In recent times, the use of AI/ML (Artificial Intelligence/Machine Learning) has taken a huge technological sweep in the medical devices and healthcare industries because of their ability to diagnose, manage and treat a wide variety of medical conditions and to enhance the patient care. But there seem to be obstacles in implementing AI/ML in daily practices, with respect to the transparency issues surrounding their software programs. Hence, it is crucial to regulate these technologies and to do so, the Regulatory bodies are effortlessly trying hard to govern the AI/ML implementation.
Basing on the same, the FDA has released a five-part action plan for its oversight of safe and patient-centric AI/ML-based SaMD. What does the action plan outline? Let us try to understand.
The FDA intends to spur SaMD development and has outlined the following five-part action plan:
- Develop a tailored Regulatory framework for the medical software by issuing a draft guidance on the pre-determined change control plan for software learning
- Good Machine Learning Practices (GMPL) for evaluation of ML algorithms
- Patient-centered approach incorporating transparency to users
- Regulatory science methods related to algorithm bias & robustness
- Real-World Performance (RWP) monitoring
The FDA has claimed that it will continue to develop and update its own proposed Regulatory framework for AI/ML-based SaMD, by issuing a draft guidance on the pre-determined change control plan. The elements specified in the guidance will support the safety and efficacy of SaMD algorithms and will also include refinement of the identification of types of modifications, appropriate under the framework.
Given the need for GMPL, the FDA plans to focus on AI/ML best practices like, data management, feature extraction, training, interpretability and evaluation and documentation, which are similar to good software engineering practices or quality system practices. Also, the Agency plans to foster harmonization of the numerous efforts to develop GMPL by leveraging already existing workstreams and involvement of other communities focused on AI/ML.
The Agency acknowledged that promoting transparency is a key aspect to a patient-centered approach and especially important for AI/ML-based medical devices, which may change over time and may incorporate algorithms exhibiting a degree of opacity. Regarding the development and utilization of AI/ML-based devices, there are unique considerations that require a proactive patient-centered approach, and it takes into account several issues, including, usability, equity, trust and accountability. The FDA is addressing these issues with an aim to build user trust around the device functionality and ensure patients’ understanding of the device’s benefits, risks and shortcomings.
The FDA notifies that given the opacity of the functioning of many AI/ML algorithms and their development using data from historical datasets, they are vulnerable to bias and prone to mirroring biases present in the data. Hence, the Agency is supporting Regulatory science research efforts and collaborating with leading researchers to develop methods for evaluating AI/ML-based medical software. The methods include the identification and elimination of bias and ensure robustness and resilience of these algorithms to withstand changing clinical inputs and conditions.
In regard to RWP, the FDA says gathering performance data on the real-world use of the SaMD may allow manufacturers to identify opportunities for improvements, understand how their products are being used and respond proactively to safety or usability concerns. As a part of the action plan, the FDA will work with stakeholders on a voluntary basis and in coordination with other ongoing FDA programs focused on the use of real-world data and thereby support the piloting of real-world performance monitoring.
On a final note, the FDA acknowledged that as AI/ML-based SaMD is rapidly progressing, the Agency anticipates that this action plan will continue to evolve and provide additional clarity. Therefore, to gain further updates on the FDA’s AI/ML-based SaMD action plan, stay tuned to our blog space. Stay informed. Stay compliant.