Integrating Large Language Models in Cancer Decision-Making: A Comprehensive Systematic Review and Meta-Analysis

In our systematic review, we analyzed 56 studies focusing on AI-based chatbots and their role in clinical decision support for cancer patients. Following rigorous screening, we identified a diverse geographic representation, with 26 studies published in 2023 and 30 in 2024, primarily from the USA. Most utilized OpenAI’s ChatGPT, with a focus on various cancer types like lung and breast cancer. Funding discrepancies were noted, particularly for lung cancer, which, despite high mortality, received proportionately less National Cancer Institute funding. The studies showed positive relationships between LLM research focus and cancer mortality, yet gaps in funding may hinder robust research. An average accuracy of 76.2% for LLM-assisted decision-making was found in our meta-analysis. Most studies relied heavily on automated evaluations, suggesting a need for a balanced approach incorporating qualitative assessments. This review emphasizes the potential and challenges of LLMs in cancer care, advocating for mixed-methods research to better inform clinical practices.

Source link