Complications Have Not Improved With Newer Generation Robots

Abstract

Study Design

Retrospective cohort study.

Objective

The purpose of this study was to see whether upgrades in newer generation robots improve safety and clinical outcomes following spine surgery.

Methods

All patients undergoing robotic-assisted spine surgery with the Mazor X Stealth Edition^TM (Medtronic, Minneapolis, MN) from 2019 to 2022 at a combined orthopedic and neurosurgical spine service were retrospectively reviewed. Robot related complications were recorded.

Results

264 consecutive patients (54.1% female; age at time of surgery 63.5 ± 15.3 years) operated on by 14 surgeons were analyzed. The average number of instrumented levels with robotics was 4.2 ± 2.7, while the average number of instrumented screws with robotics was 8.3 ± 5.3. There was a nearly 50/50 split between an open and minimally invasive approach. Six patients (2.2%) had robot related complications. Three patients had temporary nerve root injuries from misplaced screws that required reoperation, one patient had a permanent motor deficit from the tap damaging the L1 and L2 nerve roots, one patient had a durotomy from a misplaced screw that required laminectomy and intra-operative repair, and one patient had a temporary sensory L5 nerve root injury from a drill. Half of these complications (3/6) were due to a reference frame error. In total, four patients (1.5%) required reoperation to fix 10 misplaced screws.

Conclusion

Despite newer generation robots, robot related complications are not decreasing. As half the robot related complications result from reference frame errors, this is an opportunity for improvement.

Introduction

Pedicle screw fixation provides superior postoperative spinal stability. The freehand approach for screw placement relies on a combination of anatomical landmarks and fluoroscopic guidance. ¹ However, reports of screw misplacement using this method have varied widely, ranging from 8.3% to as high as 50.6%.^2,3 This is a concern, as malpositioned screws can lead to major neurological, vascular, and visceral complications. In the last decade, the use of robotics in spine surgery has become more popular for its purported increased precision and ability to be unaffected by fatigue. Additionally, robotic-assisted pedicle screw delivery in minimally invasive surgery was viewed as ideal due to the limited need for extensive tissue dissection. Recent radiographic outcome studies have reported superior radiologic accuracy, nearing an accuracy rate of 99% with the newest generation robot.^4-11 The purpose of this investigation was to see whether updates in these newest generation robots also result in improved safety and clinical outcomes.

Methods

The Cedars-Sinai Medical Center Institutional Review Board (IRB) approved this retrospective cohort study (00002314). The study does not require informed consent given the data reviewed was deidentified and collected in accordance with the institution’s standard of care. This study included consecutive patients undergoing robotic-assisted spine surgery between December 2019 to November 2022. Robots for assisted pedicle screw placement were used by 14 surgeons utilizing the Mazor X Stealth Edition^TM (Medtronic, Minneapolis, MN). Only patients undergoing pure sacroiliac joint fusions were excluded from this study. All remaining patients, regardless of age and available follow-up time, were included in the final analysis. Diagnoses indicating surgery, number of instrumented levels, number of planned robot screws per patient, robot related complications, and robot related reoperations were recorded. Only descriptive statistics (mean, standard deviation) were used as this was not a comparative study.

Results

There were 264 patients (age at time of surgery 63.5 ± 15.3 years) with an average follow-up of 1.0 ± .8 years (range: 3 day – 3.1 years) (Table 1). The mean (standard deviation) Charlson comorbidity index (CCI) was .7 (.9) and 54.1% (n = 143) of patients were female. Surgical diagnosis included: spondylolisthesis (n = 75; 28.4%), degenerative scoliosis (n = 54; 20.4%), spondylosis (n = 17; 6.4%), pseudarthrosis/implant failure (n = 15; 5.5%) and other (n = 23; 8.7%). The average number of instrumented levels was 4.2 ± 2.7, while the average number of planned robot screws was 8.3 ± 5.3 (Table 2). Open approach surgery was performed in 50.0% (n = 132), pelvic fixation was performed in 25.0% (n = 66), and interbody fusion was performed in 71.5% (n = 189) of patients. Estimated blood loss was 543.0 ± 580.7 mL, while 27.6% (n = 73) of patients received blood transfusions perioperatively. The average length of hospital stay was 6.1 ± 6.7 days.

Table 1.

Patient Demographics, Comorbidities, and Indicating Diagnoses for Robotic-Assisted Spine Surgery.

	N (%)
Total number of patients	264 (100.0%)
Female	143 (54.1%)
Age at surgery, mean (SD)	63.5 (15.3)
BMI, mean (SD)	27.37 (5.86)
Obese (BMI >30 Kg/m2)	71 (26.9%)
CCI, mean (standard deviation, SD)	.7 (.9)
Prior/Current smoker	41 (15.5%)
Follow-up time, mean (SD)	12.3 (9.2) months Min: 3 day Max: 37.5 months
Preoperative diagnosis
Spondylolisthesis	75 (28.4%)
Degenerative scoliosis	54 (20.4%)
Degenerative disc disease	44 (16.6%)
Fracture	20 (7.5%)
Spinal stenosis	16 (6.0%)
Spondylosis	17 (6.4%)
Pseudoarthrosis/Implant Failure	15 (5.5%)
Other	23 (8.7%)

Table 2.

Perioperative Data.

	N (%)
Operative data
Open (vs. minimally invasive surgery)	132 (50.0%)
Instrumented levels per patient, Mean (SD)	4.2 (2.7)
Pelvic fixation	66 (25.0%)
Interbody fusion	189 (71.5%)
Planned robot screws per patient, Mean (SD)	8.3 (5.3)
Estimated blood loss, mean (SD)	543.0 (580.7) mL
Perioperative blood transfusion	73 (27.6%)
Length of hospital stay, mean (SD)	6.1 (6.7) days

There were six patients (2.2%) who had robot related complications. Three patients had temporary nerve root injuries from misplaced screws that required reoperations (Figures 1-3), one patient had a permanent 1/5 motor deficit from the tap wrapping the L1 and L2 nerve roots that required a reoperation (Figure 4), one patient had a durotomy from a misplaced screw that required laminectomy and intra-operative repair, and one patient had a sensory L5 nerve root injury from the drill (Figure 5). Half of these complications (3/6) were due to a reference frame error. In total, four patients (1.5%) required reoperations to fix 10 misplaced screws (Table 3) (Figures 1-4), and they were all performed by four different surgeons. Years in practice for these surgeons ranged from 9 years to 27 years, and all had freehand navigation experience. These complications were not clustered in the early learning curve of robot usage of the surgeons. Average time to robotic related reoperations was 3.7 months ± 6.6 months (range 5 days – 1.1 years).

Figure 1.

(A) Medial breach of right T10 pedicle screw with lateralized left T10 pedicle screw and (B) medial breach of right T11 pedicle screw with lateralized left T11 pedicle screw in a 76 year old male presenting with worsening thoracic back pain after a left lateral T12-L1 discectomy + T10-L3 PSF. Both screws were shifted towards one side by about 10 mm (red arrows). The authors believe that the reference frame placed on the posterior superior iliac spine was too far from T10/T11, resulting in movement over the distance that was not detected. The patient later required a reoperation for screw replacement. Legend: PSF, posterior spinal fusion.

Figure 2.

(A) Medial breach of left L1 pedicle screw and (B) lateralized right L2 pedicle screw in a 71 year old female presenting with back pain and new 1/5 weakness of the right quadriceps and iliopsoas muscles after L1-Pelvis PSF. The authors believe this patient had a reference frame problem, as both screws were shifted by about 10 mm (red arrows) in the same direction. The reference frame was placed on the posterior superior iliac spine. The patient later required reoperation for screw removal.

Figure 3.

Medial breach of left L4 pedicle screw in a 70-year-old male who presented with numbness and 3/5 weakness in the left lower extremity after undergoing L4-S1 ALIF + PSF. The attending surgeon believes that the soft tissue pushed against the cannula, resulting in a medial breach of the spinal canal at L4. The patient later required a reoperation to remove screw. Of note, the reference frame was placed on the posterior superior iliac spine. Legend: ALIF, anterior lumbar interbody fusion; PSF, posterior spinal fusion.

Figure 4.

(A) Lateralized left L2 pedicle screw, (B) lateralized left L3 pedicle screw, (C) lateralized left L4 pedicle screw, and (D) superior misplacement of left L5 pedicle screw into the L4-L5 disc space in a 76-year-old female who presented with new back and left lower extremity pain after undergoing L4-S1 ALIF and T10-Pelvis PSF for long-standing scoliosis. Of note, the reference frame was docked on to the spine, rather than the posterior superior iliac spine. Patient later required a reoperation to replace screws. Legend: ALIF, anterior lumbar interbody fusion; PSF, posterior spinal fusion.

Figure 5.

This patient had a grade IV spondylolisthesis at L5-S1. When the S1-L5 transdiscal screw was placed, L5 was pushed away relative to the reference frame on the posterior superior iliac spine, which the surgeon believes resulted in the drill hitting the nerve root. This resulted in an immediate spike in the patient’s EMGs, her leg kicked, and she had went on to have 3 months of a burning sensation on her dorsal foot. (A, B) demonstrate the distance between L5 and S1 (red arrow) before and after the transdiscal screw was placed, respectively.

Table 3.

Complications.

	N (%)
Number of complications	6 patients (2.2%)
Nerve root injury from misplaced screws requiring reoperation	3 patients
Nerve root injury from tap requiring reoperation	1 patient
Durotomy from misplaced screws requiring intra-operative repair	1 patient
Nerve root injury from drill	1 patient
Presenting Symptoms
Back pain	3 patients
Sensory deficit	2 patients
Motor deficit	2 patients
Durotomy	1 patient
Number of reoperations	4 (1.5%): 10 misplaced screws; 4 surgeons
Time to reoperation, mean (SD)	3.7 (6.6) months

Discussion

Despite using the newest generation robot, the complication rate remains substantial at 2.2% and surgical reoperation rate to fix misplaced pedicle screws at 1.5% (4/268 patients).

Previous studies on earlier generation robot-assisted screw systems reported comparable accuracy results to conventional free-hand technique. Using the Gertzbein and Robbins Scale, Schatlo et al ¹² observed 83.6% of screws as perfect trajectories (Grade A) and 7.8% of screws as Grade B, which was similar to the fluoroscopic guided group. A later study reported similar results, finding 84.4% of robotic screws had trajectories of either Grade A or B. ¹³ In contrast, one report by Ringel et al ⁴ found the free-hand group to perform superiorly, with 93% of screws achieving acceptable positions compared to 85% in the robotic group. These earlier studies used a first generation robot, SpineAssist^TM (Mazor Robotics Ltd., Caesarea, Israel), introduced in 2004. Later studies reporting on an updated robotic system platform, the Mazor X^TM (Mazor Robotics Ltd., Caesarea, Israel) introduced in 2016, have shown improvements. The first study to report on the Mazor X system evaluated 20 patients and 75 screws, finding an accuracy rate of 99%. ¹⁴ When compared to CT-image guided navigation (CT-IGN) systems without robot, Mazor X achieved a similar screw accuracy of 99.5%, compared to 95.2% in the navigation group. ¹⁵ Mao et al also found a similar screw accuracy between 39 Mazor X cases and 46 CT-IGN cases, and also reported that the robot-assisted technology achieved a higher percentage of grade A type screws (robot: 86.2% vs navigation: 66.0%). ¹⁶

The clinical significance of high accuracy radiographic outcomes studies remain unclear. Floccari et al ¹⁷ surveyed 12 experienced spine surgeons and reported finding significant variability of opinion in which malpositioned screws could safely be observed in an asymptomatic patient. Thus, to evaluate clinically meaningful outcomes, the focus of our study was on clinical complications and reoperations directly attributable to the robot, and not issues such as wound complications, infections, and/or other adverse events. When compared to previous reported studies, our reoperation rate of 1.5% was the second highest reported to date (Table 4). It should be noted that out of the 12 studies reporting on robot related reoperations, the eight studies reporting a 0% reoperation rate had the fewest patients. When looking at the bigger series, the reoperation rate came closer to ours of 1.5%. Our study shows that reoperations have not necessarily been improving despite using the newest generation of robots. One of the older Mazor robots, the Mazor Renaissance^TM, demonstrated reoperation rates of .80% (3/374 patients) and 1.7% (7/406 patients), whereas our study used the newer Mazor X Stealth Edition but still found similar reoperation rates. However, it should be mentioned that another study using the newest generation Mazor X Stealth Edition found a lower reoperation rate of .63% (5/799 patients). It was noted that many of these studies were single surgeon studies and on average involved instrumentation of fewer levels per patient. In contrast, our study included 14 surgeons, consisting of a mixture of orthopedic and neurological spine surgeons, and on average involved instrumentation of 4.2 levels per patient. The fact that four different surgeons were responsible for 4 different reoperations suggests our reoperation data is not skewed by one bad surgeon.

Table 4.

Summary of Studies Reporting on Reoperations for Misplaced Pedicle Screws Using Robotic-Assistance.

Study (Year)	Country	Robotic System	Age	BMI	Length of Hospital Stay	Estimated Blood Loss	Revision Rate for Misplaced Screws
Karamian (2021) 18	USA	Globus Excelsius + Renaissance Mazor	64.9 ± 11.5	29.5 ± 5.58	NR	NR	0/85 patients (0%)
Liounako (2021) 19	USA	Mazor X + Mazor X Stealth	57.5 ± 12.8	29.5 ± 5.62	NR	NR	5/799 patients (.63%)
Good (2021) 20	USA	Mazor Renaissance	59.0 ± 12.6	31.2 ± 6.8	NR	NR	3/374 patients (.80%)
Li (2019) 21	China	Orthbot	47.4 ± 12.9	24.3 ± 1.8	13.1 ± 2.8 days	257 ± 181 mL	0/17 patients (0%)
Han (2018) 22	China	TiRobot	54.6 ± 11.3	25.7 ± 4.1	4.8 ± 1.5 days	186 ± 255 mL	0/115 patients (0%)
Keric (2017) 23	Germany	Renaissance Mazor	65.8 (20-89)	NR	13.7 ± 9.1 days	NR	7/406 patients (1.72%)
Schröder (2017) 24	Netherlands	SpineAssist	52.8 ± 11.4	25.9 ± 3.5	2.4 ± 1.1 days	336 ± 262 mL	0/72 patients (0%)
Laudato (2017) 25	Switzerland	Robotics	65	NR	NR	NR	0/11 patients (0%)
Hyun (2016) 26	Korea	Mazor Renaissance	66.8 ± 8.9	25.8 ± 3.3	6.8 ± 2.1 days	NR	0/30 patients (0%)
Kim (2016) 27	Korea	Robot (Renaissance Mazor)	65.4 ± 10.4	25.9	NR	NR	0/37 patients (0%)
Schatlo (2014) 12	Switzerland	SpineAssist	52 (27-83)	24.7 ± 3.7	9.8 ± 5.1	375 ± 263 mL	0/55 patients (0%)
Ringel (2010) 4	Germany	SpineAssist	68 (median)	26 (median)	NR	NR	0/30 patients (0%)

Legend: NR, not reported.

There was no case in which it seemed the mechanical part of the robot failed. Instead, complications in our cohort could better be described by navigation failures. In particular, half (3/6) of the patients from our study who had a robot related complication were found to have reference frame errors: likely a technical error on the part of the surgeon that occurred while using the robot, resulting in the reference frame unintentionally moving during surgery. Thus, reducing the frequency of navigation failures is a potential area for improvement to optimize robotic spine surgery. Based upon our cases, we do not recommend placing pedicle screws using only a spinous process clamp as a reference frame since it is not rigid and is prone to unintended movement. Additionally, we recommend placing the reference frame as close as possible to the levels of surgery. There were two patients who had complications from misplaced screws far from the posterior superior iliac spine where the pelvic reference frame was placed, with misplaced screws found at T10 and T11 for one patient, and at L1 and L2 for a second patient (Figures 1, 2). When using the Mazor robot, a pedicle screw based reference frame is preferred.

Limitations

There are several limitations to this study. First, this was a retrospective analysis and is thus limited by patients potentially lost to follow-up, though most of these complications were noted in the perioperative period. Secondly, intraoperative screw malposition and revision prior to exiting the operating room was not recorded. Third, the retrospective nature of the study limited the ability to analyze other variables, including bone preparation technique and general workflows of each surgeon and the frequency of their accuracy checks. However, the authors of the current study recommend frequent accuracy checks with a chicken foot prior to bone preparation for each screw, and the use of a high speed burr for those bone preparations. Lastly, this study is also limited by the absence of a non-robot comparative group.

Conclusion

Despite using the latest generation robot, we found complication rates to be 2.2% (6/268) and reoperation rates to fix misplaced pedicle screws to be 1.5% (4/268 patients), occurring amongst four different surgeons. This is a higher rate than most previous studies using earlier generation robot systems. As half the robot related complications resulted from reference frame errors, this is an opportunity for improvement.

ORCID iD

David L. Skaggs https://orcid.org/0000-0001-6137-5576