Joint Instability

How Can We Use Computational

Modeling to Improve Total Knee
Arthroplasty? Modeling Stability and
Mobility in the Implanted Knee

Paul J. Rullkoetter, PhD
Clare K. Fitzpatrick, PhD
Chadd W. Clary, PhD
Downloaded from https://journals.lww.com/jaaos by BhDMf5ePHKav1zEoum1tQfN4a+kJLhEZgbsIHo4XMi0hCywCX1AWnYQp/IlQrHD3Gamkn0m7hy6RQdEwn41RO+bmxdcZEjIhP+pWdHXXIw4= on 06/02/2018
Abstract
Validated computational models promise a virtual platform to create optimal articular
surfaces that best achieve desired implant characteristics. Today, designers can
parametrically define the primary geometric features of an implant, and automatically
modify design variables until stability/mobility performance objectives are best

achieved. This preclinical, virtual design iteration minimizes the development cycle
compared with testing physical prototypes and, by evaluating a broader scope of
design concepts, likely improves the clinical performance of the final product.
However, the scenario described is not without shortcomings and requires thorough
understanding of the capabilities and the limitations of the models used. Although
models typically represent the articular interface well, the interaction with the patient
and the surgical process includes significant variability and increase in complexity. We
present current modeling capabilities for the estimation of implant stability/mobility,
with further suggestions for answering the difficult question of how an implant might
perform throughout the population.

operative interaction between articular geometry,

From the Department of Mechanical
and Materials Engineering, University
of Denver, Denver, CO (Dr. Rullkoetter,
Dr. Fitzpatrick, and Dr. Clary).
Dr. Rullkoetter or an immediate family
member serves as a paid consultant
to and has received research or
institutional support from DePuy
Synthes. Dr. Fitzpatrick and Dr. Clary
have listed no disclosures.
Introduction
Computational modeling can improve knee ar-
throplasty through preclinical design optimization
and by advancing insight into the mechanics of the
implanted knee. Validated models can be used to
quickly evaluate many design concepts without
manufacturing physical prototypes. In essence, the
designer identifies the primary features of a new
implant, while the computer automatically iterates
design parameters until an optimal combination is
reached that best fits defined requirements.
Models also provide insight into measures not
easily obtained experimentally, such as contact
mechanics, interface shear loading, or muscle
force requirements during a dynamic activity.
Models should be thought of as complementary
tools to be integrated with, and validated against,
physical cadaveric experiments, in vivo measure-
ment of patient function, and retrieval analysis.
Model validation and good modeling practices are
a significant component of any computational
work. Truly, in-depth understanding of the
passive ligamentous constraint, and muscle con-
tribution is not well defined.
Joint stability is a critical component of suc-
cessful total knee arthroplasty (TKA); up to 25%
of patients remain dissatisfied with the outcome
of their TKA procedure, with knee instability
remaining a common complaint during high-
demand activities and muscle cocontraction fre-
quently used as a compensatory technique.1–4 In
this article, we present some of the current
computational work available to evaluate joint
stability/laxity and mobility during dynamic
activities. Although not a comprehensive liter-
ature review, the modeling progression from
simple implant-only assessments to complex
activity simulation demonstrates the current
opportunity in design development using vali-
dated computational models that can estimate
implant performance over a population. Cur-
rent traditional TKA does not re-create the
simultaneous stability and mobility of the
healthy knee, so opportunities exist for
improvement through computational design.

J Am Acad Orthop Surg 2017;25 capabilities and limitations of any model is

(suppl 1):S33-S39
DOI: 10.5435/JAAOS-D-16-00640
Copyright 2016 by the American
Academy of Orthopaedic Surgeons.
required before use in practice. Current modeling
can represent the articular surface interaction
quite well, but the interfaces with the patient and
interactions with the surgical process are more
complex and variable. Hence, representing the
implant is straightforward, while the post-
Models of Stability/Mobility
Implanted stability is a result of complex
interaction between implant design and align-
ment, compressive loading, postoperative

February 2017, Vol 25, Supplement 1 S33

Copyright ª the American Academy of Orthopaedic Surgeons. Unauthorized reproduction of this article is prohibited.
How Can We Use Computational Modeling to Improve Total Knee Arthroplasty?

Figure 1

Experimental tibiofemoral laxity/stability testing (A), finite element model representation of AP and internal–external (IE) testing with
constant compressive load (B), and model validation demonstrating successful estimation of AP and IE behavior for two implants (C).

Figure 2

A, Finite element model of tibiofemoral joint with posterior-stabilized implant and passive soft-tissue constraints. B, Compressive force required by the
total knee arthroplasty (TKA) to match natural knee stability at full extension under a tibial torque of 4.9 Nm with ligaments (solid bars) and without
ligaments (dashed bars). (Panel B reproduced with permission from Navacchia A, Rullkoetter PJ, Schütz P, List RB, Fitzpatrick CK, Shelburne KB:
Subject-specific multiscale modeling of muscle force and knee contact in total knee arthroplasty. J Orthop Res 2016;34(9):1576–1587.)

S34 Journal of the American Academy of Orthopaedic Surgeons

Copyright ª the American Academy of Orthopaedic Surgeons. Unauthorized reproduction of this article is prohibited.
 Paul J. Rullkoetter, PhD, et al

ligamentous state, and patient’s neuro-

Figure 3
musculoskeletal factors. Clinical instability is
difficult to define, but likely is either excessive
motion due to joint laxity or a high velocity/
acceleration event felt by the patient during a
dynamic activity.5 Instability may lead to a lack
of patient confidence and/or hamstrings–
quadriceps cocontraction to increase the com-
pressive load across the joint and improve
instantaneous stability. Implant designs con-
tribute to joint stability through articular surface
constraint. A variety of experiments and com-
putational models has measured the mechanical
constraint of the components in isolation (with
no soft tissue); however, many in vivo factors
contribute to joint stability that are not accounted
for in these simple simulations. Hence, more
complex models have been developed to repre-
sent passive soft tissue, musculature, and loading
conditions during dynamic activities to better
evaluate stability in the preclinical environment.

Traditional Laxity Assessment

Mechanical test standards exist for the traditional
evaluation of implant laxity (eg, ASTM F1223)
that are also performed computationally.6,7 These
implant-only experiments are typically per-
formed in a tension–torsion servohydraulic test
frame with custom fixtures (Figure 1). A constant
compressive load is applied through the femoral
component, and force–displacement or torque–
rotation data are measured as a function of
flexion angle. Computational models have been
developed to re-create the experimental condi-
tions, and validation achieved through compar-
ison with experimental data (Figure 1). Such
models can successfully estimate laxity data and
appropriately differentiate changes in implant
geometry. Although a simple assessment, the
data provide a reasonably complete estimate of
tibiofemoral (TF) stability of the isolated implant,
and these models are currently used in the
Musculoskeletal model to estimate muscle and joint loading conditions (A), finite element
development process.
model to evaluate joint stability during dynamic activity (B), and applied AP and internal–
external (IE) loading used to evaluate instantaneous stability during activity (C). D, Stability
Updated Stability Tests: evaluation was performed at key instants during activity, shown here for gait.
Determining Compressive Load
Required to Reproduce Natural
Knee Stability to enhance TF joint stability after TKA.9–14 by estimating compressive load requirements
Experimental studies have used traditional Benedetti et al reported a high level of necessary to achieve natural knee stability.
laxity assessment to compare TF stability quadriceps–hamstrings cocontraction in the Data from experimental cadaveric stability
characteristics of different implant designs stance phase of gait in patients with low tests of the natural knee were obtained from the
absent in any soft tissue, reporting substantial conformity TKA 2 years after surgery.11–12 literature.16,17 In the experimental tests, a
differences in AP and internal–external (IE) Lunderberg et al showed a similar finding; constant compressive load was applied to the
range of motion.6,8 However, interpreting a prolonged cocontraction in TKA subjects TF joint while a cyclical IE and then AP load
how the resulting differences in millimeters or compared with a healthy group.13 Mitchell were applied, and then resulting IE and AP
degrees directly impact patient function is et al15 suggested that intrinsic TKA stability motions were measured. A finite element
difficult. Adaptation of muscle recruitment may provide efficient muscle recruitment model of the implanted TF joint was devel-
patterns and forces to counteract sensations of similar to that observed with healthy controls. oped including the following primary liga-
instability is commonly reported in clinical Because the goal for TKA is to restore the ments crossing the joint (Figure 2): posterior
electromyography studies, with increased natural mechanics of the TF joint, we aim to capsule, medial collateral ligament, lateral
quadriceps–hamstrings cocontraction shown evaluate stability in a more physiological way collateral ligament, antero-lateral structure,

February 2017, Vol 25, Supplement 1 S35

Copyright ª the American Academy of Orthopaedic Surgeons. Unauthorized reproduction of this article is prohibited.
How Can We Use Computational Modeling to Improve Total Knee Arthroplasty?
Figure 4
Total muscle force requirements during gait at contralateral toe off (top) and mid-flexion
chair rise (bottom) during AP (A and B) and internal–external (IE) (C and D) stability testing
for each component (solid lines) and the baseline implant device for comparison. Insets
show relative tibiofemoral (TF) position at peak applied loads.
popliteofibular ligament, and posterior cruci-
ate ligament (with cruciate-retaining designs),
previously calibrated to reproduce measured
postoperative knee constraint.18
Subsequently, similar stability tests as the
experiment were repeated in the computational
model, but the applied compressive load was
allowed to automatically adapt so that the
same IE and AP load–displacement stability
curves measured during the cadaveric stability
tests were achieved. These simulations were
performed for different TKA implant designs
with varying levels of conformity. The com-
pressive load required for each implant was
compared with the compressive force applied
in the natural knee cadaveric tests between
implant designs, implant types (cruciate-
S36
Copyright ª the American Academy of Orthopaedic Surgeons. Unauthorized reproduction of this article is prohibited.
retaining or posterior-stabilized), and models
tested with and without ligament structures.
As anticipated, lower conformity implants
required a higher compressive force than
higher conformity geometries, regardless of
flexion angle (Figure 2). The lower confor-
mity designs required as much as two-thirds
more compressive force than that of higher
conformity designs to maintain stability
equivalent to that of the natural knee.
Simulations performed in IE with and
without soft tissues did not always show
substantial differences in the required
compressive force to match natural stability,
demonstrating that the articular surfaces are
the main contribution to joint stability.
Further modeling could be performed to
Journal of the American Academy of Orthopaedic Surgeons
determine the impact of various balance
states (eg, tight versus loose or a different
percentage of medial-lateral load distribu-
tion) on compressive loading requirements,
gaining further insight into the overall
contributions from passive soft tissue and
implant design to stability.
Muscle Force Requirements for
Stability: Comparison of
Current Implants
As the previous analysis demonstrated, com-
pressive loading requirements to produce an
equivalent level of stability as the natural knee
vary significantly among current implants.
The next step in modeling evolution is to
determine how additional compressive force
could be created through increased muscle
forces. Such an analysis would evaluate
the muscle force requirements necessary to
maintain a specified level of instantaneous
stability during key instants during dynamic
activities, for example, estimating the muscle
forces required during the weight acceptance
portion of stepdown that will maintain a
chosen level of AP stability, and comparing
predicted muscle force requirements for a
variety of current implants.
A musculoskeletal model was developed
from in vivo patient data to estimate muscle
forces for the lower limb during chair rise, gait,
and step-down activities.19,20 A computa-
tional finite element model of the implanted
lower limb (that included a detailed knee
joint, extensor mechanism, TF ligaments, and
13 muscles spanning the hip, knee, and ankle)
was virtually implanted with the same
implant as the patient was studied in vivo
(Figure 3). Instantaneous AP and IE stability
evaluations were performed at key instants
during each activity with the activity-based
muscle forces estimated from the musculo-
skeletal model applied. The model was sub-
sequently implanted with three different TKA
designs. The stability tests were repeated;
however, in these simulations, muscle forces
were allowed to adapt through cocontraction
to reproduce the same level of instantaneous
stiffness or stability of the initial design.
Muscle force requirements were compared
among designs.
A large variation in muscle forces was
found among implants, with greater differ-
ences in AP than IE tests. Differences in muscle
forces across implants were highly dependent
on activity/loading conditions (Figure 4),
although a similar ranking of forces was
demonstrated across activities tests. Given
that patients with TKA are susceptible to
muscle weakness, a likely benefit exists in
implants that reduce the amount of work the
lower limb musculature must do to maintain
 Paul J. Rullkoetter, PhD, et al

Figure 5

Experimental Kansas knee simulator (A), dynamic finite element model representation of the simulator (B), and details of the implanted knee
model (C).

Figure 6

Medial and lateral condylar low-point kinematic data for traditional dual-radius implant (A) and gradually reducing radius device (B).
Experimental curves include mean of eight cadaver tests with error bars of one SD and highlight the start of the anterior slide on the medial
condyle at the transition between femoral radii.

a stable joint. This work has provided com- equivalent levels of stability, but further during dynamic activities to provide stability
parative implant data on the amount of load research is needed to better determine the targets for implants to reproduce optimal
required through cocontraction to maintain level of stability in the healthy natural knee patient function.

February 2017, Vol 25, Supplement 1 S37

Copyright ª the American Academy of Orthopaedic Surgeons. Unauthorized reproduction of this article is prohibited.
How Can We Use Computational Modeling to Improve Total Knee Arthroplasty?

Figure 7 Figure 8

Updated lower limb finite element model with

the telemetric implant showing external
loading and boundary conditions. I–E =
internal–external.
Case Study: Addressing
Mid-Flexion Instability With
Virtual Design Iteration and
Validation
Perhaps, one of the most useful applications
of computational tools to model clinical
instability started with stability testing of a
traditional dual sagittal femoral radius
design.21 The analysis showed that stability
instantly dropped at the abrupt transition
from the larger distal femoral radius to the
smaller posterior radius. From this obser-
vation, it was hypothesized that an anterior
femoral slide (high relative velocity insta-
bility) at mid-flexion that had been observed
in clinical fluoroscopy studies was a result
of this sudden reduction from the distal
to posterior femoral radii. It was also
hypothesized that replacing the instanta-
neous radius reduction with a gradually
reducing radius would attenuate the para-
doxical slide. Finally, it was hypothesized
that the kinematic changes associated with
enhancement of the femoral “J-curve” would
be moderated by the sagittal conformity of the
tibial insert, with less-conforming inserts ex-
hibiting more condylar motion. To evaluate
these hypotheses, a combined experimental and
computational approach was used building on
Finite element model during stance-phase gait and step-down cycles (top), and
comparison of target profiles with joint loads achieved (bottom). Insets show contact
pressure distribution on the tibial insert of a right knee at instants during the dynamic
cycles.
the previous work using the Kansas knee sim-
ulator (KKS).22 The KKS is a six–degree-of-
freedom electrohydraulic mechanical simulator
used to evaluate mechanics of the knee in
natural and implanted cadaveric specimens.
Loads applied at the hip, ankle, and quadriceps
tendon create specific loading conditions at the
knee-simulating dynamic activities like a deep
knee bend. Previous experimental and compu-
tational work focused on developing and vali-
dating a finite element representation of the
KKS and implanted knee (Figure 5).18,23 The
computational model was used as an iterative
tool to predict kinematic differences due to
geometric modifications of femoral and tibial
articular surfaces. The experimental rig was
used for verification purposes. Eight cadaveric
specimens were implanted with a traditional
TKA, and a new implant system embodying
geometric enhancements was studied in the
model to verify both the predictive ability of the
model and the kinematic effects of the modified
TKA geometry (Figure 6). The iterative design
process was validated in that the computa-
tionally estimated kinematics were well repro-
duced with the subsequent cadaveric testing.
Beyond Simulation of
Simulators: Modeling Dynamic
Activities of Daily Living
Even the most complex physical simulators have
limitations when attempting to reproduce
dynamic loading conditions common in activi-
ties of daily living. To better reproduce experi-
mentally measured in vivo joint mechanics for
complex activities, an enhanced simulation was
developed, extending the model’s capabilities
beyond the experimental setup (Figure 7).24
Additional loading was incorporated to apply/
measure IE torque directly about the long axis
of the tibia, allowing AP hip motion. The
hamstrings muscle (semimembranosus, semite-
ndinosus, long and short heads of the biceps
femoris) was added, and the quadriceps muscle
was divided into four muscles (addition of
vastus lateralis and vastus medialis) using the
centroid of each muscle as the muscle path.
These measures allowed for a physiological
estimation of muscle forces required for a given
activity.
Direct in vivo measurement of TF joint loads
from patients with a telemetric implant was

S38 Journal of the American Academy of Orthopaedic Surgeons

Copyright ª the American Academy of Orthopaedic Surgeons. Unauthorized reproduction of this article is prohibited.

used to develop external loading conditions for
several patients performing a variety of activi-
ties.25 The telemetric implant geometry was
used, and loading conditions were developed
at the hip, ankle, and muscles that reproduced
TF joint loading for stance-phase gait, step-
down, squat, and chair rise.24 Subsequently,
the impact of geometry on joint mechanics (eg,
kinematics, contact mechanics, and soft-tissue
strains) can be virtually evaluated using
loading representing dynamic activities of
daily living (Figure 8).
Further Considerations
Modeling of the implanted knee has pro-
gressed significantly over the last decade. A
suite of models is available for evaluating
TKA stability and mobility characteristics,
enabling rapid evaluation of design concepts
or existing products. Primary limitations of
the current state-of-the-art include an over-
simplified passive soft-tissue representation
and a dearth of data from the postoperative
balance condition. The increased use of in-
traoperative sensors will hopefully provide
supporting in vivo data and establish balanc-
ing targets.26 These models should facilitate
an improved understanding of the relative
contribution to stability from passive con-
straint, implant design and alignment, and
joint loading. Modeling will also be further
focused by an understanding of the link
between joint mechanics and patient function
and satisfaction.27 This will only happen
through holistic collaborative studies including
clinical data, patient measurement, model-
ing, and retrieval analysis. As implants are
available that used preclinical computational
prediction, we are in the unique position of
closing the loop with clinical/retrieval data to
evaluate the quality and predictive power of
the models used. Finally, although we have a
series of dynamic loading conditions to
evaluate the impact of geometry on joint
mechanics, the modeling community has not
yet tackled how a patient may adapt because
of instability or pain to optimally perform a
task. As modeling capabilities grow in com-
plexity and supporting clinical data (eg,
telemetry and intraoperative sensors) become
widely available to validate these models, it is
evident that future TKA designs will be
optimized using preclinical computational
modeling.
February 2017, Vol 25, Supplement 1
Copyright ª the American Academy of Orthopaedic Surgeons. Unauthorized reproduction of this article is prohibited.
Acknowledgment
This work was funded in part by DePuy
Synthes, Inc., and NIH NIBIB (grant
R01EB015497).
References
1. Dawson J, Fitzpatrick R, Murray D,
Carr A: Questionnaire on the perceptions of
patients about total knee replacement. J Bone Joint
Surg Br 1998;80:63–69.
2. Noble PC, Gordon MJ, Weiss JM, Reddix RN,
Conditt MA, Mathis KB: Does total knee
replacement restore normal knee function? Clin
Orthop Relat Res 2005;(431):157–165.
3. Nam D, Nunley RM, Barrack RL:
Patient dissatisfaction following total knee
replacement: A growing concern? Bone Joint J
2014;96-B(11 suppl A):96–100.
4. Mizner RL, Snyder-Mackler L: Altered loading
during walking and sit-to-stand is affected by
quadriceps weakness after total knee arthroplasty.
J Orthop Res 2005;23:1083–1090.
5. Wright TM: Joint stability in total knee
arthroplasty: What is the target for a stable knee?
J Am Acad Orthop Surg 2017;25(suppl 1):S25–S28.
6. Haider H, Walker PS: Measurements of constraint of
total knee replacement. J Biomech 2005;38(2):
341–348.
7. Moran MF, Bhimji S, Racanelli J, Piazza SJ:
Computational assessment of constraint in total
knee replacement. J Biomech 2008;41:2013–2020.
8. Sathasivam S, Walker PS: A computer model with
surface friction for the prediction of total knee
kinematics. J Biomech 1997;30:177–184.
9. Fallah-Yakhdania HRA, Abbasi-Bafghi H,
Meijer OG, et al: Determinants of co-contraction
during walking before and after arthroplasty for
knee osteoarthritis. Clin Biomech 2012;27:
485–494.
10. Davidson BS, Judd DL, Thomas AC, Mizner RL,
Eckhoff DG, Stevens-Lapsley JE: Muscle activation
and coactivation during five-time-sit-to-stand
movement in patients undergoing total knee
arthroplasty. J Electromyogr Kinesiol 2013;23:
1485–1493.
11. Benedetti MG, Bonato P, Catani F, et al:
Myoelectric activation pattern during gait in total
knee replacement: Relationship with kinematics,
kinetics, and clinical outcome. IEEE Trans Rehabil
Eng 1999;7:140–149.
12. Benedetti MG, Catani F, Bilotta TW, Marcacci
M, Mariani E, Giannini S: Muscle activation
pattern and gait biomechanics after total knee
replacement. Clin Biomech (Bristol, Avon) 2003;
18:871–876.
13. Lunderberg HL, Rojas IL, Foucher KC, Wimmer
MA: Comparison of antagonist muscle activity
Paul J. Rullkoetter, PhD, et al
during walking between total knee replacement
and control subjects using unnormalized
electromyography. J Arthroplasty 2016;31(6):
1331–1339.
14. Thomas AC, Judd DL, Davidson BS, Eckhoff DG,
Stevens-Lapsley JE: Quadriceps/hamstrings co-
activation increases early after total knee
arthroplasty. Knee 2014;21:1115–1119.
15. Mitchell KaB S, Rawlins J, Wood SA, Hodge WA:
Strength of intrinsically stable TKA during stair-
climbing. Presented at the 51st Annual Meeting of
the Orthopaedic Research Society. Washington,
DC, May 20–23, 2005.
16. Hsieh HH, Walker PS: Stabilizing mechanisms of
the loaded and unloaded knee joint. J Bone Joint
Surg Am 1976;58:87–93.
17. Ahmed AM, Hyder A, Burke DL, Chan KH: In-vitro
ligament tension pattern in the flexed knee in passive
loading. J Orthop Res 1987;5:217–230.
18. Baldwin MA, Clary CW, Fitzpatrick CK, Deacy JS,
Maletsky LP, Rullkoetter PJ: Dynamic finite
element knee simulation for evaluation of knee
replacement mechanics. J Biomech 2012;45:
474–483.
19. Navacchia A, Rullkoetter PJ, Schütz P, List RB,
Fitzpatrick CK, Shelburne KB: Subject-specific
multiscale modeling of muscle force and knee
contact in total knee arthroplasty. J Orthop Res
2016;34(9):1576–1587.
20. Navacchia A, Myers CA, Rullkoetter P,
Shelburne K: Prediction of in vivo knee joint loads
using a global probabilistic analysis. J Biomech
Eng 2016;138(3):4032379.
21. Clary CW, Fitzpatrick CK, Maletsky LP,
Rullketter P: The influence of tibiofemoral
geometry on mid-flexion Stability: An
experimental and finite element study. J Biomech
2013;46:1351–1357.
22. Maletsky LP, Hillberry BM: Simulating dynamic
activities using a five-axis knee simulator. J Biomech
Eng 2005;127:123–133.
23. Halloran J, Clary C, Maletsky L, Taylor M, Petrella
A, Rullkoetter P: Verification of predicted knee
replacement kinematics during simulated gait in the
Kansas knee simulator. J Biomech Eng 2010;132:
081010.
24. Fitzpatrick CK, Baldwin MA, Clary CW,
Maletsky LP, Rullkoetter PJ: Evaluating knee
replacement mechanics during ADL with PID-
controlled dynamic finite element analysis.
Comput Methods Biomech Biomed Engin 2014;
17:360–369.
25. Kutzner I, Heinlein B, Graichen F, et al: Loading of
the knee joint during activities of daily living
measured in vivo in five subjects. J Biomech 2010;
43:2164–2173.
26. D’Lima DD, Colwell CW: Intraoperative
measurements and tools to assess stability. J Am
Acad Orthop Surg 2017;25(suppl 1):S29–S32.
27. Banks SA: What postoperative outcome measures
link joint stability to patient satisfaction? J Am
Acad Orthop Surg 2017;25(suppl 1):S40–S43.
S39