Integrating GLP-1 Receptor Agonists into Fertility-Centered PCOS Care: Evidence Synthesis and Bayesian Decision Framework for Metabolic-Reproductive Optimization

Background: Insulin resistance (IR) and obesity are core drivers of reproductive dysfunction in Polycystic Ovary Syndrome (PCOS). While GLP-1 receptor agonists (GLP-1RAs) have reshaped the landscape of metabolic disease management, their application in reproductive medicine faces challenges, including uncertainty in efficacy translation, risks during the preconception washout period, and a lack of standardized clinical pathways.
Objective: This study aims to construct a "fertility-centered" framework for the application of GLP-1RAs10. By combining a comprehensive narrative review with a rigorous Bayesian network meta-analysis (NMA), we quantify drug efficacy and explore future directions for AI-assisted decision-making.
Methods: We employed a dual-track evidence integration approach. First, we conducted a multi-dimensional narrative synthesis of over 200 core papers from PubMed (including RCTs, cohorts, high-quality reviews, and guidelines) covering PCOS pathophysiology, clinical outcomes, and periconception safety to build a “metabolic-reproductive-pregnancy” continuum model. Second, we performed an in-depth analysis of 22 identified core RCTs. Finally, to quantify the probability of clinical benefit, we conducted a proof-of-concept (PoC) Bayesian NMA using 11 RCTs with consistent clinical characteristics (8 nodes, 25 arms), utilizing the Region of Practical Equivalence (ROPE) to evaluate the posterior distribution of weight-loss effects.
Results: Evidence synthesis indicates that GLP-1RAs effectively reshape the metabolic foundation of PCOS by alleviating lipotoxicity and improving insulin sensitivity. A systematic evaluation of 22 RCTs suggests that metabolic outcome benefits are significant and consistent. Bayesian NMA results confirm that within a 12–16 week intensive optimization window, combination regimens (e.g., liraglutide + metformin, semaglutide + metformin) have a very high posterior probability (P > 98%) of achieving clinically meaningful weight loss. Overall, while the evidence for metabolic improvement is robust, evidence regarding hard endpoints such as live birth rates and cumulative pregnancy rates in assisted reproductive technology (ART) remains limited by high heterogeneity and small sample sizes, manifesting an evidence gradient of “proven metabolic benefits, pending reproductive translation”.
Conclusion: GLP-1RAs should be positioned as “periconception metabolic optimization tools” in PCOS fertility management. We recommend stratified management based on a “metabolic risk × fertility timing” matrix: patients with high metabolic risk should undergo 3–6 months of intensive metabolic pretreatment, with strict adherence to drug-specific washout periods and rebound prevention strategies. Future research should integrate machine learning models to more accurately predict patient-specific benefit-risk ratios.