Results from a Phase 3 randomized trial were published in 2013. Evidence of NEJM New artificial intelligence (AI) model may help determine which patients with localized prostate cancer would benefit from adding androgen deprivation therapy (ADT) to radiation therapy doing.1
Although ADT consistently improves oncologic outcomes when added to radiotherapy for prostate cancer, ADT-associated toxicities can have a significant impact on patient quality of life. Therefore, distinguishing between those who are most likely to benefit from ADT and those who are unlikely to benefit will minimize toxicity in patients whose disease is likely to respond well to radiation therapy alone. key to contain it.
Daniel Spratt, M.D., lead author of this paper.
Credit: Case Western Reserve University
The authors found that there is no valid method to determine who will benefit from ADT, and current guidelines are based on prognostic National Comprehensive Cancer Network (NCCN) risk stratification and other factors such as Gleason grade and imaging biomarkers. pointed out that it depends on the method of
“Although some of these markers have shown prognostic value, validation of randomized trials has not shown them to function as predictive biomarkers for ADT use,” the authors wrote. “A simple and reliable method to guide the individualized use of ADT with radiotherapy for men with localized prostate cancer would be of value to such patients.”
The predictive AI model validated in this study was trained using digitized data (including pathological images) from four NRG Oncology Phase 3 randomized trials: Radiation Therapy Oncology Group (RTOG)-9202 (NCT00767286), RTOG-9413 (NCT00769548), RTOG-9910 (NCT00005044), and RTOG-0126 (NCT00033631). Data from a patient in the NRG/RTOG-9408 trial (NCT00002597) to validate the model’s ability to identify a man with localized prostate cancer likely to benefit from her ADT in addition to radiation therapy used.
We trained the model using data from 5,727 patients and stratified patients based on whether they were likely to benefit from ADT or not. Most patients in the model development cohort (n = 2024) had intermediate-risk disease with a median follow-up of 10.6 years. A total of 1,050 (52%) patients received radiotherapy alone and 974 (48%) patients received short-term ADT in addition to radiotherapy.
The final model consisted primarily of histopathological features such as Gleason scores and imaging results. “Although histopathological features contribute significantly, MMAI architectures utilize deep learning to capture interaction effects as well, and models benefit from learning all features,” the authors write. .
The validation cohort included 1594 patients, of whom 806 received radiotherapy alone and 788 received radiotherapy and brief ADT. The validation cohort had a median follow-up of 14.9 years, demonstrating a significant overall improvement in outcome with ADT.
Among the 543 patients in the validation cohort whom the model judged likely to benefit from ADT, the risk of distant metastases was significantly lower than with radiation therapy alone. ADT had no effect in her 1,501 patients in whom the model predicted little or no effect.
“As patient prognosis deteriorates (i.e., moving from NCCN low-risk to high-risk NCCN), the recommendation to add ADT to radiotherapy strengthens. Despite the evidence,” the authors write. “At present, among patients with positive and negative AI model predictions, baseline PSA [prostate-specific antigen], T stage, and NCCN risk group distributions were similar. There was a small difference in Gleason scores. These results confirm that historical classification of tumor aggressiveness alone is not sufficient to determine which patients derive differential relative benefit from her ADT. ”
Study limitations included model development using retrospective data, lack of advanced molecular imaging in the era of clinical trials, and potential disease grade shifts due to changes in the Gleason grading system. However, the authors noted that any potential bias due to grade transition would likely be random and affect both trial groups.
Overall, the authors concluded that AI-based predictive models show promise in guiding the use of ADT in combination with radiotherapy in localized prostate cancer.
“This is a true breakthrough in the treatment of prostate cancer,” said lead author Vincent K. Smith, Professor of Radiation Oncology, UH Seidman Cancer Center and Case Western Reserve University. said Daniel Spratt, M.D., a professor of oncology. in a statement.2 “The first-ever AI-generated predictive biomarker of ADT efficacy in prostate cancer further unlocks our ability to create personalized approaches for cancer treatment.”
References
1. Spratt DE, Tang S, Sun Y, et al. An artificial intelligence prediction model for hormone therapy in prostate cancer. square resolution. Published online June 29, 2023. doi:10.21203/rs.3.rs-2790858/v1
2. New research published in Evidence of NEJM In collaboration with researchers at the University Hospital Sideman Cancer Center and ArteraAI, we validated the first-ever predictive AI biomarker of the efficacy of androgen deprivation therapy (ADT) in prostate cancer. news release. university hospital. June 30, 2023. Accessed July 6, 2023. https://news.uhhospitals.org/news-releases/a-new-study-published-in-nejm-evidence-with-university-hospitals-seidman-cancer-center – Researchers and Altera Eye on Androgen Deprivation Therapy Validate First-Ever Predictive AI Biomarker – Validate Prostate Cancer Benefit.htm
