How to use registry data to support FDA decision making: Agency explains in new guidance
Drugmakers looking to design a new registry or use an existing one to support a regulatory decision on a drug’s effectiveness or safety will need to consult with a new draft guidance released Monday by the FDA.
The agency’s reliance on registry data for regulatory decisions dates back more than two decades, at least, as in 1998 Bayer won approval for its anticoagulant Refludan (withdrawn from the market in 2013 for commercial reasons) based in part on a historical control group pulled from a registry.
The new 14-page draft guidance makes clear that although registries may have advantages over other real-world data sources, registries are far from perfect sources of information. FDA had to craft this guidance under the 21st Century Cures Act as part of a broader framework to evaluate the potential use of real-world evidence to help support the approval of a new indication for an existing drug.
In terms of the advantages of registries, FDA points to their ability to collect “structured and predetermined data elements” and “offer longitudinal, curated data about a defined population of patients and their corresponding disease course, complications, and medical care. In addition, registries can systematically collect patient-reported data that medical claims datasets or EHR datasets may lack,” FDA said.
Ultimately, FDA said, registry data can be used, when appropriate, to inform the design and support either interventional studies (clinical trials) or non-interventional (observational) studies.
But registries have limitations for use in a regulatory context, the agency explains.
For example, existing registries may focus on one disease, with limited information on comorbid conditions, even after linkage to other data sources. In addition, the enrolled patients may not be representative of the target population of interest due to challenges related to patient recruitment and retention. For example, patients with more severe disease may be more likely to be enrolled in a registry compared to patients with milder disease; or enrolled patients might have different self-care practices, socioeconomic backgrounds, or levels of supportive care versus the entire population of interest. These issues can potentially introduce bias into analyses.
The agency also makes clear that registries are better suited as a data source for regulatory purposes when sponsors can capture objective endpoints, such as death or hospitalization. Whereas subjective endpoints, such as pain, can be collected in a registry, FDA says, but there may be additional challenges in trying to standardize such measurements.
Factors that FDA considers when assessing the reliability of registry data (the draft digs in on other data integrity questions too) include:
- how the data were collected (data accrual)
- whether the registry personnel and processes in place during data collection and analysis minimize errors and ensure data integrity is sufficient
- whether the registry has privacy and security controls in place to ensure that the confidentiality and security of data are preserved.
The draft guidance also includes a non-exhaustive list of examples of data to include in a registry, such as demographic and clinical information, treatment information for the disease of interest, health-related outcome information and pregnancy-related information.
And FDA explains how sponsors may consider linking the data on a patient in a registry to the same patient in another data system or systems, such as another registry, an EHR, a medical claims database, or through digital health technologies such as sensors that allow for continuous or intermittent recording of physiological and/or behavioral data (e.g., blood pressure, physical activity, glucose levels) or software applications that are run on general-purpose computing platforms.
“Sponsors should use strategies to correct for redundant data, to resolve any inconsistencies in the data, and to address other potential problems, such as the ability to protect patient privacy while transferring data securely,” the draft adds.
Industry seeks more from FDA on other RWD-focused guidance
Drugmakers and Amazon also recently weighed in on another RWE-focused draft guidance from FDA, which deals with the use of EHRs and medical claims data to aid regulatory decisions.
Amgen, Janssen and Regeneron are seeking further info from FDA on how to meet with and obtain timely and efficient feedback from the agency on such RWD plans.
“To encourage advancements in RWD use and shared learnings between FDA and sponsors, Janssen encourages FDA to consider revisions to the guidance that allow for a more feasible approach to a series of focused meetings between FDA and sponsors that allows for review and feedback at milestones for designing, conducting, and reporting fit-for-purpose RWE,” Najat Khan, chief data science officer at Janssen, wrote.
Amgen is also hoping to add an appendix to the draft guidance based on further collaborations with FDA and others in industry, including data providers, on a framework for data quality.
“We believe that more specific guidance is needed to enable data providers to generate consistent assessment reports for data quality that will meet the Agency’s needs as part of a sponsor’s regulatory submissions,” Amgen said.
Regeneron said that FDA should make more clear this is a dynamic topic and highlight the agency’s flexibility to explore RWD-related concerns.
And Amazon, which doesn’t frequently weigh in on matters related to drug development, also called on FDA to indicate how sponsors can supplement data using “synthetic information, such as from artificial intelligence (AI)-based tools, natural language processing services, or digital twins (e.g. synthetic that can serve as the control arm in a study).”
Blair Anderson, director of AWS Public Policy, also sought more granular examples in the guidance:
For example, the discussion of missing data considerations (page 10) includes illustrative information on the origin of omissions and how to address the problem, such as by using a proxy for the missing data. FDA should provide similar detail in other areas. In particular, the discussion on common data elements would benefit from granularity on the data elements recommended or that would be requested by the agency based on different scenarios.