Growth, Specialty Distribution, and Regulatory Pathways of Artificial Intelligence-Enabled Medical Devices Authorized in the United States From 1995 to 2026: A Cross-Sectional Analysis.

Authors

Abstract

Introduction The United States Food and Drug Administration (FDA) publishes a list of artificial intelligence (AI)-enabled medical devices authorized for marketing. Annual authorizations have risen sharply in recent years, and an updated analysis of the complete list is needed. This study aimed to characterize regulatory authorizations on the full list through mid-2026. Methods A cross-sectional analysis of all 1,614 entries on the FDA AI-Enabled Medical Device List (final decisions from September 1995 to June 2026) was performed. Regulatory pathway, year, lead review panel, product code, and manufacturer group were derived for each entry. Growth was estimated by log-linear regression, trends were tested with the Cochran-Armitage test and logistic regression with manufacturer-clustered standard errors, and associations were tested with chi-squared and Fisher exact tests, with Holm adjustment for multiple testing. Results Annual authorizations rose from 18 in 2016 to 335 in 2025, an estimated growth of 35.3% per year (95% confidence interval (CI) 27.0-44.3), and 977 of 1,614 entries (60.5%) were authorized from 2023 onward. Radiology accounted for 1,230 entries (76.2%); its share peaked at 347 of 405 (85.7%) in 2020-2022 and fell to 750 of 977 (76.8%) in 2023-2026 (p < 0.001), while radiology authorizations kept growing in absolute terms and the number of represented specialty panels rose from 10 to 16. The 510(k) pathway accounted for 1,553 authorizations (96.2%), and radiology entries were less likely than others to use the De Novo or premarket approval pathways (18 of 1,230, 1.5% vs. 43 of 384, 11.2%; odds ratio 0.14, 95% CI 0.08-0.27 after adjustment for era). Of 40 De Novo grants, 22 created product codes that later received 181 510(k) clearances. Imaging systems such as ultrasound, computed tomography, and magnetic resonance scanners accounted for 399 entries (24.7%), and seven large manufacturers accounted for 288 of these (72.2%). Findings were similar for 1,195 deduplicated devices and for entries through December 2025. Conclusions FDA authorization of AI-enabled devices continues to accelerate. Radiology remains dominant and still growing, while other specialties are expanding alongside it; the 510(k) pathway predominates, and a growing share of authorizations concerns AI embedded in imaging equipment. These findings describe regulatory authorizations rather than clinical adoption, and they identify transparency about AI functions in conventional devices and AI-specific post-market surveillance as topics for further study.

Tags

Topics