In August 2017, the U.S. Food and Drug Administration’s (FDA) Center for Devices and Radiological Health (CDRH) finalized the guidance, Use of Real World Evidence to Support Regulatory Decision-Making for Medical Devices¹. This guidance marked a significant step by CDRH into the realm of real world evidence (RWE), recognizing RWE’s ability to enhance patient safety, streamline industry practices, and save resources by providing clearer regulatory expectations. The 2017 guidance effectively provided a “green light” for industry to engage in innovative research within this domain. At the close of 2023, CDRH released a new draft guidance on RWE which, when finalized, will replace the original 2017 version. Release of this 2023 draft guidance represents CDRH’s current thinking on RWE use, signaling shifts in regulatory approaches and setting new standards in the field.

Real world data (RWD) can generate evidence in support of expanding a device’s indications for use, as a control group for clinical trials, or as supplementary evidence in pre-market submissions. CDRH put theory into practice when it approved the groundbreaking 2023 label expansion for Johnson & Johnson’s ThermoCool SmartTouch Catheter, making it the first label expansion granted to a medical device based entirely on RWE derived from device electronic health records.² This created a landmark predicate for using RWE for a device’s label expansion in lieu of a costly and time-consuming randomized clinical trial.

The 2023 draft guidance aims to clarify the criteria for evaluating RWD quality and provides detailed information on submission requirements for medical devices. Similar to using RWE for drugs and biologics, CDRH outlines its recommendations for ensuring reliable and relevant data. However, this guidance provides specific considerations when leveraging RWD for medical device regulatory use that sponsors should take note of, including a real world dataset’s suitability for regulatory purposes, how CDRH views RWD collected by medical devices, and how RWD is used to train artificial intelligence/machine learning (AI/ML) algorithms.

Real world data needs to be fit for medical device regulatory decision making

Use of RWD for medical devices has been growing in use cases, but leveraging data properly for regulatory use can be a challenge due to a lack of use cases. FDA has published its thinking on RWD use for drugs and biologics, including recommendations on ensuring RWD relevance and reliability, and sets similar standards for medical devices. CDRH’s 2023 device draft guidance outlines many of the same considerations for assessing RWD. Though drugs and biologics RWE guidance can serve as a preliminary framework, device makers will still have to determine whether FDA finds its RWD acceptable for regulatory use without much precedent.

Innovative device technologies present different regulatory challenges surrounding RWD and RWE usage. For example, RWD is often collected by registries or other databases that meet certain regulatory standards, but there is a lack of precedent in understanding CDRH’s regulatory decision making when data is collected by software as medical devices (SaMDs). Device makers will also have to consider how unique device identifiers (UDIs) are captured in collected RWD. Though UDIs are required in device labeling³, they are not always captured in real world datasets due to their inconsistent collection in electronic health records. When tracking UDIs in datasets, sponsors will have to consider whether their data is sufficient to address study questions and if the data captures sufficient detail. They must also consider the source and technical methods used to capture UDI information when generating rationale for FDA submissions.

IDE requirements for RWD-collecting devices is based on medical practice norms

SaMDs and other medical devices have the capacity to gather RWD to demonstrate safety and effectiveness for both research and commercial purposes. However, the necessity for an investigational device exemption (IDE) becomes ambiguous when legally marketed devices are employed to collect RWD. An IDE enables the use of an investigational device in clinical studies to gather data on safety and effectiveness, typically aimed at supporting a pre-market authorization. FDA indicates that an IDE might not be required if a device is utilized within the standard scope of medical practice and the collection of data does not alter the device's administration in routine medical care. Conversely, if data collection aims to assess the device's safety and effectiveness and influences treatment decisions, an IDE might be necessary.

FDA's requirement for an IDE for devices that collect RWD is significantly influenced by the method of data collection, the healthcare environment in which it is collected, and its effect on clinical decision making. This complexity is intensified when examining SaMDs, particularly regarding how software relays data to patients and healthcare providers. The transmission and use of data by SaMDs adds another layer of consideration, as it can directly influence patient management and treatment pathways, thereby impacting the regulatory requirements for an IDE. Lacking additional details from the draft guidance, sponsors will need to understand whether their device requires an IDE by discerning their device’s impact on clinical decision making.

Training AI/ML algorithms with RWD

While AI/ML algorithms are becoming integral to the development of SaMDs and diagnostic tools, they present distinct regulatory hurdles. FDA's regulatory oversight of AI/ML algorithms as devices, along with the associated regulatory requirements, remains ambiguous and underdeveloped. This uncertainty is further compounded when RWD is utilized for algorithms designed to learn and "adapt" over time. The dynamic nature of such algorithms, which allows them to evolve based on new data, poses significant regulatory challenges, particularly in ensuring their safety and effectiveness throughout their lifecycle. FDA's current framework still requires clear guidelines on managing and monitoring these adaptive algorithms to maintain their reliability and trustworthiness in clinical settings.

In this draft guidance, FDA acknowledges that RWD can be applicable as a mechanism for re-training artificial AI/ML-enabled medical devices, but provides no further clarification on how RWD is evaluated in this context. Using RWD to train an AI/ML algorithm, and with the algorithm adapting over time, poses unique challenges in the Agency’s evaluation and for industry. Device makers will have to determine if and how their algorithms fit into FDA oversight and how RWD used as evidence will be evaluated by the Agency.

Although FDA's acknowledgment of RWD's role in AI/ML development signals a positive direction for industry sponsors, the specifics of how RWD-supported AI/ML algorithms will be integrated into the FDA’s framework for enforcement discretion and oversight are still unclear, raising more questions than providing answers. Consequently, industry is left to navigate how its algorithms will align with FDA oversight and the way RWD will be assessed as supportive evidence by the agency.

Navigating device real world evidence with the right regulatory partner

FDA's recommendations for using RWE and RWD for medical devices remain in a state of flux with a guidance yet to be finalized, and the Agency’s decision making and expectations are not always intuitive. IQVIA’s team of real world experts are ready to help you make the most of device RWE.

IQVIA is ready to help guide device makers through these new and intricate regulatory pathways, acting as a regulatory partner across the device product lifecycle. Our team of regulatory experts - including former FDA regulators, device reviewers, and policymakers - effectively minimize risks associated with innovative technologies through our comprehensive understanding the RWD’s regulatory utility for various device types. We do this by providing strategic support for leveraging device RWE and acting as a regulatory partner when communicating with the Agency.

References:

1. FDA Guidance for Industry. Use of Real World Evidence to Support Regulatory Decision-Making for Medical Devices. https://www.fda.gov/regulatory-information/search-fda-guidance-documents/use-real-world-evidence-support-regulatory-decision-making-medical-devices.
2. Yu, Yue et al. Integrating real world data to assess cardiac ablation device outcomes in a multicenter study using the OMOP common data model for regulatory decisions: implementation and evaluation. https://www.ncbi.nlm.nih.gov/pmc/articles/PMC9831049/
3. Federal Register. Unique Device Identification System. https://www.federalregister.gov/documents/2013/09/24/2013-23059/unique-device-identification-system.