Artificial Intelligence in Medicine: How AI Can Identify When to Add ADT to Prostate Cancer Radiation

Aggressive prostate cancer (PCa) comes with higher chances for recurrence—and often there’s no way to know if the PCa has started to spread at the time of treatment. It’s like a time bomb with an unanswered question: has the clock been activated? Therefore, today’s clinical standard of care treats it as if the clock is already switched on and ticking. There’s an assumption that tumor cells may already have broken free of the gland. Thus, an aggressive primary treatment (radical prostatectomy or radiation) may be accompanied by androgen deprivation therapy (ADT), sometimes referred to a hormone therapy. Designed to be a pre-emptive strike, ADT cuts off the supply of male hormones (androgens), particularly the testosterone that can fuel the tumor’s activity. Although ADT is not curative, it buys time. More importantly, there is research evidence that for some radiation patients, it boosts the effectiveness of their treatment.

The problem is, adding ADT doesn’t work that way for all radiation patients. Plus, it has unpleasant side effects such as loss of sex drive, hot flashes, breast tenderness, bone loss, etc. Why put a patient through that if it’s not going to help? That raises yet another problem. How can doctors know in advance which patients will benefit and which won’t?

Artificial Intelligence identifies biomarkers

A paper presented at the 15th Annual Prostate Cancer Congress (March 11-12, 2022) presents a solution that was developed using Artificial Intelligence (AI). A multi-disciplinary, multi-institution team used a multimodal deep learning architecture to identify predictive biomarkers based on clinical and biopsy factors.^[i] Their cohort included 5,654 patients who underwent radiation with or without ADT, and whose pretreatment biopsy slides were available. The average follow-up time for the training cohort (3,935 cases) was over 13 years, so it was possible to know which men developed distant metastasis and which did not, and whether their radiation included ADT or not.

The AI algorithm analyzed cases with and without mets, and those who had received ADT vs those who didn’t. By doing so, it identified a biomarker common to those who had received ADT and had less probability of developing mets. According to their results, those who had the AI-biomarker benefited from receiving ADT to boost their radiation, while those who did not have the AI-biomarker did not. For those who had the AI-biomarker, at 15 years after treatment, the radiation + ADT group had significantly lower rates of metastasis than the radiation alone group.

Thus, the biomarker developed utilizing deep learning methodology appears able to predict which patients with aggressive PCa (that is, unfavorable intermediate-risk and high-risk disease) will be helped by also receiving ADT. If they have the AI-biomarker, it will be worth going through ADT’s side effects.

NOTE: This content is solely for purposes of information and does not substitute for diagnostic or medical advice. Talk to your doctor if you are experiencing pelvic pain, or have any other health concerns or questions of a personal medical nature.

[i] Spratt DE, Sun Y, Van der Wal D, Huang S et al. An AI-derived digital pathology-based biomarker to predict the benefit of androgen deprivation therapy in localized prostate cancer with validation in NRG/RTOG 9408. J Clin Onc. 2022 Feb 20; 40(6):223-223.