Since August 2017, the FDA has released 11 guidance documents describing various aspects of real world data (RWD) and real world evidence (RWE) in regulatory decision-making. The 12th guidance document released this week describes expectations from clinical studies that use RWD sources, Electronic Health Records (EHRs), and medical claims data, to derive RWE. This guidance, like the previous ones, details the best practices expected from the collection, vetting, and analysis of RWD to support effectiveness and/or safety claims.
RWD is a broad term that includes EHRs, medical claims data, data from registries, patient-generated data, data from digital health technologies, etc. The current guidance document focuses only on data from EHRs and medical claims in clinical study designs. Specifically, the guidance discusses the selection, validation, traceability, and quality of data from appropriate sources to characterize study populations, exposures, outcomes of interest, and key covariates. FDA expects the source data to be accurate, complete, and traceable to be able to rely on conclusions drawn from its analysis. The data also must be relevant to the intended uses in terms of the availability of “key variables (exposure, outcome, covariates) and sufficient numbers of representative patients for the study”.
The guidance details the criteria for acceptance of the data but leaves the specific analysis chosen for a given study up to the sponsors. This guidance document like the previous documents describes the FDA’s expectations in broad terms providing discussion points for sponsors planning to collect RWD for regulatory purposes. For example, an interesting section of this guidance describes acceptable practices for handling missing data from the datasets. The section lists the two common reasons for missing data; either it was not collected, or it was not recorded. Missing data is defined as data that was intended to be collected based on the RWD collection plan and would have been used in the study analysis but were not observed, collected, or accessible. The document briefly suggests an impact analysis of the missing data but does not go into specific statistical methods or other measures to address missing data, leaving it to the sponsors to strategize, as deemed necessary. The guidance contains similar approaches for the most common discussion points in RWD studies where issues are discussed briefly and left to the sponsors to strategize how to address the same.
Overall, the guidance can be used to create checklists for RWD protocols for the collection and generation of datasets intended for an FDA-acceptable RWE. The guidance should be read along with the previous guidance documents on RWD and RWE.
