PhUSE: Retrospective vs. Proactive Anonymization of Narratives
Exerpt from whitepaper originally published by PhUSE on March 14, 2019 by project co-leads – Rashmi Dodia, MS, RAC and Gregory Campbell, BS
Ever since policy 0070 was first introduced in 2016, there have been continuing concerns and challenges among sponsors with respect to its implementation across clinical study reports (CSRs), in particular the patient narratives section. As per the International Council for Harmonization of Technical Requirements for Pharmaceuticals for Human Use (ICH) E3 (Section 12.3.2), a patient narrative should describe deaths, serious adverse events (SAEs), or adverse events (AEs) of special clinical interest and those leading to permanent discontinuation from a clinical trial and should include key information like patient identifiers, age, sex and race of the patient among other indicators like their disease/medical history, concomitant medications, current treatment outcomes and test results. Needless to say, with the increased amount of patient information, there is also an increased level of effort required to produce anonymized narratives that strike the right balance between patient confidentiality protection and data utility.
The analysis published by Khaled El Emam in April 2017 pointed towards a majority of sponsors opting to redact narratives entirely from their marketing authorization applications.1 This trend seems to have continued on since then. However, it is important to acknowledge that narratives offer further context towards understanding an event by providing key information like the nature of the event in verbatim terms, and also the timing of occurrence. Narratives also provide information on what tests were run on a particular patient (even if the actual test outcomes are anonymized), which could be another indicator of certain types of conditions that a researcher could potentially benefit from. Many researchers agree that narratives are essential in understanding related higher level concepts for events, which in turn plays an important role in any potential downstream meta-analysis. 2 Since policy 0070 is geared towards increasing data utility, sponsors should start exploring ways of anonymizing narratives instead of completely redacting them to preserve the essence of the policy by allowing academics and researchers to re-assess clinical data.
This White Paper focuses on two approaches to produce anonymized narratives – retrospective and proactive. The retrospective section sheds light on the challenges faced with qualitative methods like redaction and what impact it has on data utility. Given the limiting nature of retrospective anonymization with regards to data utility, there is clearly a need for modern solutions and enhanced skills to be developed in order to meet Policy 0070 requirements. This paper factors in the possibility of integrating a tool or software solution that supports retrospective anonymization and defines business requirements that will help sponsors make an informed decision on automating some of the anonymization efforts. The second part of this White Paper offers a perspective on proactive anonymization and how to operationalize it.