Abstract
OBJECTIVE: The rise of artificial intelligence (AI) and large language models like Llama, Gemini, or Generative Pretraining Transformer (GPT) signals a promising new era in natural language processing and has significant potential for application in medical care. This study seeks to investigate the potential of GPT-4 for automated therapy recommendations by examining individual patient health record data with a focus on gynecologic malignancies and breast cancer. METHODS: We tasked GPT-4 with generating independent treatment proposals for 60 randomly selected patient cases presented at gynecologic and senologic multidisciplinary tumor boards (MDTs). The treatment recommendations by GPT-4 were compared with those of the MDTs using a novel clinical concordance score and were reviewed both qualitatively and quantitatively by experienced gynecologic oncologists. RESULTS: GPT-4 generated coherent therapeutic recommendations for all clinical cases. Overall, these recommendations were assessed by clinical experts as moderately sufficient for real-word clinical application. Deficiencies in both accuracy and completeness were especially noted. Using a quantitative clinical concordance score, GPT-4 consistently demonstrated superior performance in managing the senologic cases compared with the gynecologic cases. Iterative prompting substantially enhanced treatment recommendations in both categories, increasing concordance with MDT decisions to up to 84% in senologic cases. CONCLUSION: GPT-4 is capable of processing complex patient cases and generates detailed treatment recommendations; however, differences persist in surgical approaches and the use of systemic therapies, and there is a tendency toward recommending excessive genetic testing. As AI-powered solutions continue to be integrated into medicine, we envision the potential for automated therapy recommendations to play a supportive role in human clinical decision-making in the future.