Study: AI-enabled pediatric medical devices rare, limited to few clinical areas

There have been a handful of artificial intelligence (AI)-enabled medical devices approved by the US Food and Drug Administration (FDA) over the last decade, and those that were have been limited to a few clinical areas, according to recent research published in JAMA Network Open.

Grzegorz Zapotoczny, PhD, of the Stanley Manne Children’s Research Institute at Ann & Robert H. Lurie Children’s Hospital of Chicago, and colleagues said that it is difficult to determine through FDA’s AI device database whether a medical device has been approved with a pediatric indication. However, what they found was that AI-enabled devices with pediatric indications lagged behind AI-enabled medical devices approved for adults.

“This cross-sectional study found that pediatric labeling among FDA-marketed AI-enabled devices was rare, recent, and concentrated in a few clinical areas, with frequent ambiguity around age indications,” Zapotoczny and colleagues wrote. “These patterns, along with inconsistent publicly available data, limit clinician ability to assess pediatric applicability and may slow or inhibit access to AI-supported care for pediatric patients.”

In a retrospective cohort study, Zapotoczny and colleagues evaluated 952 marketing submissions of AI and machine learning-enabled medical devices between November 1995 and June 2024 in FDA’s AI-enabled medical device list, identifying 42 submissions (4.4%) with specific pediatric indications, 102 submissions (10.7%) with an all-ages label, and 565 submissions (59.3%) without an age indicator.

They found clearance of the first device occurred in 2015, and devices with pediatric indications limited to three clinical areas: radiology, neurology, and cardiology. Of all medical devices evaluated, 42.9% had pediatric indications for radiology, 31.0% of devices had pediatric indications for neurology, and 9.5% had pediatric indications for cardiology. However, most of the overall medical devices in the database were cleared for radiology (95.9%) indications, while few devices were cleared for cardiovascular (10.4%) or neurologic (3.6%) indications. Researchers noted that pediatric devices were absent in ten other clinical areas (55.6%).

Devices with specific pediatric labels had significantly longer median review times, at a median of 162 days, compared to non-pediatric devices at a median of 134 days (P = .049). Researchers also found pediatric devices had a higher proportion of clinical trial registrations in FDA summaries compared to non-pediatric medical devices (14.3% vs 2.2%).

While there are no formal approval standards for these devices, the researchers said their results suggest a higher evidentiary standard for medical devices with pediatric indications, which may lead to manufacturers choosing to not invest in the development of pediatric medical devices.

“Policy changes addressing device labeling, evidence requirements, and pediatric-specific considerations could reduce friction, alongside investments in pediatric clinical research infrastructure and dataset development,” the researchers said. “Taken together, these pragmatic steps could support the safe and effective deployment of AI technologies for pediatric patients of all ages.”

Reducing barriers for pediatric medical devices

In a related editorial, R. Brandon Hunter, MD, of Baylor College of Medicine in Houston, and colleagues said that the lack of pediatric medical device innovation is well known, and the study by Zapotoczny confirms this also occurs in AI-enabled devices as well.

The findings that AI-enabled devices with pediatric indications have longer review times but a higher percentage of registered clinical trials compared with non-pediatric devices “suggest that FDA reviewers may, in practice, expect more rigorous evidence to grant pediatric approvals even in the absence of formal statutory requirements,” they said.

While a recommendation to explicitly label medical devices as having a pediatric indication and validation may help bridge the gap of uncertainty in some AI-enabled medical devices, it may introduce the problem of manufacturers experiencing longer timelines and a higher evidentiary standard in a smaller market, Hunter and colleagues said.

“Transparency mandates without accompanying support risk converting passive omission of pediatrics into active exclusion, with labels that simply state, ‘not indicated for pediatric use,’” the authors said. “To avoid this outcome, transparency requirements must be paired with mechanisms that lower barriers to pediatric AI development.”

Examples of approaches to lower barriers include streamlining validation pathways, the use of real-world evidence or smaller samples sizes, and the introduction of incentive programs like what is seen in orphan drugs and humanitarian use devices.

“Ultimately, clinicians need clear information about which AI tools are appropriate for pediatric use and a regulatory environment that makes developing those tools worthwhile,” Hunter and colleagues concluded.

JAMA Network Open Zapotoczny et al.

Study: AI-enabled pediatric medical devices rare, limited to few clinical areas

Related topics

Recommended content

Welcome to the new RAPS Digital Experience