LDH and LV218 as biomarkers for diagnosing microbiologically positive tuberculous pleural effusions

Background/objectives Tuberculous pleuritis (TP), a common manifestation of Mycobacterium tuberculosis infection, poses challenges in differentiating microbiologically positive (PEMP-MT) from negative (PEMN-MT) pleural effusions due to the limited sensitivity of traditional diagnostic methods. Methods Proteomics analysis using iTRAQ, non-targeted metabolomics, parallel reaction monitoring (PRM), and machine learning were employed to diagnose PEMN-MT or PEMP-MT. A validation cohort of 63 PEMN-MT and 28 PEMP-MT patients underwent ELISA experiments. Receiver operating characteristic (ROC) curves evaluated the predictive value of LDH and LV218 individually and in combination. Results Differentially expressed proteins (DEPs) and metabolites (DEMs) were identified using bioinformatics tools and pathway enrichment analyses. A machine learning model utilizing six biomarkers (LV218, F13A, RET4, LV321, TBA1C, and LDH) demonstrated excellent diagnostic performance with an AUROC of 0.987 and an AUPR of 0.974, distinguishing PEMP-MT from PEMN-MT. ROC curve analysis showed that both LDH and LV218, alone and in combination, provided strong predictive value for distinguishing the two groups. Conclusion LDH and LV218 are promising biomarkers for differentiating microbiologically positive and negative pleural effusions in tuberculous pleuritis. These biomarkers, particularly when combined, could improve diagnostic accuracy and clinical management.