For FDA Acceptance of Real World Data Do Better Observational Studies 
[Thursday, October 17, 2019]

Only 15% of clinical trial data could be replicated using real world data (RWD) from electronic health records and insurance claims databases by researchers trying to evaluate if RWD could replicate clinical trial evidence and potentially replace clinical trials. The report highlighted well known limitations of patient records, but also offered clues to what could be done to address the same. RWD is a considered to be very critical source of health-related information with potential applications in regulatory decision-making. FDA announced its framework for RWD last year and has since presented several guidance documents and guidelines on FDA-acceptable RWD. While FDA has been accepting RWD for medical device applications for several years, it is still a developing area for drugs and biologics. RWD is traditionally collected from electronic health records, claims and billing databases, product and disease registries, data generated from mobile devices and patient-generated data from surveys and other methods. The report published this week was limited to electronic health records and insurance claims databases and found, as expected, that for 85% of the patients the records could not provide clear information about the intervention, indication, inclusion and exclusion criteria, and primary end points comparable to those collected in clinical trials. However, this is only half the story. By pointing out the limitations of the available RWD, the report highlights what information should be collected when trying to use RWD to support regulatory decisions. Patient generated data, whether or not using mobile devices, is the basis for observational studies. By applying RWD requirements to observational studies, retrospectively and prospectively, one can bridge the deficiencies in generalized data collected from available records. FDA’s guidance documents state that acceptability of the RWD depends on details of the information collected on patient, product and indication. RWD can be used for building hypothesis, validating markers, assessing feasibility, supporting safety, or as an external control for clinical trials. The approach of collecting data randomly from electronic health records has limitations since electronic health records are still created by physicians for their patients which means that they contain only data deemed by the physicians to be needed to manage their patients and not the details expected by regulators and researchers. However, a well designed observational study protocol could preset the data requirements and focus data mining only on records that contain relevant pieces of the data. This can be further improved by prospectively designed RWD collection protocols for future patients. Using RWD elements in observational studies can be sufficiently qualified to yield data similar to that collected in clinical trials. Perhaps the answer to effectively using RWD is apply RWD data requirements to an observational study.