8/17/2019 Estimating Totalkneereplacementjointloadratiosfromkinematics
1/9
Estimating total knee replacement joint load ratios from kinematics
Clare K. Fitzpatrick n, Paul J. Rullkoetter
Center for Orthopaedic Biomechanics, University of Denver, 2390 S. York St., Denver, CO 80208, USA
a r t i c l e i n f o
Article history:
Accepted 1 July 2014
Keywords:
Total knee replacement
Finite element
Joint load ratios
Kinematics
Force prediction
Fluoroscopy
a b s t r a c t
Accurate prediction of loads acting at the joint in total knee replacement (TKR) patients is key to
developing experimental or computational simulations which evaluate implant designs under physio-
logical loading conditions. In vivo joint loads have been measured for a small number of telemetric TKR
patients, but in order to assess device performance across the entire patient population, a larger patient
cohort is necessary. This study investigates the accuracy of predicting joint loads from joint kinematics.
Specically, the objective of the study was to assess the accuracy of internal–external (I–E) and anterior–
posterior (A–P) joint load predictions from I–E and A–P motions under a given compressive load, and to
evaluate the repeatability of joint load ratios (I–E torque to compressive force (I–E:C), and A–P force to
compressive force (A–P:C)) for a range of compressive loading proles. A tibiofemoral nite element
model was developed and used to simulate deep knee bend, chair-rise and step-up activities for ve
patients. Root-mean-square (RMS) differences in I–E:C and A–P:C load ratios between telemetric
measurements and model predictions were less than 1.10e–3 Nm/N and 0.035 N/N for all activities.
I–E:C and A–P:C load ratios were consistently reproduced regardless of the compressive force prole
applied (RMS differences less than 0.53e–3 Nm/N and 0.010 N/N, respectively). When error in kinematic
measurement was introduced to the model, joint load predictions were forgiving to kinematic
measurement error when conformity between femoral and tibial components was low. The prevalence
of kinematic data, in conjunction with the analysis presented here, facilitates determining the scope of
A–P and I–E joint loading ratios experienced by the TKR population.
& 2014 Elsevier Ltd. All rights reserved.
1. Introduction
It is important to evaluate prospective total knee replacement
(TKR) devices under physiological joint loading conditions so that
in vivo mechanics can be accurately evaluated during pre-clinical
experimental or computational simulation. Loads acting at the joint
affect kinematics, wear and micromotion of the components, inu-
encing the clinical performance and longevity of these devices.
However, due to the complex nature of the knee joint (six degrees-
of-freedom (DOF), ligamentous structures, and muscle redundancy),
estimating joint forces from numerical models remains a challenge(Kinney et al., 2013; Fregly et al., 2012). The current standard in joint
force prediction from numerical models is rigid body musculoskeletal
optimization in combination with nite element (FE) modeling (Kim
et al., 2009; Shelburne et al., 2005; Taylor et al., 2004). Due to the
iterative nature of these simulations, they are computationally
intensive and can take days to solve. In recent years, direct measure-
ment of in vivo tibiofemoral (TF) joint forces during dynamic activity
has been achieved. Telemetric tibial trays have been implanted in a
number of TKR patients and TF joint loads have been recorded while
the patients perform activities of daily living (Kutzner et al., 2010;
D’Lima et al., 2011). However, in vivo joint load data has been
collected for just a handful of TKR patients. In order to develop
population-based analyses to assess device performance across the
entire TKR patient population, a larger patient cohort in conjunction
with computationally ef cient simulations is desirable.
There is a wealth of kinematic data available from TKR patients.
In vivo kinematics are acquired through a variety of techniques,
with varying levels of accuracy in joint motion measurement. Jointkinematics are frequently measured using motion capture analysis
in the gait laboratory, typically using reective or infra-red
markers attached to the skin to measure whole body motion
during dynamic activity, and applying inverse kinematics methods
to predict the motion of the underlying bones (D’Lima et al., 2012;
Lloyd and Besier, 2003). Video uoroscopy tracks the bone or
implant geometry with improved accuracy over motion capture
systems; single plane systems have reported measurement of
joint motions with accuracy in the order of 0.5 mm or 0.51 for in-
plane translations and rotations, respectively (Banks and Hodge,
1996; Mahfouz et al., 2003; Fregly et al., 2005; Prins et al., 2010;
Contents lists available at ScienceDirect
journal homepage: www.elsevier.com/locate/jbiomechwww.JBiomech.com
Journal of Biomechanics
http://dx.doi.org/10.1016/j.jbiomech.2014.07.002
0021-9290/& 2014 Elsevier Ltd. All rights reserved.
n Corresponding author. Tel.: þ 1 303 871 6435; fax: þ 1 303 871 4450.
E-mail address: [email protected] (C.K. Fitzpatrick).
Journal of Biomechanics 47 (2014) 3003–3011
http://www.sciencedirect.com/science/journal/00219290http://www.elsevier.com/locate/jbiomechhttp://www.jbiomech.com/http://dx.doi.org/10.1016/j.jbiomech.2014.07.002mailto:[email protected]:[email protected]:[email protected]://dx.doi.org/10.1016/j.jbiomech.2014.07.002http://dx.doi.org/10.1016/j.jbiomech.2014.07.002http://dx.doi.org/10.1016/j.jbiomech.2014.07.002http://dx.doi.org/10.1016/j.jbiomech.2014.07.002mailto:[email protected]://crossmark.crossref.org/dialog/?doi=10.1016/j.jbiomech.2014.07.002&domain=pdfhttp://crossmark.crossref.org/dialog/?doi=10.1016/j.jbiomech.2014.07.002&domain=pdfhttp://crossmark.crossref.org/dialog/?doi=10.1016/j.jbiomech.2014.07.002&domain=pdfhttp://dx.doi.org/10.1016/j.jbiomech.2014.07.002http://dx.doi.org/10.1016/j.jbiomech.2014.07.002http://dx.doi.org/10.1016/j.jbiomech.2014.07.002http://www.jbiomech.com/http://www.jbiomech.com/http://www.elsevier.com/locate/jbiomechhttp://www.sciencedirect.com/science/journal/00219290
8/17/2019 Estimating Totalkneereplacementjointloadratiosfromkinematics
2/9
Acker et al., 2011; Prins et al., 2011). However, some of these studies
report accuracy of the absolute position of an individual component
in space. When evaluating knee kinematics, it is the relative pose
between components that is of interest. Of single-plane studies
which have evaluated relative TF motion, errors of 0.2–1.3 mm and
1.1–1.51 have been reported in internal–external (I–E) and anterior–
posterior (A–P) motions (Banks and Hodges, 1996; Mahfouz et al.,
2003; Prins et al., 2011; Acker et al., 2011). Dual-plane systems
demonstrate improved accuracy over single-plane systems, withaccuracy of relative motion reported as 0.1–0.3 mm and 0.1–0.41 in
all DOFs (Li et al., 2004; Short et al., 2005; Hanson et al., 2006;
Bingham and Li 2006; Kaptein et al., 2007; Torry et al., 2011; Zhu and
Li, 2012). Kinematic information provides valuable insight into the
performance of different TKR component designs, but does not
describe the loads and interface mechanics at the joint.
Prior researchers have sought to establish relationships bet-
ween joint loads and joint kinematics. D’Lima et al. (2011)
investigated the concept of a unique relationship between implant
position and contact loads. They simultaneously recorded compo-
nent position and contact loads in an AMTI testing machine and
used joint loads to predict kinematics. I–E and A–P joint kine-
matics were predicted with an accuracy of 0.51 and 0.5 mm. The
current study investigates the inverse to this relationship: how
accurately joint load ratios can be predicted from more easily
obtained joint kinematics. A number of prior studies have demon-
strated that compressive and valgus-varus (V –V) loads are highly
sensitive to component position (Lin et al., 2006; Fregly et al.,
2008; Lin et al., 2010a). Fregly et al. (2008) showed that a change
of 70.1 mm in S–I o r V –V position altered S–I and V –V load
predictions by 205% and 77%, respectively. Hence, current kine-
matic measuring systems do not provide enough accuracy to
facilitate direct prediction of loading in these DOFs from kinematic
data. However, Fregly et al. (2008) and Lin et al. (2010a) also
demonstrated that when a combination of kinematics (in insensi-
tive DOFs) and assumed loads (in sensitive DOFs) were applied,
changes in pose of 70.5 mm or 0.51 have minimal changes in
resulting loads. Hence, in the current study, we hypothesize that
for most activities, I–E torque and A–P force were a function of I–E
rotation, A–P translation and compressive force. Specically, the
objective of this study was to assess the accuracy of I–E and A–P
joint load predictions from I–E and A–P motions under a given
compressive load, and to evaluate the consistency of joint load
ratios (I–E torque to compressive force (I–E:C) and A–P force to
compressive force (A–P:C) ratios) for a range of compressive
loading proles. Improved prediction of joint load ratios would
facilitate evaluation of joint mechanics over the range of potential
load ratios likely to be encountered in vivo and provide valuable
information for experimental and computational testing.
2. Methods
In-vivo joint load data, which included 6-DOF joint loads but no kinematic joint
data, was obtained from published tibial tray telemetric data from ve TKR patients
performing three activities of daily living: deep knee bend, chair-rise and step-up
(Kutzner et al., 2010). Knee exion-extension (F–E) was adopted from video
recordings of each patient performing the activity (Bergmann, 2008). These
activities were chosen as activities which are typically measured in uoroscopy
studies as knee position is reasonably stationary and stays within the eld of view
of the uoroscopy system.
A TF FE model was developed which consisted of the same femoral and tibial
components as those implanted in the telemetric patients; this was a cruciate
sacricing design with an ultracongruent tibial insert ( Heinlein et al., 2007). The
femoral component was meshed with rigid triangular shell elements, while the
tibial component was meshed using hexahedral continuum elements. Material
properties have been shown to have minimal effect on kinematics, so for
computational ef ciency, components were modeled as rigid with a pressure–
overclosure relationship (Halloran et al., 2005a; Fitzpatrick et al., 2010 ). A friction
coef cient of 0.04 was assumed between femoral and tibial components ( Godest
et al., 2002; Halloran et al., 2005b). In order to determine corresponding joint
kinematics for the telemetric loading conditions, an FE simulation was initiallyperformed where loads, as measured in the telemetric patients, were applied and
the resulting kinematics were recorded. The femoral component was kinematically
constrained in all 6-DOF, while 5-DOF loads (compressive, A–P, medial–lateral
(M–L), I–E, and V –V) and F–E kinematics were applied to the tibial insert ( Fig. 1).
The resulting I–E and A–P joint motions during the activity were recorded. This
analysis was carried out for each subject performing each of the three activities.
This analysis resulted in a matched set of joint kinematics and loads, equivalent to
data obtained from simultaneous measurement of telemetric loads and uoro-
scopic kinematics in the experimental setting, but without any kinematic
measurement error.
In order to evaluate the ability of the model to predict joint loads from
kinematics that could be measured from uoroscopy, the analysis was subse-
quently repeated by applying only compressive joint force and reproducing I –E, A–
P and F–E joint motions, while the remaining DOFs were unconstrained (Fig. 1).
Resulting I–E and A–P joint loads were compared to the original telemetric loads to
assess how accurately I–E and A–P joint loads could be predicted from applied
kinematics. While V –
V motions were thought to be too subtle to be measured withsuf cient accuracy uoroscopically, quantifying the effect of V –V motion on joint
loads was still of interest. Additional simulations were performed whereby V –V
motions were also included in the analyses; compressive load, I–E, A–P, V –V and
F–E motions were applied in the analyses and predicted I–E, A–P and V –V joint
loads were compared to the original telemetric loads. Compressive load is too
sensitive to superior–inferior (S–I) position to be accurately predicted from
kinematic joint measurements; hence this study focused on prediction of joint
load ratios rather than absolute joint load magnitudes. I–E:C and A–P:C load ratios
Fig. 1. TF model showing loading conditions for the telemetric (left) and kinematically-driven (right) analyses; applied loads shown in red, applied motions shown in green.
C.K. Fitzpatrick, P.J. Rullkoetter / Journal of Biomechanics 47 (2014) 3003– 30113004
8/17/2019 Estimating Totalkneereplacementjointloadratiosfromkinematics
3/9
were calculated for each simulation. To evaluate the consistency of joint load ratio
prediction under different compressive loads, further analysis was performed
whereby the same I–E and A–P joint motions were simulated, with a variety of
compressive force proles. Specically, compressive forces were held constant at
1000 N, 2000 N, and 3000 N, with F–E, I–E and A–P motions kinematically
prescribed and all other DOFs unconstrained for each subject during each activity.
Joint loads and joint load ratios (I–E:C and A–P:C) were compared for different
compressive force analyses.
There is uncertainty in the estimation of joint kinematics from experimental
studies. Single plane uoroscopy studies have reported accuracy of 0.51 for
rotations and 0.5 mm for in-plane translations (S–
I and A–
P), which may be largerstill when the relative pose between components, rather than a single component
position in space, is the measurement of interest (Banks and Hodge, 1996; Hoff et
al., 1998; Dennis et al., 2003; Mahfouz et al., 2003; Fregly et al., 2005; Prins et al.,
2010; Acker et al., 2011; Prins et al., 2011). Thus far, the analysis performed in this
study has assumed ideal kinematics (zero measurement error). To evaluate how
well the method outlined in the current study would perform with real experi-
mental kinematic data, rather than ideal computational data, error in kinematic
measurement was introduced into the analyses. In order to assess the effect of
measurement error on joint load prediction, I–E and A–P motions were shifted by
70.51 and 70.5 mm, and 71.01 and71.0 mm, respectively, in separate analyses.
Resulting I–E and A–P joint loads and joint load ratios were compared to analyses
with the original joint kinematics.
3. Results
Comparing telemetric I–E and A–P joint loads with model
predictions based on I–E and A–P motions and compressive loads,
root-mean-square (RMS) differences in I–E torque and A–P force
across subjects averaged 1.27 Nm and 31.0 N for deep knee bend,
1.12 Nm and 31.4 N for chair-rise, and 1.99 Nm and 68.3 N for step-
up activities, respectively. RMS differences in I–E:C and A–P:C joint
load ratios across subjects averaged 0.90e–3 Nm/N and 0.022 N/N
for deep knee bend, 0.89e–3 Nm/N and 0.027 N/N for chair-rise,
and 1.10e–3 Nm/N and 0.035 N/N for step-up activities, respec-
tively (Table 1; Figs. 2 and 3). When V –V motions were also
included in the analyses, RMS differences in I–E torque and A–P
force across subjects improved to 0.34 Nm and 6.1 N for deep knee
bend, 0.18 Nm and 4.9 N for chair-rise, and 0.45 Nm and 10.4 N for
step-up activities, respectively. RMS differences in I–E:C and A–P:C joint load ratios across subjects improved to 0.24e–3 Nm/N and
0.006 N/N for deep knee bend, 0.25e-3 Nm/N and 0.008 N/N for
chair-rise, and 0.38e–3 Nm/N and 0.015 N/N for step-up activities,
respectively. V –V:C joint loads ratios were predicted with an RMS
accuracy of 0.002e–3, 0.003e–3, 0.004e–3 Nm/N for deep knee
bend, chair-rise, and step-up activities, respectively (Fig. 4).
Comparing joint load ratios for a variety of compressive
proles, I–E:C and A–P:C joint load ratios were consistently
reproduced regardless of compressive force prole. When I–E
and A–P motions were applied for different compressive force
proles (telemetric, constant 1000 N, constant 200 0 N and con-
stant 3000 N), RMS differences in I–E:C and A–P:C ratios for all
constant proles compared to the telemetric prole were less than
0.28e–3 Nm/N and 0.010 N/N for deep knee bend, 0.53e–3 Nm/N
and 0.008 N/N for chair-rise, and 0.22e–
3 Nm/N and 0.010 N/N forstep-up activities, respectively, for all subjects (Fig. 5).
When measurement error was introduced into the kinematic
data, small changes in joint loads were observed when knee
exion was greater than 251. At knee exion greater than 251,
the worst RMS differences in I-E:C and A-P:C ratios for all
kinematic error conditions evaluated (shifts of 70.51 in I-E rota-
tions and70.5 mm in A-P translations) averaged 1.20e–3 Nm/N
and 0.049 N/N for deep knee bend, 1.34e–3 Nm/N and 0.048 N/N
for chair-rise, and 1.88e–3 Nm/N and 0.080 N/N for step-up activ-
ities. In early exion, small changes in kinematics resulted in a
large shift in the point of contact between femoral and tibial
components and as a result had a substantial impact on joint load
predictions; at low exion (less than 251) kinematic data were
unable to accurately predict joint loads (Table 2; Fig. 6).
4. Discussion
Historically, a combination of rigid body musculoskeletal mod-
els and detailed FE models has been used to predict in vivo joint
loads during dynamic activity (Kim et al., 2009; Shelburne et al.,
2005; Taylor et al., 2004). These studies require an extensive
clinical dataset (whole body kinematics, ground reaction forces,
EMG, muscle strength, dynamic imaging) and optimization simu-
lations which can take multiple days to complete (Kinney et al.,
2013; Fregly et al., 2012). Recently, the availability of measured
in vivo joint loads from patients with telemetric implants has
provided validation data for these models. Overall, medial and
lateral contact forces have been predicted with RMS accuracy in
the order of 150–300 N (Guess et al., 2014; Kinney et al., 2013; Kimet al., 2009; Lin et al., 2010b). These studies have primarily focused
on prediction of total, medial and lateral contact forces, rather
than shear forces and torques (A–P force and I–E torque) which are
important to implant design assessment, for example, in the
development of realistic loading conditions for experimental wear
and micromotion simulations. In addition, the volume of clinical
Table 1
RMS differences in joint loads and joint load ratios between telemetric measurements (taken from Kutzner et al., 2010) and model (with I–E and A–P kinematics and
compressive load) predictions for each subject during deep knee bend (top), chair-rise (center), and step-up (bottom) activities.
Patient # IE torque (Nm) IE:C load ratio (Nm/N) (e-3) AP force (N) AP:C load ratio (N/N)
1 0.40 0.26 10.7 0.006
2 0.77 0.78 32.8 0.036
3 0.23 0.18 34.1 0.018
4 4.11 2.91 47.7 0.035
5 0.82 0.40 29.7 0.015
Mean7SD 1.2771.61 0.9071.15 31.0713.3 0.02270.013
1 0.75 0.56 14.3 0.010
2 0.74 0.80 33.2 0.032
3 0.26 0.69 30.5 0.040
4 3.54 2.17 44.8 0.029
5 0.31 0.22 33.9 0.022
Mean7SD 1.1271.37 0.8970.75 31.4711.0 0.02770.011
1 2.45 0.98 90.2 0.033
2 3.29 1.77 84.8 0.038
3 1.33 0.98 96.9 0.050
4 2.13 1.19 40.9 0.036
5 0.74 0.59 28.8 0.018
Mean7SD 1.9970.99 1.1070.44 68.3731.1 0.03570.012
C.K. Fitzpatrick, P.J. Rullkoetter / Journal of Biomechanics 47 (2014) 3003– 3011 3005
8/17/2019 Estimating Totalkneereplacementjointloadratiosfromkinematics
4/9
data and computational resources required for long-running
optimizations limits the number of patient-specic simulations
which may be performed. In order to develop population-based
simulations to assess device performance across the entire TKR
patient population, a larger patient sample, in conjunction with
computationally ef cient simulations, is required.
The current study utilizes a simple TF nite element modelwhich simulates an activity cycle on a single CPU in approximately
six minutes. FE model predictions with imposed I–E and A–P
motions and compressive force demonstrated excellent agreement
in both trend and magnitude with telemetric I–E torque and A–P
force measurements, with RMS accuracy in prediction of I–E
torque and A–P force of less than 2 Nm and 70 N across three
activities. There were some portions of the analysis, notably the
step-up activity at approximately 80% of the cycle, which showed
large errors in I–E torque predictions (Fig. 2). These discrepancies
between telemetric measurements and model predictions
occurred when V –V torque from the telemetric data was large,
and exion angle was low. The curvature of the contacting
surfaces, in combination with a large V –V torque (compared to
the models with neutral V –
V torque), resulted in substantial
differences in I–E torque. When V –V motions were also applied
in the analyses, RMS accuracy in prediction of I–E torque and A–P
force improved to less than 0.5 Nm and 11 N across the three
activities. However, overall V –V motions were small (o21),
resulting from the curvature of the tibial implant rather than
condylar lift-off, and the differences in V –V motions between
these sets of analyses (V –V motion unconstrained, and V –V motion kinematically prescribed) were in the order of 0.11, which
is likely too subtle to be accurately measured with current
uoroscopy systems. Analyses where V –V motions were not
included in the simulations are perhaps a better reection of joint
load accuracy that can realistically be expected from clinically
obtained data with current uoroscopy measurement accuracy.
In the rst set of analyses, compressive force measured from the
telemetric patients was applied to the models. Unfortunately,
changes in relative TF superior–inferior position in the order of
microns have a large effect on compressive force (Fregly et al., 2008),
so S–I kinematic measurements are not accurate enough to facilitate
prediction of compressive force. Therefore, this study focused on
joint load ratios, rather than absolute joint loads – the hypothesis of
this study being that although compressive force may vary, I–
E and
Fig. 2. Comparison of I–E and A–P joint loads and joint load ratios (I–E:C and A–P:C) from telemetric loads and kinematic model predictions for one representative subject
for each activity (deep knee bend (DKB), chair-rise and step-up). Each representative subject was chosen as the subject with RMS differences closest to the mean values for
that activity.
C.K. Fitzpatrick, P.J. Rullkoetter / Journal of Biomechanics 47 (2014) 3003– 30113006
8/17/2019 Estimating Totalkneereplacementjointloadratiosfromkinematics
5/9
A–P to compressive load ratios remain relatively consistent for a
specic relative position. Our analysis conrmed this; for all tele-
metric patients and activities, when a different compressive forceprole was applied with the same set of I–E and A–P kinematics, I–E:
C and A–P:C joint load ratios were consistent regardless of applied
compressive force. More detailed analysis of joint loading conditions
such as compressive force, muscle force or muscle co-contraction
predictions would require more complex analyses such as tradition
musculoskeletal and optimization simulations.
In this study, perfect kinematics were initially assumed in the
predictive model when compared to the telemetric patient simu-
lations. In reality, there is measurement error associated with the
experimental process of obtaining joint kinematics. Accuracy of
single-plane uoroscopy systems is in the order of 0.5–1.01 and
0.5–1.0 mm (Banks and Hodge, 1996; Hoff et al., 1998; Mahfouz
et al., 2003; Dennis et al., 2003; Fregly et al., 2005; Prins et al.,
2010; Acker et al., 2011; Prins et al., 2011). This was similar to
errors reported in a complementary study from D’Lima et al. (2011)
where telemetric joint loads were used to predict I–E and A–P joint
kinematics with an accuracy of 0.51 and 0.5 mm. In the current study,the effect of measurement error was introduced by varying applied I–E
and A–P kinematics by 70.5 and 1.01 and 70.5 and 1.0 mm. In later
exion (after 251 of TF exion), errors in joint loads were small and
consistent. However, in early exion, joint load predictions varied
widely from the telemetric measurements. This was due to high
congruency between femoral and tibial components in early exion; a
small change in joint kinematics resulted in a large change in the point
of contact between the components (Fig. 7). The implant geometry
used in this study was an ultra-congruent xed-bearing design;
implant designs with lower conformity between components would
be more forgiving of kinematic measurement error. In addition, data
from dual-plane uoroscopy systems would reduce the measurement
error associated with motion tracking, facilitating more accurate joint
load ratio predictions.
Fig. 3. Comparison of average (light) and 71 standard deviation (dark) joint load ratios (I–E:C and A–P:C) from telemetric loads and kinematic model predictions across all
subjects for each activity (deep knee bend (DKB), chair-rise and step-up).
C.K. Fitzpatrick, P.J. Rullkoetter / Journal of Biomechanics 47 (2014) 3003– 3011 3007
8/17/2019 Estimating Totalkneereplacementjointloadratiosfromkinematics
6/9
No initial implant position information for the telemetricimplants was available. In this study, tibial posterior slope was
assumed to be zero, and knee exion was adopted from video
recordings, which is less accurate than uoroscopy measurement.
Overall, implant alignment does not impact the results, as a
perfectly matched set of kinematics and kinetics has been devel-
oped. An initial simulation in 5-DOF load control was performed to
determine the corresponding kinematics to this combination of
loads, implant design and implant alignment. Simulations include
a variety of patients and activities, which start in different initial
positions and under different loading conditions, with all showing
similarly accurate results in joint load ratio predictions. In practice,
when uoroscopy data rather than model data is used to predict
joint load ratios, the uoroscopy data with provide initial implant
information, reducing or eliminating this source of uncertainty.
In order for experimental and computational simulations toprovide data that is reective of the clinical performance of TKR
devices, these simulations must be representative of the loading
conditions encountered in vivo. Telemetric tibial trays have been
implanted in TKR patients and have provided valuable information
regarding in vivo TF joint loading; however, the number of patients
for which data is available is limited (Kutzner et al., 2010; D’Lima
et al., 2011). Initial results from this study demonstrate the
potential applicability of using widely available kinematic data to
predict joint load ratios. Kinematic data has been collected from
thousands of TKR patients. If this kinematic data could be
converted into joint load information, loading proles for experi-
mental and computational simulators may be greatly enhanced to
evaluate device performance under mean loading conditions, but
perhaps more importantly, under outlying loading conditions
Fig. 4. Comparison of I–E, A–P and V –V joint loads ratios from telemetric loads and kinematic model predictions for one representative subject for each activity (deep knee
bend (DKB), chair-rise and step-up). Left: model simulations applied compressive force, I–E and A–P motions. Right: model simulations included the addition of V –V motions
(applied compressive force, I–E, A–P and V –V motions).
C.K. Fitzpatrick, P.J. Rullkoetter / Journal of Biomechanics 47 (2014) 3003– 30113008
8/17/2019 Estimating Totalkneereplacementjointloadratiosfromkinematics
7/9
Fig. 5. Comparison of joint loads and joint load ratios predicted by the kinematic model for a variety of compressive force proles: the original telemetric prole, and
constant 1000 N, 2000 N, and 3000 N proles for each activity (deep knee bend (DKB), chair-rise and step-up). Shown for one representative subject for each activity. Each
representative subject was chosen as the subject with RMS differences closest to the mean values for that activity.
Table 2
RMS differences in joint loads and joint load ratios between telemetric measurements (taken from Kutzner et al., 2010) and model (with I–E and A–P kinematics and
compressive load) predictions with error imposed on kinematics (7 0.51 and70.5 mm shift (top) and711 and71 mm shift (bottom) in A–P and I–E motions, respectively)for each activity. Showing mean and standard deviations of the worst kinematic condition for the ve subjects.
Activity IE torque (Nm) IE:C load ratio (Nm/N) (e-3) AP force (N) AP:C load ratio (N/N)
Deep knee bend: full 5.5272.15 4.8571.52 405.27219.9 0.38070.175
Deep knee bend:4251 1.8071.64 1.2071.14 77.3712.9 0.04970.015
Chair-rise: full 10.4072.00 8.8072.49 739.67353.0 0.61570.320
Chair-rise:4251 1.6172.05 1.3471.08 59.4723.1 0.04870.010
Step-up: full 14.0275.22 7.1473.66 513.27278.8 0.29570.164
Step-up:4251 1.6970.26 1.8870.37 85.4725.7 0.08070.028
Deep knee bend: full 6.0972.27 5.3171.42 442.07229.3 0.40970.170
Deep knee bend:4251 2.3171.57 1.4871.14 123.8721.7 0.07770.018
Chair-rise: full 10.2671.97 8.8772.95 786.47361.2 0.64470.315
Chair-rise:4251 1.9572.06 1.5371.10 89.4728.1 0.07070.013
Step-up: full 14.7875.14 7.3673.41 594.77252.2 0.34470.138
Step-up:4251 2.0870.31 2.0970.40 129.1732.7 0.11370.039
C.K. Fitzpatrick, P.J. Rullkoetter / Journal of Biomechanics 47 (2014) 3003– 3011 3009
8/17/2019 Estimating Totalkneereplacementjointloadratiosfromkinematics
8/9
occurring in patients that are likely most vulnerable to complica-
tions and failure. While compressive load may be dif cult to
measure in vivo, computational models can effectively determine
I–E:C and A–P:C load ratios, and absolute I–E torques and A–P
forces can easily and ef ciently be calculated over a design-of-
experiments (DOE) or probabilistic range of compressive force
Fig. 6. Comparison of joint load ratios (I–E:C and A–P:C) from telemetric loads and kinematic model predictions with 70.51 and 70.5 mm (shaded dark) and711
and71 mm (shaded light) of kinematic error in I–E and A–P motions for one representative subject for each activity (deep knee bend (DKB), chair-rise and step-up). Top:
variation in joint load ratio predictions due to kinematic error for the entire activity; bottom: variation in joint load ratio predictions due to kinematic error only shown for TF
exion greater than 251.
Fig. 7. Change in contact location between femoral and tibial components with a 0.51 shift in I–E kinematics. Due to congruency between the components, small changes in
kinematics resulted in a large shift in contact location in early exion (top), while kinematics differences in later exion, when components were less conforming, resulted in
only small changes in contact location (bottom). Sagittal images show a cut-through the mid-point of the lateral condyle.
C.K. Fitzpatrick, P.J. Rullkoetter / Journal of Biomechanics 47 (2014) 3003– 30113010
8/17/2019 Estimating Totalkneereplacementjointloadratiosfromkinematics
9/9
proles. The prevalence of kinematic data, in conjunction with the
methods applied in this study, facilitates determining the scope of
A–P a n d I–E loading ratios experienced by the TKR patient
population.
Conict of interest statement
One of the authors (PJR) is a consultant to DePuy Synthes, Inc.
Acknowledgment
This work was supported in part by DePuy Synthes, a Johnson &
Johnson Company.
References
Acker, S., Li, R., Murray, H., John, P.S., Banks, S., Mu, S., Wyss, U., Deluzio, K., 2011.Accuracy of single-plane uoroscopy in determining relative position andorientation of total knee replacement components. J. Biomech. 44, 784–787.
Banks, S.A., Hodge, W.A., 1996. Accurate measurement of three-dimensional kneereplacement kinematics using single-plane uoroscopy. IEEE Trans. Biomed.Eng. 43, 638–649.
Bergmann, G., (Ed.), 2008. Charite-Universitaetsmedizin Berlin (2008) “
Orthoload”
.Retrieved July 1, 2011 from 〈http://Orthoload.com〉.Bingham, J., Li, G., 2006. An optimized image matching method for determining in-
vivo TKA kinematics with a dual-orthogonal uoroscopic imaging system. J. Biomech. Eng. 128, 588–595.
Dennis, D.A., Komistek, R.D., Mahfouz, M.R., Haas, B.D., Stiehl, J.B., 2003. Multicenterdetermination of in vivo kinematics after total knee arthroplasty. Clin. Orthop.Relat. Res. 416, 37–57.
D’Lima, D.D., Fregly, B.J., Patil, S., Steklov, N., Colwell Jr., C.W., 2012. Knee jointforces: prediction, measurement, and signicance. Proc. Inst. Mech. Eng. H 226,95–102.
D’Lima, D.D., Patil, S., Steklov, N., Colwell Jr., C.W., 2011. 'Lab ’-in-a-knee: in vivoknee forces, kinematics, and contact analysis. Clin. Orthop. Relat. Res. 469,2953–2970.
Fitzpatrick, C.K., Baldwin, M.A., Rullkoetter, P.J., 2010. Computationally ef cientnite element evaluation of natural patellofemoral mechanics. J. Biomech. Eng.132, 121013.
Fregly, B.J., Rahman, H.A., Banks, S.A., 2005. Theoretical accuracy of model-basedshape matching for measuring natural knee kinematics with single-plane
uoroscopy. J. Biomech. Eng. 127, 692–699.Fregly, B.J., Banks, S.A., D’Lima, D.D., Colwell Jr., C.W., 2008. Sensitivity of knee
replacement contact calculations to kinematic measurement errors. J. Orthop.Res. 26, 1173–1179.
Fregly, B.J., Besier, T.F., Lloyd, D.G., Delp, S.L., Banks, S.A., Pandy, M.G., D ’Lima, D.D.,2012. Grand challenge competition to predict in vivo knee loads. J. Orthop. Res.30, 503–513.
Godest, A.C., Beaugonin, M., Haug, E., Taylor, M., Gregson, P.J., 2002. Simulation of aknee joint replacement during a gait cycle using explicit nite element analysis.
J. Biomech. 35, 267–275.Guess, T.M., Stylianou, A., Kia, M., 2014. Concurrent prediction of muscle and
tibiofemoral contact forces during treadmill gait. J. Biomech. Eng. 136, 021032.
Halloran, J.P., Easley, S.K., Petrella, A.J., Rullkoetter, P.J., 2005a. Comparison of deformable and elastic foundation nite element simulations for predictingknee replacement mechanics. J. Biomech. Eng. 127, 813–818.
Halloran, J.P., Petrella, A.J., Rullkoetter, P.J., 2005b. Explicit nite element modelingof total knee replacement mechanics. J. Biomech. 38, 323–331.
Hanson, G.R., Suggs, J.F., Freiberg, A.A., Durbhakula, S., Li, G., 2006. Investigation of in vivo 6DOF total knee arthoplasty kinematics using a dual orthogonaluoroscopic system. J. Orthop. Res. 24, 974–981.
Heinlein, B., Graichen, F., Bender, A., Rohlmann, A., Bergmann, G., 2007. Design,calibration and pre-clinical testing of an instrumented tibial tray. J. Biomech.40 (1), S4–10.
Hoff, W.A., Komistek, R.D., Dennis, D.A., Gabriel, S.M., Walker, S.A., 1998. Three-dimensional determination of femoral-tibial contact positions under in vivoconditions using uoroscopy. Clin. Biomech. 13, 455–472.
Kaptein, B.L., Valstar, E.R., Stoel, B.C., Reiber, H.C., Nelissen, R.G., 2007. Clinicalvalidation of model-based RSA for a total knee prosthesis. Clin. Orthop. Relat.Res. 464, 205–209.
Kim, H.J., Fernandez, J.W., Akbarshahi, M., Walter, J.P., Fregly, B.J., Pandy, M.G., 2009.Evaluation of predicted knee-joint muscle forces during gait using an instru-mented knee implant. J. Orthop. Res. 27, 1326–1331.
Kinney, A.L., Besier, T.F., D’Lima, D.D., Fregly, B.J., 2013. Update on grand challengecompetition to predict in vivo knee loads. J. Biomech. Eng. 135, 021012.
Kutzner, I., Heinlein, B., Graichen, F., Bender, A., Rohlmann, A., Halder, A., Beier, A.,Bergmann, G., 2010. Loading of the knee joint during activities of daily livingmeasured in vivo in ve subjects. J. Biomech. 43, 2164–2173.
Li, G., Wuerz, T.H., DeFrate, L.E., 2004. Feasibility of using orthogonal uoroscopicimages to measure in vivo joint kinematics. J. Biomech. Eng. 126, 314 –318.
Lin, Y.C., Farr, J., Carter, K., Fregly, B.J., 2006. Response surface optimization for jointcontact model evaluation. J. Appl. Biomech. 22, 120–130.
Lin, Y.C., Haftka, R.T., Queipo, N.V., Fregly, B.J., 2010a. Surrogate articular contactmodels for computationally ef cient multibody dynamic simulations. Med.Eng. Phys. 32, 584–594.
Lin, Y.C., Walter, J.P., Banks, S.A., Pandy, M.G., Fregly, B.J., 2010b. Simultaneousprediction of muscle and contact forces in the knee during gait. J. Biomech. 43,945–952.
Lloyd, D.G., Besier, T.F., 2003. An EMG-driven musculoskeletal model to estimatemuscle forces and knee joint moments in vivo. J. Biomech. 36, 765 –776.
Mahfouz, M.R., Hoff, W.A., Komistek, R.D., Dennis, D.A., 2003. A robust method forregistration of three-dimensional knee implant models to two-dimensionaluoroscopy images. IEEE Trans. Med. Imaging 22, 1561–1574.
Prins, A.H., Kaptein, B.L., Stoel, B.C., Reiber, J.H., Valstar, E.R., 2010. Detecting femur-insert collisions to improve precision of uoroscopic knee arthroplasty analysis.
J. Biomech. 43, 694–700.Prins, A.H., Kaptein, B.L., Stoel, B.C., Nelissen, R.G., Reiber, J.H., Valstar, E.R., 2011.
Integrated contour detection and pose estimation for uoroscopic analysis of knee implants. Proc. Inst. Mech. Eng. H 225, 753–761.
Shelburne, K.B., Torry, M.R., Pandy, M.G., 2005. Muscle, ligament, and joint-contactforces at the knee during walking. Med. Sci. Sports Exerc. 37, 1948–1956.
Short, A., Gill, H.S., Marks, B., Waite, J.C., Kellett, C.F., Price, A.J., O ’Connor, J.J.,Murray, D.W., 2005. A novel method for in vivo knee prosthesis wearmeasurement. J. Biomech. 38, 315–322.
Taylor, W.R., Heller, M.O., Bergmann, G., Duda, G.N., 2004. Tibio-femoral loadingduring human gait and stair climbing. J. Orthop. Res. 22, 625–632.
Torry, M.R., Myers, C., Pennington, W.W., Shelburne, K.B., Krong, J.P., Giphart, J.E.,Steadman, J.R., Woo, S.L., 2011. Relationship of anterior knee laxity to kneetranslations during drop landings: a bi-plane uoroscopy study. Knee Surg.Sports Traumatol. Arthrosc. 19, 653–662.
Zhu, Z., Li, G., 2012. An automatic 2D-3D image matching method for reproducingspatial knee joint positions using single or dual uoroscopic images. Comput.Methods Biomech. Biomed. Eng. 15, 1245–1256.
C.K. Fitzpatrick, P.J. Rullkoetter / Journal of Biomechanics 47 (2014) 3003– 3011 3011
http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref1http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref1http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref1http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref1http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref1http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref1http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref1http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref1http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref2http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref2http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref2http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref2http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref2http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref2http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref2http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref2http://orthoload.com/http://orthoload.com/http://orthoload.com/http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref3http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref3http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref3http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref3http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref3http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref3http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref3http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref3http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref4http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref4http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref4http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref4http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref4http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref4http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref5http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref5http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref5http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref5http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref5http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref5http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref5http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref5http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref5http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref5http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref6http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref6http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref6http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref6http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref6http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref6http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref6http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref6http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref6http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref6http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref7http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref7http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref7http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref7http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref7http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref7http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref7http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref8http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref8http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref8http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref8http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref8http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref8http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref8http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref9http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref9http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref9http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref9http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref9http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref9http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref9http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref9http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref10http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref10http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref10http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref10http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref10http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref10http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref10http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref10http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref11http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref11http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref11http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref11http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref11http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref11http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref11http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref11http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref905http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref905http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref905http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref12http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref12http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref12http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref12http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref12http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref12http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref12http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref12http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref13http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref13http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref13http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref13http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref13http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref13http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref13http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref14http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref14http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref14http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref14http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref14http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref14http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref14http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref15http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref15http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref15http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref15http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref15http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref15http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref16http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref16http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref16http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref16http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref16http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref16http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref16http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref16http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref17http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref17http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref17http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref17http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref17http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref17http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref18http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref18http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref18http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref18http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref18http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref18http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref19http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref19http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref19http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref19http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref19http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref20http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref20http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref20http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref20http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref20http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref20http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref20http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref20http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref21http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref21http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref21http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref21http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref21http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref21http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref21http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref22http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref22http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref22http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref22http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref22http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref23http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref23http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref23http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref23http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref23http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref23http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref23http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref23http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref24http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref24http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref24http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref24http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref24http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref24http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref25http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref25http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref25http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref25http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref25http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref26http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref26http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref26http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref26http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref26http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref26http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref26http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref27http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref27http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref27http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref27http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref27http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref27http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref27http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref27http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref28http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref28http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref28http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref28http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref28http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref28http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref28http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref28http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref29http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref29http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref29http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref29http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref29http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref30http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref30http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref30http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref30http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref30http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref30http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref30http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref30http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref31http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref31http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref31http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref31http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref31http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref32http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref32http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref32http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref32http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref32http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref32http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref32http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref32http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref32http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref33http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref33http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref33http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref33http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref33http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref33http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref33http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref33http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref33http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref33http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref33http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref32http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref32http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref32http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref32http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref31http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref31http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref30http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref30http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref30http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref29http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref29http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref28http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref28http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref28http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref27http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref27http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref27http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref26http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref26http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref26http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref25http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref25http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref24http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref24http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref24http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref23http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref23http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref23http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref22http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref22http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref21http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref21http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref20http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref20http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref20http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref19http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref19http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref18http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref18http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref18http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref17http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref17http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref17http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref16http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref16http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref16http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref15http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref15http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref15http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref14http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref14http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref14http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref13http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref13http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref12http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref12http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref12http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref905http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref905http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref11http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref11http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref11http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref10http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref10http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref10http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref9http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref9http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref9http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref8http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref8http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref8http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref7http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref7http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref7http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref6http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref6http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref6http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref5http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref5http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref5http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref4http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref4http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref4http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref3http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref3http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref3http://orthoload.com/http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref2http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref2http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref2http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref1http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref1http://refhub.elsevier.com/S0021-9290(14)00382-0/sbref1