Clinical utility of large language models in metastatic prostate cancer: A multicenter expert validation for decision support - PubMed
5 hours ago
- #metastatic prostate cancer
- #large language models
- #clinical decision support
- Study evaluates the clinical utility of large language models (LLMs) in metastatic prostate cancer (mPCa) treatment planning.
- Five contemporary LLMs were tested across 238 mPCa cases from three tertiary centers (2018-2025).
- Outputs were assessed for safety and utility, with safety rates ranging from 79.0% to 84.9%.
- Mean utility scores for safety-passed outputs were in the low-to-mid 4 range on a 5-point Likert scale.
- Failures were most common in complex cases with incomplete documentation, involving errors like missing data extraction and guideline deviations.
- LLMs show potential as drafting assistants in mPCa care but require supervision due to safety concerns.