Evaluating the Accuracy of Robotic Assistance in Spine Surgery

Abstract

The rationale behind the SpineAssist surgical robot (Mazor Surgical Technologies Ltd., Caesarea, Israel) is to improve traditional image-guidance techniques by utilizing a robotic arm to guide the surgeon to a pre-planned starting point and trajectory for pedicle screw placement. This system has been FDA-approved for almost 10 years for thoracic, lumbar, and sacral pedicle screw placement. Clinical data concerning the safety and accuracy of the system has slowly accumulated. Earlier reports using cadaveric studies and then later retrospective analyses of screw placement indicated that a high degree of accuracy could be provided by robotic assistance. In the April 2012 edition of SPINE, Ringel et al report the results of the first prospective randomized trial comparing the accuracy of free-hand vs robotic-assisted placement of pedicle screws.1 In this clinical trial the authors enrolled 60 patients (298 total screws implanted) who underwent single or two-level lumbo-sacral fusion in a 1:1 randomization to free-hand or robotic pedicle screw placement. All patients underwent a post-operative thin-cut computed tomographic (CT) scan that was evaluated by a blinded neuro-radiologist to identify cortical breeches using the Gertzbein and Robbins classification system. Secondary endpoints including surgical time and radiation exposure were also recorded. Surprisingly, the authors report the free-hand technique resulted in 93% good screw position (Gertzbein and Robbins classification of A and B) compared to 85% in the robotic-assistance group (P = .019). This corresponded to a larger proportion of screws in the robotic-assistance group that were classified as sub-optimal (15% robotic vs 7% free-hand). It is also important to note that in 7 patients in the robotic group (almost 25%) there were 10 screws that required conversion to a free-hand technique after the trajectory provided by the robot missed the vertebral body laterally. Secondary measures indicated that the overall operative time per group and the time of intra-operative radiation were not significantly different per group. These results are surprising given the prior reports from cadaveric and retrospective analyses that indicated improved screw accuracy with robotic assistance. Kantelhardt et al demonstrated in a retrospective cohort study of over 500 pedicle screws that the robotic assistance increased screw accuracy to 94.5% from 91.5% using conventional techniques.2 Other retrospective reports have shown even higher accuracy with Devito et al reporting a 98.3% good placement (Gertzbein Robbins A and B) in 646 screws.3 Interestingly they also reported a successful “execution rate” of the robot only ranging from 83% to 90% depending on where in the learning curve the cases were performed. The authors of the current study also highlight limitations of the system that may help explain the low accuracy provided by robotic assistance in this study. First, there are several techniques available to fixate the robot to the patient. In this study the authors utilized the “bed mount” option where the robot is only attached to the patient by a single K-wire which they speculate may allow for movement of the system relative to the patient. Furthermore all screws were placed through a paramedian, Wiltse-type approach with blunt perforation of the paraspinal musculature by the cannula. The authors note that with this technique any firm muscle bundles could lead to deflection of the cannula. Likewise facet joint hypertrophy with a steep facet can give rise to lateral skidding of the cannula at the pedicle screw entrance point. All of these factors may influence the accuracy of the system and should be considered when utilizing the system. In conclusion, the use of robotic assistance for pedicle screw placement has been gaining clinical acceptance in select sites throughout the United States and Europe. With this increased utilization the generation of outcome data has also been accumulat