Machine learning to characterize bone biomarkers profile in rheumatoid arthritis.

Background: Bone metabolism is disrupted in rheumatoid arthritis (RA); however, the bone metabolic signature of RA is poorly known. The objective of the study is to further characterize the bone metabolic profile of RA and compare it to psoriatic arthritis (PsA), systemic sclerosis (SSc) and healthy controls. Methods: We did a cross-sectional case-control study on consecutively enrolled patients and age-matched controls. We collected clinical characteristics, serum biomarkers related to bone metabolism and Bone Mineral Density (BMD). A multiple correlation analysis using Spearman's rank correlation coefficient was conducted within the RA patient group to investigate associations between biomarker levels and clinical variables. Machine learning (ML) models and Principal Component Analysis (PCA) was performed to evaluate the ability of bone biomarker profiles to differentiate RA patients from controls. Results: We found significantly lower BMD in RA patients compared to PsA, and Systemic Sclerosis SSc groups. RA patients exhibited higher Dkk1, sclerostin and lower P1nP and B-ALP levels compared to controls. No significant differences in CTX levels were noted. Correlation analysis revealed associations between bone biomarkers and clinical variables. PCA and ML highlighted distinct biomarker patterns in RA which can effectively discriminated bone biomarkers profile in RA from controls. Conclusion: Our study helped uncover the distinct bone profile in RA, including changes in bone density and unique biomarker patterns. These findings enhance our comprehension of the intricate links between inflammation, bone dynamics, and RA activity, offering potential insights for diagnostic and therapeutic advancements in managing bone involvement in this challenging condition.