Early Clinical Features in Systemic Lupus Erythematosus: Can They Be Used to Achieve Earlier Diagnosis? A Risk Prediction Model

Abstract

Objective: To compare the primary care consulting behavior, prior to diagnosis, of people with systemic lupus erythematosus (SLE) with controls, and to develop and validate a risk prediction model to aid earlier SLE diagnosis. Methods: We included a total of 1,739 incident SLE cases practice-matched to 6,956 controls from the UK Clinical Practice Research Datalink. Using logistic regression, odds ratios were calculated for age, sex, consultation rates, selected presenting clinical features, and previous diagnoses in the 5 years preceding diagnosis date. A risk prediction model was developed from pre-selected variables using backward stepwise logistic regression. Model discrimination and calibration were tested in an independent validation cohort of 1,831,747 patients. Results: People with SLE had a significantly higher consultation rate than controls (median 9.2 versus 3.8 per year), which was in part attributable to clinical features that occur in SLE. The final risk prediction model included the variables age, sex, consultation rate, arthralgia or arthritis, rash, alopecia, sicca, Raynaud's phenomenon, serositis, and fatigue. The model discrimination and calibration in the validation sample was good (receiver operating characteristic curve 0.75, 95% confidence interval 0.73, 0.78). However, absolute risk predictions for SLE were typically less than 1% due to the rare nature of SLE. Conclusion: People with SLE consult their general practitioner more frequently and with clinical features attributable to SLE in the 5 years preceding diagnosis, suggesting that there are potential opportunities to reduce diagnostic delay in primary care. A risk prediction model was developed and validated that may be used to identify people at risk of SLE in future clinical practice.