Top Line: Can we harness the power of AI to identify prostate cancer patients who benefit from long term ADT using biopsy specimens?
The Study: Recently, several studies have used genetic assays to predict the risk of recurrence and metastasis after prostate radiation as well as the potential benefit of treatment intensification. In this study, an Artificial Intelligence model was generated to predict the benefit of long term ADT over short term ADT in reducing the risk of distant metastasis for patients with intermediate and high risk prostate cancer treated with definitive radiation. The AI model was trained using biopsy specimens and clinicopathologic data from >6000 patients treated on 7 randomized, phase 3 RTOG trials that evaluated dose escalation or various durations of ADT. The model was then validated using patients treated on RTOG 9202, which showed a significant reduction in distant metastasis with 28 months versus 4 months of ADT. In the 9202 cohort, 66% of patients were identified by the AI model to benefit from long term ADT over short term ADT. In the overall 9202 cohort, the rate of DM at 15 years was 17% with long term ADT and 26% with short term. Among those predicted to benefit from long term ADT by the AI model, the rate of DM was significantly lower with long term ADT (19% v 33%). In contrast, there was no significant difference in DM risk at 15 years among those not predicted to benefit from long term ADT (11% v 11%). As we can see, the AI model was also a strong predictor of overall distant metastasis risk regardless of ADT duration.
TBL: This AI model, which analyzes biopsy slides and clinical features for patients with intermediate and high risk prostate cancer, identified ⅔ of patients who benefit from long term ADT and ⅓ of patients who do just as well with short term ADT. | Armstrong, J Clin Oncol 2025
The Study: Recently, several studies have used genetic assays to predict the risk of recurrence and metastasis after prostate radiation as well as the potential benefit of treatment intensification. In this study, an Artificial Intelligence model was generated to predict the benefit of long term ADT over short term ADT in reducing the risk of distant metastasis for patients with intermediate and high risk prostate cancer treated with definitive radiation. The AI model was trained using biopsy specimens and clinicopathologic data from >6000 patients treated on 7 randomized, phase 3 RTOG trials that evaluated dose escalation or various durations of ADT. The model was then validated using patients treated on RTOG 9202, which showed a significant reduction in distant metastasis with 28 months versus 4 months of ADT. In the 9202 cohort, 66% of patients were identified by the AI model to benefit from long term ADT over short term ADT. In the overall 9202 cohort, the rate of DM at 15 years was 17% with long term ADT and 26% with short term. Among those predicted to benefit from long term ADT by the AI model, the rate of DM was significantly lower with long term ADT (19% v 33%). In contrast, there was no significant difference in DM risk at 15 years among those not predicted to benefit from long term ADT (11% v 11%). As we can see, the AI model was also a strong predictor of overall distant metastasis risk regardless of ADT duration.
TBL: This AI model, which analyzes biopsy slides and clinical features for patients with intermediate and high risk prostate cancer, identified ⅔ of patients who benefit from long term ADT and ⅓ of patients who do just as well with short term ADT. | Armstrong, J Clin Oncol 2025