FDA Guidelines for Leveraging Real-World Data from EHRs and Medical Claims - Guidance Regarding Study Design Elements
In the recently released FDA guidance on real-world evidence (RWE) from real-world datasets (RWD) like electronic health records (EHRs) and medical claims data, which further can be extrapolated for other types of data sources (like Real-world imaging data (RWiD), genetic data, etc.). The evaluation of RWD is crucial in regulatory decision-making for drug and biological products. This involves rigorous validation of study design elements to ensure the reliability and relevance of the data used. The design and methodology of these studies are vital to generating credible evidence to support regulatory submissions. Here, we delve into the fundamental aspects of study design elements outlined in the document.
FDA guidance specifies that the study questions of interest need to be established before the finalization of the data sources. And it should not be reversed since the limitations of the chosen RWD may restrict the options for study design and inferences that can be drawn.
The study elements need to be clearly defined before the study to reduce ambiguities:
Periods
FDA recommends clearly defining the periods pertinent to the study design in the protocol. The periods should be for identifying the study population, defining inclusion and exclusion criteria, assessing outcomes and covariates, patient follow-ups, washout, etc. The protocol should further demonstrate the data availability, accuracy, and completeness for the proposed periods and the potential impact on study validity. The justification should be provided to confirm that the data can adequately identify the study population, capture necessary covariates, and time frames for outcomes related to exposure. Additionally, it should address potential temporal changes, such as changes in the standard of care or healthcare access during events like pandemics. These considerations must be discussed with the FDA to understand their impact on the study’s internal validity.
Study Population
The FDA recommends that the protocol should include descriptions of methods for determining the implementation of inclusion and exclusion criteria. It should include operational definitions to identify eligible populations from the RWD. Identification of the study population may rely on information from multiple datasets. The operational definitions should be validated. FDA recommends including quantitative approaches in the protocol to demonstrate whether and how misclassification of criteria might occur and their potential impact on study findings.
Ascertaining and validating exposure
The FDA guidance specifies that the protocol must include the definition of the intervention/causality of interest being evaluated in the study. The definition should include the dose, formulation, strength, route, timing, frequency, and duration, and if required, manufacturer, the period between exposure and the effect, and the expected duration of effect. The sponsor should demonstrate the ability to identify the above exposure parameters in the proposed datasets. FDA recommends combining multiple RWDs to obtain the necessary information if required. While using EHRs and claims data, it is essential to consider that there could be medications not captured in those due to it being at a different point / due to it not being associated with insurance claims / obtained through other methods – cash, discount programs, samples, non-prescription drugs, dietary supplements. Validation methods should also be included in the protocol to consider misclassifications. FDA recommends including quantitative approaches in the protocol to demonstrate whether and how exposure misclassification might occur and its potential impact on the study findings.
Ascertaining and validating outcomes
One of the goals of the study is the proper ascertainment of outcomes. The protocol must properly define outcomes and the means to capture them. There should be a method to capture outcomes that occur outside the purview of medical care. Sponsor should be able to demonstrate how various outcomes will be captured. FDA recommends using multiple RWDs to capture the outcomes. The sensitivity and specificity of outcomes’ definitions should be determined and reduced to reduce outcome misclassification. FDA expects validation of the outcome variable to address outcome misclassification. The protocol should include quantitative methods to demonstrate whether and how misclassification might occur and its potential impact on study findings.
Ascertaining and validating covariates
Covariates are of 2 types – confounders and effect modifiers. Confounders are variables whose presence affects the variables being studied, so that the results do not reflect the actual relationship. Effect modifiers are variables that affect the relationship between the variables in the study. The FDA recommends identifying the possible confounders and effect modifiers that can determine the outcome and ways to capture them. A lot of these data may not be captured by EHRs and claims datasets; and even if captured, may not provide enough measurement information. FDA recommends considering potential linkages of various RWDs or the collection of additional data to capture significant covariates. The validity of the operational definitions should be determined. Quantitative approaches are encouraged to demonstrate whether and how misclassification of key covariates might impact study findings.
Conclusion
The FDA guidance document provides crucial information about the study design elements, on the cons of capturing them through RWDs, and how to mitigate them.
We at Segmed understand the importance of design elements. We can provide all the necessary information that impacts the study along with our datasets. Segmed’s fit-for-purpose regulatory grade datasets and multimodal datasets through data tokenization can support your study, by taking care of all the pitfalls mentioned above.
Get in touch with Segmed to understand our solutions and how they can support your studies.
Connect with us to explore how our datasets have supported various trials and how we have helped organizations navigate the complex web of RWE.
