introduce

On October 21, the U.S. Food and Drug Administration released draft guidance on real-world data standards for drug and biological product submissions, outlining how sponsors should submit drug and biological product study data from real-world data (RWD) sources idea. .

Previously, the team of Professor Yao Chen, Hainan Real World Data Research Institute and Vice President of Peking University First Hospital, has been engaged in data standardization research. Through previous research on different stakeholders and interpretation of the latest FDA guidelines, he summarizes the important challenges facing real-world data standardization, and puts forward corresponding solutions and suggestions.

1. FDA Considerations for Real-World Data Standards

How do sponsors submit drug and biological product study data from real-world data (RWD) sources? The considerations set out in the draft guideline are as follows:

(1) Promote the use of real-world data that meets the CDISC study data standards currently supported by the FDA.

Currently, FDA can process, review, and archive electronic submissions of clinical and nonclinical study data, including data from RWD sources, using the standards specified in the Data Standards Catalog, as described in the Study Data Guidelines. As explained in this guidance, the directory provides a list of currently supported or required standards, their purpose, the duration of FDA-supported standards, and other relevant information. FDA is issuing this guidance to provide recommendations to sponsors to use the criteria set forth in the catalog to comply with Section 745a(a) of the FD&C Act when submitting study data from RWD sources in applicable drug submissions.

FDA plans to issue further guidance and/or a standard updated catalog using study data from RWD sources. Currently, in the absence of an exemption, sponsors submitting clinical and nonclinical study data (including data from RWD sources) under Section 745a(a) of the FD&C Act must use the format described in the Study Data Guidelines and in the Catalog of Studies Listed data standards. Sponsors should refer to the quality standards, recommendations, and general considerations provided in the "Guidelines for Technical Consistency of Study Data" when submitting study data for applicable drugs to the FDA. In seeking to bring the RWD into compliance with FDA-supported data standards,

Sponsors should discuss any planned study data submissions obtained from RWD sources in applicable drug submissions as early as possible with the appropriate FDA review division, as well as their method of converting the data to current FDA-supported data standards. Sponsors should describe these methods, including in the protocol, data management plan, and/or final study report. FDA recognizes that currently supported data standards such as the Consortium for Clinical Data Interchange Standards (CDISC) Study Data Listing Model (SDTM) can be applied to RWD sources such as EHR or claims data using a range of methods. After fully documenting the consistent methodology used and its rationale, RWD-derived study data can be converted into SDTM datasets and submitted to the FDA in applicable drug submissions.

Enhance the traceability of real-world data to clinical research data by clearly articulating the difference in the meaning of the data in the metadata record:

During data management and data transformation, appropriate processes should be established to increase the reliability of the resulting data. Documentation of these processes may include, but is not limited to, electronic documentation of data additions, deletions, or changes from source data systems to final study analysis datasets (i.e., metadata-driven audit trails, quality control procedures, etc.). Sponsors should also document changes to data in their applicable drug submissions to comply with current FDA-supported data standards, as well as the potential impact of these changes.

FDA is aware that, for nearly every data domain, there are substantial differences in terms used and their precise meanings between RWD sources and FDA-supported data standards. Examples range from meaning and specific terminology for race/ethnicity, terminology systems for medications, and interpretation of healthcare records for important measurements. Even for variables that appear to be the same record (eg, male/female), there may be differences in how these variables are defined between RWD sources and FDA-supported data standards. For example, gender as a variable can be coded in CDISC's terminology as a concept based on physical characteristics, while EHR can use gender identity. In such cases, sponsors should document the potential impact on the study results of terms that map gender variables or other variables to CDISC.

It is critical to document the sponsor's rationale for selecting specific CDISC data elements for RWD and to document the differences between the two. Sponsors should provide a description of the general approach to data mapping and the expected impact in a section of the study data reviewer's guidelines or in an appendix to highlight the areas covered. In addition, sponsors should include a data dictionary that records the definition of each data element used and all relevant information about that element, such as its relationship to other data, source, usage, and format. It is recommended that the detailed mappings and processes in the relevant datasets/domains be put into a Define-XML file (see appendix), preferably not included in the Study Data Reviewer Guide.

(2) The challenges brought by standardization stem from the diversification of real-world data and the difference between the production process and the research process.

FDA understands the challenges involved in standardizing research data from RWD, including but not limited to:

(1) Diversity of RWD sources and inconsistencies in their format (eg, EHR, registration);

(2) the use of different standards, terminology and exchange formats to represent the same or similar data elements, differences in source data obtained on a regional and global scale;

(3) the various methods and algorithms used to create datasets designed to aggregate data;

(4) Many aspects of health care data that may affect the overall quality of the data, including business processes and database structures, inconsistent vocabulary and coding systems, and de-identification methods used to protect patient data when shared.

2. Understand the status quo of domestic application of data standards through interviews

Understanding multi-stakeholder barriers to standardizing real-world data is a strategy to address standardization.

Before the FDA issued the draft data standard guideline, the author has been engaged in the research of data standardization direction. Through the preliminary investigation of different stakeholders and the interpretation of the FDA guideline, the following four important challenges have been summarized and corresponding suggestions have been put forward.

1. It takes a long time to localize international standards, which makes it difficult to use in medical treatment

The FDA mentioned that challenges to standardization include: such as the diversity of RWD sources and their inconsistent formats (eg, EHR, registration); the use of different standards, terminology, and exchange formats to represent the same or similar data elements, both regionally and globally Differences in the source data obtained within the scope.

Interview findings: Standard terminology does not fully express what physicians need to describe a patient's condition, nor does the most colloquial term exist, making the process of selecting the best term more difficult. For standard data exchange within hospitals, hospital information system providers implement data exchange without using pre-specified standards (such as HL7 standard messages) and conduct case-by-case consultations based on hospital requirements.

Solution: Multiple parties should assist the construction of localization standards in combination with real-world data standards

2. The excessive structure of source data collection affects the diagnosis and treatment process

The FDA mentioned that there are many aspects of health care data that can affect the overall quality of the data, including business processes and database structures, inconsistent vocabularies and coding systems.

Interviews found that the main obstacle to standard implementation is that it increases the workload of clinicians in hospitals. A fully standardized data collection mode takes more time to ensure that each data field is collected according to the standard. Often, the use of fully normalized data collection causes clinicians to give up looking for the correct input and only use the "other" option to write their own answers. Standard data collection may further limit the range of expressions physicians can use to describe their decisions and may lead to over-collection or misclassification of data.

Solution: Keep data collection and standardization as separate as possible

3. Research datasets are difficult to reuse across multiple studies

The FDA also mentioned that the challenges to standardization come from: the various methods and algorithms used to create datasets designed to aggregate data.

Interview findings: Pharmaceutical companies and regulators want to use real-world evidence to support the development and life cycle assessment of medical products. Hospital-developed disease-specialty databases are often highly customized, based on aggregations of clinical guidelines, but cannot easily be used for secondary purposes. Pharmaceutical companies wishing to conduct research using multiple specialized databases found that aggregating data sources was not feasible due to differences in data definition and structure.

Solution: Promote more general research data models to suit different stakeholders

4. Standardized processes are difficult to audit

FDA emphasizes that, during the data management and data transformation process, appropriate processes should be established to increase the reliability of the data obtained.

Interview findings: Pharmaceutical companies and regulators find real-world data sources difficult to audit because off-the-shelf research data is often derived through a multi-step transformation process and does not allow access to source data. Big data companies also find it difficult to document data normalization processes that involve multiple departments and advanced algorithms.

Solution: Apply a management process that facilitates verification and traceability

Summary: If data standards are to be truly applied, the process of diagnosis and treatment must be adapted and aided, as well as the need to aggregate multi-party verification when collecting research data from multiple sources.