Journal of Biomechanics 70 (2018) 185–195

Contents lists available at ScienceDirect

Journal of Biomechanics
journal homepage: www.elsevier.com/locate/jbiomech
www.JBiomech.com

Effect of arm swinging on lumbar spine and hip joint forces

Lorenza Angelini a,b, Philipp Damm a, Thomas Zander a, Rizwan Arshad a, Francesca Di Puccio b,
Hendrik Schmidt a,⇑
a
Julius Wolff Institut, Charité-Universitätsmedizin Berlin, Germany
b
Department of Civil and Industrial Engineering, University of Pisa, Italy

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

Article history: During level walking, arm swing plays a key role in improving dynamic stability. In vivo investigations
Accepted 6 September 2017 with a telemeterized vertebral body replacement showed that spinal loads can be affected by differences
in arm positions during sitting and standing. However, little is known about how arm swing could
influence the lumbar spine and hip joint forces and motions during walking. The present study aims to
Keywords: provide better understanding of the contribution of the upper limbs to human gait, investigating ranges
Arm swing of motion and joint reaction forces.
Hip instrumented implant
A three-dimensional motion analysis was carried out via a motion capturing system on six healthy
Level walking
Inverse dynamic model
males and five patients with hip instrumented implant. Each subject performed walking with different
Spinal loads arm swing amplitudes (small, normal, and large) and arm positions (bound to the body, and folded across
the chest). The motion data were imported in a commercial musculoskeletal analysis software for
kinematic and inverse dynamic investigation.
The range of motion of the thorax with respect to the pelvis and of the pelvis with respect to the ground
in the transversal plane were significantly associated with arm position and swing amplitude during gait.
The hip external-internal rotation range of motion statistically varied only for non-dominant limb. Unlike
hip joint reaction forces, predicted peak spinal loads at T12-L1 and L5-S1 showed significant differences
at approximately the time of contralateral toe off and contralateral heel strike.
Therefore, arm position and swing amplitude have a relevant effect on kinematic variables and spinal
loads, but not on hip loads during walking.
Ó 2017 Elsevier Ltd. All rights reserved.

1. Introduction
Walking is one of the most important activities in daily life and
exposes the spine to cyclic loading conditions, with an average load
approximately equal to 170% the one observed in the standing
position (Rohlmann et al., 2014). The motion of the arms in oppo-
site direction to the lower limbs as well as the trunk axial rotation
are distinctive features of the human gait. Arm swing (AS)
improves the dynamic body stability (Nakakubo et al., 2014;
Ortega et al., 2008; Punt et al., 2015) by counterbalancing the
angular momentum produced by legs (Elftman, 1939; Hinrichs,
1990; Park, 2008). Additionally, it reduces the vertical displace-
ment of the centre of mass (Hinrichs, 1990; Murray et al., 1967;
Umberger, 2008) and the free vertical moment generated at the
⇑ Corresponding author at: Julius Wolff Institut, Charité – Universitätsmedizin
Berlin, Augustenburger Platz 1, 13353 Berlin, Germany.
E-mail address: hendrik.schmidt@charite.de (H. Schmidt).
contact between the foot and the ground (Angelini et al., 2016; Li
et al., 2001; Witte et al., 1991).
Current studies on the influence of arm movement during level
walking were mainly focused on the investigation of metabolic
variables (Umberger, 2008), force plate data (Angelini et al.,
2016; Li et al., 2001), gait parameters (Angelini et al., 2016) or
lower limb kinematics and kinetics (Umberger, 2008). To our best
knowledge, only Callaghan et al. (1999) examined the AS effects on
low back joint forces during walking, yet considering only two arm
conditions (i.e., free AS or arms crossed across the chest). In vivo
studies on spinal loads in patients with a vertebral body replace-
ment (VBR) showed that the arm position could affect the load
transfer mechanism in sitting and standing (Dreischarf et al.,
2010; Zander et al., 2015).
The present study aimed to investigate the effect of different
arm positions and AS amplitudes on the kinematics and on joint
reaction forces on the hip and on the upper and lower lumbar
spine. For this purpose, the research was divided into two steps:
an experimental campaign for acquiring human motion data

http://dx.doi.org/10.1016/j.jbiomech.2017.09.011
0021-9290/Ó 2017 Elsevier Ltd. All rights reserved.
186 L. Angelini et al. / Journal of Biomechanics 70 (2018) 185–195

Table 1
Subject and patient data. Group A: asymptomatic subjects, Group B: patients with instrumented hip implant. Implantation side: H2R right, H8L left, H9L left, H6R right, H5L left.

Group A Group B
Subject/Patient S1 S2 S3 S4 S5 S6 H2R H8L H9L H6R H5L
Gender (f/m) m m m m m m m m m m f
Age (years) 30 26 28 31 32 26 67 59 57 73 67
Age at surgery (years) 61 55 54 67 62
Weight (kg) 73.3 68.2 73.6 63.6 85 84 83.2 80 125.7 84 87
Body Height (cm) 181 176 189 176 184 186 172 178 181 176 168
Body Mass Index (kg/m2) 22.4 22 20.6 20.6 25.1 24.3 28.1 25.3 38.4 27.1 30.8

Fig. 1. (A) The marker set. (B) Five different walking tasks. SAS: small arm swing; NAS: normal arm swing; LAS: large arm swing; BA: bonded arms; CA: crossed arms.

during barefoot level walking, and a second step focused on mus-
culoskeletal simulations. A statistical analysis was performed to
investigate the significance of arm position and AS amplitude on
several selected output quantities, including the range of motion
(RoM) at several joints and the joint reaction forces (JRFs). Experi-
mental data from instrumented implants were used to validate the
musculoskeletal model.
2. Materials and methods
2.1. Study participants
Six healthy male volunteers (Group A) and five patients with
instrumented hip implant (Group B) were considered (Table 1).
In Group A, only subjects with no current or previous back or hip
pain and musculoskeletal disorders were included. The patients
of Group B were in good physical condition to perform all investi-
gated activities without limitations. Both, subjects of Group A and
patients of Group B received a physical exam by a clinician to
exclude the presence of spine deformities.
The implantation of the instrumented prosthesis (Hip III) was
approved by the Charité Ethics committee (EA2/057/09) and in
detail described elsewhere (Bergmann et al., 2008; Damm et al.,
2010; Graichen et al., 2007). The patients with instrumented
implant were at 70 (H2R), 43 (H8L), 38 (H9L), 63 (H6R) and 70
(H5L) months’ post-operative. At this post-operative time, patient
gait has been observed to shift towards normality (Long et al.,
1993; Murray et al., 1981).
 L. Angelini et al. / Journal of Biomechanics 70 (2018) 185–195
Fig. 2. Musculoskeletal model during an inverse dynamic analysis (AnyBody Technology A/S, Aalborg, Denmark). (A) Full-body musculoskeletal model with fifty-one
markers. Blue points represent experimental, red ones virtual markers. (B) Marker set for the spine. (For interpretation of the references to colour in this figure legend, the
reader is referred to the web version of this article.)
2.2. Experimental equipment and procedure
The Vicon Motion Capturing System (Vicon Motion Systems,
Oxford Metrics Inc., Oxford, UK), used for the three-
dimensional motion acquisition, consists of ten high-speed
(150 Hz) digital cameras. Synchronously to it, the ground reac-
tion forces (GRFs) were recorded at 900 Hz from two six compo-
nent AMTI force plates (AMTI, Model LG6-4-CE, MA, USA) fixed
into the floor.
Sixty-seven spherical markers were placed on anatomical
landmarks according to a customized full-body skin marker
set (Fig. 1A), described in Leardini et al. (2011, 2007) for the
limbs and pelvis and in Preuss and Popovic (2010) and
Chakraverty et al. (2007) for the spine. Four markers tracked
position and motion of the head. However, only fifty-one mark-
ers were considered in the musculoskeletal model, as specified
in Section 2.3.
A total of five walking tasks (Fig. 1B) were recorded, including
three with different AS amplitudes and two with different arm
positions, i.e., (1) normal arm swing (NAS), (2) small arm swing
(SAS), (3) large arm swing (LAS), (4) bonded arms (BA) and (5)
crossed arms (CA). Prior to motion recording, each subject of Group
A and each patient of Group B received verbal instructions to per-
form correctly all the investigated tasks (1–5). Specifically, they
were asked to walk barefoot at self-selected comfortable walking
speed, with the head upright and looking forward. Each one of
the five tasks was repeated six times for a total of 330 gait trials.
A gait trial was considered acceptable only when the entire foot
cleanly struck the force platform. As a reference, the upright stand-
ing position was captured before walking trial started.
187
Data from the implant, i.e. the three force and three moment
components acting on the implant head (Bergmann et al., 2016),
were recorded synchronously with Vicon and force plate data.
2.3. Musculoskeletal model
In the present study, the GaitFullBody model from the AnyBody
Managed Model Repository v.1.6.2 of the commercial software
package AnyBody Modeling System v. 6.0.4 (AnyBody Technology
A/S, Aalborg, Denmark) was employed. It includes the Twente
Lower Extremity Model for the lower limbs, the spine model pro-
posed in De Zee et al. (2007) and the shoulder and arm models
based on the data and assumptions of Van der Helm (1994) and
Veeger et al. (1991). In the GaitFullBody model, the torso is repre-
sented by seven rigid bony segments: the cervical and thoracic
spines, the five lumbar vertebrae and the sacrum. The thoracic
and cervical spines are modelled as a single lumped mass segment.
They are connected to the skull segment by a revolute joint (C1-
Head) for flexion-extension and to the lumbar spine by a spherical
joint (T12-L1). The lumbar vertebral discs, the sacrum and the hips
are modelled as spherical joints.
The body and joint coordinate reference systems were defined
in agreement with the ISB recommendations (Wu et al., 2002).
The orientations of the thorax and of the pelvis were estimated
both one with respect to the other and each one with respect to
the ground. The orientation was defined by adopting the following
sequence of three ordered rotations: flexion-extension, axial rota-
tion, and lateral bending. For the hip joint, the Euler sequence
adopted was flexion-extension, abduction-adduction, and
internal-external rotation.
188 L. Angelini et al. / Journal of Biomechanics 70 (2018) 185–195
Fig. 3. (A) The joint reaction forces in the hip determined by the musculoskeletal model (dashed lines) and measured by the instrumented hip implant (solid lines) for the
patient H2R during walking with a NAS. Positive directions of forces are medial (+Fx), anterior (+Fy) and proximal (+Fz). (B) Comparison of the resultant force predicted by the
musculoskeletal model and measured in vivo at approximately the time of contralateral toe off (CTO) and contralateral heel strike (CHS). The peak values are shown for the
patient H2R during the five different walking tasks.
Fig. 4. Median of the free vertical moment (%BWH) over one stance phase for the
subjects of Group A. NAS: normal arm swing; SAS: small arm swing; LAS: large arm
swing; BA: bonded arms; CA: crossed arms.
Model geometry and weight were initially set on the individual
data of each subject of Group A and each patient of Group B. To fit
the virtual markers (red points in Fig. 2A) placed on the muscu-
loskeletal model to the corresponding experimental markers (blue
points in Fig. 2A), the anthropometrical parameters and the local
coordinates of the virtual markers were updated over the entire
trial through an optimisation algorithm based on the least-
squared minimization (Andersen et al., 2010).
In musculoskeletal simulations, the complete marker set (see
Section 2.2) was reduced to fifty-one markers (Fig. 2A). Specifically,
since the cervical and thoracic spines were assumed as a single
rigid body, only three markers were considered for defining its
position: C7, the xiphoid process and the jugular notch of the ster-
num. Additionally, positions and motions of the lumbar spine were
detected with five markers placed on the spinous process of each
lumbar vertebra (Fig. 2B). Inter-segmental lumbar rotations were
prescribed based on a pre-defined kinematic rhythm (Hansen
et al., 2006).
The optimized and scaled model was then used in an inverse
kinematic analysis. Subsequently, the resulting joint angles from
the over-determinate kinematic solver (Andersen et al., 2009),
the experimental GRFs and moments together with mass and iner-
tia properties were used to carry out an inverse dynamic analysis.
The inverse dynamics allowed estimating the muscle forces and
the JRFs resulting from the motion. Since musculoskeletal models
are redundant systems (i.e., not enough equilibrium equations
are available to uniquely determine all the muscle forces), the
muscle forces and JRFs were calculated by solving a muscle recruit-
ment problem based on the minimization of a cost function
(Crowninshield and Brand, 1981; Pedotti et al., 1978; Praagman
et al., 2006; Rasmussen et al., 2001). In the present study, the cri-
terion of minimizing the sum of cubed muscle activities was
selected.
2.4. Data processing
Motion capture data were initially pre-processed by means of
the Vicon Nexus 1.7 software (Vicon Motion Systems, Oxford Met-
rics Inc., Oxford, UK) for labelling the markers and filling the gaps,
created by missing or occluded markers. Then, data were imported
in MatlabÒ (MATLAB R2016b, The MathWorks Inc., Natick, MA) for
filtering and detection of gait events. In particular, a second-order,
zero-phase, low-pass Butterworth filter was used. The filter cut-off
frequency was quantified by using the residual method (Winter,
2005). Gait events were yielded from the force plate recordings:
threshold values of 10 N and 5 N for the vertical GRF were selected
for the heel strike (HS) and for the toe off (TO), respectively
(Hreljac and Marshall, 2000; Mickelborough et al., 2000; Tirosh
and Sparrow, 2003).
 L. Angelini et al. / Journal of Biomechanics 70 (2018) 185–195 189

Fig. 5. (A–D) Range of motion in the transversal plane presented as box plot for the subjects of Group A. ***, ** and * indicate statistical significance at p < 0.001, p < 0.01 and
p < 0.05, respectively. NAS: normal arm swing (black); SAS: small arm swing (blue); LAS: large arm swing (red); BA: bonded arms (violet); CA: crossed arms (green). (For
interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

These pre-processing steps were necessary to prepare input
data for musculoskeletal simulations. Additionally, post-
processing analyses were performed on experimental data and
computational results, focusing the attention on: (1) the RoMs of
the thorax, lumbar spine, pelvis and femur (hip joint); (2) the free
vertical moment (Tz); (3) the lumbar and hip JRFs, both from sim-
ulations and instrumented hip implant (only for Group B).
Tz is the reaction couple about the vertical axis exerted by the
foot on the ground and applied on the centre of pressure (COP =
(xp ; yp ; 0)). The Tz value was calculated from the Eq. (1) as:
T z ¼ xp GRF y þ yp GRF x þ Mz
where GRF x is the mediolateral force, GRF y is the anteroposterior
force, and Mz is the vertical moment about the ‘‘true” origin O of
the local coordinate system of each force plate. The ‘‘true” origin
O of the local coordinate system was at a distance Z0 below the
top surface of each force platform. The RoMs were computed as
the peak-to-peak change of the Euler angles of the selected
sequence, as described in Section 2.3.
Within each arm condition, the median and range for the time
series of the selected output were estimated across the trials for
all subjects. The median of the Tz and the hip JRFs were interpo-
lated over the complete stance phase, while the spinal JRFs over
the entire gait cycle. Tz was then scaled by the percent product of
bodyweight and height (BWH), while JRFs by each subject’s body
weight (BW) (Hof, 1996).
To investigate the significance of differences among the exam-
ined five walking activities, a statistical analysis was performed
using the open-source code ‘‘R statistical package” (Version 2.13).
The non-parametric Friedman’s test was carried out to compare
the five tasks. If significant differences (p < 0.05) in RoMs and in
JRFs at local maxima were found, post hoc analyses were run.
3. Results
3.1. Model validation
As a first step in the analysis of results, hip JRFs resulting from
ð1Þ numerical simulations and recorded from instrumented implants
were compared, with the aim of validating the model. For instance,
Fig. 3A shows the resultant hip force (Fres) and the three force com-
ponents predicted by the musculoskeletal model and measured
in vivo for the patient H2R during walking with a NAS. They are
expressed in the femur coordinate reference system of the implant.
Fres showed two typical peak values at approximately the time of
contralateral toe off (CTO) and contralateral heel strike (CHS).
The first maximum was about 250%BW both for the musculoskele-
tal model and for the hip prosthesis. The second predicted peak
was about 30% higher than the measured one with the telemetric
hip implant. Similar results were also found during walking with
a SAS, a LAS, BA and CA (Fig. 3B). However, peak values predicted
by the musculoskeletal model at CHS were overestimated for all
arm conditions. Differences in peaks were smaller than about 9%
at CTO and smaller than 35% at CHS, respectively. Each force com-
ponent followed a similar trend with a comparable range of values.
R2 values for the measured and predicted force components within
190 L. Angelini et al. / Journal of Biomechanics 70 (2018) 185–195

Fig. 6. Median of the JRF components (%BW) for the subjects of Group A. (A) T12-L1. (B) L5-S1. (C) Hip. First row: coordinate reference systems according the ISB
recommendations: X anteroposterior axis, Y proximodistal axis; Z: mediolateral axis. Note that the JRFs on vertebrae are reported over the entire gait cycle while the JRFs on
the hip over the stance phase.

a complete gait cycle were between 0.7 and 0.8. The lowest R2 was
found for the anteroposterior force.
3.2. Free vertical moment
The vertical torque at COP, Tz, is affected by the different arm
conditions as shown in Fig. 4. The magnitude of Tz decreased when
the arm amplitude increased, from SAS to NAS up to LAS. The pat-
terns for SAS, BA and CA were temporally similar and the Pearson’s
correlation coefficient, r, was approximately 0.9 for all possible
combinations.
3.3. Range of motion
As described in Section 2.3, the body orientation was defined
according to the sequence flexion-extension, axial rotation, and
lateral bending, while for the hip joint, the sequence was flexion-
extension, abduction-adduction and internal-external rotation.
For the sake of brevity, only the most relevant results are reported,
in particular only the axial rotation is considered for the bodies.
RoMs in the other planes (Figs. S.1 and S.2) are provided in a sup-
plementary material file.
 L. Angelini et al. / Journal of Biomechanics 70 (2018) 185–195
3.3.1. Thorax-ground
The RoM of the thorax axial rotation with respect to the ground
(Fig. 5A) was similar within all walking conditions with a median
value of about 5°. No statistical differences were found when the
AS amplitude (SAS, NAS, LAS) or the arm position (BA and CA) were
changed during level walking.
3.3.2. Pelvis-ground
Changing in arm position and swing had an influence only on
the RoM of the axial rotation of the pelvis with respect to the
ground (Fig. 5B). Post-hoc test showed statistical differences when
LAS was compared to BA, CA and SAS. However, the axial RoM was
similar for BA, CA and SAS with a median value of about 7°. An
almost 40% and 55% increase in RoM was recorded for NAS and
LAS, respectively.
3.3.3. Thorax-Pelvis
The RoM of the axial rotation in the relative motion between
the thorax and the pelvis was significantly associated with arm
position during gait (Fig. 5C). During walking with NAS and LAS,
the median axial RoMs were about 12° and 15°, respectively while
it reduced to approximately 9° for the other cases (CA, BA, SAS). A
pairwise multiple comparison test demonstrated significant differ-
ences between walking with normal and large arm swing and
walking with suppressed or small arm swing. However, thorax
rotation with respect to the pelvis was not statistically affected
by changes in AS amplitude (p = 0.09 for the pair LAS-NAS) as well
as by restricting AS during gait (p = 0.9 for all possible combina-
tions of BA, CA and SAS).
3.3.4. Hip
The median of the hip rotation angle changed from about 14°
(external hip rotation) to 3° (internal hip rotation) for dominant
limb and from about 18° (external hip rotation) to 4° (internal
hip rotation) for non-dominant limb. Only for non-dominant limb,
a statistical difference in RoM was found when LAS was compared
to BA, CA and SAS (Fig.5D).
3.4. Joint contact forces
Figs. 6A, B and 7A, B show the three force components and the
resultant JRFs computed for the segments T12-L1 and L5-S1 in
their reference frames, over the entire gait cycle. For both vertebral
joints, the normalized anteroposterior (Fx) and proximodistal
forces (Fy) had a symmetrical trend and their magnitude increased
Fig. 7. Median of the resultant JRF (%BW) for the subjects of Group A. (A) T12-L1. (B) L5-S1. (C) Hip. HS: heel strike; CTO: contralateral toe off; CHS: contralateral heel strike;
TO: toe off. NAS: normal arm swing; SAS: small arm swing; LAS: large arm swing; BA: bonded arms; CA: crossed arms.
191
with the amplitude of the arms. The Fx direction changed from pos-
terior (negative values) to anterior (positive values) moving from
upper to lower spine level. The JRF in the mediolateral direction
was very small for the lumbar region. The resultant JRFs at T12-
L1 and L5-S1 (Fig. 7A-B) were almost equal to the Fy. They
increased from SAS to LAS, following rather similar trends in all
arm conditions. Local maxima were found at the four main gait
events, i.e. HS, CTO, CHS and TO. The resultant JRF peaks at HS
and CHS were slightly higher than those at TO and CTO, particu-
larly for NAS and their median values ranged between about 100
and 130%BW.
For the hip joint, the three force components and the resultant
JRF over the stance phase were similar for all arm conditions
(Figs. 6C and 7C). The resultant JRF showed two local maxima at
CTO and CHS with remarkable differences: the median of the first
peak ranged from about 360% to 400%BW at CTO and from about
500% to 540%BW at CHS.
Fig. 8 shows the dispersion of JRF peak values at CTO and CHS,
estimated by simulating all the trails for Group A. At T12-L1, statis-
tical differences were recorded at CTO for LAS-SAS and CA-SAS and
at CHS for LAS-SAS, LAS-BA and LAS-CA (Fig. 8A). At L5-S1, the
same statistical significance was found at CHS, but not at CTO
(Fig. 8B). The hip JRFs were not statistically influenced by arm posi-
tion and AS amplitude at CTO and CHS (Fig. 8C).
4. Discussion
In order to investigate how arm position or AS amplitude could
affect the hip and spine kinematics and JRFs during level walking,
motion data of both, six asymptomatic subjects and five patients
with instrumented hip implant were captured in Vicon system. A
total of 180 musculoskeletal simulations for Group A were run to
compare the RoMs and the JRFs during the five walking tasks,
including three different AS amplitudes and two different arm
positions.
4.1. Model validation
The musculoskeletal model used for our analyses was validated
by comparing in vivo measured and predicted hip JRFs. Although
muscles properties in our musculoskeletal model were not chan-
ged for the patients of Group B, satisfactory results were achieved
compared to the in vivo data for most of the stance phase apart
from the peak at CHS. As shown in Fig. 3, the peak at CHS was
overestimated by the model by a maximum of about 35% during
192 L. Angelini et al. / Journal of Biomechanics 70 (2018) 185–195

Fig. 8. Median and ranges of the resultant JRF (%BW) at approximately the time of contralateral toe off (CTO) and contralateral heel strike (CHS) for the subjects of Group A.
(A) T12-L1; (B) L5-S1; (C) Hip. ***, ** and * indicate statistical significance at p < 0.001, p < 0.01 and p < 0.05, respectively. NAS: normal arm swing (black); SAS: small arm
swing (blue); LAS: large arm swing (red); BA: bonded arms (violet); CA: crossed arms (green). (For interpretation of the references to colour in this figure legend, the reader is
referred to the web version of this article.)

walking with a LAS. This finding is comparable in magnitude to the
values found by Stansfield et al. (2003) and Heller et al. (2001) and
could be related to a reduction of the total hip joint work in elderly
during the double-limb support phase (Franz and Kram, 2014).
Conversely, the in vivo and model-predicted resultant JRFs at CTO
were similar in magnitude. In fact, the total work from muscles
crossing the hip does not differ between old and young adults dur-
ing single-limb support (Franz and Kram, 2014).
4.2. Free vertical moment and RoMs
The Tz (Fig. 4) and the axial RoMs in the transversal plane
(Fig. 5) are most sensitive to changes in arm position and AS ampli-
tude. Although the patterns for Tz are quite similar within the five
arm conditions, temporal differences were observed when walking
with a LAS and NAS were compared to walking with a SAS, BA and
CA (Fig. 4). Specifically, the moment Tz, produced by the foot about
 L. Angelini et al. / Journal of Biomechanics 70 (2018) 185–195
Fig. 9. Resultant hip JRF (%BW) over the stance phase for the five patients of Group B.
the vertical axis of the stance foot, increased in magnitude when AS
was restricted in order to balance the transverse angular momen-
tum created by the legs, in agreement with Elftman (1939). Con-
versely, when subjects walked with swinging arms, Tz decreased
since the lower limb swing was compensated by the moment pro-
duced by the AS. These findings are in a good agreement with pre-
vious results (Li et al., 2001; Park, 2008; Umberger, 2008).
Comparing the RoMs of the thorax axial rotation with respect to
the pelvis (Fig. 5C), higher values were obtained while walking
with a NAS and LAS and lower when AS was restricted (SAS, CA,
BA). Callaghan et al. (1999) also found a reduction of the axial twist
while walking with arms crossed across the chest, especially at the
normal and faster walking cadences. However, the estimation of
the absolute orientations of thorax and pelvis with respect to the
ground (Fig. 5A and B) highlighted that differences should be
ascribed mainly to the axial rotation of the pelvis, again higher
for NAS and LAS and lower for SAS, CA, BA. Unlike the upper body,
changing in arm position and AS amplitude had no influence on the
193
lower limb kinematics (Fig. 5D), as shown also in (Umberger,
2008). These findings provide further support for the idea that
upper body movement can produce stable walking even when
upper body inertial properties are modified.
4.3. Spinal and hip loads
Although spinal loads could not be validated as hip JRFs, the
patterns of predicted spinal loads (Figs. 6A, B and 7A, B) were very
similar to the in vivo data measured with instrumented VBR
(Rohlmann et al., 2014). For the median resultant JRF, the largest
difference was found at CHS for LAS-SAS, LAS-BA and LAS-CA
(Fig. 8A and B). However, spinal loads seem to be only slightly
influenced by arm position and AS amplitude during level walking
as found also by Callaghan et al. (1999).
As far as the hip JRF is concerned, Fig. 8C shows no statistical
differences in the force peak values changing AS amplitude or posi-
tion. Such indications, obtained simulating tasks for the subjects of
194 L. Angelini et al. / Journal of Biomechanics 70 (2018) 185–195
Group A, were also confirmed by the in vivo measurements per-
formed on the patients of Group B (Fig. 9).
4.4. Study limitations
This study has several limitations. Soft-tissue artefacts (STAs)
are the main error source in skin marker-based motion analysis
(Benoit et al., 2006; Leardini et al., 2005; Stagni et al., 2005). In
order to minimize the inaccuracies associated with STAs and mar-
ker placement on the subject, a local optimization method
(Andersen et al., 2010) was used to update the initial segment
lengths and marker locations on the model with respect to the
experimental ones.
Limitations of the model include the simplified muscle model
for the spine, neglecting the force-length-velocity relationships.
Additionally, spinal ligaments and facet articulations were not con-
sidered, while the intervertebral discs were modelled as rigid
spherical joints. The prescribed kinematic rhythm assumed to
define the lumbar intervertebral rotations could affect spinal loads.
However, for the upper body inclinations of the current study (<8°)
this effect is marginal (Arshad et al., 2016). The model loading pre-
dictions at the upper lumbar spine might be also influenced by the
rigid thorax assumption and, therefore, models with a fully articu-
lated thorax are recommended especially for large amounts of flex-
ion (Ignasiak et al., 2016). However, as walking requires a trunk
inclination angle less than 10°, this simplification could be consid-
ered acceptable in this study. The adopted muscle recruitment cri-
terion could slightly underestimate spinal loads and overestimate
the hip loads during double-limb support (Fig. 3). In fact, during
forward propulsion, the hip muscles perform mechanical work to
start leg swing. Conversely, the chosen recruitment criterion seems
to have only a marginal impact during single support (at CTO). In
fact, the hip muscle forces are lower and the hip joint reaction
behaves as an inverse pendulum force during this phase.
Finally, the small number of subjects included in this study lim-
its our findings, though it is sufficient to achieve statistical power
values of 80% (Bates et al., 1992). As only men were measured, gen-
der effects should also be investigated in the future.
In summary, free vertical moments and the axial RoMs are con-
siderably influenced by changes in AS amplitude and arm position
during walking. Conversely, AS slightly affects spinal JRFs, but not
hip JRFs.
Acknowledgments
This study was financially supported by the Bundesinstitut für
Sportwissenschaften, BiSp (MiSpEx - Network).
Conflict of interest
There are no conflicts of interest.
Appendix A. Supplementary material
Supplementary data associated with this article can be found, in
the online version, at http://dx.doi.org/10.1016/j.jbiomech.2017.
09.011.
References
Andersen, M.S., Damsgaard, M., MacWilliams, B., Rasmussen, J., 2010. A
computationally efficient optimisation-based method for parameter
identification of kinematically determinate and over-determinate
biomechanical systems. Comput. Meth. Biomech. Biomed. Eng. 13, 171–183.
Andersen, M.S., Damsgaard, M., Rasmussen, J., 2009. Kinematic analysis of over-
determinate biomechanical systems. Comput. Meth. Biomech. Biomed. Eng. 12,
371–384.
Angelini, L., Di Puccio, F., Zander, T., Schmidt, H., 2016. Influence of arm motion on
spatio-temporal gait parameters and on force data. IOSR J. Sports Phys. Edu. 3,
12–16.
Arshad, R., Zander, T., Dreischarf, M., Schmidt, H., 2016. Influence of lumbar spine
rhythms and intra-abdominal pressure on spinal loads and trunk muscle forces
during upper body inclination. Med. Eng. Phys. 38, 333–338.
Bates, B.T., Dufek, J.S., Davis, H.P., 1992. The effect of trial size on statistical power.
Med. Sci. Sports Exerc. 24, 1059–1065.
Benoit, D.L., Ramsey, D.K., Lamontagne, M., Xu, L., Wretenberg, P., Renström, P.,
2006. Effect of skin movement artifact on knee kinematics during gait and
cutting motions measured in vivo. Gait & Post. 24, 152–164.
Bergmann, G., Bender, A., Dymke, J., Duda, G., Damm, P., 2016. Standardized loads
acting in hip implants. PLoS ONE 11, e0155612.
Bergmann, G., Graichen, F., Rohlmann, A., Westerhoff, P., Heinlein, B., Bender, A.,
Ehrig, R., 2008. Design and calibration of load sensing orthopaedic implants. J.
Biomech. Eng. 130, 021009.
Callaghan, J.P., Patla, A.E., McGill, S.M., 1999. Low back three-dimensional joint
forces, kinematics, and kinetics during walking. Clin. Biomech. 14, 203–216.
Chakraverty, R., Pynsent, P., Isaacs, K., 2007. Which spinal levels are identified by
palpation of the iliac crests and the posterior superior iliac spines? J Anat 210,
232–236.
Crowninshield, R.D., Brand, R.A., 1981. A physiologically based criterion of muscle
force prediction in locomotion. J. Biomech. 14, 793–801.
Damm, P., Graichen, F., Rohlmann, A., Bender, A., Bergmann, G., 2010. Total hip joint
prosthesis for in vivo measurement of forces and moments. Med. Eng. Phys. 32,
95–100.
De Zee, M., Hansen, L., Wong, C., Rasmussen, J., Simonsen, E.B., 2007. A generic
detailed rigid-body lumbar spine model. J. Biomech. 40, 1219–1227.
Dreischarf, M., Bergmann, G., Wilke, H.J., Rohlmann, A., 2010. Different arm
positions and the shape of the thoracic spine can explain contradictory
results in the literature about spinal loads for sitting and standing. Spine
(Phila Pa 1976) 35, 2015–2021.
Elftman, H.O., 1939. The function of the arms in walking. Johns Hopkins Press,
Baltimore, Maryland, USA.
Franz, J.R., Kram, R., 2014. Advanced age and the mechanics of uphill walking: a
joint-level, inverse dynamic analysis. Gait & Post. 39, 135–140.
Graichen, F., Arnold, R., Rohlmann, A., Bergmann, G., 2007. Implantable 9-channel
telemetry system for in vivo load measurements with orthopedic implants. IEEE
Trans. Biomed. Eng. 54, 253–261.
Hansen, L., De Zee, M., Rasmussen, J., Andersen, T.B., Wong, C., Simonsen, E.B., 2006.
Anatomy and biomechanics of the back muscles in the lumbar spine with
reference to biomechanical modeling. SPINE 31, 1888–1899.
Heller, M.O., Bergmann, G., Deuretzbacher, G., Durselen, L., Pohl, M., Claes, L., Haas,
N.P., Duda, G.N., 2001. Musculo-skeletal loading conditions at the hip during
walking and stair climbing. J. Biomech. 34, 883–893.
Hinrichs, R.N., 1990. Whole Body Movement: Coordination of Arms and Legs in
Walking and Running. In: Winters, J.M., Woo, S.L.Y. (Eds.), Multiple muscle
systems: biomechanics and movement organization. Springer, New York, NY,
pp. 694–705.
Hof, A.L., 1996. Scaling gait data to body size. Gait & Post. 4, 222–223.
Hreljac, A., Marshall, R.N., 2000. Algorithms to determine event timing during
normal walking using kinematic data. J. Biomech. 33, 783–786.
Ignasiak, D., Ferguson, S.J., Arjmand, N., 2016. A rigid thorax assumption affects
model loading predictions at the upper but not lower lumbar levels. J. Biomech.
49, 3074–3078.
Leardini, A., Biagi, F., Merlo, A., Belvedere, C., Benedetti, M.G., 2011. Multi-segment
trunk kinematics during locomotion and elementary exercises. Clin. Biomech.
(Bristol, Avon) 26, 562–571.
Leardini, A., Chiari, L., Della Croce, U., Cappozzo, A., 2005. Human movement
analysis using stereophotogrammetry. Part 3. Soft tissue artifact assessment
and compensation. Gait & Post. 21, 212–225.
Leardini, A., Sawacha, Z., Paolini, G., Ingrosso, S., Nativo, R., Benedetti, M.G., 2007. A
new anatomically based protocol for gait analysis in children. Gait & Post. 26,
560–571.
Li, Y., Wang, W., Crompton, R.H., Gunther, M.M., 2001. Free vertical moments and
transverse forces in human walking and their role in relation to arm-swing. J.
Exper. Biol. 204, 47–58.
Long, W.T., Dorr, L.D., Healy, B., Perry, J., 1993. Functional recovery of noncemented
total hip arthroplasty. Clin. Orthopaed. Related Res. 288, 73–77.
Mickelborough, J., van der Linden, M.L., Richards, J., Ennos, A.R., 2000. Validity and
reliability of a kinematic protocol for determining foot contact events. Gait &
Post. 11, 32–37.
Murray, M., Gore, D.R., Brewer, B.J., Mollinger, L.A., Sepic, S.B., 1981. Joint function
after total hip arthroplasty: a four-year follow-up of 72 cases with Charnley and
Mtiller replacements. Clin. Orthopaed. Related Res. 157, 119–124.
Murray, M.P., Sepic, S.B., Barnard, E.J., 1967. Patterns of sagittal rotation of the upper
limbs in walking. Phys. Ther. 47, 272–284.
Nakakubo, S., Doi, T., Sawa, R., Misu, S., Tsutsumimoto, K., Ono, R., 2014. Does arm
swing emphasized deliberately increase the trunk stability during walking in
the elderly adults? Gait & Post. 40, 516–520.
Ortega, J.D., Fehlman, L.A., Farley, C.T., 2008. Effects of aging and arm swing on the
metabolic cost of stability in human walking. J. Biomech. 41, 3303–3308.
Park, J., 2008. Synthesis of natural arm swing motion in human bipedal walking. J.
Biomech. 41, 1417–1426.
Pedotti, A., Krishnan, V., Stark, L., 1978. Optimization of muscle-force sequencing in
human locomotion. Math. Biosci. 38, 57–76.
 L. Angelini et al. / Journal of Biomechanics 70 (2018) 185–195
Praagman, M., Chadwick, E., Van Der Helm, F., Veeger, H., 2006. The relationship
between two different mechanical cost functions and muscle oxygen
consumption. J. Biomech. 39, 758–765.
Preuss, R.A., Popovic, M.R., 2010. Three-dimensional spine kinematics during
multidirectional, target-directed trunk movement in sitting. J. Electrom.
Kinesiol.: official journal of the International Society of Electrophysiological
Kinesiology 20, 823–832.
Punt, M., Bruijn, S.M., Wittink, H., van Dieen, J.H., 2015. Effect of arm swing strategy
on local dynamic stability of human gait. Gait & Post. 41, 504–509.
Rasmussen, J., Damsgaard, M., Voigt, M., 2001. Muscle recruitment by the min/max
criterion—a comparative numerical study. J. Biomech. 34, 409–415.
Rohlmann, A., Dreischarf, M., Zander, T., Graichen, F., Bergmann, G., 2014. Loads on a
vertebral body replacement during locomotion measured in vivo. Gait & Post.
39, 750–755.
Stagni, R., Fantozzi, S., Cappello, A., Leardini, A., 2005. Quantification of soft
tissue artefact in motion analysis by combining 3D fluoroscopy and
stereophotogrammetry: a study on two subjects. Clin. Biomech. 20, 320–
329.
Stansfield, B.W., Nicol, A.C., Paul, J.P., Kelly, I.G., Graichen, F., Bergmann, G., 2003.
Direct comparison of calculated hip joint contact forces with those measured
using instrumented implants. An evaluation of a three-dimensional
mathematical model of the lower limb. J. Biomech. 36, 929–936.
195
Tirosh, O., Sparrow, W.A., 2003. Identifying heel contact and toe-off using forceplate
thresholds with a range of digital-filter cutoff frequencies. J. Appl. Biomech. 19,
178–184.
Umberger, B.R., 2008. Effects of suppressing arm swing on kinematics, kinetics, and
energetics of human walking. J. Biomech. 41, 2575–2580.
Van der Helm, F.C., 1994. A finite element musculoskeletal model of the shoulder
mechanism. J. Biomech. 27, 551–569.
Veeger, H.E.J., Van Der Helm, F.C.T., Van Der Woude, L.H.V., Pronk, G.M., Rozendal, R.
H., 1991. Inertia and muscle contraction parameters for musculoskeletal
modelling of the shoulder mechanism. J. Biomech. 24, 615–629.
Winter, D.A., 2005. Biomechanics and motor control of human movement. John
Wiley & Sons, Hoboken, N.J..
Witte, H., Preuschoft, H., Recknagel, S., 1991. Human body proportions explained on
the basis of biomechanical principles. Z. Morphol. Anthropol. 78, 407–423.
Wu, G., Siegler, S., Allard, P., Kirtley, C., Leardini, A., Rosenbaum, D., Whittle, M.,
D’Lima, D.D., Cristofolini, L., Witte, H., Schmid, O., Stokes, I., 2002. ISB
recommendation on definitions of joint coordinate system of various joints
for the reporting of human joint motion—part I: Ankle, hip, and spine. J.
Biomech. 35, 543–548.
Zander, T., Dreischarf, M., Schmidt, H., Bergmann, G., Rohlmann, A., 2015. Spinal
loads as influenced by external loads: a combined in vivo and in silico
investigation. J. Biomech. 48, 578–584.