Mapping the Sex-Dimorphic Hormonal Ecosystem in Osteoarthritis: Non-Linear Interactions Between Androgen Activity and Estrogen

Methods We performed a cross-sectional analysis of 13,848 adults from the U.S. National Health and Nutrition Examination Survey (NHANES), stratified by sex and menopausal status. Beyond complex-sampling logistic regression, we employed an advanced analytical framework integrating two-way restricted cubic splines, generalized additive models (GAMs), and interpretable machine learning (SHAP) to model non-linear interactions and identify risk thresholds. Results The FAI-OA association was markedly heterogeneous. After adjustment, higher FAI was protective in post-menopausal women (OR = 0.85, 95% CI: 0.74–0.97, P = 0.021), showed a non-significant protective trend in pre-menopausal women, and was not associated in men. Crucially, non-linear interaction analyses revealed distinct, sex-specific modification by estradiol : a “cross-over” pattern in men (protective at low estradiol, adverse at high estradiol) and a “buffering” pattern in post-menopausal women (higher estradiol mitigated risk at low FAI). Machine-learning models confirmed that the highest OA risk emerged from specific estradiol-FAI combinations (e.g., high E₂ + low FAI in men; low E₂ + low FAI in post-menopausal women), not from either hormone alone. Conclusions The influence of sex hormones on OA is governed by a context-dependent, non-linear interaction network . Our findings move beyond single-hormone paradigms, providing a dynamic “hormonal ecosystem” framework that explains sex disparities in OA through fundamentally distinct modes of steroid interaction. This approach offers novel mechanistic insights and a foundation for sex-specific, precision prevention strategies .