Evaluating AI Tools for Prostate Cancer: What Urologists Should Ask

Daniel Spratt, MD, cautions that, despite the excitement surrounding AI, there have been instances in which AI has not delivered accurate information, and that low-quality input data will produce inaccurate output. Before incorporating any AI-based risk or diagnostic tool into patient counseling, he says, urologists should look to resources such as the National Comprehensive Cancer Network guidelines to determine whether a test has met a high evidentiary bar: first, that it outperforms standard clinical tools, and second, that it has undergone rigorous evaluation, including not only clinical validation but also calibration and benchmarking studies. He urges caution with any AI test until it has robust independent validation, noting that the MMAI test has been validated in more than 10 clinical trials in addition to the real-world data being presented.

Looking ahead, Spratt acknowledges that no one can predict exactly where AI is going, but says there is no question that adoption of digital pathology and AI-based digital pathology will continue to increase across diagnosis, prognostication, and prediction. With growing validation, he expects these tools to become commonplace for patients with prostate cancer and to extend to other cancer types as comparable data emerge.

Spratt's key takeaway for community urologists is that the benchmarking analysis showed the risk estimate reported by the MMAI test appears to be accurate and concordant with outcomes observed in STAR-CAP, a cohort of patients treated at many types of centers around the world. Urologists, he says, should have confidence in that estimate when helping patients make treatment decisions.