Androgen deprivation therapy (ADT) has long been established as an effective treatment option for high-risk localized prostate cancer (PC) when combined with radiotherapy (RT). However, determining the optimal duration of ADT remains a challenge. To address this issue, researchers have developed and validated the first predictive biomarker that can guide the decision-making process and maximize the benefits of ADT while minimizing risks. The study, presented at the 2023 ASCO Annual Meeting, utilized data from multiple Phase III NRG Oncology randomized trials to train and validate the predictive biomarker for long-term (LT) versus short-term (ST) ADT (Abstract 5001).

The team used AI to develop a predictive biomarker using digitized pretreatment prostate biopsy slides from six Phase III NRG/RTOG trials. The biomarker aimed to predict the benefit of LT-ADT in terms of distant metastasis (DM). Validation was performed on data from the RTOG 9202 trial, comparing RT + ST-ADT (4 months) and LT-ADT (28 months). The biomarker's predictive utility was evaluated for the primary endpoint of DM and the secondary endpoint of PC-specific mortality (PCSM) using Fine-Gray interaction models, accounting for competing risks from other causes and estimating event rates using the cumulative incidence method.

The AI biomarker was trained on 2,641 men (median follow-up: 9.8 years) and validated on 1,192 men (median follow-up: 17.2 years) from RTOG 9202 trial. In the validation cohort with 80 percent high/very high-risk features, LT-ADT significantly improved DM. The AI biomarker was prognostic for DM. A significant biomarker-treatment interaction (P=0.04) was observed. Among AI-biomarker-positive men (66% of cohort), LT-ADT reduced DM risk (sHR: 0.55, 95% CI: 0.41-0.73, P<0.001), while no benefit was seen in AI-biomarker-negative men (34% of cohort) (sHR: 1.06, 95% CI: 0.61-1.84, P=0.84). The 10-year DM rate difference between RT+LT-ADT and RT+ST-ADT was 13 percent in AI-biomarker-positive men and 2 percent in AI-biomarker-negative men. Similar trends were observed in PCSM outcomes. Risk classification based on NCCN criteria was prognostic but not predictive of LT-ADT benefit.

To gain further understanding of the research, Oncology Times interviewed Andrew J. Armstrong, MD, ScM, FACP, the lead author of the study and Professor in Medicine, Surgery, Pharmacology, and Cancer Biology, as well as Director of Research at Duke Cancer Institute's Center for Prostate and Urologic Cancer. With a specialization in medical oncology, Armstrong is an internationally recognized expert in experimental therapeutics and the advancement of biomarkers in genitourinary cancers, notably prostate cancer.

Oncology Times: What are some of the key findings of your research, and how might they impact the way that clinicians treat patients with localized high-risk prostate cancer?

Armstrong: "Currently, men with high-risk localized prostate cancer who pursue RT are treated with RT and LT-ADT. We sought to identify patients in this setting with a low risk of metastasis after ST-ADT who could be spared the risks of LT-ADT. In this study, we successfully validated the first predictive biomarker of LT-ADT benefit with RT in localized high-risk prostate cancer using an AI-derived digital pathology-based platform in the Phase III NRG/RTOG 9202 trial. The predictive AI biomarker identified 34 percent of men that could derive similar benefit with ST-ADT, avoiding the side effects of prolonged ADT, and 43 percent of intermediate-risk men who would benefit from LT-ADT."

Oncology Times: Can you discuss any potential limitations or challenges associated with using an AI-derived digital pathology-based biomarker to guide treatment decisions in this patient population?

Armstrong: "As with any model, generalizability is critical, and we are reassured that this AI biomarker demonstrated utility based on archival digitized biopsies from community and academic sites all over North America and in a randomized trial with 20 years of follow-up that was inclusive of intermediate and high-risk men of diverse backgrounds. Further external validation of our LT-ADT biomarker in diverse patient populations and in prospective controlled trials will be important to demonstrate the performance and utility of the AI biomarker to guide therapy decisions around both treatment de-intensification in biomarker-negative men, but also treatment intensification of biomarker-positive men."

Oncology Times: What are some of the next steps in this area of research and what types of studies are needed to further validate the use of this biomarker in clinical practice?

Armstrong: "Additional validation studies in more randomized controlled trials are important to prioritize. However, it's worth noting that there is precious few such datasets available. We found that about 20 percent of AI biomarker-positive men still suffered distant metastases over time despite LT-ADT, suggesting future trials of treatment intensification with novel AR inhibitors, PET-directed radiation, or novel agents like chemotherapy or molecularly targeted therapies could be evaluated to further improve the outcomes of these very high-risk patients. The potential to deepen the validation evidence for this biomarker will depend upon additional research collaborations with more clinical trial groups to help strengthen the evidence for clinical use in prostate cancer patients."

Dibash Kumar Das is a contributing writer.