Objectives Diagnostic reasoning in systemic lupus erythematosus (SLE) is a complex process reflecting the probability of disease at a given timepoint against competing diagnoses. We applied machine learning in well-characterised patient data sets to develop an algorithm that can aid SLE diagnosis. Methods From a discovery cohort of randomly selected 802 adults with SLE or control rheumatologic diseases, clinically selected panels of deconvoluted classification criteria and non-criteria features were analysed. Feature selection and model construction were done with Random Forests and Least Absolute Shrinkage and Selection Operator-logistic regression (LASSO-LR). The best model in 10-fold cross-validation was tested in a validation cohort (512 SLE, 143 disease controls). Results A novel LASSO-LR model had the best performance and included 14 variably weighed features with thrombocytopenia/haemolytic anaemia, malar/maculopapular rash, proteinuria, low C3 and C4, antinuclear antibodies (ANA) and immunologic disorder being the strongest SLE predictors. Our model produced SLE risk probabilities (depending on the combination of features) correlating positively with disease severity and organ damage, and allowing the unbiased classification of a validation cohort into diagnostic certainty levels (unlikely, possible, likely, definitive SLE) based on the likelihood of SLE against other diagnoses. Operating the model as binary (lupus/not-lupus), we noted excellent accuracy (94.8%) for identifying SLE, and high sensitivity for early disease (93.8%), nephritis (97.9%), neuropsychiatric (91.8%) and severe lupus requiring immunosuppressives/biologics (96.4%). This was converted into a scoring system, whereby a score >7 has 94.2% accuracy. Conclusions We have developed and validated an accurate, clinician-friendly algorithm based on classical disease features for early SLE diagnosis and treatment to improve patient outcomes.

Objectives Classification criteria are biased towards classifying long-standing disease. We compared the European League Against Rheumatism (EULAR)/American College of Rheumatology (ACR)-2019, Systemic Lupus International Collaborating Clinics (SLICC)-2012 and ACR-1997 criteria in an early (median 48 months) systemic lupus erythematosus (SLE) cohort. Methods Patients diagnosed with SLE (n=690) or control diseases (n=401). Sensitivity, specificity of the criteria and time-to-classification were calculated. Modified classification algorithms were derived from a random 80% and validated in the remaining 20% of the dataset running multiple iterations. Results At last assessment, sensitivities of ACR-1997, SLICC-2012 and EULAR/ACR-2019 criteria were 85.7%, 91.3% and 88.6%, with specificities 93.0%, 93.8% and 97.3%, respectively. Both SLICC and EULAR/ACR enabled earlier classification. Only 76.7% of patients with SLE met all three criteria suggesting non-overlapping groups. Notably, unclassified patients had high prevalence of British Isles Lupus Assessment Group moderate/severe manifestations (43.3%–60%) and SLICC/ACR organ damage (30%–50%). At diagnosis, criteria missed 25.6%–30.5% of patients. Modification of EULAR/ACR and SLICC algorithms to include hypocomplementaemia and/or positive anti-phospholipid antibodies as alternative entry criterion, and/or allow classification with fewer clinical criteria from multiple organs, increased their sensitivity at diagnosis (median 82.0% and 86.2%) and overall (93.7% and 97.1%) with modest decreases in specificity. Importantly, patients who were still missed by the modified criteria had lower incidence of major organ involvement, use of immunosuppressive/biological therapies and organ damage. Conclusions The SLICC and EULAR/ACR are more sensitive than the ACR and the EULAR/ACR criteria have superior specificity in early SLE, although patients with significant disease can be missed. Combination and/or modification of the classification algorithms may enhance their sensitivity, allowing earlier classification and treatment of more patients with high disease burden.

Ecancermedicalscience is the peer-reviewed open access cancer journal founded by the European Institute of Oncology in Milan. We consider articles on all aspects of research relating to cancer, including molecular biology, genetics, pathophysiology, epide