P-544 Biochemical markers as a predictor of clinical pregnancy on day 6 after embryo transfer

Abstract Study question Is it possible to predict clinical pregnancy using Machine Learning models on day 6 after embryo transfer an acceptable degree of certainty? Summary answer Using selected biochemical markers is build a machine learning model pregnancy. What known already Human chorionic gonadotropin (β-hCG) considered the main early predictor Due fact that highest percentage spontaneous abortions occurs in first trimester predicting outcomes at stages particularly difficult. The level B-hcg measured implantation and dynamics its increase can help further development. However, specific cut-off points which be determined vary from study study, different factor often suggested certainty β-hCG-based predictions. design, size, duration 1474 single frozen transfers (1205 patients between 23 47 years age), January 2019 -December 2022 were used for data modelling statistical testing. dataset consisted endocrinal markers:β-hCG, PRG (progesterone), E (estradiol) 6th ET other practitioner-chosen data: age, PGD results, quality status, RIF or RPL history, maturation days as well outcomes. Exclusion criteria were: > 50ng/ml, β-hCG >70 mIU/ml. Participants/materials, setting, methods A was trained gradient boosting technique. 100 decision trees with 8 leaves, maximal depth 2 nodes rate 0.1. performance accuracy score metric area under ROC (AUC). Feature importance based Shapley values. Additional analysis performed testing two-sided Student’s t-test. Main results role chance naive employed chosen variables most important turned out serum embryo-transfer. final model, limited those variables, performs around 83 % (accuracy = 0.834), (AUC) 0.927. Based values, followed by are relevant model’s Low concentrations increased chances findings confirmed additional analysis. Observations divided into two groups point 24 ng/ml – median whole group. probability higher group where below even if levels same. statistically significant difference these regarding (P value < 0.001) noticed. In order more accurate should interpreted context levels. Limitations, reasons caution Since dataset, only included, fresh ETs need explored further. six hence cannot interpret any date. Wider implications accurately inform embryo-transfers, practitioners not limit themselves monitoring but also include prognostic Trial registration number Not Applicable