Oncology's AI Adoption Hinges on Weighing Benefits vs Risks and Confronting Limitations

Breast Cancer Briefing, hosted by Sara Nunnery, MD, MSCI, a breast medical oncologist and the director of Breast Cancer Research at Tennessee Oncology in Nashville, is a podcast series that breaks down the latest news in breast cancer research, one conversation at a time.

In this episode, Dr Nunnery sat down with Matthew Kurian, MD, a hematologist/oncologist at St. Elizabeth Healthcare in Edgewood, Kentucky, as well as an assistant professor of medicine at the University of Kentucky College of Medicine in Highland Heights.

The experts discussed the role of artificial intelligence (AI) in oncology, beginning with foundational concepts. Dr Kurian explained that AI comprises systems trained to recognize patterns from large datasets, distinguishing generative large language models, such as ChatGPT, from retrieval-grounded tools, such as OpenEvidence, which draw from trusted sources like ASCO and the National Comprehensive Cancer Network (NCCN) Guidelines. He cautioned about pitfalls of AI, including hallucinations and automation bias, and stressed the importance of verifying AI outputs and using effective prompts.

The conversation then addressed AI scribes, with Dr Kurian citing data showing that primary care physicians derive the most benefit from these tools, whereas surgical specialists derive the least. He noted that although scribes help clinicians be more present, these tools have not reduced time spent working after hours, which he attributed to an efficiency paradox in which saved time is reabsorbed by additional patients or tasks.

Turning to FDA-authorized AI devices, Dr Kurian traced AI’s history in mammography from computer-aided detection to Clairity Breast, which uses existing mammograms to calculate a 5-year breast cancer risk in patients starting at age 35 years. He also discussed access and reimbursement limitations associated with this technology.

Finally, Dr Kurian issued a call to action urging ASCO and the NCCN to require prospective validation before integrating AI tools, warning against reliance on outdated retrospective data and the risk of widening disparities, while emphasizing balanced, responsible adoption of these tools.