Delays between CE mark and FDA regulatory approval of AI-enabled software for radiology

In plain English

The authors compared the order and timing of European CE marking and FDA clearance for AI-enabled radiology software, then tested which product attributes were associated with delay.

What they found

• Among dual-authorized devices, the reported median lag was 17.5 months for CE-first versus 3.5 months for FDA-first products.
• Peer-reviewed evidence was not significantly associated with a shorter interval in the reported model.

The catch

• The observational analysis cannot show why an individual authorization was delayed.
• Database coverage and regulatory classifications may change over time.

Methods at a glance

Study type

Observational regulatory-database analysis

Biological system

AI-enabled radiology software

Organism

Human clinical imaging contexts

Dataset

Health AI Register and FDA sources

Sample size

239 AI-enabled radiology software devices

Model architecture

Kaplan-Meier analysis and Cox proportional-hazards models

Training objective

Estimate time to second authorization and associated factors

Code

No code repository is identified in the publisher abstract.

Data

The analysis used the public Health AI Register and FDA sources.

Tasks and metrics

Tasks: Authorization timing; Factor association analysis

Metrics: Median lag; Hazard ratio; p value

Baselines: CE-first versus FDA-first authorization sequences

Principal findings

• Median lag to second authorization was 17.5 months for CE-first and 3.5 months for FDA-first devices.

What remains unresolved

External validation: Not reported

• Observational associations do not establish the causes of delay.
• Registry completeness and changing authorization status can affect estimates.