Background: Primary angle closure glaucoma (PACG) is the leading cause of irreversible blindness worldwide, and no reliable, effective diagnostic, and predictive biomarkers are used in clinical routines. A growing body of evidence shows metabolic alterations in patients with glaucoma. we aimed to develop and validate potential metabolite biomarkers to diagnose and predict the visual field progression of PACG. Methods: Here, we used a 5-phases (discovery phase, validation phase 1, validation phase 2, supplementary phase, and cohort phase) multicenter (EENT hospital, Shanghai Xuhui central hospital), cross-sectional, prospective cohort study design to perform widely-targeted metabolomics and chemiluminescence immunoassay to determine candidate biomarkers. Five machine learning (random forest, support vector machine, lasso, K-Nearest neighbor, and Gaussian NB) approaches were used to identify an optimal algorithm. The discrimination ability was evaluated using the area under the receiver operating characteristic curve (AUC). Calibration was assessed by Hosmer-Lemeshow tests and calibration plots. Results: Studied serum samples were collected from 616 participants, and 1464 metabolites were identified. Machine learning algorithm determines that androstenedione exhibited excellent discrimination and acceptable calibration in discriminating PACG across the discovery phase (discovery set 1, areas under the receiver operating characteristic curve [AUC] =1.0 [95%CI, 1.00-1.00]; discovery set, AUC=0.85 [95%CI, 0.80-0.90]) and validation phases (internal validation, AUC=0.86 [95%CI, 0.81-0.91]; external validation, AUC=0.87 [95%CI, 0.80-0.95]). Androstenedione also exhibited a higher AUC (0.92-0.98) to discriminate the severity of PACG. In the supplemental phase, serum androstenedione levels were consistent with those in aqueous humor (r=0.82, P=0.038) and significantly (P=0.021) decreased after treatment. Further, cohort phase demonstrates that higher baseline androstenedione levels (hazard ratio=2.71 [95% CI: 1.199-6.104], P=0.017) were associated with faster visual field progression. Conclusion: Our study demonstrates serum androstenedione as a novel biomarker that can be used to diagnose PACG and predict visual field progression.