Detecting change-points in preclinical rheumatoid arthritis biomarkers using Bayesian multivariate segmented...

Area of Science:

• Immunology
• Biostatistics
• Rheumatology

Background:

• Rheumatoid arthritis (RA) has a preclinical phase marked by rising serum autoantibodies.
• Understanding the timing and extent of autoantibody changes can guide early detection and prevention.

Purpose of the Study:

• To jointly model longitudinal autoantibody trajectories in preclinical RA.
• To estimate change-points and magnitudes of autoantibody shifts using a Bayesian approach.
• To identify key autoantibodies and their timing preceding RA diagnosis.

Main Methods:

• Bayesian multivariate segmented regression applied to two large serum repository cohorts.
• Simultaneous estimation of change-points and magnitudes for multiple correlated autoantibodies.
• Utilized unstructured residual correlation matrix for modeling.

Main Results:

• Five of six biomarkers in Sample A showed pre-diagnostic shifts; RF-IgM shifted earliest (8.10 years prior).
• ACPA-IgG shifted 7.43 years before diagnosis in Sample A.
• In Sample B, anti-CCP3 (IgG) showed the earliest shift (7.00 years prior); only IgG isotypes consistently shifted.
• A composite metric integrating timing and magnitude altered biomarker rankings.

Conclusions:

• The Bayesian framework effectively estimates changes in correlated autoantibodies and quantifies uncertainty.
• This method complements existing divergence-based approaches for studying preclinical RA autoimmunity.
• Findings highlight specific autoantibodies and their temporal dynamics for potential RA screening.