Predicting autoimmune thyroiditis in primary Sjogren's syndrome patients using a random forest classifier: a...

Area of Science:

• Immunology
• Rheumatology
• Medical Informatics

Background:

• Primary Sjogren's syndrome (pSS) and autoimmune thyroiditis (AIT) share common genetic and immunological factors.
• Predicting AIT in pSS patients is crucial for comprehensive patient management.
• Thyroid-specific autoantibodies, including thyroid peroxidase antibodies (TPOAb) and thyroglobulin antibodies (TgAb), are key indicators of AIT.

Purpose of the Study:

• To evaluate the efficacy of machine learning algorithms in predicting thyroid autoantibodies in pSS patients.
• To identify key clinical and laboratory predictors for AIT in the pSS cohort.
• To assess the potential of machine learning for early AIT detection in pSS.

Main Methods:

• Retrospective analysis of clinical and laboratory data from 96 pSS patients.
• Categorization of patients into positive and negative thyroid autoantibody groups (TPOAb, TgAb).
• Application and comparison of four machine learning algorithms, with a focus on Random Forest Classifier.

Main Results:

• The Random Forest Classifier achieved the highest performance with an AUC of 0.755.
• Key predictors identified by the Random Forest model include age, IgG levels, complement component 4 (C4), and presence/absence of dry mouth.
• The study demonstrated the feasibility of using machine learning to predict AIT in pSS patients.

Conclusions:

• Machine learning, particularly the Random Forest model, shows significant promise for predicting AIT in pSS patients.
• Early identification of AIT in pSS can be facilitated by analyzing age, IgG, C4, and dry mouth status.
• This predictive approach can aid in timely intervention and improved management strategies for co-occurring AIT in pSS.