An interpretable metabolomic-driven machine learning model for early prediction of knee structural osteoarthritis (OA) progression
Authors' conclusion
This validated metabolomics DL framework (https://github.com/AFSHINJAM/KOA_Prediction_Metabolomics_Ratios) introduces a novel approach for predicting knee OA structural progression, enabling personalized risk assessment and early intervention.
Study context
- Population
- Participants categorized by likelihood of knee structural OA progression using MRI and X-ray data; two independent cohortstwo independent cohorts
- Participants
- 180 vs 137
- Comparison
- Tasmanian Older Adult vs Licofelone trial
- Tasmanian Older Adultn = 180
- Licofelone trialn = 137
- Outcome
- Prediction of knee structural OA progression (AUC, accuracy, sensitivity, specificity)
Methodological detailsexternal validation, machine/deep learning, serum metabolomics
- External validation
- Machine/deep learning
- Serum metabolomics
Performance
As reported in the abstract
Performance of final model (six metabolites plus sex and age; ANN with domain-adversarial component)
| Test or model | AUC (test) | AUC (validation) | Accuracy (test) | Accuracy (validation) |
|---|---|---|---|---|
| Tasmanian Older Adult | 0.98 [0.95, 1] | — not stated in the abstract | 0.89 [0.84, 0.94] | — not stated in the abstract |
| Licofelone trial | — not stated in the abstract | 0.89 [0.83, 0.95] | — not stated in the abstract | 0.85 [0.79, 0.91] |
AUC (test)
Tasmanian Older Adult0.98 [0.95, 1]
Licofelone trial— not stated in the abstract
AUC (validation)
Tasmanian Older Adult— not stated in the abstract
Licofelone trial0.89 [0.83, 0.95]
Accuracy (test)
Tasmanian Older Adult0.89 [0.84, 0.94]
Licofelone trial— not stated in the abstract
Accuracy (validation)
Tasmanian Older Adult— not stated in the abstract
Licofelone trial0.85 [0.79, 0.91]
— not stated in the abstract
Original publication
An interpretable metabolomic-driven machine learning model for early prediction of knee structural OA progression. Rheumatology (Oxford). Published 22 December 2025. doi:10.
Composed by AI from the published abstract and checked against it. How Briefs are made →