782P - Predicting survival in malignant struma ovarii (MSO): A machine learning approach

Background

Struma ovarii, a rare form of teratoma with mature thyroid tissue, accounts for 2-5% of ovarian teratomas, with less than 5% exhibiting malignant transformation. We aimed to develop machine-learning (ML)-based prognostic models for patients with MSO to improve treatment decisions and personalized medicine.

Methods

Data between 2000 and 2020 were obtained from the SEER database. Patients who met any of the following criteria were excluded: diagnosis not confirmed by histology, not first tumor, and unknown data. Chi-square tests were used to compare clinicopathological features, while survival rates and prognostic factors were identified using the Kaplan-Meier estimator, log-rank tests, and Cox proportional hazard regression. We constructed prognostic models using 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

A total of 329 patients were included. Most patients were 45 years or older (52.9%), and the average tumor size was 40.2 mm. years, with the predominant race being white (n =247). Most of the patients were in the localized stage (80.9%). Primary surgical resection was performed in 94.2% of patients; 5.2% had undergone chemotherapy, and 13.4% had received radiotherapy. After multivariate adjustment, older age, chemotherapy, and advanced stage of the primary tumor were found to be significant prognostic factors for adverse overall survival. In contrast, Asian ethnicity and surgery were good prognostic factors for overall survival. The random forest classifier (RFC) and K-Nearest Neighbors (KNN) were the most accurate ML models, with age, race, and total number of tumors for patients as key factors.

Conclusions

We successfully developed an ML-based prognostic model for MSO patients. These models identified key factors influencing survival, including age, race, and total number of tumors in patients. These findings can inform personalized treatment decisions and improve patient outcomes.

Legal entity responsible for the study

The authors.

Funding

Has not received any funding.

Disclosure

All authors have declared no conflicts of interest.