Identifying Plasma Biomarkers That Predict Patient-Reported Outcomes Following Treatment for Trapeziometacarpal Osteoarthritis Using Machine Learning.

Osteoarthritis (OA) in the trapeziometacarpal joint (TM) is a prevalent form of hand OA, yet research on biomarkers specific to hand OA remains limited. This study aims to identify systemic plasma biomarkers at baseline in TM OA patients that are associated with patient-reported outcomes one year post-treatment. Blood samples and clinical data were collected prospectively from 143 TM OA patients undergoing conservative therapy, fat grafting, or surgery, with one-year follow-up. Supervised machine learning with Lasso regularization analyzed associations among 10 systemic biomarkers related to cartilage turnover, bone remodeling, pain, or lipid metabolism. Generalized estimating equation models evaluated baseline biomarker associations with one-year outcomes. Patients averaged 61 years, were mostly female (69%), and were primarily treated conservatively (47%). The machine learning model identified associations between outcomes and biomarkers, including PIIANP, Visfatin, adiponectin, and leptin. Adjusted analyses revealed baseline PIIANP associated with VAS, QuickDASH, and TASD improvements, while Visfatin correlated with VAS worsening. We could identify two different plasma markers that could predict the clinical outcome of TM OA treatment. Baseline PIIANP is associated with improvement, while Visfatin is associated with worsening in TM OA outcomes up to one year post-treatment. Further prospective studies are needed to confirm and solidify these findings.