Predictive Value of Machine Learning in Knee Osteoarthritis Progression: Systematic Review and Meta-Analysis

Area of Science:

• Orthopedics
• Biomedical Engineering
• Data Science

Background:

• Machine learning (ML) is being explored for predicting knee osteoarthritis (KOA) progression.
• Existing systematic evidence on ML effectiveness for KOA progression prediction is limited, hindering precision prevention strategies.

Purpose of the Study:

• To systematically review ML applications in predicting KOA progression.
• To assess ML accuracy and compare its predictive performance against traditional methods and deep learning.

Main Methods:

• A systematic literature search was conducted across major databases (Embase, Web of Science, PubMed, Cochrane Library) following PRISMA guidelines.
• Study selection, data extraction, and risk of bias assessment were performed independently by two investigators.
• Meta-analyses focused on concordance index (C-index) and diagnostic tables, with subgroup analyses by model type, variables, and KOA progression definitions.

Main Results:

• Thirty-two studies were included, with varying risk of bias across studies.
• Pooled C-indices indicated predictive capabilities, with MRI-based models (C-index=0.798) and combined MRI+clinical features (C-index=0.806) showing strong performance.
• Clinical feature-based models using logistic regression performed comparably to other ML models; traditional ML and deep learning showed higher accuracy in image-based models.

Conclusions:

• ML models possess discriminatory power for predicting KOA progression, but results must be interpreted cautiously due to heterogeneity.
• Variations in KOA progression definitions and validation methods impact model reliability.
• Future research should standardize definitions, improve methodological rigor, and implement external validation for enhanced clinical translation.