Gait & Posture 66 (2018) 107–113

Contents lists available at ScienceDirect

Gait & Posture

journal homepage: www.elsevier.com/locate/gaitpost

The effect of center of pressure alteration on the ground reaction force T

during gait: A statistical model
⁎
Hadar Shauliana, , Deborah Solomonow-Avnona, Amir Hermanb, Nimrod Rozenc, Amir Haima,d,
Alon Wolfa
a
Biorobotics and Biomechanics Lab (BRML), Faculty of Mechanical Engineering, Technion-Israel Institute of Technology, 32000, Haifa, Israel
b
Department of Orthopedic Surgery A, Chaim Sheba Medical Center, Ramat-Gan, Israel
c
Department of Orthopedic Surgery, Ha’Emek Medical Center, Center, Afula, Israel
d
Department of Orthopedic Surgery B, Sourasky Medical Center, Tel Aviv, Israel

A R T I C LE I N FO A B S T R A C T

Keywords: Background: Foot problems and lower-limb diseases (e.g., foot ulcers, osteoarthritis, etc.), are presented with a
Ground reaction force ground reaction force (GRF) that may deviate substantially from the normal. Thus, GRF manipulation is a key
Foot center of pressure parameter when treating symptoms of these diseases. In the current study, we examined the impact of footwear-
Gait analysis generated center of pressure (COP) manipulations on the GRF components, and the ability to predict this impact
Footwear-generated biomechanical
using statistical models.
manipulations
Methods: A foot-worn biomechanical device which allows manual manipulation of the COP location was uti-
lized. Twelve healthy young men underwent gait analysis with the device set to convey seven COP conditions:
(1) a neutral condition, (2) lateral and (3) medial offset along the medio-lateral foot axis, (4) anterior and (5)
posterior offset along the antero-posterior foot axis, and (6) a dorsi-flexion and (7) plantar-flexion condition.
Changes in the magnitude and the early stance-phase impulse of the GRF components across COP conditions
were observed. Linear models were used to describe relationships between COP conditions and GRF magnitude
and impulse.
Results: With respect to ANOVA, the vertical and antero-posterior components of the GRF were significantly
influenced by the COP configuration throughout the different stages of the stance-phase, whereas the medio-
lateral components were not. The models of vertical, antero-posterior and medio-lateral GRF components were
statistically significant.
Significance: The study results are valuable for the development of a method and means for efficient treatment of
foot and lower-limb pathologies. The ability to predict and control the GRF components along three orthogonal
axes, for a given COP location, provides a strong tool for efficient treatment of foot and lower-limb diseases and
may also have relevant implications in sports shoe design. This study is a preliminary investigation for our
ultimate goal to develop an effective treatment method by developing an autonomous GRF manipulations device
based on closed-loop feedback.

1. Introduction
During gait, the foot is the point of physical contact between the
environment and the body. As a result, the force acting on the foot,
namely the three-dimensional ground reaction force (GRF), can si-
multaneously impact and be impacted by pathological disorders.
Disorders such as degenerative diseases of the joints (e.g., osteoar-
thritis), injury, or foot problems (e.g., foot ulcers secondary to diabetes
mellitus, plantar fasciitis) are presented with a GRF that may deviate
substantially from the normal [1,2]. Thus, examination of the impact of
external perturbations on the GRF may be of great interest in areas such
as shoe design and footwear orthotics designed to treat symptoms of
lower-limb joint and foot diseases.
The three-dimensional GRF is a reflection of the total mass-times-
acceleration product of all body segments [3]. The origin of the GRF
vector is the foot center of pressure (COP) which is a locus defined as
the average location of all external forces acting between the foot and
the ground [4]. Numerous studies have shown the correlation between
external perturbations applied to the foot during gait and alteration of
lower-limb biomechanics, including kinematics, kinetics, and neuro-

⁎
Corresponding author.
E-mail address: Hadar.sh@campus.technion.ac.il (H. Shaulian).

https://doi.org/10.1016/j.gaitpost.2018.08.013
Received 29 March 2018; Received in revised form 28 July 2018; Accepted 16 August 2018
0966-6362/ © 2018 Elsevier B.V. All rights reserved.
H. Shaulian et al. Gait & Posture 66 (2018) 107–113

muscular response [5–7]. The GRF, therefore, may be a key component

when treating foot problems and lower-limb diseases.
In assessing impact of the GRF on the lower extremities, the lower
limb is commonly viewed as a system of chained links, using the
standard link-segment model, comprising a functional kinetic unit [8].
Applying the well-accepted method of inverse dynamics, to calculate
moments and loads acting on the lower-limb joints, reveals that the GRF
has far-reaching influence throughout the kinematic chain [9]. Devices
currently used to treat foot and lower-limb pathologies are developed to
optimize foot and joint loads [10,11]. However, in most existing de-
vices the structure enabling modification of biomechanics is fixed, and
thus does not enable GRF adjustments. Those few that enable mod-
ifications are typically adjusted only in low-frequency cycles (i.e., every
few days due to the need of professional manual adjustment) and with
no patient-specific biomechanics indicators. These limitations give rise
to non-precise and uncontrolled fit of the device to the patient, leading
to a potentially long and less effective treatment protocol.
In previous studies from our group, a foot-worn biomechanical de-
vice, composed of two adjustable convex rubber elements attached to
the sole, allowing manipulation of the contact point between the device
and the ground (i.e., COP), was used to manipulate lower-limb joint
biomechanics. Adjustment of the elements impacts the COP trajectory
by causing it to shift correspondingly [10,12], thereby having a direct
effect on lower-limb biomechanics [5,13]. Among these studies are
those exhibiting biomechanical and clinical benefit of specific pertur-
bations in knee OA [14] and hip OA [15] populations, resulting in re-
duction of loads on the pathological joint, increase in functionality and
quality of life, and decrease in pain [16]. The device in its current
construction, however, requires manual calibration and subjective op-
timization of the device element positions by means of observational
gait analysis of a trained physical therapist.
The ultimate goal, commencing with the present study, is to develop
a footwear device for efficient treatment of lower-limb and foot
pathologies by means of autonomous objective GRF manipulation. Our
research group aims to develop a closed-loop feedback footwear device
implementing external alterations of the three-dimensional GRF, and
the ability to predict the impact of these alterations on foot and joint
loading [17].
Several studies have shown that footwear-generated manipulation
of the COP trajectory during gait leads to alteration of the GRF mag-
nitude and orientation [18,19]. However, these studies were limited to
COP modifications in the frontal plane only, and do not report the
precise resulting alterations to the GRF and its three constituent com-
ponents. Therefore, the objective of the present study was to examine
the effect of footwear-generated COP manipulations along the medio-
lateral and antero-posterior foot axes on the three-dimensional GRF
vector components. We hypothesized that COP manipulation sig-
nificantly influences the components of the GRF vector in a predictable
manner that can be statistically modelled, and thus potentially used for
precise footwear design, in general. The current study also serves as a
preliminary investigation to assess the ability to control the GRF during
gait in order to enable construction of the closed-loop feedback au-
tonomous device.

2. Methods

2.1. Participants

Data acquired in a previous study, at the Biorobotics and

Biomechanics Lab of the Faculty of Mechanical Engineering at the
Technion - Israel Institute of Technology, focusing on impact of COP
manipulation on the knee joint during gait [5] was utilized for the (caption on next page)
present analysis. The study group consisted of 12 healthy male subjects
with a similar anthropometric and demographic profile (shoe size =
French 43, dominant leg = right, age = 25.7 ± 2.1 years, height =
177.0 ± 3.8 cm, weight = 73.3 ± 4.9 kg). Subject exclusion criteria

108
H. Shaulian et al.
Fig. 1. Ensemble average of the neutral configuration: [a] vertical, [b] antero-
posterior, and [c] medio-lateral GRF components in units of % body weight.
Characteristic events of the GRF during gait cycle are presented as: FV1 - Peak
vertical GRF at early mid-stance; FV2 - Minimum vertical GRF at late mid-
stance; FV3 -Peak vertical GRF at early pre-swing; Fap1 - Peak posterior GRF at
early mid-stance; Fap2 -Peak anterior GRF at early pre-swing; Fml1 -Peak lateral
GRF at loading response; Fml2 - Peak medial GRF at mid-stance; Fml3 -Minimum
medial at terminal stance; Fml4 -Peak medial at late terminal stance; JV
-Vertical GRF impulse at loading response; Jap - Antero-posterior GRF impulse
at loading response; Jml - Medio-lateral GRF impulse at loading response.
were any orthopedic, musculoskeletal, or neurological pathology. Ap-
proval of the Ethics Sub-Committee was obtained, and informed con-
sent was given by all participants. The purpose and methods of the
study were explained to the subjects.
2.2. The biomechanical system
A foot-worn biomechanical device (APOS - Medical and Sports
Technologies Ltd. Herzliya, Israel) was used for manipulation of COP. A
detailed description of the device was previously reported [5]. The
device consists of two adjustable convex-shaped biomechanical ele-
ments attached to each foot using a platform in the form of a shoe
specially designed with two mounting rails (Fig. 2). Each element po-
sition can be adjusted individually and continuously in the transverse
plane of the foot, essentially shifting the COP in the respective direction
[10,12].
2.3. Experimental protocol
Prior to testing, functional assessment of all subjects was performed
by a single physician. Calibration positioning of the biomechanical
device elements for the neutral configuration was custom-defined and
documented by a single trained physiotherapist. The neutral config-
uration was defined for each subject as the position of the elements in
Fig. 2. Biomechanical device (a) with no elements attached, in (b) neutral, (c) medial, (d) lateral, (e) anterior, (f) posterior, (g) dorsiflexion, and (h) plantar-flexion
configuration (right shoe is shown).
109
Gait & Posture 66 (2018) 107–113
which the apparatus caused the least varus, valgus, plantar and dorsal
torque about the ankle. Prior to data acquisition, subjects were in-
structed to walk at a self-selected velocity for several minutes in order
to become accustomed to the shoes. During this period, the subject-
selected walking speed was indicated by a metronome, which was used
in all subsequent gait trials to ensure consistent cadence across all COP
configurations. After the acclimation period, gait analyses were per-
formed in the neutral configuration, and in the following 3 pairs of
device configurations (COP conditions), according to the device ele-
ment shifts along a specific foot axis:
1 “medial” and “lateral” configurations, defined as 0.8-cm medial and
1.5-cm lateral deviations, respectively, of the biomechanical ele-
ments from the neutral sagittal axis (the axis passing through the
centers of the device elements in the neutral configuration) (Fig. 2c
and d)
2 “anterior” and “posterior” configurations, defined as 1.5 cm anterior
and 1.5 cm posterior deviations, respectively, of the elements from
the neutral configuration along the neutral sagittal axis (Fig. 2e and
f)
3 “dorsi-flexion” and “plantar-flexion” configurations, set by adding
three flat disks, of 2 mm thickness each, between the shoe sole and
the forefoot and rearfoot elements, respectively (Fig. 1g and h)
Testing in each device configuration was performed in random
order on the same day.
2.4. Data acquisition and processing
The GRF vectors were recorded by two 3-dimensional AMTI OR6–7-
1000 force plates placed in tandem in the center of a 10-m walkway, at
a sampling frequency of 960 Hz. All analyses were performed for the
dominant leg.
Three components of the GRF vector (vertical, antero-posterior, and
medio-lateral) throughout the stance phase were analyzed. The sign
H. Shaulian et al.
convention used in this study was positive direction upwards, anterior,
and lateral.
For each subject, each force component was recorded for each of 6
successful trials and normalized by body weight (BW). The neutral
configuration was set as a reference in order to examine the changes in
the GRF as a result of device element shifts relative to the reference.
The magnitude of the GRF component at the stance-phase events
depicted in Fig. 1 were calculated for each device configuration. In
addition, the loading-response impulse, defined as the integral of the
GRF magnitude over time during loading response, was calculated for
all components of the GRF. Impulse was calculated specifically during
loading response since visual inspection of the GRF graphs for all COP
configurations for each subject revealed apparent differences between
COP configurations during the time of early stance. Since there are no
GRF events (maxima or minima) during this time, we chose another
representative parameter (i.e., impulse), such that we could report the
phenomenon that we observed.
In order to confirm that we successfully controlled for speed across
device configurations, we calculated the speed as the stride length di-
vided by stride time. Stride length was calculated using a Vicon motion
analysis system (Oxford Metrics Ltd., Oxford, UK) together with a re-
flective marker placed on the toe (over the second metatarsal head, on
the mid-foot side of the equinus break between the fore-foot and mid-
foot). This was part of a standard marker set used from a previous gait
analysis study from which the data of the current study was taken [5].
Specifically, stride length was calculated as the distance from the ip-
silateral toe marker’s position at the first ipsilateral foot contact to the
ipsilateral toe marker’s position at the second ipsilateral foot contact,
and stride time was calculated as the time between successive ipsi-
lateral foot strikes.
2.5. Statistical analysis
Data analysis was conducted by an experienced biostatistician
(Amir Herman) using R (Vienna, Austria) version 3.2.3. One-way ana-
lysis of variance (ANOVA) was used to confirm that speed was similar
across all COP configurations. Comparison between the means of the
various configurations was done using mixed effect ANOVA. A magni-
tude and shift-direction specific analysis was done using linear mixed
effect models, in which subject was included as a random effect, and the
three device element shift axes as fixed effect variables. In this analysis,
each of the three device element shifts along a foot axis was considered
as a separate group, including medial-lateral (medial minus, lateral
plus), anterior-posterior (anterior plus, posterior minus), and sagittal
tilt (dorsi-flexion plus, plantar-flexion minus). Shift axes variables were
used as continuous variables in which the magnitude (in cm) and di-
rection of the shift were included in the models. All reported p-values
are two-sided. A p-value below 0.05 was considered statistically sig-
nificant.
3. Results
The one-way ANOVA revealed that speed did not change with
change in COP configuration (p = 0.9882). Mean value of gait speed for
each COP configuration is available in the supplemental material [Table
S1]. With respective to the mixed effect ANOVA, the vertical and
antero-posterior components of the GRF were significantly influenced
by the COP configurations throughout the different stages of the stance
phase, whereas the medio-lateral components were not. Mean and
standard deviations of the GRF events pertaining to the vertical and
antero-posterior GRF components for the 7 COP configurations which
were statistically significant are presented in Fig. 3. Mean values and
standard deviations of the GRF at all 9 force events, and the impulse for
the 3 GRF components (Fig. 1) for the 7 COP configurations are pro-
vided as supplemental material [Table S2].
For each of the 9 force events and 3 impulse quantities previously
110
Gait & Posture 66 (2018) 107–113
defined, a linear model was calculated to predict the 3 GRF components
based on the 3 pairs of element adjustments (Eqs. (1) and (2).
GRF = GRFneutral + ∑ GRF _CoefShift *Elements _Shift
Shift (ML, AP , DF − PF ) (1)
Impulse = Impulseneutral + ∑ Impulse _CoefShift
Shift (ML, AP , DF − PF )
*Elements _Shift (2)
Where:
GRF = Magnitude of each of the 9 force events [%BW]
GRFneutral = Force w.r.t. neutral configuration [%BW]
Shift = Categorical variable representing foot axes (medio-lateral/
antero-posterior/ dorsiflexion – plantar-flexion)
GRF _CoefShift = Force coefficient per the respective categorical
variable shift [% BW /cm]
Elements _Shift = Element shifts along the respective foot axis [cm]
Impulse = Magnitude of each of the 3 GRF impulse events
[% BW × % Stance phase ]
Impulseneutral = GRF impulse w.r.t. neutral configuration
[% BW × % Stance phase ]
Impulse _CoefShift = GRF impulse coefficient per the respective ca-
tegorical variable shift [(% BW × % Stance phase )/cm].
Model terms and p-values of the force and impulse are given in
Tables 1 and 2, respectively. Each of the model variables represents the
change from the neutral configuration due to shifting of the bio-
mechanical elements along the respective foot axis. The model coeffi-
cients are the weights of each variable. The larger the absolute coeffi-
cients of a variable, the more influence it has on the event. The sign of
the coefficients multiplied by the device elements shift sign, whose
convention was described in the methods, represents the direction of
the change in force. Thus, in order to evaluate the effect of the device
element shifts relative to the neutral configuration, the total shift of the
element along the respective foot axis (with the proper sign) should be
multiple by the shift coefficient and summed with the value of the
neutral configuration. The p-value of each model represents the accu-
racy with which the model predicts the GRF magnitude for the given
event. In general, all models of the vertical, antero-posterior and medio-
lateral GRF components were statistically significant.
Analysis of the significance of each variable suggests that in loading
response (0–20% of stance phase) there is a good fit of the model for all
GRF components and all COP configurations. In mid-stance (20–51% of
stance phase) and terminal-stance (51–83% of stance phase) the ante-
rior-posterior shift is the only shift that had significant impact on all
GRF elements, while in pre-swing (83–100% of stance phase) only the
medio-lateral shift impacted all GRF events.
4. Discussion
In accordance with our hypothesis, we were able, with high statis-
tical significance, to model and predict the behavior of the 3 GRF
components throughout the entire stance phase, as well as impulse
during early stance, in response to footwear-generated COP manipula-
tions. The models show that several COP configurations impact certain
gait event parameters more than others. Specifically, antero-posterior
device element deviations impact 89% of the force events while medio-
lateral and dorsiflexion-plantar-flexion deviations each affect 33% of
the 9 force events. The models provide a continuous description of the
behavior of the GRF components, as they predict the force along each
axis for any magnitude of device element shifts, and they enable pre-
dicting the GRF for a combination of device element deviations along
several foot axes simultaneously.
Previous investigations from our research group show that when the
COP is shifted along the antero-posterior or medio-lateral foot axis,
kinetics and kinematics of all joints in the lower limbs are significantly
altered [5,10,13]. We speculate that the observed kinetic results are the
H. Shaulian et al. Gait & Posture 66 (2018) 107–113

Fig. 3. Mean and standard deviation of the GRF events that were significant for the 7 cop configurations; FV1: Vertical GRF component at the early mid-stance peak,
FV2: Vertical GRF component at the mid-stance minimum, FV3: Vertical GRF component at the early pre-swing peak, FAP1: Antero-posterior GRF component at early
mid-stance peak, FAP2: Antero-posterior GRF component at early pre-swing peak, JV: Vertical GRF impulse at loading response, JAP: Antero-posterior GRF impulse
at loading response.

outcome of changes in the kinematics and a specific neuromuscular
response, stimulated by the biomechanical device manipulations and
the induced instability. For instance, results from a previous study by
Haim at al [10]. show that in early stance phase, with the posterior
configuration, the ankle plantar-flexion was significantly increased
compared to neutral. Increasing ankle plantar-flexion may lead to a
decrease in the stride length with posterior and plantar-flexion con-
figurations [20]. Consequently, we suggest that the decrease in sagittal-
plane hip excursion causes a reduction in the magnitude of the antero-
posterior shear GRF. This would directly lead to the decrease in the
impulse along the antero-posterior axis (Jap). This phenomenon can be
seen clearly from the model which significantly predicts a decrease in

Table 1
Linear model terms for the 9 force events according to Eq. (1).
Event Neutral ML – shift AP – shift PF-DF shift Model p-value
configuration coefficients coefficients coefficients

Fv1 105.37 (p = 0) 0.27 (p = 0.465) −2.13 (p = 0) −2.43 (p = 0.001) 0.000

Fv 2 74.94 (p = 0) −0.34 (p = 0.167) 0.47 (p = 0.016) 0.45 (p = 0.350) 0.037
Fv3 105.62 (p = 0) −0.53 (p = 0.012) 0.19 (p = 0.275) 0.40 (p = 0.326) 0.037
Fap1 −15.93(p = 0) 0.24 (p = 0.249) 0.37 (p = 0.022) −0.71 (p = 0.809) 0.022
Fap2 16.51 (p = 0) 0.24 (p = 0.026) 0.67 (p = 0) −0.21 (p = 0.326) 0.000
Fml1 2.40 (p = 0) −2.04 (p = 0) −0.88 (p = 0) −0.91 (p = 0) 0.000
Fml2 −3.89 (p = 0) 0.003 (p = 0.975) −0.32 (p = 0) −0.15 (p = 0.467) 0.001
Fml3 −1.27 (p = 0) −0.14 (p = 0.115) −0.26 (p = 0.0003) −0.40 (p = 0.024) 0.0001
Fml4 −4.34 (p = 0) −0.19 (p = 0.073) −0.17 (p = 0.045) −0.29 (p = 0.163) 0.028

Each force event represented as: Fv1 - Peak vertical GRF at early mid-stance; Fv2 - Minimum vertical GRF at late mid-stance; Fv3 - Peak vertical GRF at early pre-
swing; Fap1 - Peak posterior GRF at early mid-stance; Fap2 - Peak anterior GRF at early pre-swing; Fml1 - Peak lateral GRF at loading response; Fml2 - Peak medial GRF
at mid-stance; Fml3 - Minimum medial GRF at terminal stance; Fml4 - Peak medial GRF at late terminal stance. The value of each force event for the neutral
configuration is the intercept, while each device element shift coefficient represents the weight of each element shift variable compared to neutral. The p-value of
each model represents the accuracy with which the model predicts the GRF magnitude for the given event. The p-value of each device element shift coefficient
represents the accuracy with which the coefficient predicts the GRF magnitude for the given event and given device element shift.

111
H. Shaulian et al. Gait & Posture 66 (2018) 107–113

Table 2 pedobarographic measurement system. Another gait study [10] invol-

Linear model terms for the 3 GRF impulse parameters according to Eq. (2). ving shift of the device elements along the antero-posterior foot axis,
Event Neutral ML – shift AP – shift PF-DF shift Model exhibited shift of the initial point of the GRF vector, measured as lo-
configuration coefficients coefficients coefficients p- cation coordinates on force plates, in relation to fixed reflective markers
value attached to the shoe. Although the study did not calculate the actual
COP of the foot, the calculations represented the COP of the device as a
Jv 562.80 −15.80 79.98 (p = 0) 83.73 (p = 0) 0.0000
(p = 0) (p = 0.001)
whole, and it is assumed that the COP of the foot is shifted in direct
Jap −55.36 3.68 −16.66 (p = 0) −17.96 (p = 0) 0.0000 proportion. It should however be noted that the actual COP shift re-
(p = 0) (p = 0.014) lative to the foot was not measured in this study. Additionally, for
Jml 10.98 (p = 0) −7.05 −4.01 (p = 0) −7.10 (p = 0) 0.0000 simplicity, we used a linear model for this preliminary investigation.
(p = 0)
Although model results were statistically significant, it is possible that
Each impulse event represented as: Jv - Vertical GRF impulse at loading re- more complex relationships exist between GRF magnitude and COP
sponse; JA - Antero- posterior GRF impulse at loading response; Jml : Medio- locations. Finally, it should be emphasized that the current study results
lateral GRF impulse at loading response. The value of each GRF impulse event pertain to a distinctive homogeneous cohort. These results are therefore
for the neutral configuration is the intercept, while each device element shift valid only for individuals with characteristics similar to those of the
coefficient represents the weight of each variable compared to neutral. The p- tested group. Different populations may respond differently to such
value of each model represents the accuracy with which the model predicts the interventions.
GRF impulse for the given event. The p-value of each device element shift Our long-term goal will be accomplished by developing an auton-
coefficient represents the accuracy with which the coefficient predicts the GRF omous footwear device alongside a predictive model to enable real-
impulse for the given event and given device element shift.
time, continuous GRF manipulations for efficient treatment of foot and
lower-limb pathologies. The results of the current study provide a
the impulse during loading response for the antero-posterior compo-
preliminary foundation for this goal. Upon further research in patho-
nent ( Jap ). Haim et al. [10] also reported a significant increase in knee-
logical populations, predicting the GRF components for a given COP
flexion-angle in early stance-phase with the posterior configurations.
condition may give an insight into a specific pathology's biomechanical
Previous studies have suggested that the increase in knee flexion leads
response to manipulating COP, and could enable using the GRF as a
to a more beneficial position of the antigravity muscles to absorb ki-
feedback parameter to control the foot and lower-extremity bio-
netic energy at impact, thereby reducing the magnitude of the antero-
mechanics to facilitate treatment. For example, several studies exhibit a
posterior and vertical GRF components [21,22]. This is another possible
correlation between the medio-lateral GRF component and the knee
explanation for the decrease in the impulse along the antero-posterior
lever arm [23–25]. Thus, the ability to modify the medio-lateral GRF
axis (Jap) and vertical axis (Jv ) which was observed in our models.
component in a predictable manner potentially enables reducing the
We did not observe significant changes in the medio-lateral com-
knee adduction moment, which is positively correlated with loads in
ponent of the GRF, in agreement with previous studies evaluating
the knee [7,14], in knee OA patients by shortening the knee level arm.
changes in the medio-lateral GRF component due to COP modifications
The study results may also offer clinically relevant implications in other
[23–25]. Since the medio-lateral GRF component is the component with
medical-shoe design areas. For example, people with diabetes mellitus
the greatest time-domain variance across walking trials [26], we sug-
are at high risk for foot ulceration or infection, sometime leading to the
gest that the impact of frontal-plane COP manipulation on this com-
need for amputation, which are exacerbated by shearing forces [27].
ponent may have been masked by large standard deviations for each set
Many researches have shown that specialized off-loading orthopedic
of walking trials in each test condition. Further, COP manipulations
shoes have substantial influence on the prevention and healing process
may impact the medio-lateral GRF of each subject individually, but not
of neuropathic ulcers [28,29]. Our results may help build a foundation
in a homogeneous pattern for the cohort as a whole. Although the
for designing a more effective foot-worn device to reduce the shearing
ANOVA was not significant for the medio-lateral GRF, the linear model
forces, as they include a composite analysis of the 3 GRF components
significantly predicted a change in the medio-lateral GRF component as
throughout the different stages of the stance phase. Additionally, our
result of several COP configurations, and thus medio-lateral GRF could
study may have relevance in sports shoe design for prevention of ath-
potentially be controlled by a specific combination of COP deviations
letic injuries. Podraza et al. [21] and Boden et al. [30] suggest that
not examined in this study.
propagation of the vertical and antero-posterior GRF components when
Several limitations arising from the current study should be noted.
landing with the knee near full extension implies a greater risk of non-
First, testing was performed a short time after the biomechanical device
contact ACL injuries. The ability to reduce those forces may serve as a
was first worn by the participants. Continuous usage of the device could
beneficial factor in preventing ACL injuries in certain athletes.
influence the outcome since it may lead to substantial gait adaptations.
In addition, the subjective positioning of the neutral configuration was
Conflicts of interest
defined as a control, rather than, for instance, the apparatus with no
elements attached, or a regular shoe. A curved bottom, as in the neutral
None.
and 6 other COP configurations, rather than flat bottom, as in a regular
shoe, presents an unstable contact interface. The relatively stable con-
Acknowledgment
dition of a shoe without the convex elements therefore presents an
entirely different walking condition with less demand on the neuro-
The authors thank Apos Medical and Sports Technologies Ltd. for
muscular system. However, since parameters were compared within
their generosity in contributing the devices used in the study.
subjects, and the only change in the experimental setup was the re-
positioning of the elements, a small error in the positioning of the
Appendix A. Supplementary data
neutral configuration by the physical therapist would not likely affect
the general study outcomes. Previous studies of our group demon-
Supplementary data associated with this article can be found, in the
strated that a shift in device elements causes a corresponding direc-
online version, at https://doi.org/10.1016/j.gaitpost.2018.08.013.
tional shift in the COP for medio-lateral and antero-posterior foot axes
[10,12]. Khoury et al. [12] used linear models to exhibit the relation-
References
ship between the device element manipulations along the medio-lateral
foot axis and COP shift, measured directly using an in-shoe
[1] I.A.F. Stokes, I.B. Faris, W.C. Hutton, The neuropathic ulcer and loads on the foot in

112
H. Shaulian et al.
diabetic patients, Acta Orthop. 46 (5) (1975) 839–847.
[2] A. Hreljac, R.N. Marshall, P.A. Hume, Evaluation of lower extremity overuse injury
potential in runners, Med. Sci. Sport. Exerc. 32 (9) (2000) 1635–1641.
[3] Da. Winter, Kinematic and kinetic patterns in human gait: variability and com-
pensating effects, Hum. Mov. Sci. 3 (1–2) (1984) 51–76.
[4] K.J. Chesnin, L. Selby-Silverstein, M.P. Besser, Comparison of an in-shoe pressure
measurement device to a force plate: concurrent validity of center of pressure
measurements, Gait Posture 12 (2) (2000) 128–133.
[5] A. Haim, N. Rozen, S. Dekel, N. Halperin, A. Wolf, Control of knee coronal plane
moment via modulation of center of pressure: a prospective gait analysis study, J.
Biomech. 41 (14) (2008) 3010–3016.
[6] D. Solomonow-Avnon, D. Levin, M. Elboim-Gabyzon, N. Rozen, E. Peled, A. Wolf,
Neuromuscular response of hip-spanning and low back muscles to medio-lateral
foot center of pressure manipulation during gait, J. Electromyogr. Kinesiol. 28
(2016) 53–60.
[7] S.J. Crenshaw, F.E. Pollo, E.F. Calton, Effects of lateral-wedged insoles on kinetics
at the knee, Clin. Orthop. Relat. Res. 375 (2000) 185–192.
[8] F.E. Zajac, R.R. Neptune, S.A. Kautz, Biomechanics and muscle coordination of
human walking: part I: introduction to concepts, power transfer, dynamics and si-
mulations, Gait Posture 16 (3) (2002) 215–232.
[9] Roy H. Lidtke, Carol Muehleman, Mary Kwasny, J.A. Block, Foot center of pressure
and medial knee osteoarthritis, J. Am. Podiatr. Med. Assoc. 100 (3) (2010)
178–184.
[10] A. Haim, N. Rozen, A. Wolf, The influence of sagittal center of pressure offset on
gait kinematics and kinetics, J. Biomech. 43 (2010) 969–977.
[11] A.J.M. Boulton, H. Connor, P.R. Cavanagh, A.J.M. Boulton, Andrew J.M. Boulton,
Henry Connor, Peter R. Cavanagh (Eds.), The Foot in Diabetes, 2000.
[12] M. Khoury, A. Wolf, E.M. Debbi, A. Herman, A. Haim, Foot center of pressure
trajectory alteration by biomechanical manipulation of shoe design, Foot Ankle Int.
34 (4) (2013) 593–598.
[13] D. Solomonow-Avnon, A. Wolf, A. Herman, N. Rozen, A. Haim, Reduction of
frontal-plane hip joint reaction force via medio-lateral foot center of pressure ma-
nipulation: a pilot study, J. Orthop. Res. 33 (2) (2015) 261–269.
[14] A. Haim, A. Wolf, G. Rubin, Y. Genis, Effect of center of pressure modulation on
knee adduction moment in medial compartment knee osteoarthritis, J. Orthop. Res.
(2011).
[15] D. Solomonow-Avnon, A. Haim, D. Levin, M. Elboim-Gabyzon, N. Rozen, E. Peled,
A. Wolf, Reduction of hip joint reaction force via medio-lateral foot center of
pressure manipulation in bilateral hip osteoarthritis patients, J. Orthop. Res. (2016)
1–10.
[16] A. Haim, G. Rubin, N. Rozen, Y. Goryachev, A. Wolf, Reduction in knee adduction
113
Gait & Posture 66 (2018) 107–113
moment via non-invasive biomechanical training: a longitudinal gait analysis study,
J. Biomech. 45 (1) (2012) 41–45.
[17] H. Shaulian, A. Wolf, A Robotic Shoe for Monitoring and Manipulation of the Foot
Center of Pressure for Rehabilitation and Diagnostic, (2017).
[18] K.A. Boyer, P. Federolf, C. Lin, B.M. Nigg, T.P. Andriacchi, Kinematic adaptations to
a variable stiffness shoe: mechanisms for reducing joint loading, J. Biomech. 45 (9)
(2012) 1619–1624.
[19] C.H.F. Pagani, M. Hinrichs, G.P. Brüggemann, Kinetic and kinematic changes with
the use of Valgus knee brace and lateral wedge insoles in patients with medial knee
osteoarthritis, J. Orthop. Res. 30 (7) (2012) 1125–1132.
[20] B.J. De Lateur, R.M. Giaconi, K. Questad, M. Ko, J.F. Lehmann, Footwear and
posture, Am. J. Phys. Med. Rehabil. 70 (5) (1991) 246–254.
[21] J.T. Podraza, S.C. White, Effect of knee flexion angle on ground reaction forces,
knee moments and muscle Co-contraction during an impact-like deceleration
landing: implications for the non-contact mechanism of ACL injury, Knee 17 (4)
(2010) 291–295.
[22] R. Jefferson, J. Collins, M. Whittle, E.O.J. Radin, The role of the quadriceps in
controlling impulsive forces around heel strike, Proc. IMechE Part H J. Eng. Med.
204 (1) (1990) 21–28.
[23] R.S. Hinman, K.A. Bowles, B.B. Metcalf, T.V. Wrigley, K.L. Bennell, Lateral wedge
insoles for medial knee osteoarthritis: effects on lower limb frontal plane bio-
mechanics, Clin. Biomech. Bristol Avon (Bristol, Avon) 27 (1) (2012) 27–33.
[24] T.R. Jenkyn, J.C. Erhart, T.P. Andriacchi, An analysis of the mechanisms for re-
ducing the knee adduction moment during walking using a variable stiffness shoe in
subjects with knee osteoarthritis, J. Biomech. 44 (7) (2011) 1271–1276.
[25] C.O. Kean, K.L. Bennell, T.V. Wrigley, R.S. Hinman, Modified walking shoes for
knee osteoarthritis: mechanisms for reductions in the knee adduction moment, J.
Biomech. 46 (12) (2013) 2060–2066.
[26] G. Giakas, V. Baltzopoulos, Time and frequency domain analysis of ground reaction
forces during walking: an investigation of variability and symmetry, Gait Posture 5
(3) (1997) 189–197.
[27] P. Laing, The development and complications of diabetic foot ulcers, Am. J. Surg.
176 (2 A) (1998).
[28] S.T.F.E.P. Raet, The influence of shoe design on plantar pressures in neuropathic
feet, Diabetes Care 26 (2) (2003).
[29] D.A. Nawoczenski, J.A. Birke, W. Coleman, Effect of rocker sole design on plantar
forefoot pressures, J. Am. Podiatr. Med. Assoc. 78 (1988) 455–460.
[30] B.P. Boden, J.S. Torg, S.B. Knowles, T.E. Hewett, Video analysis of anterior cruciate
ligament injury: abnormalities in hip and ankle kinematics, Am. J. Sports Med. 37
(2) (2009) 252–259.