein arrays (Oxford Gene Technology) that contain ~1600 correctly folded, full-length human proteins (Gnjatic et al (2009) J. Immunol. Methods 341:50). Autoantibodies were detected using anti-human IgG secondary antibody. Data for ANA and anti-dsDNA were generated using ELISA kits (Inova Quanta Lite). Antibodies indicative of SLE were identified and combinations of reactivities (“panels”) were developed using a suite of data classification methods. The performance of classifiers was determined by calculating the combined sensitivity and specificity using nested cross-validation to obviate over-fitting. The performance of individual biomarkers and panels was measured with respect to criteria including statistical significance, sensitivity, specificity, AUC and frequency of selection.

Results Data generated from a subset of samples (case: n=96; healthy controls: n=96) were screened for autoantibodies; panels of autoantibodies were then selected with the ability to discriminate between SLE and healthy individuals with a sensitivity and specificity of 74% and 78%, respectively (AUC = 0.86). Three independent validation studies were then undertaken using samples from patients and donors from (1) European (SLE: n=95; controls: n=86), (2) Afro-Caribbean (SLE: n=95; controls: n=99), and (3) a confounding diseases cohort (RA: n=68; connective tissue diseases: n=24). In parallel, ANA and anti-dsDNA assays were performed on all samples.

A total of 86 biomarkers were identified, representing a combination of previously identified and novel autoantibodies. A panel of biomarkers was identified which can distinguish SLE from control samples with sensitivity = 68%, specificity = 89% and an AUC = 0.84. This performance exceeds the literature performance of ANA (sensitivity = 93%; specificity = 57%). Inclusion of ANA and anti-dsDNA data reduced the specificity to 71% but increased the sensitivity to 85% with AUC = 0.88.

Conclusions A set of previously unknown autoantibodies and their cognate antigens have been identified in SLE. Using combinations of known and novel markers, panels were developed which can distinguish SLE from control samples with sensitivity = 85%, specificity = 71%. Panels comprising such biomarkers may provide additional information and offer an improvement over the existing criteria-based diagnosis.

Disclosure of Interest M. McAndrew Employee of: OGT, C. Wheeler Employee of: OGT, J. Koopmann Employee of: OGT, E. Uddin Employee of: OGT, M. Lewis: None Declared, T. Vyse: None Declared