A 510(k) ancestry of robotic surgical systems

摘要

1. Introduction Robotic surgical system usage has increased rapidly in recent years [1]. Da Vinci surgical systems (Intuitive Surgical, Sunnyvale, CA) are among the most common robotic technologies, used in more than 1.2 million procedures annually. The U.S. Food and Drug Administration (FDA) initially cleared these devices in 2000 through the 510(k) process [2], which requires manufacturers to establish that new devices are “substantially equivalent” to a predicate device. There are concerns about cascades of 510(k) device clearances, which may present an obstacle to understanding of safety and effectiveness due to insufficient clinical testing, iterative device changes over time, and use of outdated or recalled devices as predicates [3]. We investigated the 510(k) ancestry of the da Vinci robotic surgical system. 2. Methods Using the FDA’s 510(k) database, we identified the first cleared da Vinci robotic system (K990144). This device was used as the starting point to define three groups (Fig. 1 and Supplementary Material). Group 1 included devices cited as a predicate in the clearance of K990144, and predicates cited by these older predicate devices. Group 2 included devices that used K990144 as a direct predicate and devices that were later cleared based on devices that used K990144 as a predicate. Group 3 included other devices that were cited as predicates for Group 2 devices (excluding K990144). For each device, we extracted product name, clearance date, recalls, and type of evidence (clinical or non-clinical) provided. For clinical evidence, we further examined the number of patients, follow-up duration, randomization, blinding, and primary endpoint type (surrogate or clinical).Fig. 1a.: Full ancestry model graph illustrating devices related to K990144 as nodes and predicate and subsequent device relationships as lines.3. Results The ancestry model included 262 total devices (Fig. 1a). Groups 1, 2, and 3 included 5, 81, and 175 devices respectively (in addition to K990144 itself). Overall, 21 (8.0%) devices cited predicates cleared ≥10 years prior and 58 (22.1%) cleared 5–10 years prior (Fig. 1b).Fig. 1b.: Magnified partial ancestry model for K990144.19 (7.3%) clearances provided supporting clinical data (trial number, registry data, published non-registry clinical data, and/or general reference of clinical testing). Among these, 8 (3.1%) cited meta-analysis of existing publications. Inconsistent and incomplete reporting of meta-analyses, including case reports and case series, hindered comparison of data quality across these studies. Among clinical data outside of meta-analyses, 8 (3.1%) provided the number of patients, 5 (1.9%) listed follow-up duration, 3 (1.1%) employed randomization, and 3 (1.1%) included blinding. Of the 19 (7.3%) clearances including clinical data, 9 (3.4%) utilized clinical endpoints, while the remaining 10 (3.8%) listed surrogate primary endpoints. In total, 170 (64.9%) studies provided non-clinical data (performance data such as bench, animal, and cadaver testing). The remaining 73 (27.9%) cleared 510(k) devices presented no data, relying on proof of mechanical or functional similarities to previously cleared devices (most of which did not present clinical data themselves). Among 262 devices, 55 (21.0%) were recalled, all Class II recalls. 93 (35.5%) devices directly cited predicates that were eventually recalled. 4. Discussion In the ancestry model for the first da Vinci robotic system, 243 (92.7%) 510(k) clearances did not submit clinical data, including 73 (27.9%) that did not submit any supporting data. When clinical data were submitted, the quality of evidence was low. 79 (30.1%) devices cited predicates 5 or more years old and many cited recalled devices in cascades over time. Unfortunately, most postmarket studies of robot devices have been limited to single-centers and device manufacturers’ financial statements [4]. Some larger studies have used claims data, but limitations of administ