Abstract
BACKGROUND CONTEXT:
Robotic spine surgery, utilizing 3D imaging and robotic arms, has been shown to improve the accuracy of pedicle screw placement compared to conventional methods, although its superiority remains under debate. There are few studies evaluating the accuracy of 3D navigated versus robotic-guided screw placement across lumbar levels, addressing anatomical challenges to refine surgical strategies and patient safety.
PURPOSE:
This study aims to investigate the pedicle screw placement accuracy between 3D navigation and robotic arm-guided systems across distinct lumbar levels.
STUDY DESIGN:
A retrospective review of a prospectively collected registry.
PATIENT SAMPLE:
Patients undergoing fusion surgery with pedicle screw placement in the prone position, using either via 3D image navigation only or robotic arm guidance.
OUTCOME MEASURE:
Radiographical screw accuracy was assessed by the postoperative computed tomography (CT) according to the Gertzbein-Robbins classification, particularly focused on accuracy at different lumbar levels.
METHODS:
Accuracy of screw placement in the 3D navigation (Nav group) and robotic arm guidance (Robo group) was compared using Chi-squared test/Fisher’s exact test with effect size measured by Cramer’s V, both overall and at each specific lumbosacral spinal level.
RESULTS:
A total of 321 patients were included (Nav, 157; Robo, 189) and evaluated 1210 screws (Nav, 651; Robo 559). The Robo group demonstrated significantly higher overall accuracy (98.6 vs 93.9%; p<.001, V=0.25). This difference of no breach screw rate was signified the most at the L3 level (No breach screw: Robo 91.3 vs 57.8%, p<.001, V=0.35) followed by L4 (89.6 vs 64.7%, p<.001, V=0.28), and L5 (92.0 vs 74.5%, p<.001, V=0.22). However, screw accuracy at S1 was not significant between the groups (81.1 vs 72.0%, V=0.10).
CONCLUSION:
This study highlights the enhanced accuracy of robotic arm-guided systems compared to 3D navigation for pedicle screw placement in lumbar fusion surgeries, especially at the L3, L4, and L5 levels. However, at the S1 level, both systems exhibit similar effectiveness, underscoring the importance of understanding each system’s specific advantages for optimization of surgical complications.
Keywords: Mis placement, Minimally invasive spine surgery, Patient safety, Pedicle screw, Robotic screw placement, 3D Navigation, Robotic arm, Screw breach
Introduction
Robotic pedicle screw placement has emerged as a highly accurate technique, now surpassing conventional freehand and 2D fluoroscopy navigation methods. Previous systematic reviews have demonstrated its superiority, reduced bleeding, radiation exposure, and surgeon stress, attributed to the integration of advanced 3D imaging navigation [1–5].
The success of robotic technology can be primarily attributed to the integration of advanced 3-dimensional (3D) imaging navigation with the precision and stability of semi-automated robotic arms. Despite its advantages, the impact in screw accuracy of robotic arms remains a topic of debate. A meta-analysis by Mason et al showed that 3D fluoroscopic navigation without robotic arm in lumbosacral screws achieves higher pedicle insertion accuracy compared to 2D fluoroscopic imaging guidance for lumbosacral screw placement [6]. Shahi et al reported the accuracy of 3D fluoroscopic navigation was also comparable to that of robot-guided screw placement, although results of this analysis limited due to the relatively small cohort [7].
With the continuing emergence of robotic spine surgery, there remains a significant gap regarding the added value of robotic arms in conjunction with 3D imaging navigation for lumbar pedicle screw placement, especially when considering the anatomical diversity across various lumbar levels. Factors such as the narrow vertebral pedicle in upper lumbar levels, the variability in its shape, and screw entry angle across levels, necessitate a nuanced approach [8]. This study aims to bridge this gap by assessing pedicle screw placement accuracy in a comprehensive cohort, comparing the performance of 3D navigation and robotic arm-guided systems across distinct lumbar levels. Additionally, by studying impact of lumbar level-specific anatomy on robotic screw placement accuracy, this study will offer critical insights for surgeons, guiding level-specific surgical strategies and enhancing patient safety.
Methods and materials
Study design and participants selection
This retrospective study of prospectively collected multiple surgeon database at a single academic institution was conducted following approval by the Institutional Review Board. Inclusion criteria were patients who underwent lumbar fusion surgery in prone position with pedicle screw between April 2017 and May 2023 with postoperative computed tomography (CT) available for review. The assessment by CT was routinely conducted to assess the fusion at 6-month to 1-year postoperatively. Exclusion criteria were procedures not employing navigation systems, screw insertion in lateral decubitus position, surgeries for trauma or fracture, deformity surgeries, and revision surgeries requiring previous screw removal. All surgeries performed in the 3D imaging navigation group (Nav) utilized the Stryker SpineMask Tracker, SpineMap 3D Software, and NAV3i Platform (Stryker Corp.), with image acquisition via the Ziehm Vision RFD 3D fluoroscopy spin (Ziehm Imaging, Inc.). The robotic arm group (Robo) underwent procedures with the ExcelsiusGPS (Globus Medical, Inc.), a floor-mounted system, utilizing preoperative CT scans or intraoperative fluoroscopy spins, with the specific imaging system —either Ziehm Vision RFD 3D fluoroscopy spin (Ziehm Imaging, Inc.) or Excelsius3D (Globus Medical, Inc.)— chosen based on the logistics of the operating room. In the Nav group, intraoperative image acquisition was conducted via fluoroscopy spin prior to skin incision. Conversely, in the Robo group, intraoperative image acquisition was performed following skin incision. This sequence is necessitated by the requirement to place the reference frame directly on the patient, which requires invasive procedure. For all robotic procedures, the reference frame was installed on the posterior superior iliac spine at the beginning of surgery.
Demographic data, spinal alignments, and surgical variables
Data was collected and managed using REDCap (Research Electronic Data Capture) hosted at Weill Cornell Medicine Clinical and Translational Science Center supported by the National Center for Advancing Translational Science of the National Institute of Health under award number: UL1 TR002384 [9,10].
Patients’ demographic data included age, sex, race, body mass index (BMI), smoking status, and American Society of Anesthesiologists (ASA) classification. Surgical data included operative time, estimated blood loss (EBL), type of surgery, number of fusion levels, and specific surgical levels operated on. The types of surgery included transforaminal lumbar interbody fusion (TLIF), anterior lumbar interbody fusion (ALIF), lateral lumbar interbody fusion (LLIF), and posterolateral lumbar fusion (PLF). Complications related to surgery, including reoperations and new onset of leg symptoms within the first year postoperatively were documented.
Assessment of screw placement accuracy
Postoperative evaluation of screw accuracy was performed using CT with 3D reconstruction images by a board-certified spine surgeon who was blinded to the group allocation. The Gertzbein-Robbins classification was employed to assess screw accuracy as previously described in the literature [7,11]. Each screw was graded on a scale from no cortical breach (grade A) to very severe breaches over 6 mm (grade E), incrementing by 2 mm. Endplate breaches and facet violations were recorded as binary outcomes. An overall accuracy grade was assigned to each screw using a simplified grading system (Table 1), with any breach categorizing the screw as ‘Acceptable’ or ‘Poor’ if not rated as grade A. Lateral breach less than 4 mm were classified as ‘Acceptable’, while those greater than 4 mm were classified as ‘Poor’. Inferior or medial breach less than 2 mm were considered ‘acceptable’, but that exceeding 2 mm were deemed ‘Poor’. Facet violation and endplate breach affecting au unfused level was marked as ‘Poor’. In evaluating the accuracy of screw placement, the worst grading observed for a screw determined its overall grade. For example, if the assessments for a screw included grade C tip breach (Acceptable), grade D medial breach (Poor), and no other breaches or violations (Good), the final grade assigned to that screw would be ‘Poor’.
Table 1.
Simplified classification for the screw breach
| Good | Acceptable | Poor | |
|---|---|---|---|
|
| |||
| Lateral breach | Grade A | Grade B or C | Grade D or E |
| Inferior or medial breach | Grade A | Grade B | Grade C, D or E |
| Tip breach | Grade A | Grade B-E | |
| Facet violation | No | Yes | Affecting the superior unfused level |
| Endplate breach | No | Yes | Into unfused level |
Each screw was graded on a scale from no cortical breach (grade A) to very severe breaches over 6 mm(grade E), incrementing by 2mm. End-plate breaches and facet violations were recorded as binary outcomes.
Statistical analysis
All continuous and categorical variables were reported as mean ± standard deviation and n (%), respectively. Comparative analyses for demographic data and screw accuracy between the 2 groups were conducted using Welch’s t-test for continuous variables and the chi-squared test for categorical variables. Fisher’s exact test was employed instead of the Chi-Square test for 2×2 cross tables containing a cell with a small sample size (less than 10). Cramer’s V was calculated for all categorical comparison utilizing chi-square value. Cramer’s V is a measure of statistical effect size, indicating the strength of difference between the groups, with V = 0.10 representing a small effect size, 0.30 indicating a medium size, and 0.50 signifying a large effect size. Statistical significance was set as p-value <.05. All statistical analyses were performed using R (ver. 4.3.1, R Core Team (2022), Vienna, Austria)
Results
In this study, a total of 321 patients underwent prone-position lumbar pedicle screw insertion with a cumulative 1210 screws. The cohort was divided into the Nav group with 144 patients (651 screws) and the Robo group with 177 patients (559 screws). Most patients, 226 (70.4%), underwent single-level fusion, while 71 (22.1%) had 2-level fusions. Primary surgery was 242 patients (75.4%), and 79 (24.6%) were fusion conversion surgery following previous decompression surgery, with no previous screw insertion at the index level. Overall, TLIF was the most common procedure in this cohort (79.1%, 254 patients) followed by PLF and LLIF. The total percentage exceeds 100% because some patients underwent a combination of procedures. Significantly more patients in the Robo cohort were female (57% versus 44%, p=.028). They’re existed significant differences in procedures between Nav and Robo with PLF (27.8% vs 11.3%, p=.007) and LLIF (19.4% vs 8.5%, p<.001). The Robo group experienced significantly longer operative times (179.2±110.4 vs 133.4±73.3 minutes, p<.001) and greater EBL (85.6±93.2 vs 65.1±82.6 mL, p=.040) across the overall cohort (Table 2). In patients undergoing single-level TLIF, the Robo group also had longer operative times (131.6±66.7 vs 97.9±32.0 minutes, p<.001), although EBL loss was comparable (52.1±36.7 vs 46.2±39.9 mL, p=.286).
Table 2.
Demographic data
| Overall | Navigation | Robot | p-value | |||
|---|---|---|---|---|---|---|
|
| ||||||
| Number of patients | 321 | 144 | 177 | |||
| Age | 59.7 (12.3) | 58.9 (12.7) | 60.4 (12.0) | .301 | ||
| Female | 164 (51.2) | 63 (44.1) | 101 (57.1) | .028* | ||
| Race | .062 | |||||
| White | 270 (84.1) | 124 (86.1) | 146 (82.5) | |||
| Black | 18 (5.6) | 3 (2.1) | 15 (8.5) | |||
| Asian | 8 (2.5) | 3 (2.1) | 5 (2.8) | |||
| Others | 25 (7.8) | 14 (9.7) | 11 (6.2) | |||
| BMI | 28.4 (5.4) | 28.1 (5.2) | 28.7 (5.6) | .389 | ||
| ASA | ||||||
| ASA I | 25 (7.8) | 12 (8.3) | 13 (7.4) | .525 | ||
| ASA II | 274 (85.6) | 125 (86.8) | 149 (84.7) | |||
| ASA III | 21 (6.6) | 7 (4.9) | 14 (8.0) | |||
| Smoking | 14 (4.4) | 8 (5.6) | 6 (3.4) | .503 | ||
| Status | Primary | 242 (75.4) | 112 (77.8) | 130 (73.4) | .444 | |
| Revision | 79 (24.6) | 32 (22.2) | 47 (26.6) | |||
| Operative time (min) | 158.8 (98.2) | 133.4 (73.3) | 179.2 (110.4) | <.001* | ||
| EBL (ml) | 76.4 (89.0) | 65.1 (82.6) | 85.6 (93.2) | .040* | ||
| Length of hospital stay (hours) | 56.6 (46.9) | 53.7 (48.2) | 59.0 (45.7) | .315 | ||
| Number of fusion levels | .283 | |||||
| 1 | 226 (70.4) | 99 (68.8) | 127 (71.8) | |||
| 2 | 71 (22.1) | 32 (22.2) | 39 (22.0) | |||
| 3 | 18 (5.6) | 11 (7.6) | 7 (4.0) | |||
| 4 | 3 (0.9) | 0 (0.0) | 3 (1.7) | |||
| 5 | 3 (0.9) | 2 (1.4) | 1 (0.6) | |||
| Type of surgery | ||||||
| TLIF | 254 (79.1) | 110 (76.4) | 144 (81.4) | .342 | ||
| ALIF | 30 (9.3) | 8 (5.6) | 22 (12.4) | .056 | ||
| LLIF | 43 (13.4) | 28 (19.4) | 15 (8.5) | .007* | ||
| PLF | 60 (18.7) | 40 (27.8) | 20 (11.3) | <.001* | ||
| Surgical levels | ||||||
| L1 | 4 (1.2) | 2 (1.4) | 2 (1.1) | 1 | ||
| L2 | 17 (5.3) | 9 (6.2) | 8 (4.5) | .661 | ||
| L3 | 58 (18.1) | 33 (22.9) | 25 (14.1) | .059 | ||
| L4 | 222 (69.2) | 107 (74.3) | 115 (65.0) | .093 | ||
| L5 | 300 (93.5) | 130 (90.3) | 170 (96.0) | .064 | ||
| S1 | 156 (48.6) | 63 (43.8) | 93 (52.5) | .146 | ||
BMI, body mass index; ASA, American Society of Anesthesiologists class; EBL, estimated blood loss; TLIF, transforaminal lumbar interbody fusion; ALIF, anterior lumbar interbody fusion; PLF, posterolateral fusion; LLIF, lateral lumbar interbody fusion.
Revision, Primary screw insertion following previous decompression surgery.
p<.05.
Overall screw accuracy
The Robo group demonstrated an overall accuracy rated as “Good” or “Accept” in 98.6% of cases (good: 89.1%, acceptable: 9.5%). In contrast, the Nav group exhibited 93.9% (good: 69.0%, acceptable: 24.9%), indicating that there was significant difference between the groups (p<.001, V=0.25). Lateral breaches were the most frequent malposition and occurred more in the Nav group (p<.001, V=0.21). Tip breaches were also more prevalent in the Nav group (p=.015, V=0.10). There were no significant differences between groups regarding the inferior/medial breaches, endplate breaches, and facet violations (Table 3).
Table 3.
Overall accuracy
| Navigation | Robot | p-value | Cramer’s V | ||
|---|---|---|---|---|---|
|
| |||||
| n | 651 | 559 | |||
| Lateral breach | |||||
| Grade A | 534 (82.0) | 534 (95.5) | <.001* | 0.21 | |
| Grade B | 69 (10.6) | 16 (2.9) | |||
| Grade C | 24 (3.7) | 6 (1.1) | |||
| Grade D | 11 (1.7) | 3 (0.5) | |||
| Grade E | 13 (2.0) | 0 (0.0) | |||
| Inferior or medial breach | |||||
| Grade A | 612 (94.0) | 540 (96.6) | .063 | 0.07 | |
| Grade B | 33 (5.1) | 18 (3.2) | |||
| Grade C | 6 (0.9) | 1 (0.2) | |||
| Grade D | 0 (0.0) | 0 (0.0) | |||
| Grade E | 0 (0.0) | 0 (0.0) | |||
| Tip breach | |||||
| Grade A | 598 (91.9) | 535 (95.7) | .015* | 0.10 | |
| Grade B | 26 (4.0) | 11 (2.0) | |||
| Grade C | 11 (1.7) | 10 (1.8) | |||
| Grade D | 11 (1.7) | 3 (0.5) | |||
| Grade E | 5 (0.8) | 0 (0.0) | |||
| Endplate breach | Yes | 11 (1.7) | 8 (1.4) | .898 | <0.01 |
| Facet violation | 8 (1.2) | 9 (1.6) | .752 | <0.01 | |
| Overall | Good | 449 (69.0) | 498 (89.1) | <.001* | 0.245 |
| Acceptable | 162 (24.9) | 53 (9.5) | |||
| Poor | 40 (6.1) | 8 (1.4) | |||
Grade 0, No cortical breach; Grade 1, <2 mm breach; Grade 2, 2 to <4 mm breach; Grade 3, 4 to <6 mm breach; Grade 4, >6 mm breach.
p<.05.
Screw accuracy at each level
The accuracy of each lumbar level is shown in Table 4A (L1, L2, and L3) and 4B (L4, L5, and S1). Overall good placement of each screw was also summarized in the Figure. Poor screw placement accuracy, particularly at upper lumbar levels (L1-L3), was more frequent in the Nav group but revealed no significant differences in accuracy between groups at L1 and L2. At L3, the Robo group demonstrated significantly higher accuracy with fewer lateral breaches (p=.039, V=0.30), resulting in the significantly higher accuracy in the Robo group (p<.001, V=0.37).
Table 4.
Screw Accuracy in upper lumbar level (L1-L3)
| Navigation | Robot | p-value | V | Navigation | Robot | p-value | V | Navigation | Robot | p-value | V | |||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| L1 | L2 | L3 | ||||||||||||||
|
| ||||||||||||||||
| Lateral | Grade A | 1 (25.0) | 4 (100.0) | NaN | 13 (72.2) | 14 (87.5) | .294 | 0.38 | 41 (64.1) | 41 (89.1) | .039* | 0.30 | ||||
| Grade B | 2 (50.0) | 0 (0.0) | 3 (16.7) | 0 (0.0) | 11 (17.2) | 4 (8.7) | ||||||||||
| Grade C | 0 (0.0) | 0 (0.0) | 0 (0.0) | 1 (6.2) | 7 (10.9) | 1 (2.2) | ||||||||||
| Grade D | 1 (25.0) | 0 (0.0) | 1 (5.6) | 1 (6.2) | 3 (4.7) | 0 (0.0) | ||||||||||
| Grade E | 0 (0.0) | 0 (0.0) | 1 (5.6) | 0 (0.0) | 2 (3.1) | 0 (0.0) | ||||||||||
| Inferior/medial | Grade A | 4 (100.0) | 4 (100.0) | NaN | 18 (100.0) | 15 (93.8) | .952 | 0.01 | 59 (92.2) | 45 (97.8) | .398 | 0.08 | ||||
| Grade B | 0 (0.0) | 0 (0.0) | 0 (0.0) | 1 (6.2) | 5 (7.8) | 1 (2.2) | ||||||||||
| Grade C | 0 (0.0) | 0 (0.0) | 0 (0.0) | 0 (0.0) | 0 (0.0) | 0 (0.0) | ||||||||||
| Grade D | 0 (0.0) | 0 (0.0) | 0 (0.0) | 0 (0.0) | 0 (0.0) | 0 (0.0) | ||||||||||
| Grade E | 0 (0.0) | 0 (0.0) | 0 (0.0) | 0 (0.0) | 0 (0.0) | 0 (0.0) | ||||||||||
| Tip | Grade A | 4 (100.0) | 4 (100.0) | NaN | 18 (100.0) | 16 (100.0) | NaN | 64 (100.0) | 45 (97.8) | .867 | 0.02 | |||||
| Grade B | 0 (0.0) | 0 (0.0) | 0 (0.0) | 0 (0.0) | 0 (0.0) | 0 (0.0) | ||||||||||
| Grade C | 0 (0.0) | 0 (0.0) | 0 (0.0) | 0 (0.0) | 0 (0.0) | 1 (2.2) | ||||||||||
| Grade D | 0 (0.0) | 0 (0.0) | 0 (0.0) | 0 (0.0) | 0 (0.0) | 0 (0.0) | ||||||||||
| Grade E | 0 (0.0) | 0 (0.0) | 0 (0.0) | 0 (0.0) | 0 (0.0) | 0 (0.0) | ||||||||||
| Endplate | 0 (0.0) | 0 (0.0) | NaN | 0 (0.0) | 0 (0.0) | NaN | 0 (0.0) | 1 (2.2) | .868 | 0.02 | ||||||
| Facet violation | 0 (0.0) | 0 (0.0) | NaN | 0 (0.0) | 0 (0.0) | NaN | 0 (0.0) | 1 (2.2) | .868 | 0.02 | ||||||
| Overall | Good | 1 (25.0) | 4 (100.0) | .091 | 0.77 | 13 (72.2) | 16 (100.0) | .074 | 0.39 | 37 (57.8) | 42 (91.3) | <.001* | 0.37 | |||
| Acceptable | 2 (50.0) | 0 (0.0) | 3 (16.7) | 0 (0.0) | 23 (35.9) | 4(8.7) | ||||||||||
| Poor | 1 (25.0) | 0 (0.0) | 2 (11.1) | 0 (0.0) | 4 (6.2) | 0 (0.0) | ||||||||||
Each screw was graded on a scale from no cortical breach (grade A) to very severe breaches over 6 mm (grade E), incrementing by 2 mm. Endplate breaches and facet violations were recorded as binary outcomes.
NaN, Neither of Chi-square and fisher-exact test could not be performed.
V, Cramer’sV: A measure of statistical effect size, indicating the strength of difference between the groups, with V = 0.10 representing a small effect size, 0.30 indicating a medium size, and 0.50 signifying a large effect size.
p<.05.
Figure.
3D navigation vs robot: level-specific comparison of good/not good.
V, Cramer’s V: A measure of statistical effect size, indicating the strength of difference between the groups, with V = 0.10 representing a small effect size, 0.30 indicating a medium size, and 0.50 signifying a large effect size.
For at L4 and L5, the Nav group exhibited a higher incidence of lateral breaches, with the Robo group showing significantly greater accuracy (L4: p<.001, V=0.27; L5: p=.006, V=0.16). The rates of medial or inferior breaching were similar across all levels. At L5, tip breaching was significantly more frequent in the Nav group (No breaching: Nav, 12.3% vs Robo, 1.4%; p<.001, V=0.21). Overall, the Robo group achieved “good” screw placement more frequently than the Nav group at L4 (p<.001, V=0.28) and L5 (p<.001, V=0.22) while the accuracy at S1 was comparable between the groups (p=.127, V=0.10) (Figure). Tip breach rate at S1 were similar between groups (Good: Nav, 81.4% vs Robo 84.3%; p=.218, V=0.15) (Table 5).
Table 5.
Screw accuracy in lower lumbar level (L4-S1)
| Navigation | Robot | p-value | V | Navigation | Robot | p-value | V | Navigation | Robot | p-value | V | |||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| L4 | L5 | S1 | ||||||||||||||
|
| ||||||||||||||||
| Lateral | Grade A | 142 (69.6) | 140 (90.9) | <.001* | 0.27 | 222 (91.4) | 209 (98.6) | .006* | 0.16 | 115 (97.5) | 126 (99.2) | .337 | 0.09 | |||
| Grade B | 36 (17.6) | 10 (6.5) | 15 (6.2) | 2 (0.9) | 2 (1.7) | 0 (0.0) | ||||||||||
| Grade C | 13 (6.4) | 2 (1.3) | 3 (1.2) | 1 (0.5) | 1 (0.8) | 1 (0.8) | ||||||||||
| Grade D | 3 (1.5) | 2 (1.3) | 3 (1.2) | 0 (0.0) | 0 (0.0) | 0 (0.0) | ||||||||||
| Grade E | 10 (4.9) | 0 (0.0) | 0 (0.0) | 0 (0.0) | 0 (0.0) | 0 (0.0) | ||||||||||
| Inferior/medial | Grade A | 193 (94.6) | 148 (96.1) | .452 | 0.07 | 226 (93.0) | 202 (95.3) | .413 | 0.06 | 112 (94.9) | 126 (99.2) | .058 | 0.10 | |||
| Grade B | 9 (4.4) | 6 (3.9) | 13 (5.3) | 9 (4.2) | 6 (5.1) | 1 (0.8) | ||||||||||
| Grade C | 2 (1.0) | 0 (0.0) | 4 (1.6) | 1 (0.5) | 0 (0.0) | 0 (0.0) | ||||||||||
| Grade D | 0 (0.0) | 0 (0.0) | 0 (0.0) | 0 (0.0) | 0 (0.0) | 0 (0.0) | ||||||||||
| Grade E | 0 (0.0) | 0 (0.0) | 0 (0.0) | 0 (0.0) | 0 (0.0) | 0 (0.0) | ||||||||||
| Tip | Grade A | 203 (99.5) | 154 (100.0) | 1 | 213 (87.7) | 209 (98.6) | <.001* | 0.21 | 96 (81.4) | 107 (84.3) | .218 | 0.15 | ||||
| Grade B | 0 (0.0) | 0 (0.0) | 19 (7.8) | 3 (1.4) | 7 (5.9) | 8 (6.3) | ||||||||||
| Grade C | 1 (0.5) | 0 (0.0) | 5 (2.1) | 0 (0.0) | 5 (4.2) | 9 (7.1) | ||||||||||
| Grade D | 0 (0.0) | 0 (0.0) | 3 (1.2) | 0 (0.0) | 8 (6.8) | 3 (2.4) | ||||||||||
| Grade E | 0 (0.0) | 0 (0.0) | 3 (1.2) | 0 (0.0) | 2 (1.7) | 0 (0.0) | ||||||||||
| Endplate | 4 (2.0) | 4 (2.6) | .73 | <0.01 | 5 (2.1) | 2 (0.9) | .457 | 0.03 | 2 (1.7) | 1 (0.8) | .610 | <0.01 | ||||
| Facet violation | 1 (0.5) | 1 (0.6) | 1 | 2 (0.8) | 5 (2.4) | .275 | 0.03 | 5 (4.2) | 2 (1.6) | .267 | 0.05 | |||||
| Overall | Good | 132 (64.7) | 138 (89.6) | <.001* | 0.29 | 181 (74.5) | 195 (92.0) | <.001* | 0.23 | 85 (72.0) | 103 (81.1) | .143 | 0.13 | |||
| Acceptable | 54 (26.5) | 12 (7.8) | 51 (21.0) | 14 (6.6) | 29 (24.6) | 23 (18.1) | ||||||||||
| Poor | 18 (8.8) | 4 (2.6) | 11 (4.5) | 3 (1.4) | 4 (3.4) | 1 (0.8) | ||||||||||
Each screw was graded on a scale from no cortical breach (grade A) to very severe breaches over 6 mm (grade E), incrementing by 2mm. Endplate breaches and facet violations were recorded as binary outcomes.
NaN, Neither of Chi-square and fisher-exact test could not be performed.
V, Cramer’s V: A measure of statistical effect size, indicating the strength of difference between the groups, with V=0.10 representing a small effect size, 0.30 indicating a medium size, and 0.50 signifying a large effect size.
p<.05.
Within the Robo group, a comparison between levels, excluding L1, revealed that “good” screw placement at S1 were less frequent compared to L5 (81.1% vs 92.0%, p=.004, V=0.18). Additionally, the tip breach rate was significantly frequent at S1 compared to L5 within the Robo group (S1, 15.7% vs L5, 1.2%, p<.001, V=0.26). However, the rate of “poor” placement was similar between S1 and L5 (0.8% vs 1.4%, p=.604).
Complications
Five patients (3.5%) in the Nav group complained new onset symptoms in lower extremities, and 2 underwent reoperation. Eight patients (5.5%) in the Robo group reported new onset leg symptoms, and 1 underwent reoperation. However, none of the surgery was not associated with screw malposition. In the Robo group, a registration error necessitated the abandonment of screw insertion in 1 instance.
The incidence of reoperations was significantly higher in the Nav group, with 6 cases (4.1%), compared to a single case (0.6%) in the Robo group (p=.048). Specifically, within the Nav group, reoperations were performed by surgical site infections in 2 patients 2 weeks postsurgery, cage migration in 2 patients, and new onset leg pain in 2 patients occurring between 2 to 5 months postoperatively. Conversely, in the Robo group, 1 patient underwent additional decompression for recurring symptoms 9 months postsurgery, while the remaining 7 cases of new onset leg symptoms showed improvement under close monitoring.
Discussion
This study demonstrates enhanced accuracy in robotic arm-guided screw placement compared to 3D image navigated screw placement. Our findings reveal an 89.1% rate of perfect placement and 98.6% rate of clinically acceptable placement for robotic screws, compared to 69.0% and 93.9%, respectively for 3D image navigation. The differences were particularly notable in the lateral and tip breach. A detailed analysis of the lumbar levels revealed the most remarkable disparities at L3, L4, and L5, with no significant difference at S1. To our knowledge, this represents 1 of the largest single-center studies comparing the accuracy of pedicle screw placement between these 2 methods.
Previous systematic reviews have established the superiority of robotic screws over conventional freehand [12,13], and computer-assisted navigation [14]. While these meta-analyses mitigate the limitations inherent to small cohorts, they also highlight the heterogeneity of navigation platforms and robotic systems, complicating the direct comparison of robotic and navigation systems. Each system has its advantages and drawbacks, making the interpretation of robotic guide’s importance challenging. Nevertheless, studies comparing specific systems, such as TINAVI (TINAVI Medical Technologies Co., Ltd., China) to O-arm based on same 3D navigation [15], and a network meta-analysis [1], have demonstrated superiority of the robotic navigation. However, the literature lacks large-scale studies focusing solely on the ExcelsiusGPS compared to 3D navigation system [16]. Our findings align with those of a previously published case series, showing 98.7% accuracy for clinically acceptable placements with this system, which is similar to the 98.6% accuracy as we found in the present study [17].
Optimal placement without breach, grade as “good” in this study, is crucial for maximizing the pedicle screw fixation stability. The selection of Screw diameter and length significantly influences the posterior screw construct stability [18,19]. Robotic arm guidance has been demonstrated to improve pedicle screw selection by enabling the most efficient screw trajectory [7,20]. While lateral breaches are generally considered benign [21], aiming for centralized screw placement is crucial for optimal should be the most efficient way to achieve this condition for the optimal screw placement. Future studies correlation poor screw insertion with patient outcomes are necessary to further elucidate this importance.
To our knowledge, there are no previous studies investigating the level-specific accuracy analysis. Notably, our analysis highlights consistent accuracy in robotic screw placement at L3 compared to 3D navigation. The pedicle diameter of upper lumbar spine is narrower than lower spine [22]. That at L3 was the narrowest in the sagittal plane and that at L2 is the narrowest in the transvers plane [8]. Furthermore, the distance from the reference frame on the iliac crest may influence the risk of robot-guided screw malposition since the further distance from reference frame can cause reference error [23]. Here we showed that that lumbar level accuracy can remain consistently high with robotic screws regardless of specific lumbar levels operated on in patients undergoing degenerative lumbar surgery, especially at the upper lumbar levels. While this analysis did not find statistically significant differences between the Nav and Robo groups at L1 and L2, there exists potential for a type II error due to the small number of L1 and L2 screws included.
The screws at S1 did not show significant differences between systems, potentially due to the low incidence of lateral breaches in the Nav group and higher rate of tip breaches in the Robo group compared to other levels. This result is not surprising, considering the anatomical variation, particularly the broader safe zone for lateral breaches at S1. Unlike screws used in the lumbar spine, the optimal S1 screw placement strategy differs; it is recommended to target the sacral promontory or pursue bi-cortical screw placement to maximize pull-out strength, a practice our facility consistently aims to implement. Tip breaches, while generally safe if the vascular safe zone is clearly visible in imaging navigation, occur at a higher rate at S1 compared to other levels, which may indicate a distinct objective for screw placement at this site. Regarding screw insertion conducted under a concept distinct from that of conventional pedicle screws, future research needs to include a larger scale study than previous literature, focusing on the accuracy from the planning on the system [24].
The operational efficiency is another crucial point in selecting the optimal navigation system for each institution, particularly once high accuracy is achieved. Our study found longer operative times in robotic surgeries, yet there was no significant difference in LOS. Previous studies have presented mixed results concerning operative time. A meta-analysis comparing computer-assisted navigation to robotic techniques found no significant difference in operative times [1], [14]. However, accurately determining the time spent on screw placement is challenging since operative time can be influenced by other factors, such as decompression procedures. Furthermore, our approach to imaging capture and fluoroscopy spin differs, adding another layer of complexity. A multicenter study comparing using the standardized system (Mazor) with and without robotic arm showed similar durations [25]. To better understand these discrepancies, acquiring stratified time data focused specifically on navigation setting and screw placement could clarify the differences in efficacy.
Regarding EBL, this can also be influenced by factors other than the surgical technique. While a subanalysis of patients undergoing single-level TLIF showed no difference in EBL, aligning with prior studies [1]. A propensity-matched analysis highlighted the superiority of robotic arm in reducing blood fusion rates, suggesting that surgeries without robotic assistance might involve more extensive soft tissue dissection [25]. Network meta-analysis revealed 3D navigation surpass the fluoroscopic technique in blood loss, though robotic technique was comparable to 3D navigation. Our results may be a result from the mixed fusion procedure and surgical levels.
This study identified a significantly lower rate of reoperation, though it is important to note that most reoperation were not related to screw placement. Previous literature did not support the superiority of the robotic arm in terms of reducing reoperations [2,26]. The incorporation of a robotic arm has raised a concern about it potentially being and additional source of infection, but economic analysis demonstrated that robotic surgery would be beneficial by reducing potential reoperation for infection[27]. The reported infection rate in robotic spine surgery is 0.8% [26,28], whereas reoperations for infection occurred in 2 cases with navigation (1.4%) in our cohort. Large institutional data showed aligned results indicating no significant differences in overall reoperation rates, indicating that the number of reoperations for surgical site infection in free hand group (5/694, 0.7%) and in robot groups (3/347, 0.8%)[2]. Despite the comparable rates of infection, suggesting that the introduction of a robotic arm does not increase the risk of infection, the overall rarity of complications highlights the need for further research. Large-scale cohort studies could provide a more detailed examination of potential complications, enhancing our understanding of the safety profile associated with robotic technique.
There are several limitations to be addressed in this study. Firstly, its retrospective design precludes randomization in the allocation to the Nav and the Robo group. This introduces potential selection bias, as surgeons may choose a particular technology based on the complexity of the case, with robotic technique often selected for more high-technical demand or time-consuming cases. Additionally, anesthesia implication and imaging system used in each case has not been investigated in this study, which potentially have influence on the results. Although the radiographic observers for the present study were blinded to the group assignments to mitigate another source of bias, the retrospective nature limits our ability to draw definitive conclusions about other outcomes such as operative time and blood loss. Secondly, given the single center nature of the study there are potentially generalizable limitations as the patients operated at our center are likely not equivalent to those operated on across the country. Third, the data exhibited a limited number of screw placements in the upper lumbar spine. Fourth, this study was not able to account the effect from learning curve [29]. Learning curve has effect on operative time, complications, and radiation exposure in robotic spine surgeries [30,31]. However, the accuracy of the procedures is less affected by operator experience, as the system itself can compensate for inexperience [28,32]. Therefore, we did not exclude early cases from our analysis. Lastly, our study population was exclusively composed of patients undergoing surgery for degenerative spine pathology. The applicability of robotic arm technology in the context of deformity surgery remains underexplored in the current body of literature, presenting a gap that future studies should aim to address.
In conclusion, this study has demonstrated the superior accuracy of robotic arm-guided screw placement compared to 3D imaging navigated screw placement in spinal surgeries. Notably, this higher accuracy was most significant at L3, L4, and L5, with no marked difference observed at S1. These differences elucidate the pitfalls in screw placements using either system, enhancing surgical safety in lumbar spinal instrumentation surgery. This investigation not only supports the integration of robotic technology for improved surgical precision but also highlights the need for careful consideration of anatomical and technological factors across different spinal levels. Future research should focus on expanding the scope of robotic guidance in spinal surgery, exploring its benefits and limitations across a broader spectrum of spinal pathologies and surgical techniques.
Source of funding
No direct funding was received for this study. However, this study used REDCap (Research Electronic Data Capture) hosted at Weill Cornell Medicine Clinical and Translational Science Center supported by the National Center For Advancing Translational Science of the National Institute of Health under award number: UL1 TR002384.
Footnotes
Declaration of competing interest
One or more of the authors declare financial or professional relationships on ICMJE-TSJ disclosure forms.
Declaration of Generative AI and AI-assisted technologies in the writing process
During the preparation of this work the author(s) used ChatGPT-4 in order to check the grammar and enhance readability. After using this tool/service, the author(s) reviewed and edited the content as needed and take(s) full responsibility for the content of the publication.
FDA device/drug status: Approved (Stryker SpineMask Tracker, SpineMap 3D Software, and NAV3i Platform (Stryker Corp.)); Approved (Ziehm Vision RFD 3D fluoroscopy spin (Ziehm Imaging, Inc.)), Approved (ExcelsiusGPS (Globus Medical, Inc)).
Author Disclosures: TA: Nothing to disclose. TS: Nothing to disclose. CZS: Nothing to disclose. NS: Nothing to disclose. TH: Nothing to disclose. KA: Nothing to disclose. AZL: Nothing to disclose. EM: Nothing to disclose. YEK: Nothing to disclose. OT: Nothing to disclose. MRJA: Nothing to disclose. EK: Nothing to disclose. MK: Nothing to disclose. JZ: Nothing to disclose. CK: Nothing to disclose. JD: Nothing to disclose. SI: Stock Ownership: HS2, LLC(1.7%); Consulting: Stryker(B); Speaking and/or Teaching Arrangements: Globus(B); Research Support (Investigator Salary, Staff/Materials): Innovasis(C). SAQ: Royalties: Stryker (None), Globus Medical(B), Stock Ownership: HS2, LLC(2.4%), Tissue Differenciation Intelligence (D); Consulting: Stryker (C), Globus Medical (D), SpineGuard (Amount not disclosed); Consulting: Surgaliagn (Amount not disclosed), Viseon, Inc (Amount not disclosed), Scientific Advisory Board/Other Office: AMOpportunities (Amount not disclosed), Lifelink.com (Amount not disclosed), Spinal Simplicity (Amount not disclosed).
CRediT authorship contribution statement
Tomoyuki Asada: Writing – review & editing, Writing – original draft, Visualization, Software, Project administration, Methodology, Investigation, Formal analysis, Data curation, Conceptualization. Tejas Subramanian: Writing – review & editing, Writing – original draft, Data curation. Chad Z. Simon: Investigation, Data curation. Nishtha Singh: Resources, Data curation. Takashi Hirase: Writing – review & editing, Supervision. Kasra Araghi: Software, Data curation. Amy Z. Lu: Data curation. Eric Mai: Resources, Data curation. Yeo Eun Kim: Data curation. Olivia Tuma: Data curation. Myles R J Allen: Data curation. Eric Kim: Data curation. Maximilian Korsun: Data curation. Joshua Zhang: Data curation. Cole Kwas: Data curation. James Dowdell: Writing – review & editing, Supervision. Sravisht Iyer: Writing – review & editing, Supervision. Sheeraz A. Qureshi: Supervision, Project administration, Conceptualization.
References
- [1].Naik A, Smith AD, Shaffer A, et al. Evaluating robotic pedicle screw placement against conventional modalities: a systematic review and network meta-analysis. Neurosurg Focus 2022;52:E10. 10.3171/2021.10.FOCUS21509. [DOI] [PubMed] [Google Scholar]
- [2].Asada T, Simon CZ, Lu AZ, et al. Robot-navigated pedicle screw insertion can reduce intraoperative blood loss and length of hospital stay: analysis of 1,633 patients utilizing propensity score matching. Spine J 2024;24:118–24. 10.1016/j.spinee.2023.09.004. [DOI] [PubMed] [Google Scholar]
- [3].Lopez IB, Benzakour A, Mavrogenis A, Benzakour T, Ahmad A, Lemee JM. Robotics in spine surgery: systematic review of literature. Int Orthop 2023;47:447–56. 10.1007/s00264-022-05508-9. [DOI] [PubMed] [Google Scholar]
- [4].Zhang ZS, Wu Y, Zheng B. A review of cognitive support systems in the operating room. Surg Innov 2024;31:111–22. 10.1177/15533506231218962. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [5].Shahi P, Vaishnav A, Araghi K, et al. Robotics reduces radiation exposure in minimally invasive lumbar fusion compared with navigation. Spine (Phila Pa 1976) 2022;47:1279–86. 10.1097/BRS.0000000000004381. [DOI] [PubMed] [Google Scholar]
- [6].Mason A, Paulsen R, Babuska JM, et al. The accuracy of pedicle screw placement using intraoperative image guidance systems. J Neurosurg Spine 2014;20:196–203. 10.3171/2013.11.SPINE13413. [DOI] [PubMed] [Google Scholar]
- [7].Shafi KA, Pompeu YA, Vaishnav AS, et al. Does robot-assisted navigation influence pedicle screw selection and accuracy in minimally invasive spine surgery? Neurosurg Focus 2022;52:E4. 10.3171/2021.10.focus21526. [DOI] [PubMed] [Google Scholar]
- [8].Priya A, Narayan RK, Ghosh SK, Sarangi PK. Analysing lumbar pedicle morphometry observed via traditional and recent modalities. J Orthop 2023;43:17–24. 10.1016/j.jor.2023.07.022. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [9].Harris PA, Taylor R, Minor BL, et al. The REDCap consortium: Building an international community of software platform partners. J Biomed Inform 2019;95:103208. 10.1016/j.jbi.2019.103208. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [10].Harris PA, Taylor R, Thielke R, Payne J, Gonzalez N, Conde JG. Research electronic data capture (REDCap)–a metadata-driven methodology and workflow process for providing translational research informatics support. J Biomed Inform 2009;42:377–81. 10.1016/j.jbi.2008.08.010. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [11].Schatlo B, Molliqaj G, Cuvinciuc V, Kotowski M, Schaller K, Tessitore E. Safety and accuracy of robot-assisted versus fluoroscopy-guided pedicle screw insertion for degenerative diseases of the lumbar spine: a matched cohort comparison. J Neurosurg Spine 2014;20:636–43. 10.3171/2014.3.SPINE13714. [DOI] [PubMed] [Google Scholar]
- [12].Matur AV, Palmisciano P, Duah HO, Chilakapati SS, Cheng JS, Adogwa O. Robotic and navigated pedicle screws are safer and more accurate than fluoroscopic freehand screws: a systematic review and meta-analysis. Spine J 2023;23:197–208. 10.1016/j.spinee.2022.10.006. [DOI] [PubMed] [Google Scholar]
- [13].Sun WX, Huang WQ, Li HY, et al. Clinical efficacy of robotic spine surgery: an updated systematic review of 20 randomized controlled trials. EFORT Open Rev 2023;8:841–53. 10.1530/EOR-23-0125. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [14].Zhou LP, Zhang RJ, Sun YW, Zhang L, Shen CL. Accuracy of pedicle screw placement and four other clinical outcomes of robotic guidance technique versus computer-assisted navigation in thoracolumbar surgery: a meta-analysis. World Neurosurg 2021;146:e139–50. 10.1016/j.wneu.2020.10.055. [DOI] [PubMed] [Google Scholar]
- [15].Du J, Gao L, Huang D, et al. Radiological and clinical differences between Tinavi orthopedic robot and O-arm navigation system in thoracolumbar screw implantation for reconstruction of spinal stability. Med Sci Monit 2020;26:e924770. 10.12659/MSM.924770. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [16].Elswick CM, Strong MJ, Joseph JR, Saadeh Y, Oppenlander M, Park P. Robotic-assisted spinal surgery: current generation instrumentation and new applications. Neurosurg Clin N Am 2020;31:103–10. 10.1016/j.nec.2019.08.012. [DOI] [PubMed] [Google Scholar]
- [17].Vardiman AB, Wallace DJ, Booher GA, et al. Does the accuracy of pedicle screw placement differ between the attending surgeon and resident in navigated robotic-assisted minimally invasive spine surgery? J Robot Surg 2020;14:567–72. 10.1007/s11701-019-01019-9. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [18].Brantley AG, Mayfield JK, Koeneman JB, Clark KR. The effects of pedicle screw fit. An in vitro study. Spine (Phila Pa 1976) 1994;19:1752–8. 10.1097/00007632-199408000-00016. [DOI] [PubMed] [Google Scholar]
- [19].Viezens L, Sellenschloh K, Puschel K, et al. Impact of screw diameter on pedicle screw fatigue strength-a biomechanical evaluation. World Neurosurg 2021;152:e369–76. 10.1016/j.wneu.2021.05.108. [DOI] [PubMed] [Google Scholar]
- [20].Shahi P, Maayan O, Shinn D, et al. Floor-mounted robotic pedicle screw placement in lumbar spine surgery: an analysis of 1,050 screws. Neurospine 2023;20:577–86. 10.14245/ns.2346070.035. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [21].LeRoy TE, Smith IC, Kim DH, et al. Clinical significance of lateral pedicle screw malposition in lumbar spine fusion. Clin Spine Surg 2023;36:E258–62. 10.1097/BSD.0000000000001440. [DOI] [PubMed] [Google Scholar]
- [22].Lien SB, Liou NH, Wu SS. Analysis of anatomic morphometry of the pedicles and the safe zone for through-pedicle procedures in the thoracic and lumbar spine. Eur Spine J 2007;16:1215–22. 10.1007/s00586-006-0245-2. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [23].Zhang Q, Fan MX, Han XG, et al. Risk factors of unsatisfactory robot-assisted pedicle screw placement: a case-control study. Neurospine 2021;18:839–44. 10.14245/ns.2142560.180. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [24].Kisinde S, Hu X, Hesselbacher S, Satin AM, Lieberman IH. Robotic-guided placement of cervical pedicle screws: feasibility and accuracy. Eur Spine J 2022;31:693–701. 10.1007/s00586-022-07110-4. [DOI] [PubMed] [Google Scholar]
- [25].Lee NJ, Zuckerman SL, Buchanan IA, et al. Is there a difference between navigated and non-navigated robot cohorts in robot-assisted spine surgery? A multicenter, propensity-matched analysis of 2,800 screws and 372 patients. Spine J 2021;21:1504–12. 10.1016/j.spinee.2021.05.015. [DOI] [PubMed] [Google Scholar]
- [26].Yu CC, Carreon LY, Glassman SD, et al. Propensity-matched comparison of 90-day complications in robotic-assisted versus non-robotic assisted lumbar fusion. Spine (Phila Pa 1976) 2022;47:195–200. 10.1097/BRS.0000000000004288. [DOI] [PubMed] [Google Scholar]
- [27].Menger RP, Savardekar AR, Farokhi F, Sin A. A cost-effectiveness analysis of the integration of robotic spine technology in spine surgery. Neurospine 2018;15:216–24. 10.14245/ns.1836082.041. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [28].Akazawa T, Torii Y, Ueno J, et al. Safety of robotic-assisted screw placement for spine surgery: experience from the initial 125 cases. J Orthop Sci 2023. 10.1016/j.jos.2023.06.003. [DOI] [PubMed] [Google Scholar]
- [29].Alluri RK, Avrumova F, Sivaganesan A, Vaishnav AS, Lebl DR, Qureshi SA. Overview of robotic technology in spine surgery. HSS J 2021;17:308–16. 10.1177/15563316211026647. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [30].Shahi P, Subramanian T, Maayan O, et al. Surgeon experience influences robotics learning curve for minimally invasive lumbar fusion: a cumulative sum analysis. Spine (Phila Pa 1976) 2023. 10.1097/BRS.0000000000004745. [DOI] [PubMed] [Google Scholar]
- [31].Avrumova F, Morse KW, Heath M, Widmann RF, Lebl DR. Evaluation of K-wireless robotic and navigation assisted pedicle screw placement in adult degenerative spinal surgery: learning curve and technical notes. J Spine Surg 2021;7:141–54. 10.21037/jss-20-687. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [32].Shahi P, Subramanian T, Singh S, et al. Perception of robotics and navigation by spine fellows and early attendings: the impact of these technologies on their training and practice. World Neurosurg 2024;181:e330–8. 10.1016/j.wneu.2023.10.049. [DOI] [PubMed] [Google Scholar]