Western cut-offs for anti-Müllerian hormone and antral follicle count misclassify Saudi women undergoing IVF: a retrospective cohort study and locally calibrated predictive model

Abstract

Abstract Background Anti-Müllerian hormone (AMH) and antral follicle count (AFC) are routinely used to stratify women undergoing in vitro fertilization (IVF), but virtually all widely cited cut-offs were derived in European or East-Asian cohorts. Women of Arabian Peninsula ancestry have AMH and AFC values 20–30% lower than these reference populations, raising concern that imported thresholds may systematically misclassify Saudi patients. We quantified this misclassification, derived locally calibrated cut-offs, and assessed clinical utility using decision-curve analysis (DCA). Methods Single-center retrospective cohort of 186 consecutive women undergoing fresh-transfer IVF on a uniform gonadotropin-releasing hormone (GnRH)-antagonist protocol at a private fertility center in Jeddah, Saudi Arabia (June–December 2024). The primary outcome was clinical pregnancy. We applied published European and European Society of Human Reproduction and Embryology (ESHRE)/Bologna thresholds to our cohort to quantify misclassification, derived locally optimal cut-offs by Youden index, and evaluated clinical utility by DCA across a range of threshold probabilities. Results The clinical pregnancy rate was 38.2% (71/186). Applying the European “good-prognosis” AMH threshold (> 3.5 ng/mL) to our cohort would have excluded approximately 53.5% of women who actually achieved clinical pregnancy from a favorable counseling category; a comparable AFC threshold (> 14) would have excluded 50.7%. Conversely, ESHRE/Bologna “poor-responder” thresholds (AMH < 1.1 ng/mL; AFC < 7) had specificity above 93% but sensitivity below 30%, providing little discrimination of poor outcomes in this population. Locally optimal cut-offs were AMH > 2.8 ng/mL (sensitivity 74.6%, specificity 73.0%) and AFC > 11 follicles (73.2%, 79.1%). On DCA, an AMH-based decision rule produced positive net benefit across plausible threshold probabilities (10–60%) while a treat-all strategy became harmful above a threshold probability (pₜ) of approximately 0.40. Conclusion Western-derived AMH and AFC thresholds materially misclassify Saudi women undergoing IVF, with up to half of women who go on to achieve clinical pregnancy falling below European “good-prognosis” cut-offs. Locally calibrated thresholds (AMH > 2.8 ng/mL; AFC > 11) better reflect this population and add net clinical benefit over default-treatment strategies on DCA. These findings should be interpreted as derivation-cohort estimates requiring multicenter external validation before clinical implementation. Because clinical pregnancy is a surrogate for live birth, these thresholds should be confirmed against cumulative live-birth rates in future work.