 KNEE
The learning curve associated with robotic-
B. Kayani,
S. Konan,
J. R. T. Pietrzak,
S. S. Huq,
J. Tahmassebi,
F. S. Haddad
From Princess Grace
Hospital and
University College
Hospital, London,
United Kingdom
 B. Kayani, MRCS, MBBS, BSc
(Hons), Specialist Registrar
Trauma and Orthopaedics
 J. Tahmassebi, BSc,
Extended Scope Practitioner
Department of Trauma and
Orthopaedics, University
College Hospital, London, UK
and Princess Grace Hospital,
London, UK.
 S. Konan, MBBS, MD (Res),
MRCS, FRCS (Tr&Orth),
Consultant Orthopaedic
Surgeon
 J. R. T. Pietrzak, MBBCh, FC
Orth (SA), Senior Clinical
Fellow
 S. S.Huq, MRCS. MBBS,
Clinical Research Fellow
Department of Trauma and
Orthopaedics, University
College Hospital and Princess
Grace Hospital, London, UK.
 F. S.Haddad, BSc MD (Res),
FRCS (Tr&Orth), Professor of
Orthopaedic Surgery
University College London
Hospitals, The Princess Grace
Hospital, and The NIHR
Biomedical Research Centre at
UCLH, London, UK.
Correspondence should be sent
to B. Kayani; email:
babar.kayani@gmail.com
©2018 The British Editorial
Society of Bone & Joint Surgery
doi:10.1302/0301-620X.100B8.
BJJ-2018-0040.R1 $2.00
Bone Joint J
2018;100-B:1033–42.
VOL. 100-B, No. 8, AUGUST 2018
arm assisted unicompartmental knee
arthroplasty
A PROSPECTIVE COHORT STUDY
Aims
The primary aim of this study was to determine the surgical team’s learning curve for
introducing robotic-arm assisted unicompartmental knee arthroplasty (UKA) into routine
surgical practice. The secondary objective was to compare accuracy of implant positioning
in conventional jig-based UKA versus robotic-arm assisted UKA.
Patients and Methods
This prospective single-surgeon cohort study included 60 consecutive conventional jig-
based UKAs compared with 60 consecutive robotic-arm assisted UKAs for medial
compartment knee osteoarthritis. Patients undergoing conventional UKA and robotic-arm
assisted UKA were well-matched for baseline characteristics including a mean age of 65.5
years (SD 6.8) vs 64.1 years (SD 8.7), (p = 0.31); a mean body mass index of 27.2 kg.m2 (SD 2.7)
vs 28.1 kg.m2 (SD 4.5), (p = 0.25); and gender (27 males: 33 females vs 26 males: 34 females,
p = 0.85). Surrogate measures of the learning curve were prospectively collected. These
included operative times, the Spielberger State-Trait Anxiety Inventory (STAI) questionnaire
to assess preoperative stress levels amongst the surgical team, accuracy of implant
positioning, limb alignment, and postoperative complications.
Results
Robotic-arm assisted UKA was associated with a learning curve of six cases for operating
time (p < 0.001) and surgical team confidence levels (p < 0.001). Cumulative robotic
experience did not affect accuracy of implant positioning (p = 0.52), posterior condylar offset
ratio (p = 0.71), posterior tibial slope (p = 0.68), native joint line preservation (p = 0.55), and
postoperative limb alignment (p = 0.65). Robotic-arm assisted UKA improved accuracy of
femoral (p < 0.001) and tibial (p < 0.001) implant positioning with no additional risk of
postoperative complications compared to conventional jig-based UKA.
Conclusion
Robotic-arm assisted UKA was associated with a learning curve of six cases for operating
time and surgical team confidence levels but no learning curve for accuracy of implant
positioning.
Cite this article: Bone Joint J 2018;100-B:1033–42.
Unicompartmental knee osteoarthritis is a widely stock, better restoration of kinematics, increased
occurring problem which has been considered to patient satisfaction, and improved functional out-
affect between 6% and 40% of the population.1 comes.4-7 However, UKA is associated with
End-stage disease that is refractory to conservative reduced implant survivorship and increased revi-
treatment may be treated with unicompartmental sion rates compared with TKA.8,9 Accuracy of
knee arthroplasty (UKA), which currently implant positioning and limb alignment are impor-
accounts for between 8% and 10% of all knee tant prognostic factors for implant survivorship
arthroplasty procedures performed in the United following UKA.9-11 Accordingly, techniques that
Kingdom and United States.2,3 Advantages of improve component positioning may reduce the
UKA over total knee arthroplasty (TKA) include burden of revision surgery in UKA.
reduced operating time, decreased intraoperative Robotic-arm assisted UKA uses a preoperative
blood loss, reduced periarticular soft-tissue CT scan to create a patient-specific computer
trauma, improved preservation of native bone aided design (CAD) model of the patient’s knee
1033
1034 B. KAYANI, S. KONAN, J. R. T. PIETRZAK, S. S. HUQ, J. TAHMASSEBI, F. S. HADDAD

Table I. Baseline characteristics, operating times, and Spielberger State-Trait Anxiety Inventory (STAI) scores in patients undergoing conven-
tional jig-based unicompartmental knee arthroplasty (UKA) versus robotic-arm assisted UKA

Characteristic Conventional jig-based UKA (n = 60) Robotic-arm assisted UKA (n = 60) p-value
Mean age, yrs (SD) 65.5 (6.8) 64.1 (8.7) 0.31*
Mean body mass index (kg/m2) 27.3 (2.7) 28.1 (4.5) 0.25*
Gender, n (%)
Female 27 (45) 26 (43.3) 0.85†
Male 33 (55) 34 (56.7)
ASA grade, n (%) 0.86†
I 28 (46.7) 31 (51.7)
II 31 (51.7) 28 (46.7)
III 1 (1.7) 1 (1.7)
Side intervention, n (%) 0.86†
Left 29 (48.3) 30 (50.0)
Right 31 (51.7) 30 (50.0)
Mean change in Hb concentration, g/l (sd) -13.5 (7.2) -12.7 (5.7) 0.52*
Mean operative time, mins (sd) 62.0 (6.4) 64.2 (8.8) 0.12‡
Mean preoperative STAI score (sd)
Operating surgeon 9.7 (4.1) 10.2 (2.8) 0.67§
Anaesthetist 9.1 (4.5) 10.4 (3.2) 0.71§
Scrub nurse 11.1 (2.7) 13.3 (3.2) 0.43§
Circulating nurse 14.7 (4.9) 13.6 (2.9) 0.32§
Operating department practitioner 12.6 (3.2) 12.0 (1.7) 0.48§
*Unpaired t-test
†Chi-squared test
‡Mann-Whitney test
§Paired t-test
ASA, American Society of Anesthesiologists; Hb, haemoglobin; STAI, Spielberger State-Trait Anxiety Inventory

joint. The surgeon is able to select the size and position of
implants based on virtual changes in bone coverage, limb align-
ment, and kinematics through the arc of motion. An intraopera-
tive surgeon-controlled robotic arm is used to resect predefined
haptic windows to achieve this plan to a high degree of accu-
racy.12 Robotic-arm assisted UKA is associated with increased
accuracy of implant positioning and alignment compared with
conventional jig-based UKA.13-16 Previous studies have shown
well-established learning curves for UKA, with the introduction
of new component designs, minimally invasive surgery, com-
puter navigation, and patient-specific implants.17-19 However, to
our knowledge, the learning curve for implementation of
robotic-arm assisted UKA has not been previously investigated.
The primary objective of this prospective single-surgeon
cohort study was to establish the surgical team’s learning curve
for robotic-arm assisted UKA through assessment of operating
times, surgical team confidence levels, accuracy of implant
positioning, limb alignment, and postoperative complications.
Our hypothesis was that cumulative experience with robotic
UKA would reduce operating times and improve surgical team
confidence levels but there would be no learning curve for accu-
racy of implant positioning and postoperative limb alignment.
The secondary objectives were to compare accuracy of femoral
and tibial implant positioning in patients undergoing robotic-
arm assisted UKA versus conventional jig-based UKA.
Patients and Methods
Patient selection. This study included 120 patients with symp-
tomatic medial compartment osteoarthritis undergoing primary
UKA between April 2016 and September 2017. This included
60 consecutive patients undergoing conventional jig-based
UKA followed by 60 consecutive patients receiving robotic-
arm assisted UKA. Patients were not randomized but this ena-
bled assessment of learning curves associated with complete
transition from conventional jig-based UKA to robotic-arm
assisted UKA. All operative procedures were performed by the
senior author (FSH) using the standard medial parapatellar
approach for UKA. Patients who qualified for medial unicom-
partmental knee replacement were included in this study. To be
treated with this option the following criteria were met: diagno-
sis of osteoarthritis or osteonecrosis limited to the medial com-
partment; preservation of the other compartments of the knee
joint; passively correctible varus deformity of less than 10°;
fixed flexion deformity less than 15°; maximum knee flexion
greater than 90°; and patient between 18 and 80 years of age.
Exclusion criteria comprised the following: diagnosis of inflam-
matory arthritis, haemochromatosis, chondrocalcinosis, or hae-
mophilia; symptomatic knee instability or anterior cruciate
ligament deficiency; multi-compartment disease; previously
failed correctional osteotomy or ipsilateral UKA; immobility or
other neurological condition affecting musculoskeletal func-
tion. Patients in both conventional and robotic groups were
well-matched for baseline characteristics including American
Society for Anesthesiologists20 (Table I). Two fellowship-
trained surgeons (SK and JRTP) collected all operative and radi-
ological outcomes and both observers were blinded to each
other’s recordings. Institutional review board approval was
gained prior to commencement of this study.
The Oxford Phase 3 mobile-bearing UKA (Zimmer Biomet,
Bridgend, United Kingdom) was implanted using standard man-
ual instrumentation for patients undergoing conventional jig-
based UKA. The RESTORIS MCK (Mako Surgical Corpora-

Follow us @BoneJointJ THE BONE & JOINT JOURNAL

 THE LEARNING CURVE ASSOCIATED WITH ROBOTIC-ARM ASSISTED UNICOMPARTMENTAL KNEE ARTHROPLASTY
tion, Kalamazoo, Michigan) fixed-bearing UKA system was
implanted using the Mako RIO robotic interactive orthopaedic
system (Mako Surgical Corporation) in patients undergoing
robotic-arm assisted UKA.
Surgical technique. Conventional jig-based UKA was per-
formed using standard instrumentation with extramedullary ref-
erencing to guide tibial bone resection and intramedullary
referencing for femoral bone resection. Standard instrumenta-
tion jigs were used to obtain fixed target values for all patients,
in accordance with the guidance of the Biomet operating techni-
cal manual.21 Limb alignment was guided by on-table assessment
of soft-tissue tension and native alignment restored. No further
intraoperative adjustments or tailoring of implant positioning
were performed to account for individual patient anatomy.
Robotic-arm assisted UKA was planned using the preopera-
tive CT scan and patient-specific CAD model. Implant size and
positioning were selected to optimize bone coverage, restore
native alignment, and minimize bone resection. The surgical
plan was therefore individualized to each patient’s unique joint
anatomy. The Mako software calculated haptic femoral and tib-
ial bone resection windows to achieve the planned implant posi-
tions and overall limb alignment. Stab incisions were performed
in the distal femur and proximal tibia, to enable bicortical regis-
tration pins to be inserted, and fixed arrays mounted to these to
enable intraoperative dynamic referencing. Bone registration
was performed by intraoperatively mapping radiological land-
marks displayed on the computer screen to verify anatomy and
establish bone geometry. Joint balancing captured femoral and
tibial poses with corrective forces, assessed kinematics through
the arc of motion, and enabled fine tuning of implant position-
ing based on laxity of the soft-tissue envelope to restore native
limb alignment. Bone resection was performed within the stere-
otactic boundaries of the haptic window using a high-speed,
water-cooled burr with tactile, visual, and audio feedback. Opti-
cal motion capture technology was used to assess limb align-
ment, range of movement, flexion and extension gaps, and arc
of motion with trial implants prior to definitive selection and
cemented implantation of final components.
Outcome measures. Operative time. Operative time was
defined as time from initial surgical incision to final wound clo-
sure. In robotic-arm assisted UKA, surgical times for the fol-
lowing parts of the procedure were also recorded: surgical tray
and instrument setup; surgical approach and insertion of regis-
tration pins; bone registration; joint balancing; bone prepara-
tion; implant trialling; and implantation of final prosthesis.
Surgical-team confidence. The Spielberger State-Trait Anxi-
ety Inventory (STAI) questionnaire is a validated subjective
assessment tool for quantifying an individual’s stress levels with
individual traits arising from the clinical environment.22 The
six-item questionnaire has a four-point rating scale and total
scores range from 6 to 24, with higher values indicating higher
levels of stress. The STAI questionnaire was completed by each
member of the surgical team prior to the surgical time-out for all
study patients. This included the operating surgeon, two con-
sultant anaesthetists, two senior scrub nurses, one operating
department practitioner (ODP), and one circulating nurse.
VOL. 100-B, No. 8, AUGUST 2018
1035
Accuracy of implant positioning and postoperative limb
alignment. Patients in both treatment groups underwent pre-
and postoperative anteroposterior knee, lateral knee, and full
leg-length standing radiographs. The two observers (JRTP and
SK) determined the accuracy of implant positioning by compar-
ing the value achieved in the postoperative radiograph to the
planned value in the corresponding preoperative radiograph. All
measurements were performed using the picture archiving and
communication system (PACS) software. Agreement between
the observers was investigated by interclass correlation coeffi-
cient (ICC). Femoral and tibial axes were used as reference
markers to assess accuracy of all positioning and measure align-
ment in degrees as described by Bell et al.15 The femoral coronal
implant alignment was measured as the angle subtended by the
femoral mechanical axis and the longitudinal axis of the femoral
condylar implant (Fig. 1).23 The femoral sagittal implant align-
ment was measured as the angle subtended by the femoral
implant peg axis and the femoral mechanical axis. The tibial
coronal implant alignment was assessed as the angle subtended
by the tibial mechanical axis and the medial to lateral axis of the
tibial implant. The tibial sagittal alignment was calculated as the
angle between the tibial mechanical axis and tibial implant.
Anteroposterior plain knee radiographs were used to measure
the joint line height by calculating the perpendicular distance
from a line extending through the distal points of the femoral
condyles and a parallel line extending to the fibular head. True
lateral knee radiographs were used to calculate the posterior
condylar offset ratio (PCOR) and posterior tibial slope using the
methods described by Guadnini et al16 and Johal et al.24
Complications. All patients were reviewed in outpatients at 30
days following surgery by the independent observers for clinical
assessment and full weight-bearing radiographs performed.
Any postoperative complications and their respective treat-
ments during this follow-up period were recorded for analysis.
Power calculation. Previous studies comparing operating
times between conventional jig-based and computer navigation
knee arthroplasty have shown that the mean difference in oper-
ating time is five minutes.25 The minimum detectable difference
in operating time in this study was one minute. Assuming simi-
lar differences in operating time between conventional jig-
based and robotic-arm assisted UKA with standard deviation of
10 minutes, we needed a sample size of 120 patients to detect a
difference of at least five minutes using a two-tailed, two-sample
t-test with a power of 80% and a significance level of 5%.
Statistical analysis. The cumulative summation (CUSUM)
analysis method26 was used to assess learning curves in robotic-
arm assisted UKA for operating time and surgical team stress
levels as assessed using the STAI questionnaire. The standard-
ized target values for the CUSUM analyses were set using the
mean values for these outcome measures from the conventional
jig-based UKA group. CUSUM values represent a running total
of the differences between the value of each data point and the
standardized target values for each outcome. The results of the
CUSUM analysis were presented on a chart with chronologi-
cally ordered case numbers on the x-axis and the corresponding
CUSUM score on the y-axis. This enabled performance over
1036 B. KAYANI, S. KONAN, J. R. T. PIETRZAK, S. S. HUQ, J. TAHMASSEBI, F. S. HADDAD

Fig. 1

Postoperative anteroposterior and lateral knee radiographs showing radiographic technique for assessing
coronal and sagittal implant positioning. Images used with permission from Koh IJ, Kim JH, Jang SW, et al.
Are the Oxford(®) medial unicompartmental knee arthroplasty new instruments reducing the bearing dislo-
cation risk while improving components relationships? A case control study. Orthop Traumatol Surg Res 2016
Apr;102:183-7.

consecutive procedures to be visualized and inflexion points
showing transition points in the learning curve to be identified.
The learning curves for accuracy of implant position and post-
operative limb alignment in robotic-arm assisted UKA were
assessed by dividing patients undergoing robotic surgery into
six groups of ten patients based on chronology of surgery. The
root mean square error (RMSE) values for accuracy of implant
position in each group were calculated and assessed for progres-
sion with each consecutive group. Categorical data were com-
pared using Fisher’s exact test and chi-squared test. Normally
distributed continuous variables were compared using inde-
pendent t-tests for unpaired data sets, paired t-tests for related
(paired) data sets, and one-way analysis of variance (ANOVA)
for multiple data sets. The Mann–Whitney test was used for
non-parametric data. Statistical significance was set at a p-value
< 0.05 for all analyses. All statistical analysis was performed
using SPSS software version 21 (IBM Corp., Armonk, New
York).
Results
Interclass correlation coefficient. Operative and radiological
outcomes recorded all had an ICC > 0.80 (range 0.86 to 0.92),
indicating good agreement on all parameters assessed by the
two independent observers.
Operating times. In robotic-arm assisted UKAs, CUSUM
analysis for operative time revealed a sharp inflexion after the
sixth case, which helped to identify two distinct phases in the
learning curve (Fig. 2). Phase 1 represents the initial learning
period, whereas Phase 2 represents the proficiency phase. Com-
parison of the two phases demonstrated phase 1 procedures to
be significantly longer (p < 0.001) with no differences in base-
line characteristics (Tables II and III). Overall, robotic-arm
assisted surgery was not associated with increased operating
times compared with conventional jig-based UKA (p = 0.12).
Surgical-team confidence levels. CUSUM analysis of preop-
erative stress levels as assessed using the STAI questionnaire
revealed a significant inflexion after six cases in a pattern simi-
lar to the operative time learning curve (Fig. 3). Further analysis
revealed STAI scores to be significantly higher in phase 1 than
in phase 2 for every member of the surgical team (Fig. 4). There
was no difference in the overall STAI scores amongst members
of the surgical team between the two treatment groups (Table I).
Implant positioning and limb alignment. There was no learn-
ing curve effect in robotic-arm assisted UKA for accuracy of
achieving the planned femoral (p = 0.80) and tibial (p = 0.52)
implant positioning, PCOR (p = 0.71), posterior tibial slope
(p = 0.68), joint line height (p = 0.55), and postoperative align-
ment (p = 0.65) (Table IV). There was a statistically significant

Follow us @BoneJointJ THE BONE & JOINT JOURNAL

 THE LEARNING CURVE ASSOCIATED WITH ROBOTIC-ARM ASSISTED UNICOMPARTMENTAL KNEE ARTHROPLASTY 1037

200 160
y = 19.53 + 23.23 * x
R2 Linear = 0.987
150 140
CUSUM (operative time)

CUSUM (operative time)

120
100

100
50
80
0
60

–50
40

–100 20
0 20 40 60 80 100 120 0 1 2 3 4 5 6
Case number Case number
Fig. 2a Fig. 2b

160

y = 1.59E2 + -0.79 * x + 5.87E - 3 * x2

R2 Quadratic = 0.613

150
CUSUM (operative time)

140

130

120
7 17 27 37 47 57
Case number
Fig. 2c

Charts displaying cumulative summation (CUSUM) analysis for study patients undergoing unicompartmental knee arthroplasty (UKA). a) Chart plot-
ting CUSUM analysis for operative times in all 120 consecutive UKA procedures. The solid vertical line represents transition from conventional jig-
based UKA to robotic-arm assisted UKA. The dashed vertical line represents transition between Phase 1 (learning) and Phase 2 (proficiency) of the
learning curve for robotic-arm assisted UKA. b) Chart plotting CUSUM analysis for phase 1 of operative times in consecutive robotic-arm assisted
UKA procedures and c) chart plotting CUSUM analysis for phase 2 of operative times in consecutive robotic-arm assisted UKA procedures.

improvement afforded by the robotic-arm assisted UKA in toms in both patients resolved with conservative treatment prior
achieving the planned femoral coronal (p < 0.001) and sagittal to discharge. No other postoperative complications were identi-
(p < 0.001) implant positioning, tibial coronal (p < 0.001) and fied in either treatment group within 30 days follow-up.
sagittal (p < 0.001) implant positioning, posterior tibial slope
(p < 0.001), and joint line height (p < 0.001) compared with Discussion
conventional jig-based UKA (Table V). There was no difference This prospective single-surgeon cohort study showed that
between the two treatment groups relating to achieving the transitioning from conventional jig-based UKA to robotic-
planned PCOR (p = 0.54). arm assisted UKA is associated with a learning curve of six
Complications cases for achieving operating times and surgical-team confi-
Two patients in the conventional jig-based UKA group devel- dence levels equivalent to those of conventional jig-based
oped increasing pain and swelling in the operated leg at day 2 UKA. Robotic-arm assisted UKA was not associated with a
following surgery. Both of these patients had Doppler ultra- learning curve for accuracy of implant positioning, limb
sound scans that were negative for deep vein thrombosis. Symp- alignment, PCOR, and posterior tibial slope, and there was

VOL. 100-B, No. 8, AUGUST 2018

1038 B. KAYANI, S. KONAN, J. R. T. PIETRZAK, S. S. HUQ, J. TAHMASSEBI, F. S. HADDAD

Table II. Operative data in patients undergoing robotic-arm assisted unicompartmental knee arthroplasty (UKA)

Characteristic Cases 1 to 10 Cases 11 to 20 Cases 21 to 30 Cases 31 to 40 Cases 41 to 50 Cases 51 to 60 p-value

Change in Hb, g/l -13.5 (5.6) -12.0 (6.5) -13.1 (4.2) -12.7 (6.9) -12.0 (5.6) -12.9 (6.4) 0.99*
Tray/instrument setup 13.2 (2.0) 7.3 (1.8) 6.9 (1.4) 8.5 (3.7) 7.5 (1.6) 6.8 (1.5) < 0.03*
Surgical approach 10.2 (3.1) 8.1 (2.6) 7.6 (1.0) 8.0 (1.5) 6.9 (1.2) 7.3 (0.5) < 0.001†
Bone registration 11.6 (2.5) 7.6 (1.3) 7.9 (1.2) 8.5 (1.1) 7.7 (1.1) 8.2 (1.7) < 0.001†
Joint balancing 10.8 (4.4) 6.5 (1.2) 6.7 (1.2) 7.4 (1.2) 6.5 (0.7) 6.8 (1.2) < 0.001†
Bone preparation 15.3 (6.8) 8.9 (2.0) 9.1 (1.4) 9.8 (1.6) 8.8 (0.9) 9.0 (1.6) < 0.001†
Implant trialling 6.0 (1.6) 6.2 (1.0) 6.4 (1.5) 6.4 (1.8) 6.7 (1.8) 6.9 (1.6) 0.93*
Cement implantation 15.7 (1.2) 15.8 (1.3) 15.6 (1.8) 15.6 (1.0) 16.4 (1.2) 16.7 (3.0) 0.59*
Overall operating time 77.3 (14.6) 60.6 (2.5) 61.5 (3.9) 63.0 (4.1) 60.8 (3.6) 62.1 (2.7) < 0.001†
*One-way analysis of variance (ANOVA)
†One-way ANOVA with Welch test

Table III. Comparison of operative time cumulative summation (CUSUM) learning curve phases in
patients undergoing robotic-arm assisted unicompartmental knee arthroplasty (UKA)

Characteristic Phase 1 (n = 6) Phase 2 (n = 54) p-value

Mean age, yrs (SD) 66.8 (9.1) 63.8 (8.7) 0.42*
Mean BMI, kg/m2 (SD) 27.8 (4.0) 28.2 (4.6) 0.85*
ASA grade II or III, n (%) 3 (50) 26 (48.1) 1.00†
Male gender, n (%) 3 (50) 31 (57.4) 1.00†
Mean preoperative Hb, g/l (SD) 130.7 (11.7) 135.3 (9.8) 0.28*
Mean postoperative Hb change, g/l (SD) 11.3 (6.2) 12.9 (5.7) 0.54*
Mean operative time, mins (SD) 87.3 (8.6) 61.7 (3.4) < 0.01*
*Unpaired t-test
†Fisher exact test
BMI, body mass index; ASA, American Society of Anesthesiologists

no additional risk of complications compared with conven-
tional jig-based UKA.
Operative time is often used as a surrogate marker of surgical
proficiency. In this study, there was a rapid decline in operating
times during the initial six cases as the surgeon became increas-
ingly familiar and adept with robotic technology. Marked
improvements were observed in time for bone preparation,
which decreased by over 40% after the initial learning phase,
with more moderate improvements in time for bone registration
and joint balancing. Intraoperative anatomical landmarks for
bone registration were standardized in all patients and so, with
increasing experience, the surgeon was able to proactively posi-
tion the tracker tip and leg holder into the correct position for the
next step of the registration. Furthermore, as the surgeon
became more practiced with fine movements of the robotic arm
and developed greater sensitivity to inhibitory feedback from
the robotic milling burr, time for bone resection also decreased
markedly. The most important change here was the pressure
with which the surgeon held the bone burr used to prepare the
bone surface. Our findings showing improved operating times
with increasing robotic experience are supported by other stud-
ies on robotic-guided total hip arthroplasty and dry-bone studies
with inexperienced surgeons.27,28 In our study, operating times
after the initial six cases of robotic-arm assisted UKA were
equivalent to those of conventional jig-based UKA. However,
these findings should be interpreted with caution as these might
in part reflect the general experience derived from the operating
surgeon’s high-volume practice, and therefore his learning
curve may not be directly transferrable to other surgeons.
Increasing levels of stress and mental strain are associated
with diminished operative performance, poor decision-making,
and reduced technical skills.22 Implementation of new technol-
ogy leads to heightened sympathetic nervous system function
with increased subjective and objective markers of stress
amongst the surgical team.23 In robotic-arm assisted UKA, there
was a learning curve of six cases for achieving baseline surgical
anxiety and stress levels as assessed using the STAI question-
naire. During this initial learning phase, the surgical team
became more proficient with setting up the new trays and instru-
ments, positioning the robot in theatre, attaching the burr to the
robotic arm, and proactively preparing the registration pins,
check points, and arrays. As the team became more confident
with these steps, the subjective anxiety and stress levels also
diminished. Importantly, the increased subjective stress levels
and operating times over the first six cases did not result in any
compromise in implant positioning or limb alignment in robotic
surgery.
Conventional jig-based UKA is associated with a steep learn-
ing curve that is associated with poorer outcomes whilst surgical
experience is being acquired.29,30 Our study did not show any
learning curve effect in robotic-arm assisted UKA for achieving
the planned implant position or postoperative alignment. The
native joint line was also preserved to within 2 mm in all
patients undergoing robotic surgery, which is important as pre-
vious studies have shown that overstuffing or increasing the
joint line by more than 2 mm can lead to increased valgus,
greater strain on the medial collateral ligament, and poorer out-
comes in medial UKA.31 There was no trend or impact of cumu-

Follow us @BoneJointJ THE BONE & JOINT JOURNAL

 THE LEARNING CURVE ASSOCIATED WITH ROBOTIC-ARM ASSISTED UNICOMPARTMENTAL KNEE ARTHROPLASTY 1039

Operating surgeon Scrub nurse Circulating nurse

Anaesthetist ODP

50
Phase 1 Phase 2

40

30
Value

20

10

–10
1 6 11 16 21 26 31 36 41 46 51 56
Case number
Fig. 3

Chart displaying cumulative summation (CUSUM) analysis for STAI scores amongst surgical team mem-
bers in all robotic-arm assisted unicompartmental knee arthroplasty (UKA) procedures. ODP, operating
department practitioner

lative robotic experience in the ability to reproduce the native
PCOR. The mean PCOR was 0.45 with conventional UKA and
0.46 with robotic UKA, which are similar to the previously
reported values of 0.48 for accurate restoration of the PCOR in
UKA16 and 0.47 in TKA.24 Accurate reproduction of the native
PCOR is important for preventing flexion contractures and opti-
mizing postoperative movement.16 Robotic-arm assisted UKA
did not show any learning curve for achieving the planned pos-
terior tibial slope, which has been previously shown to influence
postoperative pain, functional outcomes, and tibial implant sur-
vivorship.16,17 Blyth et al32 conducted a randomized controlled
trial in 139 patients comparing postoperative functional out-
comes in patients undergoing UKA using either the conven-
tional jig-based technique or robotic-arm assistance. The study
showed that robotic-arm assisted UKA was associated with
improved early postoperative pain scores and functional out-
comes as assessed with the American Knee Society Score
(AKSS) for the three months following surgery, but there was
no difference in functional outcomes observed between the two
groups at one year after surgery.
Our findings are consistent with previous studies showing
improved accuracy of implant positioning in robotic-arm assisted
UKA compared with conventional jig-based UKA.13-16 Bell et al15
conducted a prospective randomized study on 140 patients under-
going UKA and showed robotic-arm assisted UKA was associ-
ated with improved accuracy and reduced outliers in implant
positioning compared to conventional jig-based UKA. The
reported RMSE values for femoral and tibial implant positioning
using CT scan were similar to the RMSE values observed using
plain radiographs in our study. Cobb et al14 conducted a prospec-
tive randomized study on 27 patients and showed that all patients
undergoing robotic-arm assisted UKA had tibiofemoral align-
ment in the coronal plane within 2° of the planned position com-
pared with only 40% in those undergoing conventional jig-based
UKA. Robotic-arm assisted UKA is associated with statistically
significant improved accuracy in implant positioning but the
long-term clinical and functional significance of these radiologi-
cal differences still needs to be established.
The findings of this study will aid clinicians and healthcare
policymakers in the safe implementation of robotic-arm assisted

VOL. 100-B, No. 8, AUGUST 2018

1040 B. KAYANI, S. KONAN, J. R. T. PIETRZAK, S. S. HUQ, J. TAHMASSEBI, F. S. HADDAD

Operating surgeon Scrub nurse

Circulating nurse Anaesthetist
ODP

25

20
Mean STAI score

15

10

0
1 2
STAI learning curve phase
Fig. 4

Chart comparing Spielberger State-Trait Anxiety Inventory (STAI)

scores between learning phases for all members of the surgical team in
robotic-arm assisted unicompartmental knee arthroplasty (UKA) proce-
dures. Error bars represent 95% CI.

Table IV. Radiological outcomes in patients undergoing robotic-arm-assisted unicompartmental knee arthroplasty (UKA)

Characteristic Cases 1 to 10 Cases 11 to 20 Cases 21 to 30 Cases 31 to 40 Cases 41 to 50 Cases 51 to 60 p-value*

Femoral coronal RMSE, ° 2.03 1.71 2.09 1.97 1.95 2.01 0.80
Femoral sagittal RMSE, ° 3.13 3.08 3.14 3.14 3.02 3.08 0.85
Tibial coronal RMSE, ° 1.20 1.31 1.24 1.06 1.23 1.05 0.52
Tibial sagittal RMSE, ° 2.05 2.74 2.78 2.54 2.29 2.57 0.72
Posterior condylar offset ratio RMSE 0.03 0.03 0.04 0.02 0.03 0.03 0.71
Posterior tibial slope RMSE, ° 1.94 2.14 1.86 1.95 1.99 1.81 0.68
Joint line RMSE, mm 0.84 0.97 0.94 0.97 0.84 0.87 0.55
Postoperative mechanical alignment, ° 1.77 1.85 1.43 1.70 1.62 1.55 0.65
*One-way analysis of variance (ANOVA)
RMSE, Root mean square error

Table V. Root mean square implantation errors (RMSE) in patients undergoing conventional jig-based unicompartmental knee
arthroplasty (UKA) versus robotic-arm assisted UKA

Characteristic Conventional jig-based UKA (n = 60) Robotic-arm assisted UKA (n = 60) p-value*
Posterior condylar offset ratio 0.04 0.03 0.54
Posterior tibial slope, ° 2.77 1.94 < 0.001
Joint line height, mm 1.94 0.91 < 0.001
Femoral coronal implant, ° 5.88 1.96 < 0.001
Femoral sagittal implant, ° 5.12 3.09 < 0.001
Tibial coronal implant, ° 3.05 1.18 < 0.001
Tibial sagittal implant, ° 4.06 2.50 < 0.001
Unpaired t-test

UKA into clinical practice. Theatre planning and scheduling of assisted UKA into the work flow increases surgical anxiety and
operative cases should account for increased operating times stress levels amongst all members of the surgical team during
during the initial learning curve. Introduction of robotic-arm this learning phase. As team members become more familiar

Follow us @BoneJointJ THE BONE & JOINT JOURNAL

 THE LEARNING CURVE ASSOCIATED WITH ROBOTIC-ARM ASSISTED UNICOMPARTMENTAL KNEE ARTHROPLASTY
and adept with robotic technology, confidence levels improve
and theatre efficiency increases thereafter. There is no impact of
cumulative experience with the robotic device on accuracy of
implant positioning or postoperative limb alignment. This will
help the safe incorporation of this technology into surgical prac-
tice and minimize implant failure due to surgeon-controlled
technical errors in component positioning.
There are several limitations of this study that need to be con-
sidered when interpreting its findings. Firstly, it was not possi-
ble to blind patients or observers recording radiological
outcomes as different implant designs were used in each treat-
ment group. Secondly, accuracy of implant positioning and
alignment were measured using plain radiographs, which is less
accurate than CT. There was good interobserver agreement on
all radiological outcomes recorded but we could not demon-
strate the accuracy of our individual measurements. Thirdly, fol-
low-up time was limited to 30 days and so only early
complications have been captured in this study. Fourthly, the
surgical team in this study are all experienced in working with
both conventional and navigated UKA in a high-volume arthro-
plasty centre with different operating surgeons. The impact of
their previous experience and dynamics on the learning curve
may not be directly transferrable to other less experienced
teams. Conventional UKA was associated with a very high
accuracy of implant positioning and therefore the clinical signif-
icance of the statistically significant difference in measured
positions between the two treatment techniques remains uncer-
tain. Fifthly, patients undergoing robotic-arm assisted UKA
received a fixed-bearing implant whilst those undergoing con-
ventional jig-based UKA received a mobile bearing implant.
The mobile-bearing (Oxford) UKA was chosen as the compar-
ator in this study as this is the most commonly used implant for
UKA as per the National Joint Registry of England and Wales2
and the implant that the operating surgeon was most familiar
with for UKA.
In conclusion, robotic-arm assisted UKA was associated with
a learning curve of six cases for attainment of operating times
and reaching surgical team confidence levels equivalent to that
of conventional jig-based UKA. There was no learning curve in
robotic-arm assisted UKA for accuracy of femoral or tibial
implant positioning, PCOR, posterior tibial slope, joint line res-
toration, and postoperative limb alignment. Robotic-arm assisted
UKA was associated with statistically significant improvement
in the accuracy in implant positioning with no additional risk of
complications at 30 days follow-up compared with conven-
tional jig-based UKA.
Take home message:
- Robotic-arm assisted UKA does not have a learning curve for
accuracy in achieving the planned femoral and tibial implant
positioning.
- There is no additional risk of postoperative complications during the
learning phase of robotic-arm assisted UKA compared to conventional
jig-based UKA.
- Robotic-arm assisted UKA is associated with a learning curve of six oper-
ative cases for achieving operating times and surgical-team confidence
levels comparable to conventional jig-based UKA.
VOL. 100-B, No. 8, AUGUST 2018
1041
References
1. Ackroyd CE. Medial compartment arthroplasty of the knee. J Bone Joint Surg [Br]
2003;85-B:937–942.
2. No authors listed. National Joint Registry for England, Wales, Northern Ireland and
the Isle of Man. 14th Annual Report, 2017. http://www.njrreports.org.uk/Portals/0/
PDFdownloads/NJR%2014th%20Annual%20Report%202017.pdf (date last accessed
16 May 2018).
3. Riddle DL, Jiranek WA, McGlynn FJ. Yearly incidence of unicompartmental knee
arthroplasty in the United States. J Arthroplasty 2008;23:408–412.
4. Chassin EP, Mikosz RP, Andriacchi TP, Rosenberg AG. Functional analysis of
cemented medial unicompartmental knee arthroplasty. J Arthroplasty 1996;11:553–559.
5. Isaac SM, Barker KL, Danial IN, et al. Does arthroplasty type influence knee joint
proprioception? A longitudinal prospective study comparing total and unicompart-
mental arthroplasty. Knee 2007;14:212–217.
6. Jeer PJ, Cossey AJ, Keene GC. Haemoglobin levels following unicompartmental
knee arthroplasty: influence of transfusion practice and surgical approach. Knee
2005;12:358–361.
7. Price AJ, Webb J, Topf H, et al. Rapid recovery after Oxford unicompartmental
arthroplasty through a short incision. J Arthroplasty 2001;16:970–976.
8. Koskinen E, Eskelinen A, Paavolainen P, Pulkkinen P, Remes V. Comparison of
survival and cost-effectiveness between unicondylar arthroplasty and total knee
arthroplasty in patients with primary osteoarthritis: a follow-up study of 50,493 knee
replacements from the Finnish Arthroplasty Register. Acta Orthop 2008;79:499–507.
9. Hernigou P, Deschamps G. Alignment influences wear in the knee after medial uni-
compartmental arthroplasty. Clin Orthop Relat Res 2004;423:161–165.
10. Emerson RH Jr, Higgins LL. Unicompartmental knee arthroplasty with the Oxford
prosthesis in patients with medial compartment arthritis. J Bone Joint Surg [Am]
2008;90-A:118–122.
11. Collier MB, Eickmann TH, Sukezaki F, McAuley JP, Engh GA. Patient, implant,
and alignment factors associated with revision of medial compartment unicondylar
arthroplasty. J Arthroplasty 2006;21(Suppl 2):108–115.
12. Conditt MA, Roche MW. Minimally invasive robotic-arm-guided unicompartmental
knee arthroplasty. J Bone Joint Surg [Am] 2009;91-A(Suppl 1):63–68.
13. Lonner JH, John TK, Conditt MA. Robotic arm-assisted UKA improves tibial com-
ponent alignment: a pilot study. Clin Orthop Relat Res 2010;468:141–146.
14. Cobb J, Henckel J, Gomes P, et al. Hands-on robotic unicompartmental knee
replacement: a prospective, randomised controlled study of the Acrobot system.
J Bone Joint Surg [Br] 2006;88-B:188–197.
15. Bell SW, Anthony I, Jones B, et al. Improved accuracy of component positioning
with robotic-assisted unicompartmental knee arthroplasty: data from a prospective,
randomized controlled study. J Bone Joint Surg [Am] 2016;98-A:627–635.
16. Gaudiani MA, Nwachukwu BU, Baviskar JV, Sharma M, Ranawat AS. Opti-
mization of sagittal and coronal planes with robotic-assisted unicompartmental knee
arthroplasty. Knee 2017;24:837–843.
17. Lubowitz JH, Sahasrabudhe A, Appleby D. Minimally invasive surgery in total
knee arthroplasty: the learning curve. Orthopedics 2007;30(Suppl):80–82.
18. Jenny JY, Miehlke RK, Giurea A. Learning curve in navigated total knee replace-
ment. A multi-centre study comparing experienced and beginner centres. Knee
2008;15:80–84.
19. Kashyap SN, Van Ommeren JW, Shankar S. Minimally invasive surgical tech-
nique in total knee arthroplasty: a learning curve. Surg Innov 2009;16:55–62.
20. Saklad M. Grading of patients for surgical procedures. Anesthesiol 1941;2:281–284.
21. No authors listed. Oxford unicompartmental knee manual of the surgical technique
http://www.biomet.se/resource/17723/Oxford%20ST.pdf (date last accessed 18
June 2018).
22. Marteau TM, Bekker H. The development of a six-item short-form of the state scale
of the Spielberger State-Trait Anxiety Inventory (STAI). Br J Clin Psychol 1992;31:301–
306.
23. Koh IJ, Kim JH, Jang SW, et al. Are the Oxford medial unicompartmental knee
arthroplasty new instruments reducing the bearing dislocation risk while improving
components relationships? A case control study. Orthop Traumatol Surg Res 2016
Apr;102:183–187.
24. Johal P, Hassaballa MA, Eldridge JD, Porteous AJ. The posterior condylar off-
set ratio. Knee 2012;19:843–845.
1042 B. KAYANI, S. KONAN, J. R. T. PIETRZAK, S. S. HUQ, J. TAHMASSEBI, F. S. HADDAD

25. Seon JK, Song EK. Navigation-assisted less invasive total knee arthroplasty com- Author contributions:
pared with conventional total knee arthroplasty: a randomized prospective trial. B. Kayani: Hypothesis generation, Data collection and presentation,
J Arthroplasty 2006;21:777–782.
Manuscript preparation.
26. Yap C- H, Colson ME, Watters DA. Cumulative sum techniques for surgeons: a
brief review. ANZ J Surg 2007;77:583–586. S. Konan: Data collection, manuscript preparation.
J. R. T. Pietrzak: Data collection.
27. Redmond JM, Gupta A, Hammarstedt JE, et al. The learning curve associated
with robotic-assisted total hip arthroplasty. J Arthroplasty 2015;30:50–54. S. S. Huq: Data analysis.
28. Karia M, Masjedi M, Andrews B, Jaffry Z, Cobb J. Robotic assistance enables J. Tahmassebi: Data analysis.
inexperienced surgeons to perform unicompartmental knee arthroplasties on dry bone F. S. Haddad: Manuscript preparation.
models with accuracy superior to conventional methods. Adv Orthop
2013;2013:481039. Funding statement:
29. Rees JL, Price AJ, Beard DJ, Dodd CA, Murray DW. Minimally invasive Oxford The author or one or more of the authors have received or will receive ben-
unicompartmental knee arthroplasty: functional results at 1 year and the effect of sur-
gical inexperience. Knee 2004;11:363–367. efits for personal or professional use from a commercial party related directly

30. Zhang Q, Zhang Q, Guo W, et al. The learning curve for minimally invasive Oxford
phase 3 unicompartmental knee arthroplasty: cumulative summation test for learning
curve (LC-CUSUM). J Orthop Surg Res 2014;9:81.
31. Chatellard R, Sauleau V, Colmar M, et al. Medial unicompartmental knee arthro-
plasty: does tibial component position influence clinical outcomes and arthroplasty
survival? Orthop Traumatol Surg Res 2013;99(Suppl):S219–S225.
32. Blyth MJG, Anthony I, Rowe P, et al. Robotic arm-assisted versus conventional
unicompartmental knee arthroplasty: exploratory secondary analysis of a randomised
controlled trial. Bone Joint Res 2017;6:631–639.
or indirectly to the subject of this article. In addition, benefits have been or
will be directed to a research fund, foundation, educational institution, or
other non- profit organization with which one or more of the authors are
associated.
This research/study/project was supported by the National Institute for Health
Research University College London Hospitals Biomedical Research Centre.
This article was primary edited by G. Scott.

Follow us @BoneJointJ THE BONE & JOINT JOURNAL