Osteoporosis, a disease characterized by low bone mineral density (BMD), increases the risk for fractures. Conventional risk factors alone do not completely explain measured BMD or osteoporotic fracture risk. Metabolomics may provide additional information. We aim to identify BMD-associated metabolomic markers that are predictive of fracture risk. We assessed 209 plasma metabolites by LC-MS/MS in 1,552 Framingham Offspring Study participants, and measured femoral neck (FN) and lumbar spine (LS) BMD 2-10 years later using dual energy x-ray absorptiometry. We assessed osteoporotic fractures up to 27-year follow-up after metabolomic profiling. We identified twenty-seven metabolites associated with FN-BMD or LS-BMD by LASSO regression with internal validation. Incorporating selected metabolites significantly improved the prediction and the classification of osteoporotic fracture risk beyond conventional risk factors (AUC=0.74 for the model with identified metabolites and risk factors vs AUC=0.70 with risk factors alone, p=0.001; Net reclassification index=0.07, p=0.03). We replicated significant improvement in fracture prediction by incorporating selected metabolites in 634 participants from the Hong Kong Osteoporosis Study (HKOS). The glycine, serine, and threonine metabolism pathway (including four identified metabolites: creatine, dimethylglycine, glycine, and serine) was significantly enriched (FDR p-value=0.028). Furthermore, three causally related metabolites (glycine, Phosphatidylcholine [PC], and Triacylglycerol [TAG]) were negatively associated with FN-BMD while PC and TAG were negatively associated with LS-BMD through Mendelian randomization analysis. In summary, metabolites associated with BMD are helpful in osteoporotic fracture risk prediction. Potential causal mechanisms explaining the three metabolites on BMD are worthy of further experimental validation. Our findings may provide novel insights into the pathogenesis of osteoporosis.

Download data

• Downloaded 35 times
• Download rankings, all-time:
  • Site-wide: 126,906
  • In epidemiology: 5,114
• Year to date:
  • Site-wide: 31,353
• Since beginning of last month:
  • Site-wide: 31,353

Altmetric data

Downloads over time

Distribution of downloads per paper, site-wide