Over the past decade, growth in data and computing power has led to a boom in AI-related patent filings, with life and medical sciences emerging as top application areas. His more than 100 applications filed with the U.S. Food and Drug Administration in 2021 included an artificial intelligence and machine learning (AI/ML) component.
According to Brigid Bondock, partner at Morrison Foerster in Washington, D.C., the FDA shares the industry’s optimism and, in the words of Dr. Patrizia Cavazzoni, “AI/ML is safe and effective. This allows us to provide patients with high-quality care more quickly.” Director of the FDA Center for Drug Evaluation and Research (CDER).
For more information on how the agency evaluates and manages emerging AI/ML technologies in pharmaceutical and biologics applications, see the panel, “Artificial Intelligence in the Life Sciences,” presented by Morrison Foerster (MoFo) on May 11, 2023. Intelligence: An Evolving Regulatory Landscape”. At the Association for Corporate Counsel (ACC) live CLE event “2023 Life Sciences Conference: Future-Ready Resilience: Preparing to Face Challenges in the Uncertain Future.”
A panel composed of Bondoc. Anna Yuan, MoFo Senior Associate. Wendy Chou, Treasury Advisor. and Privacy Officer, Senior Director Lauren Wu Regulatory and Compliance Legal, U.S. Privacy Officer, and Global DPO at Evidation Health, said she understands evolving FDA policies and regulations regarding AI and its impact on data privacy. , discussed the protection of AI technology in life sciences.
FDA regulation
Bondoc provided an overview of the FDA’s AI regulation. The FDA actively monitors AI/ML software in medical device and clinical development. The agency could expand its coverage to FDA-regulated activities such as medical device automation and learning, diagnostic and therapeutic development efficiency, regulatory review, and postmarket surveillance.
21st Century Therapeutics excludes five types of software from the FDA’s definition and regulation of medical devices, including administrative support software, wellness, and electronic patient records. Still, the law allows the FDA to “withdraw the exception if it identifies risks associated with software as a medical device,” Wu said. With the advent of AI technology, wellness, and electronic records, “the water will get muddy,” Wu commented. For example, what to do with a pedometer that monitors sleep and blood pressure. It may move beyond just wellness devices and into the realm of regulated medical devices.
The FDA appears to be flexible and cautious in developing its AI/ML regulatory framework, offsetting the need to foster innovation while protecting public health. For example, Bondoc points to the agency’s draft guidance on Predetermined Change Control Plans (PCCPs) for AI/ML-enabled device software capabilities, which seeks to address critical issues facing emerging technologies.
Historically, submission of a Premarket Notice 510(k) was required for approval or changes to approved medical devices that could have a significant impact on the safety or efficacy of the device. However, the notification program is “fundamentally incompatible with his use of AI/ML in medical devices, which encourages continuous improvement and modification of devices based on data collected during use,” said Bondoc. rice field. The FDA aims to develop a regulatory approach tailored to AI/ML-enabled devices that allows for safe and rapid changes in response to new data, while ensuring safety and efficacy.
The FDA considers the use of AI/ML in drug development to have ethical, privacy, and security issues. The agency is also concerned about the lack of transparency in algorithms that can lead to amplification errors and pre-existing biases. “We are aware of the problem, but no potential regulatory solution has yet been provided,” said Bondoc. Still, the agency aims to prevent algorithmic discrimination and promote the use of AI/ML techniques. The company recently published a discussion paper, “Using Artificial Intelligence and Machine Learning in Drug and Biologics Development.” And to further address the issue of AI in drug manufacturing, the FDA announced “Artificial Intelligence in Drug Manufacturing.”
Protecting AI Technologies in Life Sciences
Citing WIPO Technology Trends 2019: Artificial Intelligence and Mondac, Yuan observed an eight-fold increase in AI patents from 2017 to 2020 and observed changes in the types of patents granted. Previously, patents were more theoretical, such as new AI algorithms for speech recognition. Currently, there are a growing number of patents on specific, practical applications of machine learning in drug discovery and medical diagnostics.
Patent applications include novel practical applications of AI algorithms, AI-developed medicines, and computer hardware configurations and optimizations to facilitate AI training and inference processes. Patents he falls into two buckets. One is the application of known AI to specific fields and sectors, and the other two are new AI models and algorithms. Mr. Yuan said the former are generally more worthy of patenting, including broader and more detectable claims, than the latter, which are narrower in scope and contain mathematically oriented claims.
Privacy and AI
Uses of AI/ML in life sciences include data from individuals and even protected health information (PHI). At some point, whether the training input, inference, or output of the process, the use of the data violates state privacy laws, involves new laws regulating AI and automated decision-making, medical It could lead to the activation of federal legislation such as insurance interoperability and medical insurance interoperability. The Liability Act (HIPAA) or Federal Trade Commission law prohibits unfair and deceptive practices, such as the sale or use of racially biased programs.
Chow also elaborated on states with overarching privacy laws, including California, Virginia, and Tennessee, while Connecticut, Indiana, Iowa, Montana, and Washington have numerous state laws. said it was pending. State legislatures complement privacy laws with laws regulating AI and automated decision-making, such as the New York State AI Bias Act. Still, with privacy laws in the background, Wu said, “We are not necessarily a driver of AI, we are a stakeholder.”
