TOPLINE:

A digital pathology multimodal AI test predicted a benefit from adding abiraterone to standard long-term androgen deprivation therapy (ADT) and radiotherapy in very high-risk, nonmetastatic prostate cancer. In very high-risk patients identified through the test, adding abiraterone to standard therapy improved 5-year metastasis-free survival from 62% to 81%, whereas patients identified as standard high-risk showed no statistically significant improvement with abiraterone.

METHODOLOGY:

• The multimodal AI test has previously been shown to identify patients with very high-risk localized prostate cancer, but whether it can predict which patients are most likely to benefit from treatment intensification with abiraterone remains less clear. Given the toxicity and cost of abiraterone, “a predictive biomarker to refine patient selection is needed,” the authors said.
• The current retrospective study included 1137 patients who had been randomized to standard therapy (n = 583) or standard therapy with abiraterone (n = 554) in two sequential STAMPEDE phase 3 trials.
• Multimodal AI scores were generated using digital pathology images, PSA, tumor stage, and age. The AI classified patients as very high-risk (n = 268, upper quartile) or standard high-risk (n = 869) using a previously established 75th percentile threshold. Median follow-up was 6.1 years.
• Researchers assessed whether multimodal AI-defined risk predicted differential benefit from abiraterone, with a primary endpoint of metastasis-free survival and secondary outcomes including overall survival, prostate cancer-specific mortality, and time to distant metastases.

TAKEAWAY:

• Adding abiraterone to standard therapy was associated with a significant improvement in metastasis-free survival in the very high-risk group (hazard ratio [HR], 0.47) but not in the standard high-risk group (HR, 0.83; 95% CI, 0.63-1.09). Patients in the very high-risk group had significantly longer metastasis-free survival with abiraterone, with an estimated 5-year rate of 81% vs 62% with standard therapy alone.
• In very high-risk patients, the abiraterone benefit was consistent across nodal subgroups (HR, 0.45 for node-negative disease and HR, 0.48 for node-positive disease); there was no evidence of an abiraterone benefit across nodal subgroups in standard high-risk patients.
• At 8 years, adding abiraterone to standard therapy reduced prostate cancer-specific mortality from 36% to 12% in the very high-risk group, but showed no noticeable improvement in the standard high-risk group, where rates remained at approximately 13% in both treatment arms.
• Digital pathology image features accounted for 77.4% of the model's predictive signal for abiraterone benefit; the number of clinical high-risk factors alone showed no significant differential treatment effect (P for interaction = .37 for three vs two factors).

IN PRACTICE:

“This is the first biomarker study to predict abiraterone efficacy in very high-risk localized prostate cancer,” the study authors concluded. “Incorporating [multimodal AI] into treatment decision-making could potentially improve personalization of care and optimize the balance between therapeutic benefit and treatment burden for very high-risk prostate cancer.”

SOURCE:

The study, led by CTA Parker, PhD Researcher, University College London Cancer Institute, London, England, was published online in Annals of Oncology.

LIMITATIONS:

This was a retrospective, post hoc biomarker analysis and the predictive utility of the AI model has not yet been prospectively validated. The study did not collect race or ethnicity data, limiting assessment of generalizability across demographic groups. The STAMPEDE trial recruited patients prior to widespread adoption of next-generation imaging modalities such as prostate-specific membrane antigen positron emission tomography-computed tomography and magnetic resonance imaging, so metastatic status was determined using conventional imaging alone, which has lower sensitivity and specificity for detecting pelvic nodal disease and distant metastases. According to the authors, the long-term utility of combining newer staging techniques with multimodal AI-based risk stratification to guide treatment decision-making in high-risk localized prostate cancer will need evaluation in future studies. 

DISCLOSURES:

The study received support from Cancer Research UK’s Clinical Research Committee with educational grants from Novartis, Sanofi-Aventis, Pfizer, Janssen Pharma NV, Astellas, and Clovis Oncology. Artera, Inc, contributed the multimodal artificial intelligence algorithm and scoring for this study. University College London outlicensed the clinical and pathology data to Artera, Inc, for commercial use, and both organizations could gain commercially from clinical implementation. Parker received funding from a Clinical PhD Fellowship from the Jean Shanks Foundation and the Pathological Society of Great Britain & Ireland. Additional disclosures are noted in the original article.

This article was created using several editorial tools, including AI, as part of the process. Human editors reviewed this content before publication.