An artificial intelligence (AI) tool is more effective than clinical prognostic markers for predicting long-term outcomes in patients with prostate cancer, according to a study presented at the ASCO Genitourinary Cancers Symposium 2022.
“AI tools have significant advantages over traditional risk stratification tools for patients with prostate cancer,” said Osama Mohamad, MD, PhD, of the University of California, San Francisco, when presenting this study at the meeting.
“They can perform across different needs…, they are quick, they are robust, they can be developed on thousands of patient samples without consuming any tissues. Thus, we think that AI tools represent a very good solution for global expansion of very useful prognostic markers for prostate cancer.”
Continue Reading
With their study, Dr Mohamad and colleagues used data from 5 phase 3 prostate cancer trials. To develop a multimodal AI tool, the researchers used pathology image data as well as clinical data, including T-stage, baseline prostate-specific antigen (PSA) level, combined Gleason score, primary Gleason, secondary Gleason, and age.
The researchers then developed 6 models for predicting 6 outcomes — biochemical failure at 5 years and 10 years, distant metastasis at 5 years and 10 years, 10-year prostate cancer-specific survival (PCSS), and 10-year overall survival (OS).
The researchers used 16,204 pretreatment histopathology slides from 5654 patients to train and validate the models. The team used 80% of the samples to train the models and 20% to validate them.
The researchers then compared the AI tool to National Comprehensive Cancer Network (NCCN) guidelines, which use PSA, digital rectal examination, and Gleason score to risk stratify patients.
The AI tool outperformed the NCCN model for all endpoints, including:
- 5-year distant metastasis (area under the curve [AUC], 0.837 vs 0.735)
- 10-year distant metastasis (AUC, 0.781 vs 0.701)
- 5-year biochemical failure (AUC, 0.670 vs 0.585)
- 10-year biochemical failure (AUC, 0.657 vs 0.602)
- 10-year PCSS (AUC, 0.765 vs 0.677)
- 10-year OS (AUC, 0.652 vs 0.585).
Based on these results, Dr Mohamad concluded that the AI tool can successfully predict long-term, clinically relevant outcomes for patients with prostate cancer.
Disclosures: Some study authors declared affiliations with biotech, pharmaceutical, and/or device companies. Please see the original reference for a full list of disclosures.
Reference
Esteva A, Feng J, Huang S-C, et al. Development and validation of a prognostic AI biomarker using multi-modal deep learning with digital histopathology in localized prostate cancer on NRG Oncology phase III clinical trials. Presented at ASCO GU 2022; February 17-19, 2022. Abstract 222.
This article originally appeared on Cancer Therapy Advisor
