Skip to main content
ASTRO 2024 Main banner
Icon Legend
Presentation Icons
Event Add-on
Event Add-on
In Person Only
In Person Only
Off-site Event
Off-site Event
Location: Marriott Marquis
Location: Marriott Marquis
Back
Main Session View Poster

PQA 08 - PQA 08 Genitourinary Cancer, Patient Safety, and Nursing/Supportive Care Poster Q&A

3280 - Multi-Class Prediction Using Machine Learning to Predict Pathologic Stage in Prostate Cancer

Tuesday, October 1, 2024
2:30 PM – 3:45 PM ET
Location: Hall C

Screen: 22

H. Semwal1, A. Sabbagh2, S. Maroongroge3, Y. R. Li3, S. M. Glaser3, A. Amini3, B. Yuh4, J. Yoshida5, S. Liu6, O. Mohamad7, P. Lee8, J. Y. C. Wong3, S. V. Dandapani3, and C. J. Ladbury3; 1Department of Integrative Biology and Physiology, University of California Los Angeles, Los Angeles, CA, 2University of California San Francisco, San Francisco, CA, 3Department of Radiation Oncology, City of Hope National Medical Center, Duarte, CA, 4Department of Surgery, City of Hope National Medical Center, Duarte, CA, 5Department of Surgery, Lennar Foundation Comprehensive Cancer Center, City of Hope National Medical Center, Irvine, CA, 6Department of Medical Oncology, Lennar Foundation Comprehensive Cancer Center, City of Hope National Medical Center, Irvine, CA, 7Department of Radiation Oncology, The University of Texas MD Anderson Cancer Center, Houston, TX, 8Department of Radiation Oncology, Lennar Foundation Comprehensive Cancer Center, City of Hope National Medical Center, Irvine, CA

Purpose/Objective(s): Prediction of pathologic stage in patients with clinical node-negative (cN0) prostate cancer is critical for determination of optimal treatment options. Though several nomograms exist (with only the Partin tables capable of multi-class prediction [or classification within more than two categories]), machine learning (ML) approaches might allow for improved performance through their ability to capture and model complex interactions. Herein, we developed a multi-class ML model to predict pathologic stage and explained how variables were used to produce model predictions using an explainable artificial intelligence (XAI) framework. Materials/

Methods: Patients with cN0 prostate adenocarcinoma diagnosed from 2018-2020 were identified in the National Cancer Database. The query was limited to patients with known clinical staging and biopsy results who did not receive neoadjuvant therapy prior to pelvic nodal examination. Patients were classified as having organ confined disease (OC), extracapsular extension (ECE), seminal vesicle invasion (SVI), or lymph node involvement (LNI). These were considered mutually exclusive, so patients were classified as the highest stage they met criteria for. Variables incorporated into the model were age, core biopsy results, Gleason score, preoperative prostate-specific antigen (PSA), and clinical T-stage. Model performance was measured using one-versus-rest area under the receiver operator characteristic curve (AUC) on a holdout testing dataset. The best-performing model was explained using SHapley Additive exPlanation (SHAP) values. For model interpretation, SHAP values were converted into odds ratios (OR), confidence intervals (CI), and p-values.
Results: A total of 23,131 patients met inclusion criteria; 8517 (49.5%), 4898 (28.4%), 1660 (9.6%), and 2140 (12.4%) had OC, ECE, SVI, and LNI, respectively. Random Forest model outperformed all other models with an AUC of 0.706 (95% CI: 0.691-0.720) compared to 0.691 logistic regression (95% CI: 0.676-0.705; p<.001) and the Partin Tables 0.674 (95% CI: 0.659-0.688; p<.001). The most important overall model features were primary Gleason, percentage of positive cores (PPC), and PSA. The most important features for each class are summarized in Table 1.
Conclusion: Our ML achieved improved performance relative to logistic regression and the Partin Tables at predicting pathologic stage. XAI provided insight into the inner workings of the ML model. ML can be used to identify patients at risk for pathologic upstaging and therefore inform clinical decision-making. Abstract 3280 - Table 1
Most Important Features OC ECE SVI LNI
1st Primary Gleason (4 vs 3; OR: 0.886; p</span><.001) PPC (OR: 1.001; p<.001) Primary Gleason (5 vs 3; OR: 1.053; p<.001) Primary Gleason (5 vs 3; OR: 1.117; p<.001)
2nd PPC (OR: 0.997; p</span><.001) PSA (OR: 1.001; p</span>=.013) PPC (OR: 1.001; p<.001) PSA (OR: 1.003; p<.001)
3rd PSA (OR: 0.995; p</span><.001) Positive cores (OR: 1.003; p<.001) PSA (OR: 1.001; p<.001) Total Gleason (10 vs 5; OR: 1.029; p<.001)