Related Experiment Video
Updated: Sep 19, 2026
Radiation Planning Assistant - A Web-based Tool to Support High-quality Radiotherapy in Clinics with Limited Resources
Published on: October 6, 2023
Radiation Oncology Device Recalls (2015-2025): Software and Artificial Intelligence as Emerging Drivers of Postmarket
1Department of Radiation Oncology, Royal Brisbane and Women's Hospital, Brisbane, Queensland, Australia; Faculty of Medicine, University of Queensland, Brisbane, Queensland, Australia.
Purpose:
Radiation oncology (RO) relies on complex devices spanning simulation, planning, delivery, and verification, making postmarket surveillance critical. Prior analyses of U.S. Food and Drug Administration (FDA) recalls (2002-2015) demonstrated a high recall burden, predominantly driven by software-related failures. Following recent technological advances, including rapid expansion of software systems and artificial intelligence (AI)-enabled tools, we analyzed RO device recalls from 2015 to 2025 to investigate if this has resulted in a proportional increase in recalls.
Methods And Materials:
FDA recall data from 2015 to 2025 were queried, and RO devices were identified using product codes and device descriptions. Recalls were analyzed by device category (external beam, brachytherapy, software, or simulation), recall severity, FDA-determined root cause, and quantity in commerce. AI-enabled devices were identified by cross-referencing the FDA AI-enabled medical device list with product codes. RO-specific FDA approvals per year were assessed using the 510(k)-clearance database.
Results:
Of the 30,648 FDA recalls from 2015 to 2025, 426 (1.39%) were RO-related. External beam systems accounted for 61.6% of recalls, and planning/software systems of 34.4%. Software issues were the most common root cause (45%), exceeding 50% in 6 of 11 years. Most recalls were class II (97.7%), indicating potential temporary or reversible health consequences. From 2018 onward, 1357 AI-enabled devices entered the market, with 70 (5.16%) identified as RO-related. Over this period, 68 AI-enabled RO devices were recalled, with 62 (91.18%) of those involving a single planning system.
Conclusions:
RO has rapidly adopted software-based and AI-enabled technologies, yet recall frequency has remained stable, which may potentially obscure software-related safety risks. These findings underscore the need for continued postmarket surveillance, rigorous commissioning and revalidation, lifecycle oversight, and clinician education. This study also provides a contemporary benchmark to inform patient safety strategies as the field continues to evolve.
