Reproduce “Using Large Language Models to Generate Clinical Trial Tables and Figures” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Yumeng Yang, Peter Krusche, Kristyn Pantoja, Cheng Shi, Ethan Ludmir, Kirk Roberts, Gen Zhu
Researchers demonstrated that large language models (LLMs) can accurately generate clinical trial tables, figures, and listings (TFLs) by translating natural language prompts into Python code, achieving 100% accuracy on 16 tables and 1 Kaplan-Meier plot from a CDISC pilot benchmark dataset. This work from Novartis and the University of Texas also developed a two-layer agent to make TFL generation accessible to non-expert users.
Key insights
- LLM-generated executable code is more accurate for TFL creation than direct table population by the LLM, enhancing transparency and traceability.
- Prompt structures demonstrated promising generalizability, requiring only minor modifications to generate TFLs across different clinical trials and varying endpoints.
- The developed agent effectively bridges the gap between natural language requests and precise code generation, offering a practical solution for operationalizing LLMs in a regulated environment.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.