In the wake of COVID-19, this year’s PHUSE EU Connect saw significant emphasis on our responsibility to share high utility data sets using anonymization for research.
d-wise was honored to demonstrate Blur Anonymization software and conduct an on-demand recording around quantified risk assessment. Separately, our team was able to educate transparency leaders on what is actually happening inside “the black box” of Blur anonymization software, by using Microsoft Excel to anonymize datasets by hand in a hands-on workshop. Key session summaries are available below.
Hands-on De-identification Workshop Summary
d-wise clinical trial transparency expert, Cathal Gallagher, started the workshop with a demonstration on how to anonymise data in Excel. Throughout the workshop attendees learned about quantitative vs qualitative risk, anonymisation vs redaction, and regulatory requirements surrounding EMA Policy 0070 & Health Canada PRCI. With recent focus on the importance of sharing clinical trial data for research, it’s unsurprising that there seemed to be more interest in data anonymisation than document anonymisation throughout the conversation.
Considerable discussion occurred during the workshop around sponsor transparency strategies regarding the desired risk level for private vs public sharing as well as multi-dimensional risk assessment and guidance on which is most appropriate and when (e.g. a 5 year age band on one group and a 10 year age band on another).
d-wise is evolving along with transparency needs. Stay tuned for Blur 2.5 updates in 2021 to assist you in conducting your own multi-dimensional risk assessments.
Our team really enjoyed the following sessions from our clinical trial transparency peers.
DH01: Data Anonymisation and Risk Assessment Automation - PHUSE WG Deliverable (On-Demand)
This presentation, led by Lukasz Kniola of Biogen, introduced a recent white paper released by the PHUSE Data Transparency Working Group. The team behind this deliverable identified the steps to get from original data to the anonymized version and listed individual tasks and problems associated with each step - resulting in 10 major and 60 minor tasks total - that were merged into one cohesive document.
The paper also identified opportunities for automation and illustrated those possibilities with code snippets created in Python, R, SAS and even a few in Julia. Lukasz then conducted a high-level demo of applying automation when anonymizing clinical trial data and a brief comparison of languages.
Follow-On: If you’re interested in using validated open-source languages, check out TT06: The Validation of Open-source Tools for Clinical Statistics and Programming on-demand presentation presented with our partner, Mango Solutions to learn more about validating R in your environment.
Applications of Privacy-Enhancing Technology to Data Sharing at a Global Pharmaceutical Company
In this live, standing-room-only presentation Stephen Bamford, PHUSE founder and Head of Transparency at Janssen, took attendees through the evolving transparency landscape and addressed different sharing approaches, privacy enhancing techniques and how those can affect data controllers ability to balance risk and data utility.
We found the conversation around using synthesized data versus actual, anonymized patient data for sharing or software testing to be forward thinking in moving sponsors down the pathway for comfort in anonymization.