Advancing Early Detection of Osteoarthritis Through Biomarker Profiling and Predictive Modelling: A Review

Osteoarthritis (OA) is a multifactorial chronic musculoskeletal disorder characterised by cartilage degradation, synovial inflammation, and subchondral bone remodelling. Conventional diagnostic modalities, including radiographic imaging and symptom-based assessments, primarily detect disease in its later stages, limiting the potential for timely intervention. Inflammatory biomarkers, particularly Interleukin-6 (IL-6), Tumour Necrosis Factor-alpha (TNF-α), and Myeloperoxidase (MPO), have emerged as biologically relevant indicators of disease activity, with potential applications as companion diagnostics in precision medicine. This review examines the diagnostic and prognostic relevance of IL-6, TNF-α, and MPO in OA, focusing on their mechanistic roles in inflammation and joint degeneration, particularly through the activity of fibroblast-like synoviocytes (FLSs). The influence of sample type (serum, plasma, synovial fluid) and analytical performance, including enzyme-linked immunosorbent assay (ELISA), is discussed in the context of biomarker detectability. Advanced statistical and computational methodologies, including rank-based analysis of covariance (ANCOVA), discriminant function analysis (DFA), and Cox proportional hazards modelling, are explored for their capacity to validate biomarker associations, adjust for demographic variability, and stratify patient risk. Further, the utility of synthetic data generation, hierarchical clustering, and dimensionality reduction techniques (e.g., t-distributed stochastic neighbour embedding) in addressing inter-individual variability and enhancing model generalisability is also examined. Collectively, this synthesis supports the integration of biomarker profiling with advanced analytical modelling to improve early OA detection, enable patient-specific classification, and inform the development of targeted therapeutic strategies.