BACKGROUND: No index for non-invasive diagnosis of subclinical pelvic inflammatory disease (PID) is available at this time. Here we carried out a plasma metabolomic study to search for potential biomarkers to facilitate its non-invasive diagnosis. METHOD: The metabolites in plasma were detected by using an LC-Q-TOF-MS method. The metabolic profiles of subclinical PID patients and healthy controls were discriminated by multivariate analysis. 30 patients and 28 controls were enrolled for PLS-DA model construction, and further 8 patients and 8 controls were employed for model validation. Univariate analysis was performed to evaluate potential biomarkers. RESULTS: The metabolic profiles of subclinical PID patients were different from those of healthy controls in a PLS-DA model, and this model was validated by permutation test and could accurately classify further 16 samples in T-prediction. Eleven differentiating metabolites, with the variable importance in the project >1 and corrected P < 0.05, were found as potential biomarkers. These metabolites included eight lipids, p-cresol, 3-indolepropionic acid and indoxylsulfuric acid. Among them, lysophosphatidic acid (16,0/0:0) showed a highest AUC value of receiver operating characteristic curve (0.855), with sensitivity of 89.3% and specificity of 73.3%. CONCLUSION: Through an LC-Q-TOF-MS based metabolomic analysis on subclinical PID, this study reports the potential plasma biomarkers which may be helpful for its non-invasive diagnosis.
BACKGROUND: No index for non-invasive diagnosis of subclinical pelvic inflammatory disease (PID) is available at this time. Here we carried out a plasma metabolomic study to search for potential biomarkers to facilitate its non-invasive diagnosis. METHOD: The metabolites in plasma were detected by using an LC-Q-TOF-MS method. The metabolic profiles of subclinical PID patients and healthy controls were discriminated by multivariate analysis. 30 patients and 28 controls were enrolled for PLS-DA model construction, and further 8 patients and 8 controls were employed for model validation. Univariate analysis was performed to evaluate potential biomarkers. RESULTS: The metabolic profiles of subclinical PID patients were different from those of healthy controls in a PLS-DA model, and this model was validated by permutation test and could accurately classify further 16 samples in T-prediction. Eleven differentiating metabolites, with the variable importance in the project >1 and corrected P < 0.05, were found as potential biomarkers. These metabolites included eight lipids, p-cresol, 3-indolepropionic acid and indoxylsulfuric acid. Among them, lysophosphatidic acid (16,0/0:0) showed a highest AUC value of receiver operating characteristic curve (0.855), with sensitivity of 89.3% and specificity of 73.3%. CONCLUSION: Through an LC-Q-TOF-MS based metabolomic analysis on subclinical PID, this study reports the potential plasma biomarkers which may be helpful for its non-invasive diagnosis.