Health Canada digital health head says AI/ML guidance imminent

TORONTO – Health Canada is very close to publishing a much-anticipated guidance on artificial intelligence/machine learning (AI/ML) medical devices and has incorporated pre-determined change control plans (PCCP) into the document, according to Marc Lamoureux, manager of the Digital Health Division at Health Canada.

Lamoureux spoke on 15 October during a panel on artificial intelligence at the 2024 Medtech Conference, hosted by AdvaMed. During his talk, he updated attendees on how the agency is addressing the emergence of AI/ML and using PCCPs to allow devices on the market.

Health Canada has observed several challenges in the AI/ML space over the past six years, according to Lamoureux, who focused on three specific challenges. First, he noted that algorithmic drift and degradation over time are a major challenge, and said AI/ML models tend to work best when they are continuously updated.

“This is a fact,” said Lamoureux. “Obviously, it depends on the model itself, some can deteriorate faster than others, but performance today isn’t necessarily performance for tomorrow.”

Transparency is another key challenge, Lamoureux noted, adding that it is important that regulators, users, and all other stakeholders are transparent about their roles in the AI/ML ecosystem.

“Speaking as a regulator, not only is transparency a risk mitigation, making sure that the user knows what they’re doing and what they’re using, it also is a tool to build trust and increase adoptions in the [health care] systems,” he said.

Postmarket monitoring is also crucial, according to Lamoureux. He noted that a broad range of stakeholders, including patients, health care providers, and medtech experts, have emphasized to Health Canada the need for postmarket monitoring.

Lamoureux said that Health Canada published draft guidance on premarket machine learning-enabled medical devices in September 2023, which takes a total product lifecycle (TPLC) approach. He said the agency has reviewed stakeholder comments on the draft version, and a final version is expected to be published in the next few months.

“The machine learning-enabled medical device guidance document covers things like bias, representativeness of the training data to our Canadian population, it covers transparency… in addition to some comments around the need for postmarket monitoring,” said Lamoureux.

He also noted that Health Canada has regulatory provisions allowing it to put terms and conditions on some of the agency's licensing decisions. For conventional medical devices, that tends to involve requiring a final study report from a clinical trial to get more data on the long-term safety of the licensed device. But now, Lamoureux added, regulators are also instituting conditions that higher-risk devices are monitored in the postmarket setting to ensure the data collected from those products are applicable to the broader Canadian population.

Lamoureux said that while the US Food and Drug Administration (FDA) has taken the lead on developing policies on predetermined change control plans (PCCP), Health Canada has also been working on the issue for several years. He added that the concept has been inserted into the machine learning-enabled device guidance, and the agency is accepting PCCPs.

“PCCP is a new tool, we're very excited to use it, and we look forward to the benefits that it may provide to our native patients,” said Lamoureux.

“We're definitely not going to take the credit for this particular concept, we're happy to follow the US lead as well as the lead from the Japanese [Pharmaceuticals and Medical Devices Agency],” he added.

Medtech Conference

Health Canada digital health head says AI/ML guidance imminent

Related topics

Recommended content

Welcome to the new RAPS Digital Experience