Anti-Müllerian Hormone and Metabolic Dysfunction in Polycystic Ovary Syndrome: A Cross-Sectional Study with Age-Stratified Analysis and BMI Mediation

Background. Anti-Müllerian hormone (AMH) is a well-established marker of ovarian reserve and is characteristically elevated in polycystic ovary syndrome (PCOS). Whether AMH carries independent information about metabolic risk in PCOS remains debated, with prior studies yielding conflicting results. The present study sought to clarify the association between AMH and metabolic parameters, with particular attention to age-specific patterns and the potential mediating role of body mass index (BMI). Methods. This single-center cross-sectional analysis included 123 women diagnosed with PCOS (43 adolescents aged 10–19 years and 80 reproductive-aged women aged 20–36 years). Fasting blood samples were assayed for AMH, reproductive hormones, fasting insulin, fasting glucose, lipid panels, liver enzymes, and uric acid. Insulin resistance was quantified using the homeostatic model assessment (HOMA-IR). The analytical strategy encompassed bivariate correlation, multivariable linear regression, AMH tertile comparison, receiver operating characteristic (ROC) curve analysis, and bootstrap-validated mediation analysis. A multivariable logistic regression prediction model was additionally constructed and evaluated by calibration curves and decision curve analysis. Results. In the overall cohort, AMH correlated inversely with BMI (r = − 0.198, p = 0.029) and LDL-cholesterol (r = − 0.418, p = 0.007), and positively with testosterone (r = 0.250, p = 0.005) and the LH/FSH ratio (r = 0.332, p < 0.001). An inverse association between AMH and HOMA-IR was evident among reproductive-aged women (r = − 0.302, p = 0.033) but not in adolescents (r = 0.020, p = 0.935). Upon adjustment for BMI, the independent association between AMH and HOMA-IR was substantially attenuated (β = −0.038, p = 0.488). Mediation analysis attributed 94.1% of the total AMH–HOMA-IR association to the indirect pathway through BMI (bootstrap 95% CI: −0.133 to − 0.030). AMH alone showed negligible discriminative ability for insulin resistance (AUC = 0.503), whereas BMI alone performed well (AUC = 0.798). The full prediction model incorporating AMH, BMI, age, testosterone, SHBG, and LH/FSH achieved an AUC of 0.800 (95% CI: 0.742–0.918), though DeLong’s test confirmed no significant improvement over BMI alone (p = 0.881). Conclusions. AMH tracks with a more favorable lipid profile and higher androgen levels in PCOS, yet its apparent association with insulin resistance is almost entirely accounted for by BMI. AMH does not appear to be an independent predictor of insulin resistance, and adding it to BMI-based risk stratification yields minimal incremental value. These findings underscore that metabolic risk assessment in PCOS should remain anchored in anthropometric measures, while AMH’s clinical utility lies primarily in the reproductive domain. The prediction model and nomogram presented here may serve as a preliminary tool for risk stratification, pending external validation.