We present the development of a methodology enabling the use of Large Language Models (LLMs) to assist in the creation of information sheets and informed consent forms for clinical trials. This study is conducted within the context of the research project REPO4EU (Precision drug REPurpOsing For EUrope and the world). REPO4EU is committed to adhering to both national and international legislation and ethical and privacy best practices. The project follows an ethics and privacy-by-design approach, addressing all ethical, legal, social, and safety concerns arising from its research. As part of its commitment to research ethics, this study aims to create project-specific templates for information sheets and informed consent forms and explore the potential of LLMs to facilitate this process.
For this purpose, we used a multi-step methodological approach and two pilot exercises have been conducted for REPO-HYPER II clinical trial. A bibliographic review to identify the essential elements of information sheets and informed consent forms for clinical trials has been performed (Tam et al., 2015; O'Sullivan et al., 2020; Pietrzykowski, et al., 2021; Solomon et al., 2021; Wu et al., 2024). This included peer-reviewed articles and relevant guidelines. Additionally, "grey literature" sources, such as reports and white papers from regulatory bodies, were explored. The next step was the development of a preliminary checklist to capture the key information required for both documents based on the bibliographic review. This checklist was then validated through analysis of 15 information sheets and informed consent forms of ongoing clinical trials registered in the EU Clinical Trials database. The focus of this analysis was on how these trials addressed the information needs of potential participants. The initial checklist was subsequently adjusted following the analysis of existing clinical trial documents.
Furthermore, an additional task involved the development of a general template for the information sheet and informed consent forms for this clinical trial [1] . These templates have been based on a template provided by the Swedish regulatory authority and supplemented with all the requirements identified in the previous exercise. Moreover, the recommendations outlined by Coleman et al. (2021) were taken into account in order to improve readability and, consequently, enhance the decision-making capacity of the participants.
However, the analysis and the design of these documents can be time-consuming, as each item and document must be scrutinised in detail. Therefore, we aim to leverage LLMs (Large Language Models) to assist in the creation and/or analysis of both the information sheet and the informed consent documents. LLMs are a type of deep learning model that are trained on large amounts of data to learn and generate text that mimics the natural language (Chang et al., 2024). As such, LLMs can be used in the future REPO4EU platform to automatically validate all items in information sheets and informed consent forms and to help the creation of these documents (Clusmann et al., 2023). This could lead to significant time savings, allowing clinical practitioners and researchers to focus more on their relationship with research participants and to have more flexible documents for participants with different levels of literacy. The implemented methodology could be transferable to other contexts, beyond REPO4EU and represents an ethics by design outcome of this project.
[1] The research team is grateful to the researchers of the Research Institute, particularly Markus Kastelitz, for their contributions to this part of the study.
Smart Citations