Immunohistochemistry as a reliable predictor of remission in patients with endometrial cancer: Establishment and validation of a machine learning model

• Published in: Oncology letters 2025-01, Vol.29 (1), Article 59
• Main Authors: Wang, Ruiqi, Wang, Jingyuan, Wu, Yuman, Zhu, Aoxuan, Li, Xingchen, Wang, Jianliu
• Format: Article
• Language: English
• Subjects: Algorithms; Endometrial cancer; Estrogens; Hysterectomy; Immunohistochemistry; Lymphatic system; Machine learning; Medical prognosis; Mortality; Patients
• ISSN: 1792-1074; 1792-1082
• DOI: 10.3892/ol.2024.14805

Endometrial cancer (EC) is the most common gynecologic cancer. Unfortunately, its prognosis remains poor due to limited screening and treatment options. To address this issue, the present study evaluated the predictive value of four immunohistochemical (IHC) indicators for overall survival (OS) and recurrence-free survival (RFS) in patients with EC. A total of 834 patients diagnosed with EC were included at Peking University People's Hospital between January 2006 and December 2020. These patients were randomly divided into training and validation cohorts at a 2:1 ratio, collecting data on clinicopathological information and IHC indicators. A total of 92 combinations of algorithms were assessed using the Leave-One-Out Cross-Validation framework to identify the one with the highest C-index. To estimate the accuracy of the factors and four IHC indicators for predicting both OS and RFS, survival curves and receiver operating characteristic (ROC) curves were used. Independent predictors included estrogen receptor, progesterone receptor, body mass index, P53, FIGO stage, histology, grade, Ki67, ascites and lymph node metastasis. Both the training and validation cohorts exhibited excellent predictive performance for OS and RFS, as demonstrated by ROC curves at 1-year, 3-year and 5-year follow-ups. By introducing a model based solely on clinicopathological information as model 1 and adding four IHC indicators in model 2, a significant improvement was observed in the area under the curve (AUC) values across the entire sample. The AUC value for OS curves increased from 0.765 to 0.872, and the AUC for RFS curves rose from 0.791 to 0.882. Thus, the present study's model effectively predicts patients' probability of OS and RFS using these factors. This prediction capability can guide postoperative treatment plans and follow-up intervals, potentially enhancing long-term survival for patients with EC.