Development of a predictive model for systemic lupus erythematosus incidence risk based on environmental exposure factors.

Abstract

Objective: Systemic lupus erythematosus (SLE) is an autoimmune disease characterised by a loss of immune tolerance, affecting multiple organs and significantly impairing patients' health and quality of life. While hereditary elements are essential in the onset of SLE, external environmental influences are also significant. Currently, there are few predictive models for SLE that takes into account the impact of occupational and living environmental exposures. Therefore, we collected basic information, occupational background and living environmental exposure data from patients with SLE to construct a predictive model that facilitates easier intervention.

Methods: We conducted a study comparing 316 individuals diagnosed with SLE and 851 healthy volunteers in a case-control design, collecting their basic information, occupational exposure history and environmental exposure data. Subjects were randomly allocated into training and validation groups using a 70/30 split. Using three-feature selection methods, we constructed four predictive models with multivariate logistic regression. Model performance and clinical utility were evaluated via receiver operating characteristic, calibration and decision curves. Leave-one-out cross-validation further validated the models. The best model was used to create a dynamic nomogram, visually representing the predicted relative risk of SLE onset.

Results: The ForestMDG model demonstrated strong predictive ability, with an area under the curve of 0.903 (95% CI 0.880 to 0.925) in the training set and 0.851 (95% CI 0.809 to 0.894) in the validation set, as indicated by model performance evaluation. Calibration and decision curves demonstrated accurate results along with practical clinical value. Leave-one-out cross-validation confirmed that the ForestMDG model had the best accuracy (0.8338). Finally, we developed a dynamic nomogram for practical use, which is accessible via the following link: https://yingzhang99321.shinyapps.io/dynnomapp/.

Conclusion: We created a user-friendly dynamic nomogram for predicting the relative risk of SLE onset based on occupational and living environmental exposures.

Trial registration number: ChiCTR2000038187.