Regulatory Strategies for AI and Emerging Technologies

Insights on working with regulatory bodies on novel devices such as artificial intelligence, machine learning and algorithms.

When powerful new technologies emerge, unbridled excitement often reigns. The possibilities are endless and markets are undoubtedly huge. Artificial intelligence and machine learning in medicine are at an early stage and the potential to improve medical care is solidifying. Many opportunities are being pursued including enhancing and supporting the decision-making process of physicians, individualizing patient care with precision medicine and using near real-time information to improve care. Early ideas such as wearables that monitor activity levels or heart rate have been on the market for some time and have mixed results as tools for patient health management. Technology platforms such as Watson from IBM have shown amazing capabilities. AI- related applications have progressed to a point where formal medical device development is occurring, but challenges remain.

Innovators are often fearless and can be blind to the real risks and business difficulties to commercialize emerging technology. On the other hand, the mainstream medical device community can perceive that innovation exposes threat of the long-timelines, high-costs and painful failures. The reality is probably in between both extremes and risk can be mitigated with a deliberate approach to regulatory strategy.

Regulatory Challenges

Regulatory requirements require transparency and regulators need to understand how and why a result came about. Many AI applications are a black box with little transparency and daunting complexities. AI applications can produce valid conclusions that are counter intuitive to those which individuals or even teams of experts derive. Traditionally to gain marketing clearance for a medical device, substantial equivalence to a predicate device needs to be demonstrated. Regulators review and clear a device as a system that will not change without considerable deliberation; in fact, manufacturers go to great lengths to ensure that a device manufactured years after initial clearances demonstrates the same performance as the original device. However, one of the more powerful aspects of some AI applications is that the accuracy of output information can continually improve with continued use. This will produce a perpetually dynamic system by design. There are solutions to this dynamic situation. Making applications as seamless as possible is where the challenges and opportunities are. Independent demonstration of the safety and effectiveness of the dynamic system through the Pre-Market Approval (PMA) process could be necessitated, albeit this regulatory path is much more costly, complex and difficult.

Digital Health and the FDA

On 12/2/2016, the FDA and industry finalized a new medical device user fee agreement that lays out the application process and expectations regarding medical devices being submitted for regulatory approval. The agreement calls for:

• A new digital health unit overseeing Software as a Medical Device (SaMD) and software inside of medical devices (SiMD)
• Opportunities to support 510(k), premarket approval and clearance pathways tailored to SaMD, SiMD, and take into account real world evidence
• Participation in international harmonization efforts for digital health initiatives

Regulatory Strategies to Address these Challenges and Opportunities

1. Build your case, if regulators cannot or do not yet understand a technology, they will struggle to establish ways to assess the safety and effectiveness of your product.

• Find clinical and regulatory information throughout the world that is supportive of what you are trying to achieve. If negative information is uncovered, do not ignore it; instead, address it.
• Do not plan on submitting a “black box.” Spend the time to develop ways to communicate how and why a particular result comes about.
• Seek related credible sources, publications, guidance documents and experts, reference them, and utilize them.

2. Plan on early and frequent meetings with regulatory authorities. Gaining regulatory clearances for emerging technologies benefits from building solid working relationships with regulatory bodies. These relationships are built over a series of meetings that begin early in the development process, aiding both parties as they learn together. These early collaboration meetings provide a venue for industry to explore new ideas with regulators. Based on working with many novel products, we recommend the following considerations, particularly in working with the FDA.

• Start with a solid pre-submission package. Make certain to follow established guidelines for pre-submissions, setting the tone that your organization has done its homework. Additionally, a complete pre-submission package enables regulators to prepare for the meeting and identify the appropriate experts needed.
• Be ready for questions surrounding mechanism of action. In traditional medical devices, a key component of the submission surrounds the mechanism of action, or how the device is perceived to work. Be prepared to explain what your novel technology does and how it does it.
• Prepare for the meeting. Before the meeting or teleconference, identify your question areas and structure your questions to get the specific information needed without using open-ended formats. Avoid introducing new information that the FDA is not prepared to discuss.
• Set an open tone during the meeting. Be transparent about the information you have and the areas that are lacking. In early development cycles, gaps exist and it is important to be candid so you get accurate early feedback. Set aside any fears that your organization is setting precedent by not having complete information and instead, work to close those gaps as you progress through the product development process.
• Set specific but realistic expectations for meetings. Your device is novel and that means regulators have not seen anything like it before. Expect regulators will have questions that will be challenging to answer. Remember, agencies can only provide feedback on what you have presented--what your device is currently-- and not what you hope it will be in the future.
• At all times, be professional and seek to understand their questions. Early collaboration around emerging technologies is a two-way street. When you provide information the FDA needs, they will provide direction to guide your product development. And remember, the agency is a resource. Be open to the answers you receive, so your processes can evolve and improve.

Cultural Challenges

There are cultural differences that exist between the Silicon Valley visionaries that are pushing novel technologies forward versus the deliberately cautious, stepwise approach taken by traditional medical device development teams. In advance of nearly any new, disruptive technology being adopted by broader medical establishments and nearly always in advance of solid profitability, teams of brilliant visionaries have been embraced and included by traditional medical product development organizations.

Team Building Strategies to Address Culture Challenges and Opportunities

• Build effective multidisciplinary teams by addressing cultural divides. When building regulated medical products based on AI, organizations will benefit from vertical structures and need to bring in key leaders from the industry early in the business. Building medical device products requires deep subject matter expertise. Combining subject matter with technical expertise and embracing the multidisciplinary approach drives innovation, making it possible to accurately model the vision, answer the key questions and realistically hold leaders accountable for results. Team evolution needs to be intentional and designed. If you are unable to set the joint domain early, one side dominates, and it becomes a challenge to bring in world-class experts from the other side. Top people recognize such unbalance and the byproduct that they will not achieve the parity in authority and respect.
• Create solid regulatory affairs teams with prior experience in addressing uncertainty surrounding quality systems and regulatory approval, specifically:

• Members with a focused subject matter expertise in the AI/machine learning space and the mindset/ability required to work in a highly complex multidisciplinary environment.
• Regulatory expertise introducing and gaining FDA clearance or approval for new technologies and with the intention and ability to communicate with the developers.
• Direct expertise developing software for medical devices and developing and managing quality systems for software development.
• Where clinical trial work is required, the CRO and strategic regulatory experts need to work closely together to generate the type of clinical data that supports the safety and effectiveness of the underlying product. Also the regulatory team needs to develop and conduct the studies in ways that bring transparency and simplify the “black box” aspects of the product.

In Summary

• Embrace and enforce the multidisciplinary approach.
• Break down the problem and bring subject matter experts to the table to solve each.
• Beware of thinking you can do all. Inventors and software developers are uniquely brilliant but not necessarily the best at navigating regulatory challenges.
• Don’t assume the person with the most RA experience should lead the FDA strategy. The RA and especially QA mindset can limit creativity and curiosity that propels novel technologies forward. QA is a support mechanism to the RA strategy.
• Leverage working with RA SMEs experienced in emerging technologies and with prior nontraditional submissions.