Policy 0070― published by the European Medical Agency (EMA) in October 2014― has made the world of regulatory writing a more complicated place. The policy requires specified submission documents to be made public for all marketing authorization applications (MAA’s) submitted as of January 1, 2015 and for all indication extensions and line extensions submitted as of July 1, 2015. The Policy 0070 document primarily addresses company confidential information (CCI) rather than personal protected data (PPD). The EMA published the “External guidance on the implementation of the European Medicines Agency policy on the publication of clinical data for medicinal products for human use” (the guidance) in March 2016; this document details policy execution. It specifies additional requirements around PPD. And then in December of 2016, EMA issued an updated version of the Guidance, with a few significant changes. In this blog post, I’ll discuss the complexities created by Policy 0070 and suggest some solutions to manage those challenges.
Deliverables required by EMA Policy 0070
Policy 70 is a two-round process. In the first round, the “Redacted Proposal Version”, the deliverables include:
- A cover letter with templated language provided in the guidance document
- A document list of all the components submitted in the redacted proposal version
- The submission documents that include proposed anonymization changes and proposed company confidential information redactions
- The justification table listing the CCI proposed for redaction
- The anonymization report outlining the method by which the submission documents were anonymized
The second round includes a slightly shorter list of deliverables that will be made public. This list includes:
- The cover letter
- The document list
- The previously listed submission documents with finalized anonymization techniques.
- The anonymization report
What is involved in anonymizing submission documents? Submission documents contain personal data, which the EMA defines as “Any information relating to an identified or identifiable natural person (‘data subject’).” Recital 26 of the EU General Data Protection Regulation which drives privacy handling states, “To anonymize any data, the data must be stripped of sufficient elements such that the data subject can no longer be identified by all likely and reasonable means.”
Sponsors must ensure that re-identification cannot occur using linkability, singling out, or inference. Previously, we thought that just removing basic information like clinical trial ID numbers or names from medical records could prevent re-identification of individuals. Through malicious attacks and through demonstration attacks by researchers, we now understand that more must be done to protect privacy. The more information is provided about an individual (not including names), the easier it is to re-identify someone using the information in the document and outside sources. Preventing re-identification is difficult and impacts data utility. So, the EMA requires sponsors to assess the re-identification risk of their anonymization methods.
The options to anonymize documents include:
- Masking― otherwise known as redaction― obscures personal data with black boxes, or in the case of the EMA, blue boxes.
- Other techniques include:
- Noise addition
- Differential privacy
These options have different impacts on the risk of re-identification and data utility.
Why redaction is the dominant anonymization technique (for now)
Currently, most sponsors prefer using redaction. The first six Policy 0070 documents made public at different times since October 20th by the EMA used redaction techniques. The EMA acknowledged in the Guidance that this would be the industry’s preferred route in the initial implementation of Policy 0070.
Redaction is appealing for several reasons. First, we are dealing with retrospective redaction. Many trials in the submissions currently being prepared for Policy 0070 and some of the clinical study reports written for those trials were prepared before Policy 0070 was published. Thus, patients who participated in these trials were not informed that their information was going to be made public, the documents were not written with publication in mind, and contracts with study administrators were not written with publication in mind. Likewise, some of the other techniques require reverting from PDFs to Microsoft Word documents to modify content. Re-opening documents for modification has quality control and cost implications.
Who is entitled to privacy under Policy 0070?
PPD in regulatory documents typically covers the following groups:
- Patients plus spouses, partners, or children referred to in studies
- Study administrators including investigators, sponsor employees, committee members, and vendors
What goes into the anonymization report?
The anonymization report outlines how the sponsor anonymized the documents. Additionally, sponsors must explain how they maximized data utility. The EMA has suggested that they will not review the submission documents for all proposed anonymization; they will review the anonymization report and provide feedback on it. We have learned that in these early phases the EMA is reviewing the anonymization report and the proposed anonymized documents.
Keeping CCI under wraps
CCI is information not in the public domain and where disclosure may undermine the sponsor’s legitimate economic interest. Sponsors must list all proposed CCI redactions in the justification table. This table must also include the location of the CCI and the rationale for requesting its removal. During the consultation phase, the EMA will both validate what has been proposed for removal and provide feedback on CCI in the justification table.
Transparency and disclosure prior to Policy 0070
Prior to Policy 0070, most transparency efforts made by sponsors were a result of EMA Policy 0043, sponsors’ own transparency policies, and clinical study report (CSR) documents submitted with publications. Under the sponsor’s own transparency policies, the sponsor had complete control over how to anonymize their documents for publication.
The brave new post-Policy 0070 world
In this post-Policy 0070 era, the practices are different. Rather than establishing overall anonymization policies, sponsors must assess every submission to determine unique anonymization techniques that will maximize data utility and protect patient privacy. Sponsors can no longer have complete control over how to anonymize their documents. The EMA is weighing in on how anonymization should be performed.
The timeline for submissions under Policy 0070
The EMA has proposed the following submission process under Policy 0070. The redaction proposal versions are to be supplied to the EMA between 181 – 220 days after submission. The consultation phase is supposed to take around 42 days. Sponsors will then have 27 days to implement and respond to the EMA’s feedback. Finally, the documents will be published within 60 days of the European Commission issuing its decision.
But, we are not there yet. Currently, the EMA is playing catch-up. They are working back from the earliest opinions in September of 2015. Based on our clients’ experiences, the EMA requests that the redacted proposal version be provided within 30 days, but this deadline appears to be negotiable. The sponsor delivers the redacted proposal version at the agreed upon date. For now, sponsors are taking several months to get through consultation. Obviously, the EMA is aiming to expedite this portion of the process. Once the consultation is complete, the final redacted version is submitted to the EMA with the agreed upon updates.
Developing a rule set for your redaction policy
With the anonymization technique of redaction, sponsors must determine the rules that will govern their redaction policy. Then, they must apply these rules across the documents and explain them in the anonymization report. In the pre-Policy 0070 days, one fairly conservative rule set could be applied across all trials. That is no longer the case in the Policy 0070 era. Different rule sets are needed for different trial types. The characteristics driving the different rule sets we have created for sponsors include:
- Disease prevalence
- Population size
- Number of study sites
- Number of patients per site
- Sites per country
Who approves these rules for each sponsor? A group of stakeholders participates in transparency and disclosure decisions, particularly around Policy 0070 because of the broad publication of the documents. The stakeholders typically include representatives from:
- Medical writing because they are closest to the sponsor’s documents
- Statistics because they typically have experience preparing de-identified data sets
- Regulatory because they communicate with the EMA
- Therapeutic area representatives because they know the particular compound, can represent the complexities of the submission, and understand what should be considered CCI
Why is Policy 0070 compliance so complex?
Many steps are required to prepare a Policy 0070 submission. What makes its requirements so complex? Here are some areas for consideration.
What has the sponsor posted on clinicaltrials.gov? Sponsors provide investigator information, including location, on clinicaltrials.gov to aid recruitment. It lets the public know if the trial is taking place locally and provides volunteering information. However, what if this trial is a rare disease study or small population in which at a later date, under Policy 0070, the sponsor wants to redact investigator information to provide greater anonymization for patients? In small studies with small patient populations, investigator location information and even investigator identities can be a proxy for patient location. Providing patient location increases the risk of patient re-identification.
Another complexity is around scope. In the initial version of the Guidance document, only trials provided in submissions made after January 1, 2015 needed to have their corresponding CSRs included in the submission. However, in the updated Guidance published in December of 2016, any trials referenced within a submission might be considered in-scope for the purpose of Policy 0070. This becomes a significant issue for pediatric submissions. In the EMA, CSRs for pediatric studies must be delivered within six months of trial close. For a pediatric submission made after January 1, 2015, all the corresponding studies may have been provided separately prior to that date. Per the updated Guidance, those trials are now in scope.
For other submissions such as line extensions or indication extensions, the “cross-referenced” studies within the submission may not be clearly in-scope for this new rule causing confusion and potential delays in preparing a Policy 0070 submission in advance of the deadline.
Study characteristics also pose issues. Many of the sponsors we support in Policy 0070 preparation have chosen to redact gender for patient information. What happens when a trial is one gender and that gender has already been posted on ClinicalTrials.gov during the registration process?
Some of the sponsors with whom we work have also chosen to redact race and ethnicity. Similar to the gender scenario, what happens when a study is conducted in a single race and that race is posted on ClinicalTrials.gov? Should a sponsor attempt to redact it from documents if the information is already in the public domain? Should sponsors make earlier efforts to keep this information confidential both for the registration process as well as the publication of the document?
Policy 0070 and redaction
Redaction is more complicated under Policy 0070. The redaction complications concern the following:
- Sections within the document: What do you do with listings, narratives, mini-narratives for non-standard trials?
- Demographic complications like the issues we proposed with single gender or single ethnicity studies.
- Other identifier type complications: For non-standard studies, does the medical information create too high a risk of re-identification. If that information is removed, does any data utility remain?
Creating the anonymization report and justification table
Developing the anonymization report is complicated. First, sponsors must choose either a qualitative or quantitative method to assess the risk of re-identification for each submission. Sponsors must then establish their desired risk threshold. Corresponding with this decision, they must next assess the likelihood of different attacker scenarios. Sponsors must then assess the risk of re-identification against the pre-determined threshold. They must also justify how data utility was maintained. Lastly, sponsors must make global redaction decisions as well as decisions specific to each submission and, potentially, each trial.
The justification table complications concern the fact that each proposed CCI redaction requires a unique rationale. The EMA does not accept “canned answers” for CCI requests.
Learn more about streamlining document redaction
Over time, many of these complexities will resolve. Sponsors will gain experience, incorporate best practices into writing these documents, and have access to technologies that handle the complexities in a more sophisticated manner. We are innovating to bring sponsors greater options to address their transparency and disclosure needs.
Our partnership with PleaseTech is part of our innovation strategy. PleaseTech’s newly announced PleaseReview’s 6.0 software release provides users with redaction capabilities. While the complexities remain today, PleaseTech’s future 6.1 release of PleaseReview will enable Synchrogenix to create a seamless redaction process that streamlines collaboration in the Policy 0070 era. To learn more, please watch our archived webinar on this topic.