FDA - Real World Data draft guidance and RBQM
Updated: Nov 9, 2021
Last month (Sept 2021) the FDA published a guidance document for comment entitled “Real-World Data: Assessing Electronic Health Records and Medical Claims Data To Support Regulatory Decision Making for Drug and Biological Products”. This blog summarizes the key points related to Risk-Based Quality Management (RBQM).
This guidance discusses the following topics related to the potential use of Electronic Health Records (EHRs) and medical claims in clinical studies to support regulatory decisions:
Selection of data sources that appropriately address the study question and sufficiently characterize study populations, exposure(s), outcome(s) of interest, and key covariates
Development and validation of definitions for study design elements (e.g., exposure, outcomes, covariates)
Data provenance and quality during data accrual, data curation, and into the final study specific dataset.
Those three areas relate to the quality management approach set out in ICH GCP E6(R3) and E8(R1). For example, the FDA guidance stresses that protocols should identify all data sources proposed for a study and document any potential limitations related to that data source. All part of good quality planning and building in quality by design.
The guidance also has some interesting points on data capture and quality. We know from our work with Sponsors and CROs that data capture and quality can present significant challenges. The section on ‘Data Linkage and Synthesis’ discusses the potential problems with integrating data from different data sources or study sites to deliver “acceptable quality”. The guidance also covers the increasing use of Common Data Models (CDMs) for product safety surveillance and research. The benefit of CDMs comes from the ability to combine data from multiple sites to execute identical queries on multiple data sets.
The use of Artificial Intelligence (AI) to process unstructured data is discussed, and while the FDA is explicit in not endorsing any specific AI technology, it’s clear that AI will have an increasingly significant impact on clinical trials. From a quality perspective the protocol needs to include the assumptions being made about AI and the impacts on data quality.
Section VI on Data Quality includes a useful diagram showing an example of a Life Cycle of EHR Data.
The concept of the data life cycle illustrates one of the key principles of quality management, namely that the process for examining the quality of the data is iterative, it’s not a one-time assessment. The guidance includes an interesting section on ‘Characterizing Data’ which covers data accrual, curation, and transformation to ensure completeness and accuracy of the data.
The two sections on “Documentation of the QA/QC Plan” and “Documentation of Data Management Process” are surprisingly short from an RBQM perspective, perhaps because they are dealt with in greater detail in other guidance. They do however stress the need for clear documentation and descriptions in the protocol, and for a multidisciplinary approach that includes clinical input to ensure that the capture and handling of data inherently incorporates the nuances and intricacies of health care delivery.
So while not directly related to RBQM, the guidance reinforces the key RBQM principles, and gives an interesting perspective on a big challenge for the sector - ensuring you have good quality data to support regulatory decision making for your drugs and biological products.