Exploratory analysis of anti-Müllerian hormone and androgen profiles across historical ovarian–menstrual categories defined by antral follicle count in an infertility-associated reproductive endocrine cohort

Polycystic ovary syndrome (PCOS) is heterogeneous. Ovarian morphology, menstrual function, anti-Müllerian hormone (AMH), androgen excess, and metabolic features may not align. We compared classical and 11-oxygenated androgens across historical ovarian–menstrual categories defined by antral follicle count (AFC) and examined whether selected androgens remained associated with group assignment after adjustment for AMH, age, and body mass index (BMI). Separate analyses explored their relation to metabolic context. This retrospective secondary analysis used an existing single-center database of 495 women evaluated at the Reproductive Medicine Center of Peking University Third Hospital. Ovarian morphology was represented by the historical overall AFC categories recorded in the source database: ≤12, 13–24, and > 24 follicles. The latter two categories were designated as AFC-positive for this exploratory analysis. Because per-ovary follicle counts were not uniformly available, this study-specific designation is not equivalent to the current polycystic ovary morphology (PCOM) criterion of ≥ 20 follicles in at least one ovary, and reliable reclassification was not possible. Of 433 women with available historical AFC and menstrual-cycle categories, 431 had complete data for age, BMI, AMH, inhibin B, and a validated eight-steroid mass spectrometry panel. The isolated menstrual dysfunction (MD) group included only two women and was retained for descriptive reporting only. The three-group analysis cohort therefore remained 429 women: the reference group ( n = 27), isolated AFC-positive group ( n = 69), and combined AFC-positive/MD group ( n = 333). Main analyses included three-category multinomial regression, androgen residuals conditional on AMH, age, and BMI, and endpoint-resolved models. K-means clustering and PCOS–BMI subgroup analyses addressed separate, non-prespecified exploratory questions. AMH showed the clearest gradient across the three groups, with medians of 0.89, 3.85, and 5.76 ng/mL, respectively ( p = 2.51 × 10⁻¹⁷). For each 1-standard deviation (SD) increase in AMH, the adjusted odds ratios (ORs) and 95% confidence intervals (CIs) were 5.66 (2.48–12.90) for the isolated AFC-positive group versus the reference group, 8.95 (3.99–20.11) for the combined AFC-positive/MD group versus the reference group, and 1.58 (1.07–2.34) for the combined AFC-positive/MD group versus the isolated AFC-positive group. Testosterone (T) and androstenedione (A4) differed between groups before adjustment but were not independently associated with group assignment in the multivariable model. The corresponding residual distributions did not differ after adjustment for AMH, age, and BMI. None of the four 11-oxygenated androgens differed across the three historical categories. In non-prespecified analyses, 11β-hydroxytestosterone (11OHT) and 11-ketotestosterone (11KT) were higher in the cluster with higher homeostasis model assessment of insulin resistance (HOMA-IR) and BMI ( p = 6.78 × 10⁻⁴ and p = 8.51 × 10⁻⁴, respectively). In the extreme PCOS–BMI comparison, only 11OHT differed significantly. The four-marker model had an apparent area under the receiver operating characteristic curve (AUC) of 0.724 and a leave-one-out cross-validated AUC of 0.664. These analyses were not adjusted for multiple comparisons or externally validated. AMH showed the strongest adjusted association with the historical AFC-defined categories. T and A4 differed only in unadjusted comparisons, and the four 11-oxygenated androgens did not distinguish the groups. Associations with metabolic context arose only in non-prespecified analyses and require independent confirmation.