Predictive analysis of placental site trophoblastic tumors (PSTT) and epithelioid trophoblastic tumors (ETT) using machine learning.

e17608 Background: The characteristics and outcomes of PSTT and ETT have not been well documented. To improve personalized care, machine learning was used to assess treatment effectiveness and develop accurate prognostic models. Gestational choriocarcinoma (GCC) was used as a control to emphasize the clinical features of PSTT and ETT. Methods: The Surveillance, Epidemiology, and End Results database provided the data used for this study’s analysis (2000–2019). Patients who met any of the following criteria were excluded: diagnosis not confirmed by histology, previous history of cancer or other concurrent malignancies, or unknown data. To identify prognostic variables, we conducted Cox regression analysis and constructed prognostic models using machine learning (ML) algorithms to predict the 5-year survival. Patient records were randomly divided into training (70 %) and validation (30 %) sets. A validation method incorporating the area under the curve (AUC) of the receiver operating characteristic curve was used to validate the accuracy and reliability of the ML models. Results: The study population comprised 725 patients. Among them, 139 patients had ETT, 107 had PSTT, and 479 had GCC. The median patient age was 37 years. Most of the patients were of white race (62.3%) and more than half were married (52.7%). The most common stage of primary tumor histology was distant metastasis, which occurred in 330 patients. The average tumor size was 6.13 cm. More than two-thirds of patients underwent chemotherapy (75.9%), 57.9% underwent surgery, and 4.3% underwent radiotherapy. There were no significant differences in survival between the different tumor groups. Multivariate Cox regression analysis revealed that older age, marital status, metastasis, and radiotherapy were poor prognostic factors, while surgery was a good prognostic factor. ML models revealed that the random forest classifier (RFC) accurately predicted the outcomes. The most important characteristics that influenced the model prediction performance were tumor size, years of diagnosis, race, and age. The performance metrics for all ML algorithms are summarized in Table. Conclusions: Our research sheds light on crucial aspects of clinical decision-making, emphasizing the significance of key factors affecting patient outcomes and offering hope for personalized treatment. Further research should explore the molecular markers and validate the generalizability of our findings. [Table: see text]