A quantitative accuracy assessment of the use of a rigid robotic arm in navigated placement of 726 pedicle screws

Minimally invasive (MI) techniques have been integrated into spine surgery in a number of ways, including the placement of percutaneous pedicle screws. As the most common anchor point in posterior spinal fusion procedures, high level of accuracy in screw placement is paramount. Despite relatively low clinically adverse complications associated with malposition, a misplaced screw may still result in neurological/neurovascular injury [1, 2], dural tears [3, 4], sub-optimal biomechanics [5] or other visceral involvement [6]. Furthermore, the number of underreported cases of pedicle malposition are estimated to be as high as 15.7% [3], while the range of pedicle-screw based complications is extremely high (between 1 and 54%), demonstrating that patient conditions and other anatomical factors play a large role in successful screw placement [7]. Several identified patient factors such as challenging deformities, osteoporosis, and tumor have been described [8].

The major modalities for MI spine surgery are fluoroscopic guidance, computed tomography navigation (CTnav), and robot-assisted surgery. The level of evidence in all 3 fields has grown substantially, with several systematic reviews available [9‐11]. Fluoroscopy alone, while effective (91.3% accuracy [9]), suffers from obvious radiation and ease-of-use considerations [12, 13], as well as a level of unpredictability based on surgeon experience. Both robotic-guided and CTnav have higher predictability irrespective of experience levels (15.7% [3] vs 7.0% [14] and 5.1% [15] respectively), and most literature supports that robotic-guided surgery has still higher accuracy than fluoroscopic guidance [14]. The use of navigation and robotic assistance results in less radiation exposure than the use of traditional fluoroscopic guidance [16]. Patient factors, such as minimal muscle disruption, quicker recovery times, and less postoperative pain have led to more prevalent uses of this technology [16], while accuracy, safety, and cost appear to be the primary drivers of adoption.

Further data are needed to describe the accuracy and safety of such systems, particularly in a field in which the robotics technologies operate with different hardware, software, and clinical implementation. The aim of present study was to quantitatively assess the accuracy of pedicle screws placed with the guidance and navigation of a robot.

This retrospective chart review was exempt from the Italian Ethics Committee. Data were collected from 3 surgeons at a single site. All three surgeons are experienced neurosurgeons who have been doing robotic-assisted surgery since its adoption at their surgical center. Patients who were included in the analysis were between 21 and 85 years of age and required surgery that includes screws to be inserted in thoracic, lumbar spine or sacrum. Demographic data (including age, gender, BMI and diagnosis), operative data (including set up time, screw insertion time, operative time, blood loss, radiation time), preoperative/postoperative computerized tomography (CT) scans, and complication rates of 127 patients treated with lumbosacral pedicle screws through a minimally invasive robotic assisted technique were analyzed. The methods for this publication are similar to those described in a previously published manuscript by the same principal investigators [17].

The surgical technique employed in this study utilized a robotic positioning system (ExcelsiusGPS®, Globus Medical, Inc., Audubon, PA, USA) (Fig. 1) in which patient radiographs may be uploaded and registered through one of three modalities: preoperative CT, intraoperative CT, or fluoroscopy. For this study, only the preoperative CT workflow was used. In this workflow, a CT scan is taken prior to the surgery, and screw placement planning can be performed. In the operating room (OR), fluoroscopy is used to “merge” the CT and plan to the patient’s positioning on the surgical table. The robotic system uses a dynamic reference base and positioning camera to track the position of instrumentation in real time and 3-dimensional (3D) space, while the rigid robotic arm guides the surgeon to the planned screw trajectory.

Accurate placement of pedicle screws is essential to avoid complications directly related to screw placement, including dural tears, vascular injury, and compression of the neural elements leading to neurological deficit, among others [23]. Malposition rates of screws implanted without robotic guidance vary widely in the literature from 5.9 to 20.4% [10, 18, 24]. A systematic literature review of robotic screw placement by Joseph et al. [25] found accuracy ranged from 85 to 100% of screws across 25 studies, using a GRS grading of A or B as accurate. In the current study 97.9% of screws were graded either A or B on GRS. This puts the current study in line, but in the 50th percentile of studies compiled in the review done by Joseph et al. [25]. Laudato et al. in a retrospective radiological study from a single academic surgical center, [26] reported a rate of 1.56% of screws determined to be greater than 4 mm breach in a cohort of patients treated with the aid of robot. The current study reports 0.6% (4/726) of screws at this rating. The results of this analysis are similar to other published studies of robotic assisted pedicle screw placement and add to the growing body of literature on the subject.

Placement of the screw within the boundaries of the pedicle is dependent on the accuracy of the screw to the plan developed by the surgeon. Utilizing preoperative CT, the surgeon creates a plan for the placement of each pedicle screw which is registered to the intraoperative fluoroscopy. Close matching to the preoperative plan indicates that the navigation system is accurate in guiding the surgeon during the procedure. Jiang et al. [27] have shown in a similar study, published recently, that accurate screw placement can be achieved with robot with mean deviation of only 1 mm from the planned screw trajectory. Also, in a prospective randomized controlled trial, Han et al. reported the mean deviation of placed screws from planned trajectories was 1.4 ± 0.9 mm for the entry point and 1.6 ± 1.0 mm for the end point. The mean deviation for each screw was 1.5 ± 0.8 mm [28]. In the current study, the average offset from preoperative plan to final screw placement was 1.9 ± 1.5 mm at the tip, 2.2 ± 1.4 mm at the tail and 2.9 ± 2.3° of angulation. Slight deviations from the plan can be expected due to shifting of the patient anatomy intraoperatively, however the values demonstrate that this impact is low. A surprising finding of this study was that neither patient age nor BMI was correlated to screw offset values. Obesity has been shown to be a risk factor of screw misplacement with traditional methods. In one study [8], it was found that obesity significantly increased the odds of misplacement to 3.4. A retrospective analysis of 874 screws [29] determined that BMI was a risk factor for malposition of screws implanted with robotic assistance. The odds ratio for screw misplacement in patients with obesity was found to be 5.4 and statistically significant. Similarly, a comparative study [24] found no difference in the odds ratio for screw misplacement in case of obesity between robotic assisted and freehand placed screws. Absence of a correlation in this study suggests that robotic screw placement in this cohort was not greatly impacted by patient obesity. As obesity may complicate surgical procedures leading to an increased risk of complication [30], this is a finding of note, which requires further investigation. Querying the National Outcomes Database, Onyekwelu et al. found that the obese patients had greater blood loss in surgery, longer surgery times and longer length of stay in hospital postoperatively [31]. Robotic surgery has been shown to reduce operative time [32], potentially mitigating these issues.

The strengths of this study include the large patient size and the addition of surgical/accuracy data to a relatively disparate amount of data that exists for the 3 currently marketed robotic systems, of which the Mazor system (Medtronic, Denver, CO, USA) has been most heavily studied. Limitations of this study are the lack of long-term clinical outcomes to demonstrate the impact of robotic surgery and pedicle screw accuracy on patient health. The current study is from a single location, which impacts the generalizability of the results. Future studies with long-term follow ups from multiple sites under multiple surgical workflows are required to resolve these limitations. Also, future studies can investigate the factors affecting the accuracy of screw placement with the robots.

Robotic-assisted surgery is highly accurate for the placement of pedicle screws within the patient’s pedicles and according to preoperative planning. In this study, 97.9% of screws were rated either A or B on the Gertzbein–Robbins Scale and tip and tail offset values were an average of 1.9 ± 1.5 mm and 2.2 ± 1.4 mm respectively. Mean intraoperative radiation time was 18.3 ± 13.3 s. Screw placement accuracy is unaffected by patient BMI, a potential benefit of this procedural approach.