Osteonecrosis in individuals with systemic lupus erythematosus: A predictive model

Area of Science:

• Rheumatology
• Pharmacology
• Data Science

Background:

• Osteonecrosis (ON) is a serious complication in patients with systemic lupus erythematosus (SLE).
• Predictive models for ON in SLE patients are crucial for early intervention and management.
• Pharmacological, demographic, and psychoactive factors may influence ON development in SLE.

Purpose of the Study:

• To develop and validate a predictive model for osteonecrosis (ON) in individuals with systemic lupus erythematosus (SLE).
• To identify key pharmacological, demographic, and psychoactive factors associated with ON development in SLE patients.

Main Methods:

• A literature review informed a survey administered to SLE patients in Chile.
• Data from 46 de-identified records were analyzed using Bayesian logistic regression (non-informative and informative priors) and a random forest model.
• All models underwent cross-validation to assess performance.

Main Results:

• Mean daily corticosteroid dosage and tobacco use were identified as significant predictors of ON.
• The random forest model demonstrated good accuracy and sensitivity but limited specificity.
• Bayesian logistic regression with informative priors improved model specificity.

Conclusions:

• Corticosteroid use and tobacco consumption are significant factors for predicting ON in SLE.
• Incorporating prior information into Bayesian logistic regression enhances prediction accuracy, specificity, and sensitivity.
• Further research is recommended to validate the predictive model with an independent testing set.