NarromaticAI

Abstract

Medical writers play a critical role in the development of clinical study report (CSR) documents by synthesizing and organizing data from clinical trials. They collaborate closely with researchers, statisticians, and other experts to ensure that the CSR accurately and comprehensively presents the study findings in adherence to regulatory guidelines. This involves incorporating the study protocol, statistical analyses, safety and efficacy results, and other pertinent information. Medical writers also prioritize clarity, conciseness, and adherence to the highest standards of scientific and medical accuracy in the CSR. In the past, the medical writing process was largely manual. Previous attempts to automate aspects of the writing process relied on rule-based approaches, which were precise but inherently time-consuming to maintain, extend, and adjust. However, recent advancements in Generative Artificial Intelligence (GenAI) and Large Language Models (LLMs) have significantly improved the potential for generating text from data. We have developed an end-to-end Shiny application to support Medical Writers in drafting CSRs. This application provides secure access to clinical data tables, interactive configuration and filtering of data tables, and the generation and adjustment of scientific texts using LLMs. It also facilitates the storage of generated content in document templates which include Quarto reports. The application is currently in productive use, demonstrating robustness and modularization, making it easily extendable.