The Journal of Arthroplasty 37 (2022) 795e801

Contents lists available at ScienceDirect

The Journal of Arthroplasty

journal homepage: www.arthroplastyjournal.org

Basic Science

Image-Free Robotic-Assisted Total Knee Arthroplasty Improves

Implant Alignment Accuracy: A Cadaveric Study
Gary W. Doan, MS a, R. Patrick Courtis, PhD b, Joseph G. Wyss, BS b,
Eric W. Green, MD c, Chadd W. Clary, PhD a, *
a
Center for Orthopaedic Biomechanics, University of Denver, Denver, CO
b
DePuy-Synthes, Raynham, MA
c
St. Cloud Orthopedic Associates, Sartell, MN

a r t i c l e i n f o a b s t r a c t

Article history: Background: Improving resection accuracy and eliminating outliers in total knee arthroplasty (TKA) is
Received 16 September 2021 important to improving patient outcomes regardless of alignment philosophy. Robotic-assisted surgical
Received in revised form systems improve resection accuracy and reproducibility compared to conventional instrumentation.
10 December 2021
Some systems require preoperative imaging while others rely on intraoperative anatomic landmarks. We
Accepted 28 December 2021
Available online 1 January 2022
hypothesized that the alignment accuracy of a novel image-free robotic-assisted surgical system would
be equivalent or better than conventional instrumentation with fewer outliers.
Methods: Forty cadaveric specimens were used in this study. Five orthopedic surgeons performed 8
Keywords:
total knee arthroplasty
bilateral TKAs each, using the VELYS Robotic-Assisted System (DePuy Synthes) and conventional
robotic surgery instrumentation on contralateral knees. Pre-resection and postresection computed tomography scans,
accuracy along with optical scans of the implant positions were performed to quantify resection accuracies
alignment relative to the alignment targets recorded intraoperatively.
conventional instrumentation Results: The robotic-assisted cohort demonstrated smaller resection errors compared to conventional
instrumentation in femoral coronal alignment (0.63
± 0.50
vs 1.39
± 0.95
, P < .001), femoral sagittal
alignment (1.21
± 0.90
vs 3.27
± 2.51
, P < .001), and tibial coronal alignment (0.93
± 0.72
vs 1.65
±
1.29
, P ¼ .001). All other resection angle accuracies were equivalent. Similar improvements were found
in the femoral implant coronal alignment (0.89
± 0.82
vs 1.42
± 1.15
, P ¼ .011), femoral implant
sagittal alignment (1.51
± 1.08
vs 2.49
± 2.10
, P ¼ .006), and tibial implant coronal alignment (1.31
±
0.84
vs 2.03
± 1.44
, P ¼ .004). The robotic-assisted cohort had fewer outliers (errors >3
) for all angular
resection alignments.
Conclusion: This in vitro study demonstrated that image-free robotic-assisted TKA can improve align-
ment accuracy compared to conventional instrumentation and reduce the incidence of outliers.
© 2022 The Authors. Published by Elsevier Inc. This is an open access article under the CC BY license
(http://creativecommons.org/licenses/by/4.0/).

As total knee arthroplasty (TKA) surgeons refine alignment phi-

losophies to improve patient satisfaction and reduce the risk of
Funding: This study was funded in part by DePuy Synthes Joint Reconstruction, a revision, accurate execution of the planned resections is crucial
Johnson & Johnson Company. The funding source provided recommendations on regardless of alignment philosophy. Some clinical evidence suggests,
the study design and assisted in data collection. The funding source did not but is not overwhelmingly compelling, that patient-specific align-
contribute to the analysis or interpretation of the data, writing of the report, or in
ment may improve outcomes [1e3]. However, there is strong evi-
the decision to submit the article for publication.
dence that established alignment bounds reduce the risk of revision
One or more of the authors of this paper have disclosed potential or pertinent [4,5]. Computer-aided surgery (CAS) systems and patient-specific
conflicts of interest, which may include receipt of payment, either direct or indirect, resection guides have been introduced to improve the accuracy of
institutional support, or association with an entity in the biomedical field which TKA resections [6e9]. More recently, multiple reports demonstrate
may be perceived to have potential conflict of interest with this work. For full
that robotic-assisted total knee arthroplasty (RATKA) systems
disclosure statements refer to https://doi.org/10.1016/j.arth.2021.12.035.
* Address correspondence to: Chadd W. Clary, PhD, University of Denver, 2155 E. improve resection accuracy, particularly in the coronal plane, when
Wesley Avenue, Room 279, Denver, CO, 80126. compared to conventional surgical instrumentation [10e14].

https://doi.org/10.1016/j.arth.2021.12.035
0883-5403/© 2022 The Authors. Published by Elsevier Inc. This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/4.0/).
796 G.W. Doan et al. / The Journal of Arthroplasty 37 (2022) 795e801

RATKA systems enable surgical resections by either constraining robotic system on the first knee and conventional instrumenta-
the saw to a specified anatomic plane during the resection tion (CTKA) on the contralateral knee. For the RATKA surgeries,
[10,12,15e17], positioning a resection guide relative to the bone optical tracking arrays were mounted to the bones and anatomic
similar to previous CAS systems [13,14,18,19], or positioning a landmarks were identified to construct anatomic coordinate sys-
burring tool during milling of the resection surface [10,20]. All tems. Surgeons planned their resection angles in real time using
these approaches have proven efficacious for improving resection the system’s software. The target resection alignments were
accuracy. Some systems require preoperative imaging to enable recorded, including the femoral sagittal angle (FSA), femoral cor-
surgical planning and execution [12,21], which increases cost, time, onal angle (FCA), femoral internal-external rotation angle (FRA),
and radiation exposure for the patient. Other systems are tibial sagittal angle (TSA), and tibial coronal angle (TCA), along
image-free, relying on intraoperative anatomic landmarking and with the thicknesses for the distal and posterior resections of the
real-time planning [18]. It is unclear if preoperative imaging is medial femoral condyle [femoral distal resection (FDR) and
necessary to achieve the desired level of accuracy. femoral posterior resection (FPR)] and tibial resection (TR). For the
The VELYS Robotic-Assisted Solution (DePuy Synthes, Warsaw, CTKA surgeries, alignment targets were recorded from manual
IN) for TKA was recently released. The system consists of an optical instrument settings where possible (FCA, FRA, FDR, FPR, TSA, TCA,
tracking system with bone-mounted arrays and a bed-mounted and TR). The femoral sagittal resection angle was influenced by
robotic arm that positions a surgical saw to perform bony re- the fit of the intermedullary rod, thus the anatomic sagittal angle
sections (Fig. 1). This system differs from some previous robotic of the femoral shaft was measured from the CT scans and used as
systems in that it does not require pre-operative imaging. The the target.
purpose of this study is to quantify the accuracy of the system After primary resections, bone remnant thicknesses were
compared to conventional instrumentation during simulated use. measured using a digital caliper. The thicknesses of femur re-
Both resection and implant positioning accuracy were evaluated. sections were measured from the resection plane to the most
Our study hypothesized that all resection and implant alignment prominent point on the articular surface. The tibia thickness was
metrics for the RATKA cohort would be equivalent to or better than measured to an electrocautery mark on the articular surface
conventional instrumentation with fewer outliers. marking the point used in planning. After surgery, CT scans were
repeated using the same imaging protocol. The proximal 300 mm of
Materials and Methods the tibia and the distal 300 mm of the femur were extracted from
the specimen and denuded of soft tissue. Optical scans of the
Data Collection extracted femur and tibia were performed using a scanner with a
reported accuracy of 0.05 mm (Space Spider; Artec 3D,
Forty cadaveric specimens were used in this study (age: 70.4 ± Luxembourg, Luxembourg). The extracted femurs were implanted
8.2 years, height: 67.1 ± 4.1 in., body mass index: 20.6 ± 4.9). Five with an ATTUNE cruciate-retaining cementless femur and the
board-certified orthopedic surgeons participated in the experi- extracted tibiae were implanted with an ATTUNE cemented tibial
ment. All surgeons were proficient with the ATTUNE INTUITION base. A second optical scan was performed to capture the orienta-
conventional instrumentation, experienced with either CAS or tion of the implant geometry relative to the extracted bone.
robotic-assisted surgical systems, and trained in the robotic
system. Resection Alignment Measurements
Preoperative computed tomography (CT) scans were per-
formed with a 0.6-mm interslice distance at the knee to enable the The intact bony anatomy was segmented from the preoperative
accuracy assessment but were not used in surgical planning. Each CT scans and anatomic coordinate systems were constructed. The
surgeon performed bilateral TKAs on 8 specimens using the superior-inferior (S-I) axis of the femoral was aligned to the femoral

Fig. 1. The VELYS Robotic-Assisted solution is an imageless system that consists of an optical tracking system with bone-mounted tibial and femoral arrays, a stylus to identify
anatomic landmarks, and a bed-mounted robotic arm that dynamically positions a surgical saw relative to the bony anatomy to perform the tibial and femoral bony resections.
 G.W. Doan et al. / The Journal of Arthroplasty 37 (2022) 795e801 797

Fig. 2. The relative orientation of the femoral anterior, distal, and posterior resections was calculated in both the (A) sagittal and (B) transverse planes using models generated from
optical scans. A-P, anterior-posterior.

mechanical axis and rotationally aligned to the posterior condylar Statistical Analysis
axis. Likewise, the S-I axis of the tibial coordinate system was
aligned to the tibial mechanical axis and rotationally aligned to the Alignment errors were calculated as the difference between the
medial third of the tibial tubercle. Geometries of the resected bones measured alignment metric and the corresponding target. Sum-
were segmented from the postoperative CT scans and solid models mary statistics, including the mean and standard deviation of the
of the resected and implanted bones were fused from the optical errors, the median errors, and the mean and standard deviation of
scans. These solid models were registered to the intact models the absolute errors were calculated for each alignment metric
using an iterative closest point algorithm in their respective across the RATKA and CTKA cohorts. Conditional 2-sample t-tests
anatomic coordinate systems [22]. were performed. The first t-test evaluated the null hypothesis that
Planes were fit to the distal femoral, posterior femoral, and the RATKA cohort had superior accuracy to the CTKA cohort based
proximal tibial resections from both the CT and optical scans. on absolute errors. If RATKA was not statistically superior, a sec-
Angles between the distal femoral resection and the S-I axis in ondary t-test was performed to determine if the RATKA cohort was
the sagittal and coronal planes were used to calculate the FSA noninferior to the CTKA cohort within a margin of 0.5
or 0.5 mm
(distal reference) and FCA, respectively. Another measurement of (P  .05) where appropriate. The proportion of specimen with ac-
the FSA was performed using the posterior femoral resection and curacy errors less than 1
or 1 mm, 2
or 2 mm, and 3
or 3 mm
the S-I axis (posterior reference). The angle between the poste- were calculated. Since accuracy of the primary resections was
rior femoral resection and the anterior-posterior (A-P) anatomic calculated using both CT and optical scans, mean absolute differ-
axis in the transverse plane was used to quantify the FRA ences and correlation coefficients were calculated between the
alignment. Like the femur, the angles between the proximal TR redundant measures. Finally, mean absolute differences and cor-
and the S-I axis in the sagittal and coronal plane were used to relation coefficients were calculated between the resection align-
calculate the TSA and TCA, respectively. To calculate the implant ments and final implant alignments.
alignment, equivalent calculations were performed using the
implant’s bone interface surfaces to calculate the femoral
implant sagittal, coronal, and internal-external rotation angles Results
(FISA, FICA, and FIRA) and the tibial implant sagittal and coronal
angles (TICA, TISA). During 2 of the 40 RATKA surgeries, the femoral array was
dislodged intraoperatively. The damaged array was replaced, and
no further issues were observed. The femoral alignment metrics for
these 2 specimens were excluded from the analysis. Two TR depth
Femoral Relative Resection Accuracy Measurements measurements were inadvertently not recorded during surgery,
one each from the RATKA and CTKA cohorts, and were likewise
The relative alignment of the anterior, distal, and posterior excluded from the analysis. Finally, optical scans for 1 femur
femoral resections was also characterized which affects fit of the implant and 1 tibial implant were corrupted, both from the CTKA
femoral component. A femoral implant coordinate system was cohort, and excluded. All other alignment measurements were
constructed relative to the femoral resections with the S-I axis successfully collected.
perpendicular to the distal femoral resection and rotationally For the RATKA cohort, absolute mean errors in resection align-
aligned to the anterior resection plane. Angles between the pos- ment ranged from 0.63
± 0.50
(FCA) to 1.71
± 1.31
(FSA, distal
terior resection and the S-I axis and between the anterior resection reference). For the CTKA cohort, absolute mean errors ranged from
plane and the S-I axis in the sagittal plane were calculated, and 1.00
± 0.70
(FRA) to 3.27
± 2.51
(FSA, posterior reference). The
between the anterior and posterior resections in the transverse RATKA cohort demonstrated statistical superiority to the CTKA
plane (Fig. 2). Likewise, the A-P distance between the anterior and cohort in accuracy of the FCA (P < .001); FSA, posterior reference
posterior resection surfaces was calculated 3 mm distal from the (P < .001); and TCA (P ¼ .001), and noninferiority for all other
most proximal point on the posterior resection. Equivalent target resection metrics (Table 1). RATKA had fewer outliers (errors >3
)
resection angles and A-P distances were extracted from the for all angular resection alignment measures (Fig. 3, Table 2). In the
appropriately sized femoral implant geometry. RATKA cohort, 100% of the specimen had FCA, FRA, and TCA
798 G.W. Doan et al. / The Journal of Arthroplasty 37 (2022) 795e801

Table 1
Summary Accuracy Measurements for Differences Between Target and Measured Resection and Implant Alignments of the RATKA and CTKA Cohorts.

Accuracy Metric CT Scans RATKA CTKA RATKA Accuracy P-Value

Compared to CTKA
Mean ± SD Median Absolute Mean ± SD Median Absolute
Mean ± SD Mean ± SD

Femur resection
Coronal angle (
) 0.01 ± 0.82 -0.01 0.63 ± 0.50 0.20 ± 1.69 0.25 1.39 ± 0.95 Superior <.001
Sagittal angle, distal reference (
) 1.23 ± 1.78 1.20 1.71 ± 1.31 0.47 ± 2.41 0.56 1.93 ± 1.50 Noninferior .014
Sagittal angle, posterior reference (
) 0.65 ± 1.37 0.71 1.21 ± 0.90 0.15 ± 4.15 0.19 3.27 ± 2.51 Superior <.001
I-E rotation angle (
) 0.38 ± 1.27 0.02 1.04 ± 0.81 0.04 ± 1.23 0.16 1.00 ± 0.70 Noninferior .004
Tibial resection
Coronal angle (
) 0.04 ± 1.18 0.11 0.93 ± 0.72 0.92 ± 1.90 0.97 1.65 ± 1.29 Superior .001
Sagittal angle (
) 0.69 ± 1.86 1.07 1.62 ± 1.13 0.59 ± 2.07 0.65 1.63 ± 1.39 Noninferior .036

Accuracy Metric Optical Scans RATKA CTKA RATKA Accuracy P-Value

Compared to CTKA
Mean ± SD Median Absolute Mean ± SD Median Absolute
Mean ± SD Mean ± SD

Femur resection
Coronal angle (
) 0.12 ± 0.75 0.14 0.59 ± 0.47 0.35 ± 1.70 0.48 1.39 ± 1.02 Superior <.001
Sagittal angle, distal reference (
) 1.42 ± 0.99 1.58 1.50 ± 0.86 0.25 ± 2.22 0.36 1.70 ± 1.42 Noninferior .005
Sagittal angle, posterior reference (
) 0.89 ± 1.23 0.85 1.14 ± 1.00 0.02 ± 4.00 -0.18 2.93 ± 2.69 Superior <.001
I-E rotation angle (
) 0.02 ± 1.13 0.12 0.97 ± 0.57 0.29 ± 1.33 0.37 1.10 ± 0.79 Noninferior <.001
Posterior resection sagittal angle (
) 0.56 ± 1.24 0.46 0.95 ± 0.96 0.23 ± 2.93 0.15 2.08 ± 2.05 Superior .001
Anterior resection sagittal angle (
) 1.91 ± 1.05 1.90 1.91 ± 1.05 1.75 ± 1.51 1.81 1.92 ± 1.27 Noninferior .028
A-P resections transverse angle (
) 0.03 ± 1.11 0.10 0.85 ± 0.70 0.33 ± 0.86 0.19 0.71 ± 0.58 Noninferior .007
A-P resections A-P distance (mm) 0.16 ± 0.83 0.16 0.56 ± 0.63 0.17 ± 1.27 0.28 0.97 ± 0.82 Superior .006
Tibial resection
Coronal angle (
) 0.47 ± 1.47 0.89 1.28 ± 0.85 1.48 ± 1.91 1.48 1.97 ± 1.38 Superior .004
Sagittal angle (
) 0.14 ± 1.61 0.09 1.21 ± 1.05 0.04 ± 2.24 0.12 1.65 ± 1.49 Noninferior .001
Femur implant
Coronal angle (
) 0.27 ± 1.19 0.30 0.89 ± 0.82 0.62 ± 1.73 0.42 1.42 ± 1.15 Superior .011
Sagittal angle (
) 0.53 ± 1.79 0.36 1.51 ± 1.08 2.22 ± 2.39 1.75 2.49 ± 2.10 Superior .006
I-E rotation angle (
) 0.24 ± 1.36 0.35 1.11 ± 0.80 0.13 ± 1.33 0.01 0.98 ± 0.90 Noninferior .031
Tibial implant
Coronal angle (
) 0.47 ± 1.50 0.51 1.31 ± 0.84 1.54 ± 1.97 1.68 2.03 ± 1.44 Superior .004
Sagittal angle (
) 0.10 ± 1.77 0.37 1.37 ± 1.11 0.13 ± 2.25 0.36 1.65 ± 1.51 Noninferior .005

Accuracy Metric Calipers RATKA CTKA RATKA Accuracy P-Value

Compared to CTKA
Mean ± SD Median Absolute Mean ± SD Median Absolute
Mean ± SD Mean ± SD

Femur resection
Distal resection depth (mm) 0.01 ± 0.87 0.02 0.62 ± 0.60 0.56 ± 1.18 0.30 0.88 ± 0.96 Noninferior <.001
Posterior resection depth (mm) 0.08 ± 0.69 0.28 0.54 ± 0.43 0.69 ± 0.89 0.46 0.76 ± 0.83 Noninferior <.001
Tibial resection
Tibial resection depth (mm) 0.44 ± 0.86 0.30 0.67 ± 0.69 1.21 ± 1.79 1.30 1.66 ± 1.38 Superior <.001

Results of the statistical analysis are reported for each alignment metric with the associated P-values for the reported outcome.
CTKA, conventional total knee arthroplasty; RATKA, robotic-assisted total knee arthroplasty; A-P, anterior-posterior; I-E, internal-external; CT, computed tomography; SD,
standard deviation.

resections within 3
of the target alignments, compared to 92.5%, RATKA resulted in fewer outliers in the relative orientation of the
97.5%, and 80%, respectively, for the CTKA cohort. Strong agree- femoral resections (errors >3
or 3 mm), except for in the transvers
ments between the CT and optical-based measurements were angle between the anterior and posterior resections (Table 2).
observed for all metrics except the FSA (distal reference), with The accuracy of the femur implant coronal and sagittal align-
correlation coefficients ranging from 0.91 to 0.97, and absolute ment and the tibial implant coronal alignment in RATKA was su-
mean differences ranging from 0.31
± 0.27
to 0.91
± 0.95
perior to CTKA (P ¼ .011, P ¼ .006, and P ¼ .004, respectively). In both
(Table 3). A weaker correlation (0.80) and increased absolute mean cohorts, the femoral component was flexed relative to the
difference (0.94
± 0.73
) was observed for the FSA (distal refer- target alignment and the measured femoral sagittal resection
ence) between the 2 measurement modalities. angle. In RATKA, the femoral implant was flexed an average of 0.53

Femoral resection depth accuracy for RATKA was noninferior to ± 1.79
relative to the target, while in CTKA the femoral implant
CTKA with mean absolute errors less than 0.62 mm for RAKTA and was flexed 2.22
± 2.39
(Table 1). Strong correlations, ranging from
0.88 mm CTKA (Table 1). RATKA TR mean absolute error (0.67 ± 0.88 to 0.97, were observed between the resection alignment and
0.69 mm) was superior to CTKA (1.62 ± 1.39 mm, P < .001). Outliers resulting implant alignment for angles other than the femoral
were reduced for all resection depth measures in the RATKA cohort implant sagittal angle (Table 3).
(Table 2).
Accuracy of the posterior femoral resection relative to the distal Discussion
resection in the sagittal plane was superior in RATKA (P ¼ .001),
along with the A-P distance between the anterior and posterior The current study represents one of the largest and most
resections (P ¼ .006; Table 1). The sagittal orientation of the ante- comprehensive preclinical assessments of RATKA system accuracy
rior resection and the transverse angle between the anterior and to date. The VELYS Robotic-Assisted Solution evaluated in this study
posterior femoral resections in RATKA were noninferior to CTKA. demonstrated statistically superior accuracy for many resection
 G.W. Doan et al. / The Journal of Arthroplasty 37 (2022) 795e801 799

Fig. 3. Histograms of femoral and tibial angular resection errors for RATKA and CTKA cohorts. CTKA, conventional total knee arthroplasty; RATKA, robotic-assisted total knee
arthroplasty.

and implant alignments, and noninferiority for the remaining [14,16,18e20], which neglects potentially significant errors in the
measures when compared to CTKA. In particular, the RATKA system registration process [23]. Other studies use preoperative and
yielded superior accuracy for the coronal alignment of the femur postoperative CT scans to measure resection or implant alignment
and tibia resections and implant alignments. angles [11,17], which allow independent quantification of the sys-
With the improved accuracy and precision of modern RATKA tem accuracy and reduce uncertainty in establishing anatomic co-
surgical systems, detecting meaningful differences in resection ordinate systems. Limitations in the CT slice thickness induce errors
accuracy is challenging. Some previous studies have relied on the in segmentation of the resection surfaces which are commonly
robot’s optical tracking system to quantify resection accuracy parallel to the axial slices. This study used an interslice distance of
800 G.W. Doan et al. / The Journal of Arthroplasty 37 (2022) 795e801

Table 2
Summary of Alignment Outliers for the RATKA and CTKA Cohorts.

Accuracy Metric RATKA CTKA

Error From Target (% Specimen) Error From Target (% Specimen)

1
or 1 mm 2
or 2 mm 3
or 3 mm 1
or 1 mm 2
or 2 mm 3
or 3 mm

Femur resection
Coronal angle (
) 86.8% 97.4% 100.0% 42.5% 85.0% 92.5%
Sagittal angle, distal reference (
) 34.2% 63.2% 89.5% 35.0% 60.0% 85.0%
Sagittal angle, posterior reference (
) 44.7% 92.1% 97.4% 10.0% 35.0% 57.5%
I-E rotation angle (
) 52.6% 78.9% 100.0% 60.0% 90.0% 97.5%
Distal resection depth (mm) 78.9% 94.7% 100.0% 77.5% 90.0% 95.0%
Posterior resection depth (mm) 84.2% 97.4% 100.0% 75.0% 90.0% 95.0%
Posterior resection sagittal angle (
) 73.7% 92.1% 92.1% 35.0% 67.5% 82.5%
Anterior resection sagittal angle (
) 15.8% 55.3% 97.4% 27.5% 57.5% 80.0%
A-P resections transverse angle (
) 71.1% 97.4% 97.4% 75.0% 97.5% 100.0%
A-P resections A-P distance (mm) 86.8% 100.0% 100.0% 70.0% 87.5% 95.0%
Tibial resection
Coronal angle (
) 60.0% 90.0% 100.0% 37.5% 72.5% 80.0%
Sagittal angle (
) 35.0% 70.0% 85.0% 42.5% 70.0% 80.0%
Tibial resection depth (mm) 79.5% 94.9% 97.4% 41.0% 71.8% 92.3%
Femur implant
Coronal angle (
) 68.4% 89.5% 94.7% 46.2% 71.8% 87.2%
Sagittal angle (
) 42.1% 68.4% 92.1% 20.5% 56.4% 69.2%
I-E rotation angle (
) 52.6% 92.1% 94.7% 66.7% 84.6% 94.9%
Tibial implant
Coronal angle (
) 35.0% 85.0% 95.0% 28.2% 56.4% 82.1%
Sagittal angle (
) 42.5% 77.5% 90.0% 46.2% 61.5% 84.6%

CTKA, conventional total knee arthroplasty; RATKA, robotic-assisted total knee arthroplasty; A-P, anterior-posterior; I-E, internal-external.

0.6 mm, which is finer than most clinical imaging protocols. Like- system which positions a resection guide. Errors were measured
wise, implant-induced metal artifacts cause uncertainty in align- using the CAS component of the robotic system which excluded
ment measures [24,25]. A study by Sires and Wilson [17] found errors in the registration process. Even so, the reported errors were
absolute differences between resection angles measured intra- on par with the resection accuracy found in the current study,
operatively and the corresponding implant alignments measured ranging from 0.59
to 1.50
. Other studies have documented similar
using a CT scan ranged from 1.09
± 0.75
to 1.97
± 1.41
for tibial statistically significant improvements in coronal implant alignment
coronal and sagittal alignment, respectively, with 9% of the accu- when using RATKA [10e14].
racy measurements deviating by >3
. The small differences be- Our study leveraged redundant imaging modalities to improve
tween resection angles and implant angles observed in our study confidence in our findings. Strong correlations were observed be-
suggest that much of this error may be attributed to the anatomic tween the resection angles measured from CT scans and the optical
registration step in the RATKA workflow. scans, with the notable exception of FSA (distal resection). The
Comparisons to other robotic systems are difficult due to dif- distal femoral resection surface has a small A-P width and aligns
ferences in measurement methods and reported accuracy metrics. closely with the axial slice of the CT scan, which induced errors in
A multicenter, prospective, nonrandomized study compared the the segmentation and subsequent CT-based measurement. The
postoperative implant alignment between image-based RATKA and optical scan generated higher fidelity resection surfaces, but accu-
CTKA cohorts using CT scans [12]. The study found mean absolute racy metrics using this surface did not indicate significant differ-
errors of 1.0
(FICA), 1.7
(FIRA), 1.3
(TICA), and 1.4
(TISA) for the ences in FSA between RATKA and CTKA. RATKA superiority was
RATKA cohort. These errors were similar to those reported in this observed when comparing the FSA measured from the posterior
study using the image-free RATKA system (0.89
± 0.82
, 1.11
± resections from both the CT and optical scans. Future CT-based
0.80
, 1.31
± 0.84
, and 1.37
± 1.11
for FICA, FIRA, TICA, and TISA, accuracy studies should consider using the anterior or posterior
respectively). Parratte et al. [18] reported resection errors ranging femoral resections to quantify the FSA due to the improved imaging
from 0.50
to 1.29
in a cohort of 30 cadaveric knees using a RATKA accuracy.

Table 3
Correlations Between the CT and Optically Measured Resection Accuracies and Between the Resection Angles and Corresponding Implant Alignments.

Accuracy Metric CT Measured vs Optically Measured Resection Angles Optically Measured Resection Angles vs Implant Angles

Correlation Mean Difference ± SD Absolute Mean Correlation Mean Difference ± SD Absolute Mean
Difference ± SD Difference ± SD

Femur angles
Coronal angle (
) 0.97 0.13 ± 0.40 0.31 ± 0.28 0.88 0.22 ± 0.90 0.64 ± 0.66
Sagittal angle, distal reference (
) 0.80 0.01 ± 1.19 0.94 ± 0.73 0.74 2.21 ± 1.53 2.24 ± 1.48
Sagittal angle, posterior reference (
) 0.91 0.03 ± 1.32 0.91 ± 0.95 0.53 1.82 ± 2.68 2.58 ± 1.95
I-E rotation angle (
) 0.95 0.34 ± 0.65 0.58 ± 0.45 0.92 0.18 ± 0.83 0.64 ± 0.56
Tibial angles
Coronal angle (
) 0.96 0.53 ± 0.56 0.61 ± 0.47 0.97 0.03 ± 0.45 0.37 ± 0.26
Sagittal angle (
) 0.91 0.73 ± 0.81 0.88 ± 0.65 0.93 0.26 ± 0.75 0.61 ± 0.50

CT, computed tomography; SD, standard deviation; I-E, internal-external.

 G.W. Doan et al. / The Journal of Arthroplasty 37 (2022) 795e801 801

Small absolute mean differences (<0.64
) and strong correla- arthroplasty: a cohort study with long-term radiostereometric analysis re-
sults. J Bone Joint Surg Am 2019;101:1203e12. https://doi.org/10.2106/
tions (>0.88) were observed between the resection alignment and
JBJS.18.00691.
implant alignment, except for the FSA. Cementless femoral com- [6] Pfitzner T, Abdel MP, von Roth P, Perka C, Hommel H. Small improvements in
ponents were implanted in our study which have an interference fit mechanical axis alignment achieved with MRI versus CT-based patient-spe-
with the resected bone. The interference fit induced an average of cific instruments in TKA: a randomized clinical trial. Clin Orthop Relat Res
2014;472:2913e22. https://doi.org/10.1007/s11999-014-3784-6.
2.2
± 1.5
of femoral implant flexion during impaction relative to [7] De Vloo R, Pellikaan P, Dhollander A, Vander Sloten J. Three-dimensional
the distal resection surface across both surgical cohorts. The analysis of accuracy of component positioning in total knee arthroplasty with
observed component flexion was smaller in the RATKA cohort than patient specific and conventional instruments: a randomized controlled trial.
Knee 2017;24:1469e77. https://doi.org/10.1016/j.knee.2017.08.059.
the CTKA, likely due to the improved accuracy of the anterior and [8] An VVG, Sivakumar BS, Phan K, Levy YD, Bruce WJM. Accuracy of MRI-based
posterior femoral resections in the RATKA group. It should be noted vs. CT-based patient-specific instrumentation in total knee arthroplasty: a
that implantation of the final components took place after the meta-analysis. J Orthop Sci 2017;22:116e20. https://doi.org/10.1016/
j.jos.2016.10.007.
bones had been extracted from the specimen. This increased the [9] Cheng T, Zhao S, Peng X, Zhang X. Does computer-assisted surgery improve
stiffness of the impaction process and improved visibility of postoperative leg alignment and implant positioning following total knee
the implant bone interface to ensure full seating of the implants. arthroplasty? A meta-analysis of randomized controlled trials? Knee Surg
Sports Traumatol Arthrosc 2012;20:1307e22. https://doi.org/10.1007/
Impactions performed in a clinical setting may not result in such s00167-011-1588-8.
strong correlations between the resection and implant alignments. [10] Bollars P, Boeckxstaens A, Mievis J, Kalaai S, Schotanus MGM, Janssen D. Pre-
The small errors between resection alignments and implant liminary experience with an image-free handheld robot for total knee arthro-
plasty: 77 cases compared with a matched control group. Eur J Orthop Surg
alignments observed in this study (excluding femoral sagittal
Traumatol 2020;30:723e9. https://doi.org/10.1007/s00590-020-02624-3.
alignment) lend credence to retrospective clinical studies which [11] Scholl LY, Hampp EL, De Souza KM, Chang TC, Deren M, Yenna ZC, et al. How
quantify system accuracy through postoperative CT scans, where does robotic-arm assisted technology influence total knee arthroplasty
differences between intraoperative target alignments and the implant placement for surgeons in fellowship training? J Knee Surg 2022;35:
198e203. https://doi.org/10.1055/s-0040-1716983.
resulting implant alignment greater than 1
could likely be attrib- [12] Mahoney O, Kinsey T, Sodhi N, Mont MA, Chen AF, Orozco F, et al. Improved
uted to the robotic system performance. component placement accuracy with robotic-arm assisted total knee arthro-
This study had several notable limitations. Inherent variation plasty. J Knee Surg 2020. https://doi.org/10.1055/s-0040-1715571.
[13] Seidenstein A, Birmingham M, Foran J, Ogden S. Better accuracy and repro-
between cadaveric and in vivo tissue properties limits clinical ducibility of a new robotically-assisted system for total knee arthroplasty
applicability of these results. Cadaveric specimens were age- compared to conventional instrumentation: a cadaveric study. Knee Surg
matched to the TKA patient population but only 33% of specimen Sports Traumatol Arthrosc 2021;29:859e66. https://doi.org/10.1007/s00167-
020-06038-w.
had cartilage degeneration typical of TKA patients. Due to the [14] Koulalis D, O’Loughlin PF, Plaskos C, Kendoff D, Cross MB, Pearle AD. Sequential
bilateral comparison of RATKA and CTKA, this likely did not cause versus automated cutting guides in computer-assisted total knee arthroplasty.
significant bias between the cohorts. Likewise due to the cadaveric Knee 2011;18:436e42. https://doi.org/10.1016/j.knee.2010.08.007.
[15] Scho €llhorn W. Applications of artificial neural nets in clinical biomechanics. Clin
tissue, we were not able to measure weight-bearing postoperative Biomech 2004;19:876e98. https://doi.org/10.1016/j.clinbiomech.2004.04.005.
alignment. Finally, while resection accuracy is a fundamental [16] Sires JD, Craik JD, Wilson CJ. Accuracy of bone resection in MAKO total knee
component of implant longevity, appropriate soft tissue balance is robotic-assisted surgery. J Knee Surg 2019;1:13e6. https://doi.org/10.1055/s-
0039-1700570.
also critical. The current study did not assess knee stability or soft
[17] Sires JD, Wilson CJ. CT validation of intraoperative implant position and knee
tissue balance, which should be the focus of future studies. alignment as determined by the MAKO total knee arthroplasty system. J Knee
Surg 2020;34:1133e7. https://doi.org/10.1055/s-0040-1701447.
[18] Parratte S, Price AJ, Jeys LM, Jackson WF, Clarke HD. Accuracy of a new
Conclusions
robotically assisted technique for total knee arthroplasty: a cadaveric study.
J Arthroplasty 2019;34:2799e803. https://doi.org/10.1016/j.arth.2019.06.040.
In conclusion, the results of this study demonstrate that use of [19] Figueroa F, Wakelin E, Twiggs J, Fritsch B. Comparison between navigated
reported position and postoperative computed tomography to evaluate ac-
image-free RATKA can improve implant alignment, especially in the
curacy in a robotic navigation system in total knee arthroplasty. Knee
coronal plane, compared to conventional instrumentation and 2019;26:869e75. https://doi.org/10.1016/j.knee.2019.05.004.
reduce the occurrence of outliers. [20] Casper M, Mitra R, Khare R, Jaramaz B, Hamlin B, McGinley B, et al. Accuracy
assessment of a novel image-free handheld robot for Total Knee Arthroplasty
in a cadaveric study. Comput Assist Surg 2018;23:14e20. https://doi.org/
References 10.1080/24699322.2018.1519038.
[21] Marchand KB, Salem HS, Mathew KK, Harwin SF, Mont MA, Marchand RC. The
[1] Dossett HG, Estrada NA, Swartz GJ, LeFevre GW, Kwasman BG. A randomised accuracy of computed tomography-based, three-dimensional implant plan-
controlled trial of kinematically and mechanically aligned total knee re- ning in robotic-assisted total knee arthroplasty. J Knee Surg 2021. https://
placements: two-year clinical results. Bone Joint J 2014;96-B:907e13. https:// doi.org/10.1055/s-0041-1729548.
doi.org/10.1302/0301-620X.96B7.32812. [22] Besl PJ, McKay ND. A method for registration of 3-D shapes. IEEE Trans Pattern
[2] Swartz GJ, Dossett HG, Kwasman BG, Estrada NA, LeFevre GW. Kinematically Anal Mach Intell 1992;14:239e56. https://doi.org/10.1109/34.121791.
versus mechanically aligned total knee arthroplasty. Orthopedics 2012;35: [23] Davis ET, Pagkalos J, Gallie PAM, Macgroarty K, Waddell JP, Schemitsch EH.
e160e9. https://doi.org/10.3928/01477447-20120123-04. Defining the errors in the registration process during imageless computer
[3] Yaron BZ, Ilan S, Tomer K, Eran B, Gabriel A, Noam S. Patients undergoing navigation in total knee arthroplasty: a cadaveric study. J Arthroplasty
staged bilateral knee arthroplasty are less aware of their kinematic aligned 2014;29:698e701. https://doi.org/10.1016/j.arth.2013.06.034.
knee compared to their mechanical knee. J Orthop 2021;23:155e9. https:// [24] Jonkergouw F, Alle  F, Chellaoui K, Vander Sloten J, Vangeneugden D. Three-
doi.org/10.1016/j.jor.2020.12.032. dimensional measurement technique to assess implant position and orien-
[4] Teeter MG, Naudie DD, McCalden RW, Yuan X, Holdsworth DW, MacDonald SJ, tation after total knee arthroplasty. Med Eng Phys 2016;38:1513e7. https://
et al. Varus tibial alignment is associated with greater tibial baseplate migration doi.org/10.1016/j.medengphy.2016.09.006.
at 10 years following total knee arthroplasty. Knee Surg Sports Traumatol [25] Campanelli V, Lozano R, Akhlaghpour H, Brar AS, Maislin D, Nedopil AJ, et al.
Arthrosc 2018;26:1610e7. https://doi.org/10.1007/s00167-017-4765-6. Implant placement accuracy in total knee arthroplasty: validation of a CT-
[5] van Hamersveld KT, Marang-van de Mheen PJ, Nelissen RGHH. The effect of based measurement technique. Quant Imaging Med Surg 2020;10:475e84.
coronal alignment on tibial component migration following total knee https://doi.org/10.21037/qims.2020.01.02.