Artificial Intelligence Predictive Model for Hormone Therapy Use in Prostate Cancer.

Area of Science:

• Oncology
• Radiotherapy
• Artificial Intelligence

Background:

• Androgen deprivation therapy (ADT) combined with radiotherapy offers benefits for localized prostate cancer.
• ADT use is limited by negative quality of life impacts and a lack of predictive models.
• Validated predictive models are needed to guide ADT use in prostate cancer treatment.

Approach:

• Developed and validated an artificial intelligence (AI)-derived predictive model using digital pathology and clinical data from 5,727 patients across five phase III trials.
• The model assessed the benefit of radiotherapy with or without ADT, focusing on distant metastasis as the primary endpoint.
• Validated the AI model on the NRG/RTOG 9408 trial (1,594 patients) to evaluate its predictive accuracy for ADT benefit.

Key Points:

• In the validation cohort, ADT significantly improved time to distant metastasis (sHR=0.64, p=0.01).
• A significant interaction between the predictive model and treatment effect was observed (p-interaction=0.01).
• AI model-positive patients (34%) showed a significant reduction in distant metastasis risk with ADT (sHR=0.34, p<0.001), while model-negative patients did not (sHR=0.92, p=0.71).

Conclusions:

• An AI-based predictive model effectively identifies prostate cancer patients, particularly those with intermediate-risk disease, who benefit from short-term ADT.
• The model's validation on completed phase III trials supports its utility in clinical decision-making.
• This AI tool enables personalized treatment strategies, optimizing ADT use in conjunction with radiotherapy.