PulseExploreJournal ClubDebatesTrendingResearchersJournals
Instagram
HomeExploreJournal ClubTrending
Synapse
⌘+K
Synapse
April 30, 2026Lara D. Veeken0 citations

OA04 Metabolomic predictors of osteoarthritis onset in the UK Biobank

View Full Paper
PSPratyasha SahaMDMarianne DefernezGGGwenaelle Le Gall

Key Points

  • This research aims to identify metabolomic predictors of osteoarthritis onset by analyzing metabolic profiling data.
  • Nested case-control analysis within the UK Biobank, matching 30,490 OA cases to controls on key demographics.
  • Analysis of serum metabolomic data using random survival forests (RSF) to model time-to-event outcomes.
  • Utilized up to 249 metabolomic measures from blood samples through NMR-based assays.
  • RSF model achieved a Brier score of 0.14 at 5 years, indicating good accuracy.
  • Higher glycoprotein acetyls and amino acid alanine levels were linked to increased OA risk, while polyunsaturated fatty acids correlated with reduced risk.
  • Identified three main metabolic pathways connected to OA: impaired energy metabolism, chronic inflammation, and adverse lipid composition.

Abstract

Abstract Background/Aims Osteoarthritis (OA) is a major cause of disability worldwide, yet the systemic processes that precede disease onset remain ill-defined. Metabolomic profiling offers an opportunity to identify antecedent biological pathways linked to OA risk beyond traditional factors. The UK Biobank is a large prospective longitudinal research database ( 500,000 participants) with detailed clinical data and nuclear magnetic resonance (NMR)-based metabolomics assays from blood samples at enrolment. Methods A nested case-control analysis was conducted within UK Biobank. Participants who developed hip, knee or hand OA (n = 30,490) after metabolomic sampling were matched 1:1 to controls on age, sex, body mass index (BMI) and smoking status. Serum metabolomic data (up to 249 measures) were analysed using random survival forests (RSF), a machine learning approach for time-to-event data, with 500 trees per model. RSF accommodates high-dimensional inputs and non-linear effects, moving beyond single metabolite analyses. Results Participants had a median age of 68.6 years (IQR 62.9-73.3) and 57.9% were female, with a mean BMI 29.2 kg/m2 (SD 5.2) and 8.8% were current smokers, 40.3% previous smokers, and 50.4% never smokers. There was a median interval of 7.3 years (IQR 4.1 - 10.3) between metabolomic blood sampling and subsequent OA diagnosis. The RSF model had a Brier score at 5 years of 0.14 (Brier scores range from 0 to 1, with lower values indicating better accuracy) and highlighted three broad domains of metabolic variation associated with subsequent OA diagnosis: 1) Markers of energy metabolism (lactate, citrate, amino acids including alanine and valine); 2) Markers of inflammation (glycoprotein acetyls); and 3) Lipid measures (apolipoprotein B, VLDL cholesterol, omega-6 fatty acids). Increased risk of OA was associated with higher levels of glycoprotein acetyls, the amino acid alanine, and an increased ratio of saturated fatty acids to total fatty acids; whilst higher proportions of polyunsaturated fatty acids, particularly linoleic acid, were associated with reduced risk of OA. Conclusion This analysis demonstrates that systemic metabolic perturbations are detectable years before OA onset, reducing the likelihood of reverse causation, and supports the potential future role of metabolomic biomarkers. The findings converge on three mechanistic pathways: 1) impaired energy metabolism, 2) chronic low-grade inflammation, and 3) adverse lipid composition. These processes provide biologically coherent explanations for how systemic metabolism may influence joint vulnerability, whilst protective associations with polyunsaturated fatty acids highlight potentially modifiable risk factors. Previous cross-sectional studies have reported metabolite-OA associations, while a recent longitudinal study incorporated metabolomics into a predictive OA diagnosis model but focussed on short-term prediction and patient stratification. In contrast, our analysis places metabolomics at the centre, uses a larger incident OA cohort with longer follow-up, and applies RSF to model time-to-diagnosis whilst capturing non-linear interactions with clinical covariates Disclosure P. Saha: None. M. Defernez: None. G. Le Gall: None. P. Cardenas-Canto: None. J. Dainty: None. G. Wortley: None. M. Yates: None. R. Davidson: None. I. Clark: None. M. Traka: None. K. Kemsley: None. A. MacGregor: None.

Ask AI
Helpful
Bookmark
Share
View Full Paper

Cite This Study

Saha et al. (2026) studied this question.

synapsesocial.com/papers/69f2f1771e5f7920c638730ehttps://doi.org/10.1093/rheumatology/keag121.004
Ask AI
Helpful
Bookmark
Share
View Full Paper