AI Tool Accurately Predicts Prostate Cancer Outcomes Across Racial Groups

Mack Roach III, MD, professor of radiation oncology, medical oncology, and urology at the University of California, San Francisco, discusses the development and evaluation of a multimodal artificial intelligence (AI) algorithm designed to predict prostate cancer outcomes and its performance across racial subgroups.

According to Roach, the motivation behind the study stemmed from a longstanding controversy in prostate cancer: the role of race in determining outcomes. While prostate cancer incidence is 1.5 to 2 times higher among Black men compared with other racial groups, the question of whether biological differences contribute to disparities in outcomes has been widely debated.

“For many years, there have been different opinions about whether there is an inherent biologic factor or not, and so it is important for us to distinguish between the incidence of the disease and the biologic behavior of a disease once it is diagnosed,” he explains.

To address this, researchers turned to a high-quality data set derived from prospective phase 3 randomized clinical trials. These trials ensured uniform treatment protocols, patient stratification, and systematic follow-up. “The value of that resource,” Roach explains, “is that the quality of care and the eligibility are controlled in such a way that biases do not really enter into the quality of treatment.” Importantly, variables such as insurance coverage, treatment type, and dose were standardized, which allowed for more accurate analysis of outcomes by race.

This dataset provided a unique opportunity to test whether AI could not only enhance prognostic accuracy, but also maintain fairness across racial groups. The AI model analyzed digitized biopsy slides along with clinical features like prostate-specific antigen, tumor grade, and stage, uncovering subtle prognostic indicators not readily visible to human pathologists.

The study, published in JCO Clinical Cancer Informatics, demonstrated that the algorithm accurately predicted outcomes such as recurrence and metastasis and did so without introducing racial bias. The results support the broader application of AI in oncology while reinforcing the importance of diverse, well-controlled data in developing equitable predictive tools.

REFERENCE:
Roach M 3rd, Zhang J, Mohamad O, et al. Assessing algorithmic fairness with a multimodal artificial intelligence model in men of African and non-African origin on NRG oncology prostate cancer phase III trials. JCO Clin Cancer Inform. 2025;9:e2400284. doi:10.1200/CCI-24-00284

Continue Reading