Regulatory News | 25 July 2024 | Ferdous Al-Faruque
FDA headquarters in White Oak, MD. (Source: Ferdous Al-Faruque)
The US Food and Drug Administration (FDA) has finalized its guidance on evaluating real-world data (RWD) from electronic health records (EHR). The final guidance offers additional clarification on selecting study variables and validation and includes recommendations for the use of quantitative approaches to aid in interpreting study results and analyzing potential misclassifications.
On 24 July, the FDA finalized the RWD guidance, which had been published in draft form in September 2021. The guidance received significant feedback from the medtech industry to provide more flexibility, but the agency has mostly kept it close to the draft version. A glossary of terms that appeared in the draft version has been incorporated into the text of the guidance. (RELATED: Industry calls for flexibility on RWD data sources, validation, Regulatory Focus, 26 January 2024)
The final guidance also includes several new recommendations, including what data sources sponsors can use. The agency repeated that medical claims data may not accurately reflect a particular disease, but also added that medical claims data may change over time and lead to variations in reported diagnoses and procedures.
When looking at data sources outside the US, the agency added that sponsors should include additional information in their protocol to support using OUS data. It recommended including “a discussion on how factors relating to the health care system, including its practices, might affect the generalizability of the study findings from the selected data sources.”
“When non-U.S. data sources are proposed, additional explanation (e.g., demographic factors, standard of care) should be provided to support the generalizability to the U.S. population,” the agency added.
When selecting the study population from EHRs, FDA has added that it recommends including quantitative approaches, such as quantitative bias analyses, in the protocol to show potential misclassification of inclusion and exclusion criteria that may affect the findings.
When determining the ascertainment of outcomes, the agency has added that if the study is used to evaluate a risk ratio, picking high specificity may be more important than using high sensitivity because it may not bias the risk ratio in such situations.
“Focusing on very high specificity in this scenario will help ensure the study result is correct even if data are imperfect, while high sensitivity is still important to ensure the precision of the estimates,” the agency added.
Covariates in RWD studies may include confounders and effect modifiers, and FDA added more detail about what that means. The agency repeated that information on potential confounders might be collected through nonrandomized studies but added that the sponsor’s protocol should include all known unmeasured confounders in the proposed data sources, approaches to supplement the data on unmeasured confounders and provide justifications for picking any proxy measures.
Final guidance