Objective To investigate the metabolism characteristics and the potetial biomarker candidates of osteonecrosis of the femoral head (ONFH) using metabolomic technology. Methods The femoral head specimens from 23 ONFH patients (25 necrotic femoral heads) and 18 normal femoral heads from femoral neck fracture patients were collected for histopathological examination to confirm the diagnosis of all samples. All the metabolites of bone trabecula were extracted for ultrahigh performance liquid chromatography-MS/MS analyzed. The measured variables was pretreat, and PCA (principal component analysis), PLS-DA (partial least squares-discriminant analysis) and OPLS-DA (orthogonal-partial least squares-discriminant analysis) models were employed to confirm the difference between these two groups after UPLC-MS/MS (ultra-high performance liquid chromatography-mass spectrometry/mass spectrometry) analysis. At last, the differential variables were screened out by PLS-DA and variate analysis (Kruskal-Wallis H test). The changed metabolites were confirmed by MS and MS/MS aligned in HMDB (human metabolomic database) and Massbank. The changed metabolites with the most obviously changed peak abundance, D-arginine, L-proline and L-glutamine, were picked out as the potential diagnostic biomarkers. After binary logistic regression analysis, the combined biomarkers candidates were further analyzed by receiver operating characteristic (ROC) curve to evaluate the significance of the combined biomarkers. Results Significant distinction of metabolites expression mode can be seen in PCA, PLS-DA and OPLSDA models scoring plots between ONFH and control groups. Twelve changed metabolites in ONFH bone trabeculas were confirmed by multi-variate statistical analysis and variate statistical analysis. Compared with the femoral neck fracture patients, the increased metabolites included D-arginine, L-proline, L-glutamine, creatine, uracil, uridine, LysoPC(20∶4(5Z, 8Z, 11Z, 14Z)), LysoPC(16∶0), PC(20∶1(11Z)/18∶3(6Z, 9Z, 12Z)) and PE(P-16∶0e/0∶0). The decreased metabolites were reticulataxanthin and β-cryptoxanthin. According to the change fold of peak abundance and variable weight projection in PLS-DA, the most obviously differential metabolites were picked out as the biomarker candidates of ONFH. The potential biomarkers candidates were identified as D-arginine, L-proline and L-glutamine. The area under the curve of D-arginine, L-proline and L-glutamine ROC were 0.873, 0.712 and 0.862. The area under the curve of ROC was 0.946 after combining D-arginine, L-proline, L-glutamine using binary logistic regression analysis. Conclusion PCA, PLS-DA and OPLS-DA models were used to find out the differential variables in the metabolites of bone trabeculas in ONFH and femoral neck fracture patients. Twelve metabolites were identified by MS/MS, and 3 obviously changed metabolites, D-arginine, L-proline, L-glutamine, were indicated as biomarker candidates. These 3 obviously changed metabolites showed a good diagnostic significance. Key words: Femur head necrosis; Mass spectrometry; Metabolomics; Biological markers