Error Types and Associations of Clinically Significant Events Within Food and Drug Administration Recalls of Linear Accelerators and Related Products.

ABSTRACT

PURPOSE: Medical devices in radiation therapy undergo a complex process of Food and Drug Administration (FDA) approval. Little is known about which processes within the radiation therapy medical device industry are most prone to events involving wrong dose, volume, or targeting in radiation therapy treatment. METHODS AND MATERIALS We carried out a retrospective analysis of the United States FDA Medical Device Recalls database for recalls of products classified as "Accelerator, Linear, Medical" from 2010 to 2016. Each recall event was classified using a modified Delphi method among 3 experts in safety according to product type, error category, and severity score. Error categories included inconvenience; suboptimal plan or treatment; incorrect dose, volume, or targeting; and nonradiation injury risk. Variables investigated were product type, recall year, FDA-determined cause, and quantity of units recalled. Univariate and multivariate logistic regression were used to identify factors prognostic of incorrect dose, volume, or targeting. RESULTS: We identified a total of 250 recall events between 2010 and 2016, with 165 eligible for analysis. Linear accelerators (LINACs) (28%) and LINAC control software (19%) were the most frequently recalled products. The most common FDA-determined causes for recalls were software design (42%) and device design (26%). On univariate analysis (P < .05), LINAC control software (odds ratio [OR] 5.4) and oncology information system or treatment management system (OR 3.9) versus LINACs and software design (OR 3.4) versus device design were associated with wrong dose, volume, or targeting events. On multivariate analysis, only the association with LINAC control software (OR 3.7) persisted for wrong dose, volume, or targeting events. CONCLUSIONS: Review of these data shows that problems with LINAC control software were associated with incorrect dose delivery at a 4-fold higher rate than errors with LINACs. Manufacturers should focus on improvements in software design to minimize dose- and targeting-related errors to patients.