This contribution shows how unstructured medical device safety data can be transformed into actionable knowledge to support early-stage product development. Focusing on national recall and safety reports from the German Federal Institute for Drugs and Medical Devices (BfArM), the authors developed a structured approach for extracting, analysing, and applying real-world failure data. Using Large Language Models (LLMs), unstructured reports were converted into a searchable, categorized dataset highlighting root causes and common failure patterns. To evaluate the practical value of this structured data for future product development of medical devices, a live lab study was conducted. The participants used the data to model product profiles and assessed the usefulness of the database. Results show that the data has potential to support the generation of specific product profile elements. The research demonstrates that regulatory safety data, when properly processed, can offer value in guiding product development actions.
Link et al. (Wed,) studied this question.