– Written by the Submitting Real World Data project within the Real World Evidence Working Group.
One of the objectives of the Submitting Real World Data project within the RWE Working Group is to explore possible approaches for submission of real-world data to regulatory bodies. Real-world data, as defined by the Food and Drug Administration in their guidance, comprises data collected from non-interventional study designs. (1) From project group discussions, ‘claims’ and ‘electronic health records’ comprise most real-world data used in providing clinical evidence. Fast Health Interoperability Resources, or FHIR, is a platform used by HL7 for health information exchange. (2) This platform is widely used to exchange healthcare information, as far as claims and EHR data is concerned. As a result, a significant portion of real-world data used by the pharma/biotech industry follows HL7-FHIR. Sponsors using this data for clinical evidence may need to submit it to regulatory bodies. As of now, as stated in their guidance, the FDA expects the submission of real-world data in CDISC format unless granted a waiver by the FDA. (3)
Our project started conducting preliminary assessments of the possibility of using HL7-FHIR as a submission standard. As of now, submission of clinical data follows standardised processes driven by data models and standards defined by the Clinical Data Standards Interchange Consortium. (4) The pharmaceutical industry is very much familiar with the ecosystem developed through use of the CDISC standards. This ecosystem comprises processes derived from standards guidance documents issued by CDISC, eCTD guidance issued by the FDA, and expectations pertaining to compliance of clinical data with CDISC standards. One of the approaches to assess the feasibility of other standards, such as HL7-FHIR, with submission requirements is to view such standards through the current data operations processes followed by the pharma industry.
It is important to first assess the benefits of using HL7-FHIR as a standard for submission. First and foremost, most real-world data coming from claims and electronic health records is likely to follow HL7-FHIR. Acceptance of HL7-FHIR as a submission standard would therefore mean no need for data transformation to CDISC standards. Second, FHIR follows JSON format. This would likely have good control over traceability of the data if such data is received in JSON format. If traceability aspects are proven with JSON format, the question is, would that need Define-XML? The focus on traceability may shift from drafting Define-XML to ensuring adherence of data to JSON format and traceability through it. Another important factor is the ability of FHIR with JSON format to insert additional variables as tags or instances. This flexibility may help address the need for additional variable definitions in the process.
So, one may wonder, why not consider FHIR a platform for submitting data to regulatory bodies? There are no formalised data submission processes using FHIR. Although FHIR profiles come with rules, there are no specifically defined compliance rules pertaining to submission of such data to regulatory bodies. We mentioned earlier about embedding new variables as a tag or instances in JSON format. But we are unclear about how to conduct data analysis or aggregation processes while using FHIR. Analysis Data Model (ADaM) and Study Data Tabulation Model (SDTM) are built by CDISC to address some of these challenges while the data is submitted in CDISC format. But when it comes to FHIR, we do not have such formalised data models or processes. The broader question here is, what type of data models does the industry need while analysing, assessing and submitting real-world data? The feasibility assessment of FHIR in this context could reveal the need for data models which may or may not align with current CDISC data models. Fundamentally, CDISC-based data model and submission processes are built to support randomised controlled trials. The context of real-world data and its usage for clinical evidence may demonstrate a need for new standardised data models. Similarly, would there be a strong need for standardised models such as SDTM and ADaM based on randomised clinical trials when submitting real-world data?
The PHUSE RWE Working Group will continue to conduct the assessment usage of real-world data and the impact of data sources and usage on submission processes.
Contributors: Parag Shiralkar, Matt Baldwin, Joe Xi, Jeff Abolafia, Ingeborg Holt, Deepthi Errabelly, Lorraine Fang, Yanhui Yang, and members of the Submitting Real World Data project in the RWE Working Group.
- Framework of the FDA’s Real World Evidence Program: FDA’s Sentinel Initiative | FDA
- FHIR-HL7: Index – FHIR v5.0.0 (hl7.org)
- Data Standards for Drug and Biological Product Submissions Containing Real-World Data: Guidance for Industry: Data Standards for Drug and Biological Product Submissions Containing Real-World Data | FDA
- Clinical Data Interchange Standards Consortium: CDISC | Clear Data. Clear Impact.
- PHUSE Project Dataset-JSON as Alternative Transport Format for Regulatory Submissions, under the Optimizing the Use of Data Standards Working Group.
