Baseline endocrine factors influencing live birth outcomes in Chinese infertile women undergoing their first fresh IVF cycle: A multistate model-based cohort study.

Conventional in vitro fertilization (IVF) outcome prediction is limited by static, single-endpoint analyses. We aimed to overcome this by using a multistate model to dissect the stage-specific and, crucially, the non-linear influence of endocrine factors across the entire pregnancy continuum in a large-scale cohort. We applied multistate regression models to a large cohort of 12,674 women undergoing their first fresh IVF cycle. This advanced method allowed us to analyze three sequential transitions (from infertility to biochemical pregnancy, from biochemical pregnancy to clinical pregnancy, and ultimately to live birth) and test for non-linear effects of baseline hormones, including anti-Müllerian hormone (AMH), luteinizing hormone (LH), and antral follicle count (AFC), on the hazard of success at each stage. The principal finding was a significant non-linear relationship between baseline AMH, LH, and AFC and pregnancy success (P < 0.05 for non-linearity). This directly challenges the "higher is better" paradigm, revealing that optimal hormonal "windows", not just maximum levels, are linked to clinical success. The multistate model further distinguished AMH and LH as robust predictors across all stages, while AFC's predictive power was confined to achieving initial pregnancy. The predictive value of baseline hormones in IVF is fundamentally non-linear. Our use of a multistate model demonstrates that while AMH and LH are consistent predictors for the entire pregnancy journey, their clinical interpretation must shift from a linear scale to identifying optimal ranges. This finding provides a more precise scientific basis to personalize ART treatment and improve live birth rates.