Data Privacy and Anonymisation in Clinical Trial Data Sharing

Abstract

Transparent sharing of clinical trial data is now recognised as essential for evidence-based medicine and earning trust in the pharmaceutical industry. Data sharing initiatives have developed rapidly over the past decade, but the protection of patient privacy remains a critical consideration in the context of sharing patient-level data. There are therefore dual ethical considerations: on the one hand, to share data in service of advancing science, and on the other, to safeguard the privacy of research participants. This thesis explores the evolving ecosystem of clinical data sharing with a focus on the challenges and opportunities surrounding data privacy and anonymisation. The outputs aim to balance transparency with the ethical duty to safeguard patients’ privacy. The thesis examines and highlights tensions between ethical frameworks, regulatory policies and stakeholder perspectives that influence data sharing practices. A particular focus is placed on the technical complexities of anonymising clinical trial data, particularly free text format, in a way that preserves data utility. To investigate the challenges in depth, this research focusses on a case study of patient safety narratives from clinical study reports (CSRs) in the pharmaceutical industry. These are used to illustrate the challenges and opportunities of anonymising rich and disclosive patient data. A mixed methods approach is used comprising of qualitative interviews with pharmaceutical industry professionals revealing practical real-world challenges, content analysis of patient safety narratives in the context of natural language processing (NLP) for anonymisation, and application of quantitative statistical disclosure control (SDC) methods to examine their effectiveness in reducing the risk of re-identification while maintaining data utility in patient safety narratives. The findings show that despite the development of regulatory guidance and technical innovation, significant limitations remain in current anonymisation practices for clinical data sharing when applied to unstructured data. Key insights include that: (1) current guidance for anonymising structured data does not map effectively to unstructured components of a CSR (i.e patient safety narratives), (2) standard NLP metrics need to be tailored for anonymisation workflows, (3) current SDC techniques and the regulatory threshold for risk of re-identification dramatically reduce data utility of unstructured text in small clinical trials. The thesis proposes practical recommendations for improving anonymisation methodologies and advocates for a more standardised but context-sensitive approach to clinical data sharing. This research contributes towards bridging the areas of data privacy protection, clinical data transparency and NLP to offer insights for data processors and regulators, whether they are based in academia, industry or governance agencies so that the ecosystem can continue to work towards responsible and ethical clinical data sharing.

Date of Award	19 Dec 2025
Original language	English
Awarding Institution	The University of Manchester
Supervisor	Mark Elliot (Main Supervisor), Elaine Mackey (Co Supervisor) & Goran Nenadic (Co Supervisor)