L26/P-167 AI-based evaluation of gamete quality reveals additive effects on embryo development in ICSI cycles

Résumé fourni par la source

Abstract Study question Does early embryo development depend on independent gamete quality or on synergistic oocyte–sperm interactions when assessed using AI-based scoring systems? Summary answer AI-based analyses indicated that oocyte quality primarily drives fertilization and embryo development, with no evidence of synergistic interaction with sperm quality, supporting an additive model. What is known already Gamete quality is a central determinant of reproductive outcomes, but the relative contribution of each gamete and the possibility of a synergistic interaction remain debated. Traditional assessments rely on subjective morphological criteria with limited predictive accuracy. Recent advances in artificial intelligence (AI) allow objective, quantitative evaluation of oocyte and sperm features, offering unprecedented precision in predicting early embryo development. However, despite their growing use, it is still unclear whether AI-based scores capture independent or interacting effects of both gametes on embryogenesis. Understanding this relationship is crucial for improving patient counselling, embryo selection strategies, and individualized approaches to assisted reproduction. Study design, size, duration Retrospective, single-center cohort study including 200 infertile couples undergoing ICSI cycles (137 using autologous oocytes and 63 using donated oocytes), between March and December 2024. Oocyte and sperm quality were quantified using AI-based scoring algorithms. Fertilization rate, blastocyst development and usability were compared across four combined gamete-quality groups. Interaction analyses evaluated whether the effect of one gamete varied according to the other’s quality, and additional predictors of embryo competence were examined using multivariate logistic regressions. Participants/materials, setting, methods Oocyte quality was assessed using Magenta (Future Fertility) (score 0-10) and individual sperm quality with SiD v2.0 (IVF2.0;Ltd)(score 0-100); before microinjection. Four groups were established: G1 (AI-oocyte≥5+AI-sperm≥85; both high quality), G2 (AI-oocyte≥5+AI-sperm<85; high oocyte, low sperm), G3 (AI-oocyte<5+AI-sperm≥85; low oocyte, high sperm), and G4 (AI-oocyte<5+AI-sperm<85; both low quality). The interaction between gametes and their influence on reproductive outcomes was examined using nested logistic regression models, with comparisons performed through likelihood ratio tests (LRT) and ΔAIC. Main results and the role of chance Interaction models evaluating the combined contribution of both gametes did not detect a significant interaction between AI-derived oocyte and sperm scores, across fertilization (OR = 0.75, [0.38–1.45]; LRT p = 0.388), blastocyst formation (OR = 1.06, [0.76–1.50]; LRT p = 0.727), or usable blastocyst outcomes (OR = 0.84, [0.59–1.19]; LRT p = 0.334), indicating absence of synergistic effects and supporting an additive model of gamete contribution to embryo development. Consistent with this framework, sperm-related effect sizes were small (Cohen’s h ≤ 0.15), resulting in limited statistical power (≤52%), whereas oocyte quality showed a strong and consistent impact across reproductive outcomes (power >99%). In line with these findings, the descriptive results derived from the group-based analysis of the raw data showed that high-quality oocytes (G1, G2) consistently outperformed lower-quality groups (G3, G4) across all endpoints. Fertilization rates were higher in G1 and G2 (87.6% and 88.8%) compared with G3 and G4 (71.4% and 67.7%), with similar patterns observed for blastocyst formation (59.4–63.1% versus 36.3–41.5%) and usable blastocyst rates (46.3–53.6% versus 27.0–29.5%). All differences between high and low-quality oocyte groups were statistically significant (p < 0.001). Finally, multivariate models identified additional independent predictors of embryo quality and usability, including day of blastocyst development and oocyte source (autologous versus donor). Limitations, reasons for caution This study has limitations, including its retrospective, single-center design, which may limit external validity. Sperm gross morphology was not included in the individual AI sperm score. Moreover, pregnancy and implantation outcomes could not be assessed due to the short interval between study completion and abstract preparation. Wider implications of the findings These results highlight the predominant role of oocyte competence in early embryo development and support the integration of AI-based oocyte assessment into clinical decision-making. While sperm quality contributes additively, AI-derived sperm scores may remain useful in IVF or severe male-factor scenarios. Trial registration number No

Ce résumé expose les affirmations des auteurs. BNTIC ne l’interprète pas comme une validation indépendante des résultats.