Multi-Parametric Analysis of Clinical, Hormonal, BMP-4, and AI Machine Learning Approaches as Predictors in Fertility Treatment Outcomes

Abstract

Background: Current improvements in reproductive medicine underscore the critical role of innovative biomarkers to predict outcomes of assisted reproductive technology. Infertility impacts 10-15% of couples worldwide, representing significant challenges in reproductive health.

Objectives: The current study inspects the predictors of fertility treatment outcomes, concentrating on bone morphogenetic protein-4 (BMP-4), which is associated with folliculogenesis and embryo quality.

Methods: A multi-parametric inclusive analysis was executed, incorporating clinical data, hormonal profiles, serum BMP-4 levels, and ovarian response metrics from 100 infertile females undergoing ART.

Results: Serum BMP-4 levels were significantly greater in patients attaining positive clinical pregnancy outcomes compared to those with negative consequences (126.4 ± 85.9 pg/mL vs 58.7 ± 42.3 pg/mL, p =0.001), respectively. The receiver operating characteristic analysis exhibited BMP-4's strong predictive capacity for clinical pregnancy, with an accuracy of 85% (CI: 0.77– 0.92). The optimal cutoff for BMP-4 was over 95 pg/mL, yielding 82% sensitivity and 84% specificity. Multivariate logistic regression identified significant independent predictors of successful pregnancy: BMP-4 >95 pg/mL (aOR 3.2), younger age (aOR 1.18), the presence of ≥2 Grade 1 embryos (aOR 2.9), and progesterone <1 ng/mL (aOR 1.9). Age stratification highlighted reduced ovarian reserve and embryo quality in patients aged ≥35 years, relating to lower BMP-4 levels and pregnancy rates.

Conclusion: the findings support the incorporation of circulating BMP-4 measurement alongside traditional clinical and hormonal assessments to enhance ART prognostication and optimize individualized fertility treatment protocols.