Development and External Validation of Machine Learning-based Models for Predicting Survival Outcomes in Endometrial...

Most endometrial cancers are early-stage cancers with a good prognosis, but the prognosis for recurrent endometrial cancer is poor. Accurate prognostication is essential for the management of patients with cancer. This study aimed to assess the survival outcome of endometrial cancer using machine learning.

We used data from the Surveillance, Epidemiology, and End Results (SEER) database, constructing machine learning models to predict the 5-year overall survival (OS) and cancer-specific survival (CSS). The variables included patient demographics, pathological factors, and therapeutic factors.

The OS rates of 71,506 patients and the CSS rates of 66,368 patients were included. For the prediction of OS, the best machine learning model achieved a class accuracy of 0.86 (95% CI=0.85-0.87) and an area under the curve (AUC) of 0.83 (95% CI=0.82-0.84) in the internal validation set (SEER dataset). In the external validation set of 149 patients, the best model achieved a class accuracy of 0.85 (95% CI=0.86-0.86) and an AUC of 0.85 (95% CI=0.85-0.86). The model predicted CSS more accurately than OS.

Using machine learning, we were able to predict the prognosis of patients with endometrial cancer.