The Data Transparency Winter Event 2025 was a fantastic three days of discussion, collaboration, and insight, bringing together 857 attendees from across the industry.

From emerging trends in AI to innovative approaches in data sharing, the level of engagement and knowledge exchange was truly inspiring! All recordings and presentation slides are available on the PHUSE Advance Hub.

Continue reading for reflections from our Data Transparency Leads.

Day 1

– Jean-Marc Ferran, Qualiance

The Data Transparency Winter Event Day 1 delved into the complexities of data anonymisation in clinical trials and healthcare, focusing on regulatory compliance and emerging methodologies. Discussions highlighted the challenges of balancing data utility with privacy protection, particularly in light of regulatory frameworks such as EMA Policy 0070 and the EU Clinical Trials Regulation. Anju Bhatia and Swetha Ram (Novo Nordisk) shared their experiences with anonymised document publication. The session provided an in-depth look at techniques for handling direct and indirect identifiers, including pseudonymisation, suppression and generalisation, and detailed the submission life cycle, from pre-submission meetings with health authorities to the final publication of anonymised documents.

Lisa Pilgram (University of Ottawa)’s presentation explored the regulatory uncertainty surrounding anonymisation, with discussions on the legal definitions of identifiability across jurisdictions in Canada and the perspectives of privacy regulators across provinces on anonymised data and AI/ML models trained on such data. Recommendations included developing codes of practice for anonymisation, creating incentives for implementing and improving anonymisation, and addressing the ethical implications of anonymisation decisions.

The event also introduced new methodologies aimed at improving data quality and explainability, responding to concerns that conservative anonymisation approaches often compromise data utility. A novel approach based on risk analysis methods (such as EBIOS) from the cybersecurity space was presented by Louis Philippe Sondeck (Clever Identity), considering both exposure and severity of re-identification risk. This approach aims to provide greater clarity on reference populations and heighten transparency in anonymisation decisions.

We concluded the event with a presentation from Lora Killian (Pfizer) on EFPIA anonymisation gradient, illustrating the contexts clinical data is being shared in and the measures to be taken.

Overall, the event underscored the need for ongoing innovation and collaboration to tackle the evolving challenges of data transparency. Experts stressed the importance of regulatory clarity, risk-based anonymisation techniques and balancing privacy with data utility for meaningful and ethical data sharing in clinical research.

Day 2

– Devaki Thavarajah, Independent

The event, held on 5 February 2025, brought together experts from various fields to discuss the integration of artificial intelligence (AI) in clinical trial transparency, lay summary creation, and the enhancement of ClinicalTrials.gov. The presentations showed the potential of AI to streamline processes, increase data consistency, and improve patient and caregiver experiences, while addressing the challenges and risks associated with AI adoption.

Kim Edwards and Julie Holtzople discussed the need for responsible AI use in creating lay summaries to avoid inaccuracies, inconsistencies and culturally insensitive language. They stressed that AI should augment, not replace, human involvement, particularly in areas such as health literacy refinement and review of drafts. The importance of transparency, data privacy and maintaining trust through accurate and culturally sensitive communication was also highlighted.

Samantha Toscano and Megan Larkin presented strategies for enhancing ClinicalTrials.gov, with a focus on addressing health literacy disparities and improving data consistency. They noted that less than 1% of clinical trials have summaries written at an 8th-grade reading level, which is crucial for empowering patients and caregivers. Recommendations included simplifying language, improving search and navigation features, and using tools such as the Flesch—Kincaid readability test for clarity and accessibility.

Shalini Dwivedi analysed compliance with Annex IV of EU CTR 536/2014 in clinical trial results posted on the Clinical Trials Information System (CTIS). Her findings revealed that only 8.1% of trials had published result summaries, with 24.3% of those excluded from evaluation due to non-English language or early termination. The analysis drew attention to the need for more awareness, clarity and validation in the disclosure of clinical trial results to ensure meaningful and consistent information for healthcare professionals and researchers.

Woo Song from Xogene explored the concept of AI agents and their application in clinical trial transparency. AI agents, characterised by autonomy, goal orientation and learning capacity, were demonstrated in tasks such as protocol registration, results posting and plain language summary writing. The presentation showcased an AI agentic network in action, with the potential for increased autonomy, broader tool use and smarter decision-making. The integration of AI in these processes aims to boost efficiency and accuracy while maintaining human oversight.

The event demonstrated the transformative potential of AI in clinical trial transparency and communication, and the need for responsible implementation, human involvement and continuous improvement to address challenges and guarantee ethical and effective use of AI technologies.

Day 3

– Muhammad Oneeb Rehman Mian, Privacy Analytics

The final day of the Winter Data Transparency Event centred on the transformative role of artificial intelligence (AI) and advanced data management techniques in addressing data privacy, compliance and anonymisation challenges in the pharmaceutical and healthcare industries. Key themes included the use of synthetic data, AI-augmented anonymisation, and the sharing of medical imaging data, for strengthening data utility while ensuring privacy and regulatory compliance.

Zixin Nie explored the challenges of sharing and de-identifying medical imaging data. The NHLBI BioData Catalyst® initiative was presented as a cloud-based ecosystem that facilitates data sharing and analysis while maintaining privacy. A Python-based open-source pipeline for DICOM file de-identification was showcased, combining optical character recognition (OCR) and natural language processing (NLP) to detect and remove personal identifying information (PII) from medical images. This approach allows for selective de-identification, to preserve data utility while protecting patient privacy.

Synthetic data emerged as a critical tool for overcoming privacy barriers in clinical trials and research. Jaskaran Singh Saini’s presentation covered replicating real-world data without exposing sensitive information and how synthetic data allows researchers to conduct experiments and develop models securely. Techniques such as bootstrapping, rule-based generation and AI-driven methods were highlighted, alongside Python libraries such as NumPy, Pandas and synthcity, which facilitate synthetic data creation. Jaskaran also discussed the significance of measuring synthetic data accuracy through methods such as the chi-square test and Wasserstein distance.

AI’s role in data privacy and anonymisation was the focus of Zach Weingarden’s presentation, with discussions on how AI can advance traditional anonymisation methods. AI models can assess re-identification risks, recommend risk assessment methodologies and optimise data utility within privacy thresholds. Use cases included augmented medical writing, where AI significantly reduced the time required to draft informed consent forms, and the identification of commercially confidential information (CCI) in regulatory documents. AI’s ability to understand context and industry-specific jargon improved the accuracy and efficiency of CCI redaction, balancing privacy with data utility.

The importance of context in anonymisation featured throughout the event, particularly during the final presentation by Nastazja Laskowski, in free-text patient data from clinical study reports. NLP techniques were discussed as effective tools for identifying sensitive information, but challenges remain in achieving foolproof anonymisation due to the nuanced nature of language and the high stakes of missing even a single instance of sensitive data. Intruder scenarios looked at the risks posed by different levels of knowledge and motivation, and emphasised the need for robust anonymisation strategies.

In summary, the event showcased how AI, synthetic data and advanced data management techniques are revolutionising data privacy and compliance in the pharmaceutical and healthcare sectors. By leveraging these technologies, organisations can accelerate scientific discoveries, improve operational efficiency and ensure regulatory compliance while safeguarding sensitive information. The discussions identified both the potential and the challenges of these approaches, calling for continued innovation and collaboration to achieve optimal outcomes.

Mark your calendars for the Data Transparency Autumn Event, which will take place on 16–18 September.

For companies looking to make an impact, sponsorship opportunities are open.

The earlier you secure your package, the more visibility and engagement you will gain. Put your organisation at the forefront of the data transparency conversation!

View the Return on Investment Summary and prospectus for more information. Apply here.

Thank You to Our Autumn Event Sponsors Already Signed Up!