University of Southern California Dissertations and Theses

Sex differences in hip adduction during running: influence of hip abductor strength, muscle activation, and pelvis & femur morphology

(USC Thesis Other)

Sex differences in hip adduction during running: influence of hip abductor strength, muscle activation, and pelvis & femur morphology

PDF

Download

Open document

Flip pages

Download a page range

Download transcript

Copy asset link

Request this asset

Transcript (if available)

Content

SEX DIFFERENCES IN HIP ADDUCTION DURING RUNNING:
INFLUENCE OF HIP ABDUCTOR STRENGTH,
MUSCLE ACTIVATION, AND
PELVIS & FEMUR MORPHOLOGY

by

Jia Liu

A Dissertation Presented to the
FACULTY OF THE USC GRADUATE SCHOOL
UNIVERSITY OF SOUTHERN CALIFORNIA
In Partial Fulfillment of the
Requirements for the Degree
DOCTOR OF PHILOSOPHY
(BIOKINESIOLOGY)

May 2021

Copyright 2021 Jia Liu
ii

Dedication

This dissertation is dedicated to my grandmother, Shulan Zhang, who has passed away on
February 16, 2018, for teaching me kindness, hard-working spirits, and never giving up on my
dreams.
iii

Acknowledgments
My deepest gratitude goes to my parents and my younger brother for their unconditional
support of my academic pursuit aboard for all of these past years. Especially during the COVID-
19 pandemic, I know they are extremely worried about me being away from home, but they never
forgot to provide reassuring words so that I could fully concentrate on my dissertation. My love
for them is far deeper than any insights I obtained from my research.
Throughout the completion of this dissertation, I received a great deal of support from my
dissertation committee. I would like to thank my Ph.D. advisor Dr. Christopher Powers, for every
guidance he has provided me, not only for this dissertation but at each stage of my academic
advancement. I am grateful for his willingness to impart his knowledge, keen perspectives, and
clear writing skills. I also would like to express my gratitude towards Dr. Kristi Lewton, for those
countless long hours of discussions to help me overcome so many research puzzles. Her strong
enthusiasm, rigorous attitude, and perceptive thinking towards scientific research always impress
me. This dissertation project will not have progressed without the support of Dr. Patrick Colletti.
It is under his guidance that the low-dose computed tomography protocol could be developed and
used in my dissertation. And I will always remember and forever grateful that at a critical stage,
he even volunteered to become my pilot subject for testing the effectiveness of the protocol before
I applied it to my research participants. Working with Dr. Salem has been full of positivity and
encouragement. He never held back his compliments when he saw the progress I had made.
Throughout these years, his generous support and kindness helped me become a confident
researcher. Last but not least, Dr. Baker has given me so much guidance with electromyography
signal processing and noise handling. I consider myself lucky to have the opportunity to learn so
much from her.
iv

I am also grateful for the support I have obtained from the faculty and students in the
Division of Biokinesiology and Physical Therapy of USC. In particular, I would like to thank all
of my good friends in the Division of Biokinesiology and Physical for helping me improve and for
building such a supportive research environment. I would also like to thank the American College
of Sports Medicine, the International Society of Biomechanics, the American Society of
Biomechanics, and the USC CTSI department for their generous financial supports that made this
dissertation possible. Finally, I would like to thank all of my participants for their time and interest
in my dissertation.
v

Table of Contents
Dedication ....................................................................................................................................... ii
Acknowledgments.......................................................................................................................... iii
List of Tables ................................................................................................................................. vi
List of Figures ............................................................................................................................... vii
Abstract ........................................................................................................................................ viii
Chapter 1. Overview ....................................................................................................................... 1
Chapter 2. Background and Significance........................................................................................ 4
Chapter 3. Sex Difference in Hip Adduction During Stance Phase of Running: A Swing Phase
Problem? ......................................................................................................................................... 9
Abstract ......................................................................................................................... 9
Introduction ................................................................................................................. 10
Methods....................................................................................................................... 11
Results ......................................................................................................................... 14
Discussion ................................................................................................................... 15
Conclusion .................................................................................................................. 18
Chapter 4. Sex Differences in Hip Adduction during Running: Influence of Hip Abductor Strength,
Activation, and Pelvis & Femur Morphology............................................................................... 19
Abstract ....................................................................................................................... 19
Introduction ................................................................................................................. 20
Methods....................................................................................................................... 22
Results ......................................................................................................................... 29
Discussion ................................................................................................................... 32
Conclusion .................................................................................................................. 34
Chapter 5. Classification of Runners with High versus Low Hip Adduction using Pelvis and Femur
Morphology Measures .................................................................................................................. 35
Abstract ....................................................................................................................... 35
Introduction ................................................................................................................. 36
Methods....................................................................................................................... 38
Results ......................................................................................................................... 47
Discussion ................................................................................................................... 50
Conclusion .................................................................................................................. 53
Chapter 6. Summary and Conclusions .......................................................................................... 54
References ..................................................................................................................................... 60
vi

List of Tables
TABLE 3-1. Participant demographics......................................................................................... 12
TABLE 4-1. Participant demographics......................................................................................... 22
TABLE 4-2. Sex differences in hip abductor muscle strength, muscle activation, and pelvis and
femur morphology variables. ................................................................................ 30
TABLE 4-3. Predictions of hip abductor muscle strength, muscle activation, and pelvis & femur
morphology variables to peak hip adduction angles during running. ................... 30
TABLE 5-1. Description of linear morphology measurements of pelvis and femur. ................... 44
TABLE 5-2. Classifier prediction performance using cross-validated datasets ........................... 48
TABLE 5-3. Group differences in hip abductor strength and pelvis & femur morphology variables.
............................................................................................................................... 50
vii

List of Figures
FIGURE 2-1. Dissertation theoretical framework. ......................................................................... 8
FIGURE 3-1. Sex difference in hip adduction during running. Data are mean ± SD. ................. 14
FIGURE 3-2. Sex difference in peak hip adduction angles during late swing phase (a) and stance
phase (b) of running. Data are mean ± SD. *: p<0.05 .......................................... 15
FIGURE 3-3. Correlation plot showing the association between peak hip adduction during the late
swing and stance phases of running. Grey shading represents the standard error of
the slope. ............................................................................................................... 16
FIGURE 4-1. Pelvis morphology measurements. (a) pelvis coordinate system, (b) acetabulum face
plane and acetabulum orientation vector, (c) acetabular anteversion angle, (d)
acetabular abduction angle. ................................................................................... 26
FIGURE 4-2. Proximal femur morphology measurements. (a) femoral coordinate system, (b)
femoral head-neck and shaft axes, (c) femoral neck-shaft angle, (d) femoral
anteversion angle. ................................................................................................. 27
FIGURE 4-3. Sex difference in frontal plane hip motion during running. Data are mean ± SD. 29
FIGURE 4-4. Correlations of femoral anteversion angle to peak hip adduction during the late
swing phase (a) and stance phase (b) of running. Red dots represent females, and
green dots represent males. ................................................................................... 31
FIGURE 5-1. Femoral orientation measurements. (a) femoral coordinate system, (b) femoral neck-
shaft angle, (c) femoral valgus angle, (d) femoral head-neck anteversion angle, (e)
femoral neck anteversion angle, (f) femoral head anteversion angle. .................. 43
FIGURE 5-2. Classification Outcome using the dataset from all 29 participants. Red region
represents the SVM-classified high adduction group, and the dark blue region
represents the SVM-classified low adduction. Black dots represent runners with
kinematics-classified low hip adduction and yellow dots represent runners with
kinematics-classified high hip adduction. Only 1 participant was false-positively
classified as a high adduction runner (outlined by a white rectangular box). ....... 49
FIGURE 5-3. Importance of femoral head anteversion and femur length in the best SVM classifier.
............................................................................................................................... 49
viii

Abstract
Excessive hip adduction is a common movement impairment that has been proposed to
underlie various lower extremity injuries (1-3). Compared to males, females are more likely to
exhibit excessive hip adduction during weight-bearing activities such as running (4-7). Despite
decades of research investigating the potential roles of hip abductor muscle strength and activation
as being potential causes of excessive hip adduction (8-13),

studies have reported that these
variables have little influence on frontal plane hip motion (5,8,11). Another factor that has not
been systematically explored in the context of sex differences in frontal plane hip kinematics is
the potential influence of pelvis and femur bony morphology. The premise that bony morphology
may play a role in hip joint kinematics is supported by 2 underlying assertions: 1) males and
females differ in various aspects of femur and pelvis morphology and 2) sex differences in frontal
plane hip kinematics have been reported at initial contact and the stance phase of running. The
latter suggests that hip kinematics during late swing (a time point at which there is little muscular
demand) may have an influence on hip adduction during stance. Given the limited work in this
area, the objective of this dissertation was to comprehensively examine the roles of hip abductor
strength, neuromuscular activation, and pelvis and femur bony morphology in contributing to
frontal plane hip kinematics during running. To achieve this objective, 3 studies were undertaken.
The purpose of Chapter III was to compare sex differences in frontal plane hip kinematics
during both late swing and stance phases of running and determine the relationship between late
swing and stance phase kinematics. Fifteen female and 16 male runners (all heel strikers) ran
overground at a speed of 4 m/s. Hip joint kinematics during running were quantified using a 3D
motion capture system. Sex differences in peak hip adduction during the late swing and stance
phases were compared using independent sample t-tests. Linear regression analysis was used to
ix

determine the relationship between late swing and stance phase peak hip adduction. Compared to
males, females exhibited significantly greater peak hip adduction during both the late swing (8.5
± 2.6 vs 6.1 ± 2.8°, p = 0.019) and stance phases of running (13.3 ± 4.2 vs 9.6 ± 3.4°, p = 0.011).
Furthermore, late swing peak hip adduction was predictive of subsequent stance phase peak hip
adduction (r = 0.63, p < 0.001).
The purpose of Chapter IV was twofold: 1) to compare sex differences in hip abductor
strength, activation, and pelvis and femur bony morphology, and 2) to determine the best
combination of predictor(s) of peak hip adduction during late swing and stance phases of running.
Fifteen female and 14 male runners underwent strength testing, instrumented analysis of
overground running (e.g., kinematics and muscle activation), and computed tomography (CT)
scanning of pelvis and femur. Morphologic measurements included the bilateral hip width to femur
length ratio, acetabulum abduction, acetabulum anteversion, femoral anteversion, and femoral
neck-shaft angles. Sex differences for all variables (e.g., kinematics, strength, activation, and
morphology) were examined using independent t tests. Linear regression was used to assess the
ability of each independent variable of interest to predict peak hip adduction during the late swing
and stance phase of running. Compared to males, females exhibited significantly lower hip
abductor strength (1.8 ± 0.3 vs 2.0 ± 0.3 Nm/kg, p=0.04), greater femoral neck-shaft angles (134.1
± 5.0° vs 129.9 ± 4.1°, p=0.01), and a greater hip width to femur length ratio (0.44 ± 0.02 vs 0.42
± 0.03, p=0.03). However, femoral anteversion was found to be the only significant predictor of
peak hip adduction during late swing (r=0.36, p=0.05) and stance (r=0.41, p=0.03).
The purpose of Chapter V was to determine the most relevant pelvis and femur
morphological characteristics for differentiating runners with high versus low hip adduction during
the stance phase of running. Fifteen female and 14 male runners underwent instrumented
x

kinematics analysis of overground running and CT scanning of pelvis and femur. Using a hybrid
approach of support vector machine classification and best subset feature selection techniques, the
combination of femoral head anteversion and femur length was shown to be the best performing
variables for distinguishing runners with high versus low peak hip adduction during running.
Together, these two variables achieved a prediction accuracy of 0.93, sensitivity of 1 and
specificity of 0.88. Also, the results of a sensitivity analysis revealed that femoral head anteversion
was more important than femur length in contributing to hip adduction group classification.
Taken together, the results of this dissertation highlight the role of bony morphology as
opposed to muscle strength and activation as being contributory to hip adduction during running.
With respect to morphology, there appears to be a ‘localized’ influence on frontal plane hip
kinematics related to femoral head orientation. While it was beyond the scope of this dissertation
to determine the mechanisms underlying this ‘localized’ morphology influence, 2 potential
mechanisms may be contributory: 1) the function of ligamentum teres in limiting excessive hip
adduction, and/or 2) interaction between frontal and transverse plane hip kinematics in
contributing to joint congruency during running. Further investigation is needed to fully
understand the complex relationship between proximal femur morphology and frontal plane hip
kinematics during dynamic tasks.
1

Chapter 1. Overview
Excessive hip adduction is a common movement impairment that has been proposed to
underlie various lower extremity injuries (i.e., patellofemoral pain (14,15), and iliotibial band
syndrome (16,17)). Compared to males, females are more likely to exhibit excessive hip adduction
during weight-bearing activities such as running (4-7). Despite decades of research efforts
investigating the potential roles of hip abductor muscle strength and activation as being causes of
excessive hip adduction (8-13),

studies have reported that these variables have limited influence
on frontal plane hip motion (5,8,11). To date, the underlying causes of sex differences in frontal
plane hip kinematics during weight-bearing activities are not fully understood.
One factor that has not been systematically explored in the context of sex differences in
frontal plane hip kinematics is the potential influence of pelvis and femur morphology. For
example, a greater bilateral hip width relative to femur length ratio theoretically would require
larger degrees of hip adduction to position the foot under the center of mass at ground contact.
Also, hip adduction during running could be influenced by structural features that affect the
orientation of the acetabulum and proximal femur such as acetabulum abduction, acetabulum
anteversion, femoral neck-shaft angle, and femoral anteversion. The premise that bony structure
may play a role in affecting lower limb motions is supported by preliminary data showing that hip
adduction during the late swing phase of running was highly predictive of peak hip adduction
during stance. As such, it is reasonable to speculate that relevant bony morphology characteristics
may bias the hip joint into greater adduction prior to ground contact, and, in turn, affect the
following stance phase kinematics.
To date, sex differences in frontal plane kinematics of the hip have not been
comprehensively studied from the neuro-muscular-skeletal perspective. To this end, this
2

dissertation sought to examine the roles of hip abductor strength, neuromuscular activation, and
pelvis and femur morphology in contributing to frontal plane hip kinematics during running. To
accomplish this objective, 3 studies were undertaken with the following aims:

Aim 1: To compare sex differences in peak hip adduction during both late swing and stance
phases of running, and determine the relationship between late swing and stance phase
kinematics.
Hypothesis:
1a) When compared to males, females will exhibit greater peak hip adduction during both
late swing phase and stance phase of running.
1b) Peak hip adduction during late swing will be predictive of peak hip adduction during
the stance.

Aim 2: To compare sex differences in modifiable and non-modifiable risk factors associated
with hip adduction, and to determine the best combination of predictor(s) of peak hip
adduction during running.
Hypothesis:
2a) When compared to males, females will exhibit diminished hip abductor strength and
gluteus medius activation.
2b) When compared to males, females will exhibit smaller acetabulum abduction and
femoral neck-shaft angles, greater femoral anteversion and acetabulum anteversion angles,
and greater bilateral hip width to femur length ratio.
2c) Measures of pelvis and proximal femur morphology would explain a greater amount
3

of variance in hip adduction during the late swing and stance phases of running compared
to hip abductor strength and neuromuscular recruitment.

Aim 3: To determine the most relevant pelvis and femur morphological characteristics for
discriminating runners with high versus low hip adduction during the stance phase of
running.
Hypothesis:
A subset of pelvis and femur morphologic measures variables will discriminate runners
with high and low hip adduction with high prediction accuracy.

4

Chapter 2. Background and Significance
STATEMENT OF PROBLEM
While running is a popular form of physical activity, injuries such as patellofemoral pain
and iliotibial band syndrome are highly prevalent among runners in the United States and beyond
(6,18-22). The development of these running injuries has been linked to a commonly observed
movement impairment, i.e., excessive hip adduction (1,3,14-17,23,24). For example, repetitive hip
adduction during physical activity has been shown to result in high contact stresses between lateral
patella and femoral trochlea, thereby inducing the development of patellofemoral pain syndrome
(3) and, possibly, future development of patellofemoral osteoarthritis (1). In addition, there is
evidence that increased hip adduction can increase the strain in the iliotibial band (17) and thus
contribute to iliotibial band syndrome (16).
Sex Difference in Lower Extremity Injury Rates
Evidence suggests that females sustain higher rates of lower-limb injuries compared to
males (18,19). For example, females are highly predisposed to patellofemoral pain (19,25). Boling
et al reported that females in United States Naval Academy were 2.23 times more likely to develop
patellofemoral pain than their males during a 2.5-year follow-up (25). A retrospective search from
the PearlDiver Patient Record Database, a national database containing orthopedic patient records,
revealed that females had a greater incidence of patellofemoral pain than males (55.4% vs 44.6%)
(19). In a study of 2002 running-related injury cases, females accounted for 62% of patellofemoral
pain (6). Females are also more likely to be diagnosed with iliotibial band syndrome than males
(6,20-22). Furthermore, a retrospective study conducted by Clement et al (1981) of 1650 patients
revealed that females were more often diagnosed with iliotibial band syndrome than males (21).
5

Even though Marti et al (20) and Macera et al (22) have concluded that females and males showed
a similar incidence of iliotibial band syndrome after controlling factors such as running style,
duration, and intensity, iliotibial band syndrome has become the second most common running
injury with a larger proportion observed in females (6).
Sex Differences in Hip Adduction During Running
Compared to males, females are more likely to exhibit excessive hip adduction during
running (26-28). Multiple studies have reported that females demonstrate significantly greater hip
adduction than males during the stance phase of running, regardless of speed (26,27), inclination
(27), and type of running (i.e., overground or treadmill) (28). On average, females tend to have 4-
6 degrees greater hip adduction than males during running with moderate to vigorous intensities
(26-28). This difference is considered clinically meaningful as literature has suggested that females
who developed iliotibial band syndrome or patellofemoral pain demonstrated 4 or 5 degrees greater
hip adduction than males, respectively (2,3).
Influence of Hip Abductor Muscle Strength on Frontal Plane Hip Kinematics
Hip abductor muscle weakness has been proposed as being contributory to excessive hip
adduction during dynamic tasks (8,29,30). However, existing evidence is conflicting. While some
studies have reported lower hip abductor torque production in females compared to males (8,30,31),
others have not (32-34). The influence of hip abductor strength on hip adduction during weight
activities also is ambiguous. For example, Phol et al found that participants exhibited no
differences in hip adduction during walking after their hip abductor muscle strength was
significantly reduced by a superior gluteal nerve block injection when compared with before the
injection (35). Jacobs et al reported a weak correlation between hip abductor strength with peak
hip adduction angles in females, however, no association was found in males (8). Collectively,
6

these studies suggest that hip abductor muscle strength may to a small extent, influence the hip
adduction motions during weight-bearing activities, and that this influence may be more evident
in females than in males.
Influence of Hip Abductor Muscle Activation on Frontal Plane Hip Kinematics
Another widely studied risk factor with respect to excessive frontal plane hip kinematics is
gluteus medius activation. Despite the fact that females exhibit greater degrees of hip adduction
compared to males, a majority of studies have reported that gluteus medius activation is similar
between males and females during various tasks such as running (9,27), single-leg drop jump (4),
kicking (13), and single-leg squatting (10). In addition, one study reported that greater gluteus
medius activation was higher in males compared to females during single-leg squatting (32).
Furthermore, gluteus medius activation has not been reported to be predictive of hip adduction
during running (27). Taken together, existing literature suggests that gluteus medius activation has
a limited influence on hip adduction.
Influence of Pelvis and Femur Morphology on Frontal Plane Hip Kinematics
One factor that may be contributory but has not been systematically explored in the context
of sex differences in frontal plane hip kinematics is the potential influence of pelvis and femur
morphology. For example, a greater bilateral hip width relative to femur length ratio theoretically
would require larger degrees of hip adduction to position the foot under the center of mass at
ground contact. Also, hip adduction during running could be influenced by structural features that
affect the orientation of the acetabulum and proximal femur such as acetabulum abduction,
acetabulum anteversion, femoral neck-shaft angle, and femoral anteversion. To date, few studies
have examined the potential influence of pelvis and/or femur morphology on hip adduction during
running. Studies examining the hip width to leg length ratio (27) and bilateral greater trochanteric
7

distance to femur length ratio (36) have reported non-significant or very low associations with hip
adduction during running, respectively. Similarly, a study by Baggaley et al. (5) reported that the
femoral neck-shaft angle was not associated with the magnitude of hip adduction in females during
running. However, a limitation of previous work in this area is that measures of pelvis and femur
morphology were either estimated from external markers (27) or were obtained from 2D
radiographs (5). Such measurements are prone to error (37,38) and may not accurately represent
structural morphology.
INNOVATION
A limitation of research that has attempted to understand underlying causes of hip
adduction, particularly in females, is that the potential influences of hip abductor strength,
activation, and pelvis femur morphology have been studied in relative isolation. In addition,
previous related research involving the influence of morphology have relied on marker-based or
2D radiograph measurements which are of questionable accuracy and are susceptible to
measurement error. Lastly, there lacks a systematic and integrative approach to this research
question. Therefore, the purpose of this dissertation was to comprehensively examine the roles of
hip abductor strength, neuromuscular activation, and pelvis and femur morphology (using
sophisticated 3D-based measurements) in contributing to frontal plane hip kinematics during
running (FIGURE 2-1).
In addition, this dissertation approached the research question from not only the stance
phase but also the late swing phase of running. Literature cited above has not investigated the
relationship between the late swing phase and stance phase of running with respect to the frontal
plane hip motions. Hip kinematics during weight-bearing could be affected by the joint motion
prior to ground contact. This premise is supported by a previous study that reported females who
8

demonstrated greater hip adduction during stance phase also exhibited greater hip adduction prior
to ground contact while running (9).
FIGURE 2-1. Dissertation theoretical framework.
9

Chapter 3. Sex Difference in Hip Adduction During Stance Phase of
Running: A Swing Phase Problem?
ABSTRACT
Purpose: The purpose of the current study was two-fold: 1) evaluate sex differences in peak hip
adduction during the late swing and stance phases of running and 2) determine if peak hip
adduction during late swing is predictive of peak hip adduction during stance.
Methods: 15 female and 16 male runners (all heel strikers) ran over ground at a speed of 4 m/s.
Hip joint kinematics during running were quantified using a 3D motion capture system. Sex
differences in peak hip adduction during the late swing and stance phases were compared using
independent sample t-tests. Linear regression analysis was used to determine the relationship
between late swing and stance phase hip adduction.
Results: Compared to males, females exhibited significantly greater peak hip adduction during
both the late swing (8.5 ± 2.6 vs 6.1 ± 2.8°, p = 0.019) and stance phases of running (13.3 ± 4.2 vs
9.6 ± 3.4°, p = 0.011). Furthermore, late swing peak hip adduction was predictive of subsequent
stance phase peak hip adduction (r = 0.63, p < 0.001).
Conclusion: Sex differences in hip adduction during stance are influenced in part by late swing
phase hip adduction. Further studies are needed to identify potential causes of excessive hip
adduction during the late swing phase of running.
10

INTRODUCTION
Excessive hip adduction is a common movement impairment that has been reported to
underlie various running injuries (2,3,15-17). Compared to males, females exhibit greater degrees
of hip adduction regardless of speed (26,27), inclination (27), or mode of running (i.e., overground
or treadmill) (28). The higher magnitude of hip adduction observed in female runners is thought
to contribute to the reported sex differences in the incidence of patellofemoral pain (15,25) and
iliotibial band syndrome (16,17).
Most studies that have examined sex differences in frontal plane hip kinematics have
focused primarily on the stance phase of running (9,26,28,34,39-41). However, females also have
been reported to exhibit higher degrees of hip adduction at initial contact (9,26,39). This suggests
that the tendency of female runners to exhibit greater degrees of hip adduction is not just limited
to stance, but also may be present during the late swing phase of running.
During the late swing phase of running, hip adduction serves to align the base of support
with the center of mass in preparation for ground contact (28). Given that females have distinct
differences in pelvis morphology compared to males (i.e., acetabulum orientation and bilateral hip
joint width (42-45)), it is possible that greater degrees of hip adduction would be required to
position the foot relative to the center of mass to achieve stance-phase stability. An increase in
swing phase hip adduction however may bias the hip towards greater degrees of hip adduction
during stance.
Diminished hip abductor strength is commonly believed to contribute to excessive hip
adduction during the stance phase of running. However, studies that have examined this
relationship have reported weak associations (34) or no associations at all (5,8,46,47). It is
conceivable that stance phase hip adduction, may be influenced to a greater degree by the frontal
11

plane orientation of the hip in late swing phase hip adduction. To further explore this premise, the
purpose of the current study was two-fold: 1) to evaluate sex differences in peak hip adduction
during the late swing and stance phases of running and 2) determine if peak hip adduction during
late swing is predictive of peak hip adduction during stance. Based on previous reports in this area,
we hypothesized that, compared to males, females would demonstrate greater peak hip adduction
during the stance and late swing phases of running. In addition, we hypothesized that peak hip
adduction during the late swing phase of running would be predictive of stance phase values for
both males and females. Information gained from this study may provide additional insight into
potential causes of sex differences in lower limb kinematics that are related to running injuries.
METHODS
Participants
Thirty-one recreational runners participated (16 males and 15 females; TABLE 3-1). To
be eligible for the study, participants must have been between 18-45 years of age, and currently
running at least 16 km per week. All participants were natural heel strikers (e.g., runners who
contacted the ground with the rear third of their foot), which was verified using sagittal plane
images from high-speed video (120 Hz). Only heel strike runners were recruited owing to the
known biomechanical differences among runners with varying foot-strike patterns (48).
Potential participants were excluded if they reported any of the following: (1) current lower
extremity or low back pain; (2) previous history of lower extremity surgery, fracture, osteoarthritis,
or hip dysplasia, or (3) any lower extremity pathology that caused pain or discomfort during
running within 6 months prior to participation. An a priori power analysis using data previously
obtained from healthy young female and male runners indicated that 11 subjects per group would
12

be sufficient to detect differences and correlations in variables of interest with a statistical power
of 80% (using an alpha level of 0.05).
TABLE 3-1. Participant demographics.

Females
(n=15)
Males
(n=16)
Age (years) 25.2 ± 3.4 25.6 ± 4.6
Height (m)* 1.7 ± 0.1 1.8 ± 0.1
Weight (kg)* 59.2 ± 8.5 77.5 ± 9.9
Running distance (km/week) 25.3 ± 9.2 34.4 ± 29.2
Stride Length (m) 2.34 ± 0.18 2.44 ± 0.20
Stride Width (m) 0.07 ± 0.05 0.10 ± 0.05
Values are mean ± SD. *: significant sex difference, p<0.05
Instrumentation
Three-dimensional lower extremity kinematics during running were collected using an 11-
camera motion capture system (Qualisys, Gothenburg, Sweden) at a sampling rate of 250 Hz.
Ground reaction force data were obtained at a rate of 1500 Hz using a single force plate (AMTI,
Newton, MA). Kinematic and ground reaction force data were collected and synchronized using a
motion capturing software (Qualisys Track Manager version 2.12).
Procedure
Data were collected at the Jacquelin Perry Musculoskeletal Biomechanics Research
Laboratory at the University of Southern California. Prior to data collection, participants were
informed as to the objectives, procedures, and potential risks of participation in the study and
provided written informed consent as approved by the Health Science Institutional Review Board
of the University of Southern California.
Data were obtained from each participant's dominant leg which was defined as the leg they
13

preferred to use when kicking a ball. Prior to data collection, 21 anatomical markers (reflective
14-mm diameter) were placed on the following bony landmarks: distal phalange of second toes,
first and fifth metatarsal heads, medial and lateral malleoli, medial and lateral epicondyles of
femurs, greater trochanters, iliac crests, L5-S1 junction, and anterior superior iliac spines (ASISs).
In addition, tracking marker clusters mounted on semi-rigid plastic plates were placed on the
posterior sacrum and the lateral surfaces of the participant's thighs, shanks, and heel counters of
the shoes. A standing calibration trial was obtained to define the segmental coordinate systems and
joint axes. After the calibration trial, anatomical markers were removed, except for those at the
ASISs, iliac crests, and L5-S1 junction.
Participants were instructed to run over ground at a controlled speed of 4 m/s along a 14-
meter runway. Running speed was monitored using infrared light-switches place at the ends of the
runway. A successful trial was defined when the running speed was within ± 5% of the target speed
and the foot of the dominant leg fell within the borders of the force plate. A total of 3 successful
trials were collected from each participant.
Data Analysis
Kinematic data were low pass filtered at 20Hz using a fourth-order Butterworth filter (49).
Visual 3D software (C-Motion, Rockville, MD) was used to quantify 3-D hip joint kinematics.
using a Cardan rotation sequence of flexion/extension, abduction/adduction, and internal/external
rotation (50). Peak hip adduction during the late swing phase of running (i.e., last 1/3 of swing
phase) and the deceleration phase of stance (i.e., from heel strike to maximum knee flexion during
stance) were identified for each trial. Data obtained from the 3 trials were averaged for statistical
analysis.
Statistical Analysis
14

Regarding the first hypothesis, sex differences in peak adduction during the late swing
phase and stance phase of running were evaluated using independent sample t-tests. To evaluate
whether peak hip adduction during late swing was predictive of peak adduction during stance
(second hypothesis) multiple linear regression analysis was used. First, we tested for the presence
of a sex interaction. Since no interaction was present, males and females were combined in the
final simple linear regression model. All statistical analyses were performed using R Statistical
Software (R Foundation for Statistical Computing, Vienna, Austria), using significance level
α=0.05.
RESULTS
Hip adduction time series data for males and females are presented in FIGURE 3-1.
Compared to males, females exhibited significantly greater peak hip adduction during both the late
swing phase (8.5 ± 2.6 vs 6.1 ± 2.8°, p = 0.019, FIGURE 3-2a) and stance phase of running (13.3
± 4.2 vs 9.6 ± 3.4°, p = 0.011, FIGURE 3-2b).
FIGURE 3-1. Sex difference in hip adduction during running. Data are mean ± SD.

15

Results of the multiple regression analysis did not reveal a significant sex interaction
(p>0.05). As such, data from males and females were pooled in the final model. Peak hip adduction
during late swing was found to predict peak hip adduction during stance (r = 0.63, p < 0.001,
FIGURE 3-3).
DISCUSSION
The purpose of the current study was to evaluate sex differences in peak hip adduction
during the late swing and stance phases of running and determine whether peak hip adduction
during late swing was predictive of peak hip adduction during stance. Consistent with our
hypotheses, significant sex differences in peak hip adduction were found during both stance and
late swing phases. In addition, peak adduction during late swing was found to be predictive of peak
hip adduction during stance.
FIGURE 3-2. Sex difference in peak hip adduction angles during late swing phase (a) and
stance phase (b) of running. Data are mean ± SD. *: p<0.05
16

Our finding of sex differences in peak hip adduction during the stance phase of running is
consistent with previous studies (9,26-28,40,41). On average, the females in the current study
exhibited approximately 4 degrees greater peak hip adduction during stance than males. This
difference is similar in magnitude to previous studies that have reported sex differences peak hip
adduction ranging from 4 to 6 degrees (9,26-28,40,41). Although small, differences in hip
adduction in this range are considered clinically meaningful as prospective studies involving
female runners have reported that peak hip adduction differences of 4-5 degrees can discriminate
among runners who developed iliotibial band syndrome (2) and patellofemoral pain (3) from those
who do not.
Regarding peak hip adduction during late swing, females exhibited 2 degrees greater hip
adduction on average compared to males. Although statistically significant, this finding is smaller
in magnitude to previously reported sex differences in hip adduction at initial contact, which have
FIGURE 3-3. Correlation plot showing the association between peak hip adduction during
the late swing and stance phases of running. Grey shading represents the standard error of
the slope.
R
2
=0.40, p<0.01

17

averaged about 4 degrees (9,26,39). Differences among studies may be due to the fact that the
current investigation evaluated peak hip adduction during the last one-third of the swing phase as
opposed to initial contact. However, our results related to late swing phase hip adduction are in
agreement with graphical results of Chumanov et al. (27) and Schache et al. (51).
The novel finding of the current study was that peak hip adduction during late swing was
predictive of peak hip adduction during stance. Specifically, 40% of the variance in stance phase
peak hip adduction could be explained by late swing values. Interestingly this percent of explained
variance for peak hip adduction during the stance phase of running is considerably higher than
what has been reported for hip abductor strength. For example, studies that have reported
significant associations between hip abductor strength and hip adduction during running have
reported R
2
values up to 16% (5,8,34,46,47). Taken together, our results suggest that sex difference
in hip adduction during the stance phase of running has its origin, at least in part, during late swing.
Although determining reasons for increased hip adduction during late swing of running
was beyond the scope of the current study, it is interesting to speculate on potential causes. One
potential contributing factor may be related to known sex differences in pelvis morphology. For
example anatomical differences such as pelvis width, bilateral hip joint center distance, acetabular
orientation, and femoral neck-shaft angle could result in the need for greater hip adduction during
late swing to adequately position the foot relative to the center of mass to achieve stance phase
stability. Although previous studies have evaluated the association between measures of pelvis and
femur morphology with hip adduction during the stance phase of running (e.g. femoral neck shaft
angle (5), hip joint center width (27), bi-greater trochanteric distance (36)), the reported amount
of variance in hip adduction explained by these morphologic measures have been non-significant
(5,27) or very low (R
2
=0.05) (36). It is possible that pelvis and femur morphology may be more
18

predictive of hip adduction during swing, as motion during this phase is less likely to be influenced
by additional variables such as external demand (i.e., adduction moment) and hip abductor strength.
Several limitations need to be considered when interpreting the results of this study. First,
the participants from our study were young, healthy recreational runners. Caution is needed when
generalizing results of this study to injured runners or older populations. In addition, we only
examined heel strike runners. As such, our findings may not apply to forefoot or midfoot strike
runners. Lastly participants were instructed to run at 4 m/s which is a moderate speed. Whether or
not our results would be relevant to faster speed running or sprinting remains to be determined.
CONCLUSION
When compared to males, females demonstrated greater amounts of hip adduction during
the stance and swing phases of running. In addition, peak hip adduction during late swing was
predictive of respective peak angles during the stance phase of running. As such, the hip position
before ground contact appears to be an important determinant of stance phase hip adduction. Future
studies should consider causes of late swing kinematics during running including the influence of
bony morphology and hip abductor muscle activation as well as the potential impact of spatial-
temporal characteristics such as cadence, stride length, and stride width. In addition, future
research should be directed towards understanding if hip re-positioning during late swing can be
used as a potential method to affect peak hip adduction during stance.
19

Chapter 4. Sex Differences in Hip Adduction during Running: Influence
of Hip Abductor Strength, Activation, and Pelvis & Femur Morphology
ABSTRACT
Purpose: To examine the influence of hip abductor strength, neuromuscular activation, and pelvis
& femur morphology in contributing to sex difference in hip adduction during running.
Methods: Fifteen female and 14 male runners underwent strength testing, instrumented analysis
of overground running (e.g., kinematics and electromyography)), and computed tomography of
the pelvis and femur. Morphology measurements included bilateral hip width to femur length ratio,
acetabulum abduction, acetabulum anteversion, femoral anteversion, and femoral neck-shaft
angles. Sex differences for all variables of interest were examined using independent t tests. Linear
regression was used to assess the ability of each variable of interest to predict peak hip adduction
during the late swing and stance phase of running.
Results: Compared to males, females exhibited significantly greater peak hip adduction during
both the late swing (8.5 ± 2.6° vs 6.2 ± 2.8°, p = 0.04) and stance phases of running (13.4 ± 4.2°
vs 10.0 ± 3.2°, p = 0.02). In addition, females exhibited significantly lower hip abductor strength
(1.8 ± 0.3 vs 2.0 ± 0.3 Nm/kg, p=0.04), greater femoral neck-shaft angles (134.1 ± 5.0° vs 129.9
± 4.1°, p=0.01), and greater hip width to femur length ratios (0.44 ± 0.02 vs 0.42 ± 0.03, p=0.03)
than males. The femoral anteversion was found to be the only significant predictor of peak hip
adduction during late swing (r=0.36, p=0.05) and stance (r=0.41, p=0.03).
Conclusion: Only a small amount of variance in peak hip adduction could be explained by
variables examined in this study. Our findings highlight the difficulty in predicting hip kinematics
during complex motor tasks using simple measures of muscle performance and skeletal structure.
20

INTRODUCTION
Compared to males, females have been reported to exhibit higher degrees of hip adduction
during running (4,19,26,40,52). Greater hip adduction has been shown to be contributory to several
running injuries (2,3,15-17) and may explain the higher incidence of certain running injuries in
women compared to men (i.e., patellofemoral pain) (14,15). For this reason, there has been interest
in understanding the underlying causes of excessive hip adduction in female runners.
Sex differences in hip adduction have long been postulated to be the result of diminished
strength and/or activation of the hip abductors. While some studies have reported lower hip
abductor torque production in females compared to males (8,30,31), others have not (32-34). In
addition, previous investigations have found that hip abductor strength is either not correlated with
hip adduction (9,10,12,13), or weakly associated in females only (34). With respect to
neuromuscular recruitment, previous studies have reported similar levels of gluteus medius
activation between males and females during various activities, even though females in these
studies demonstrated significantly greater hip adduction than their male counterparts (9,10,13).
Furthermore, gluteus medius activation has not been reported to be predictive of hip adduction
during running (27).
Another potential contributor to sex differences in hip adduction during running may be
pelvis and femur skeletal morphology. For example, a greater bilateral hip width relative to femur
length ratio would theoretically require larger degrees of hip adduction to position the foot under
the center of mass at ground contact. In addition, smaller degrees of acetabulum abduction and/or
femoral neck-shaft angles could bias the hip into greater hip adduction. Furthermore, there is
evidence that femoral anteversion and acetabulum anteversion may affect frontal plane hip
kinematics (53). The potential influence of femoral anteversion and acetabulum anteversion on hip
21

adduction is highlighted by a study that reported individuals who exhibited greater femoral
anteversion also exhibited greater degrees of frontal plane knee motion during a single-leg landing
task (53). Taken together, the premise that bony structure (as opposed to strength) may play a role
in influencing hip kinematics is supported by the findings from Chapter III that hip adduction
during the late swing phase of running was highly predictive of peak hip adduction during stance.
To date, few studies have examined the potential influence of pelvis and/or femur
morphology on hip adduction during running. Studies examining the hip width to limb length ratio
(27) and bilateral greater trochanteric distance to femur length ratio (36) have reported non-
significant or very low associations with hip adduction during running, respectively. Similarly,
Baggaley et al. (5) reported that femoral neck-shaft angle was not associated with the magnitude
of hip adduction in females during running. However, a limitation of previous work in this area is
that measures of pelvis and femur morphology were either estimated from external markers (27)
or were obtained from 2D radiographs (5). Such measurements are prone to error and may not
accurately represent true structural morphology (37,38).
Given that joint kinematics are likely influenced by the complex interaction of muscular
actions and skeletal structure, the purpose of the current study was to examine the role of hip
abductor strength, neuromuscular activation, and pelvis and femur morphology (as obtained from
computed tomography (CT)) in contributing to sex differences in hip adduction during running.
Based on existing literature, we hypothesized that females would exhibit lower hip abductor torque
production, but similar hip abductor activation during running when compared to males. In
addition, we hypothesized that when compared to males, females would exhibit differences in
pelvis and femur morphology that would be suggestive of higher degrees of hip adduction during
running (i.e., greater pelvis width relative to femur length, lower acetabulum abduction and
22

femoral neck-shaft angles, and greater degrees of acetabulum anteversion and femoral anteversion).
Lastly, we hypothesized that measures of pelvis and proximal femur morphology would explain a
greater amount of the variance in hip adduction during the late swing and stance phases of running
compared to hip abductor strength and activation.
METHODS
Participants
Twenty-nine recreational runners participated in this study (14 males and 15 females;
TABLE 4-1). To be eligible for this study, participants must have been between 18-45 years of
age, and currently running at least 16 km per week. All participants were natural heel strikers (e.g.,
runners who contacted the ground with the rear third of their foot), which was verified using
sagittal plane images from high-speed video (120 Hz). Only heel strike runners were recruited
owing to the known biomechanical differences among runners with varying foot-strike patterns
(48).
TABLE 4-1. Participant demographics.

Females
(n=15)
Males
(n=14)
Age (years) 25.2 ± 3.4 26.2 ± 4.5
Height (m)* 1.7 ± 0.1 1.8 ± 0.1
Weight (kg)* 59.2 ± 8.5 76.7 ± 8.6
Running distance (km/week) 25.3 ± 9.2 36.3 ± 30.7
Values are mean ± SD. *: significant sex difference, p<0.05
Potential participants were excluded if they reported any of the following: 1) women who
were pregnant or breastfeeding;(2) current lower extremity pain; 3) previous history of lower
extremity surgery, fracture, osteoarthritis, hip dysplasia, or 4) any lower extremity condition that
23

resulted in pain or discomfort during running within 6 months prior to participation. An a priori
power analysis indicated that 29 participants was sufficient to achieve statistical power of 80% for
a moderate effect size (expected correlations of 0.5) using an alpha level of 0.05.
Procedure
After screening and enrollment, each participant completed 3 phases of data collection: 1)
CT of the pelvis and femur, 2) hip abductor strength testing, and 3) instrumented gait analysis.
Prior to data collection, participants were informed as to the objectives, procedures, and potential
risks of participation in the study and provided informed consent as approved by the Health
Science Institutional Review Board of the University of Southern California.
CT Scanning
Pelvis and femur morphology data were obtained using a 320 detector Toshiba Aquilion
One CT scanner. Participants were positioned supine in the CT system with the knees extended.
The lower limbs were stabilized in a neutral position to avoid hip external rotation and abduction.
Axial plane images of the pelvis and both femurs were obtained for each participant (0.5 mm slice
thickness, zero tilt, 100 mA, and 80 KVP).
Hip Abductor Strength Testing
Hip abductor strength of each participant’s dominant leg (defined as the leg preferred to
kick a ball) was measured using a Cybex dynamometer (Computer Sports Medicine Inc.,
Stoughton, MA). The dynamometer provided torque values in Nm, with a precision of 0.02% (full
scale). The sampling frequency was 100 Hz.
Electromyographic (EMG) signals of the gluteus medius were obtained during the strength
testing trials for normalization purposes. Surface electrodes were placed over the gluteus medius
at the midpoint between iliac crest and greater trochanter (54). EMG signals were collected at a
24

sampling rate of 3000 Hz using a Noraxon Telemyo DTS EMG system (Noraxon Inc, AZ, USA).
The system had a differential input impedance of greater than 100 MOhm and a common-mode
rejection ratio greater than 100 dB.
Participants were positioned in side-lying with the axis of the dynamometer aligned with
the hip joint center. The resistance pad was positioned just proximal to the lateral knee joint line
and secured to the distal thigh with straps. Three trials of 5-second maximal efforts were performed
with a one-minute break between trials. Verbal encouragement was given throughout testing. Hip
abductor strength was quantified by averaging the peak hip abduction torques obtained from the 3
trials. Torque values were normalized to body mass.
Instrumented Running Analysis
Three-dimensional lower extremity kinematics during running were collected using an 11-
camera motion capture system (Qualisys, Gothenburg, Sweden) at a sampling rate of 250 Hz.
Ground reaction force data were obtained at a rate of 1500 Hz using a single force plate (AMTI,
Newton, MA).
Prior to data collection, 21 anatomical markers (reflective 14-mm diameter) were placed
on the following bony landmarks: distal phalanges of second toes, first and fifth metatarsal heads,
medial and lateral malleoli, medial and lateral epicondyles of femurs, greater trochanters, iliac
crests, L5-S1 junction, and anterior superior iliac spines (ASISs). In addition, marker clusters
mounted on semi-rigid plastic plates were placed on the back of sacrum and the lateral surfaces of
the participant's thighs, shanks, and heel counters of the shoes. A standing calibration trial was
first obtained to define the segmental coordinate systems and joint axes. After the calibration trial,
anatomical markers were removed, except for those at the ASISs, iliac crests, and L5-S1 junction.
Following marker placement participants were instructed to run overground along a 14-
25

meter runway at a controlled speed of 4 m/s. A successful trial was defined when the running speed
was within ± 5% of the target speed and the foot of the dominant leg fell within the borders of the
force plate from initial contact to toe-off. A total of 3 successful trials were collected from each
participant’s dominant leg. Kinematic, kinetic, and EMG data were collected and synchronized
using a motion capturing software (Qualisys Track Manager version 2.12).
Data Analysis
Morphology Measurements
Using a commercial software package (Avizo, FEI Visualization Sciences Group, USA),
CT slices of pelvis and femur were separately segmented, reconstructed, and smoothed to derive
3D subject-specific pelvis and femur geometry models. Manually selected landmarks were
identified on each model (anterior superior iliac spines, pubic symphysis, 32 evenly spaced points
along the acetabulum rim, medial and lateral femoral epicondyles, and the point immediately
inferior to lesser trochanter). In addition, regions of femoral head and femoral head-neck junction
were also manually selected from the 3D femur model for purposes of defining femoral head center
and femoral head-neck axis, respectively. Next, the models and selected landmarks were then
imported into MATLAB software (MathWorks Inc., MA, USA) to obtain the structural
measurements of interest from the dominant leg.
To obtain measures of acetabulum orientation, the pelvis coordinate system (FIGURE 4-
1a) was first defined. The frontal plane was determined as the plane passing through the bilateral
ASISs and pubic symphysis. The transverse plane was defined as the plane containing bilateral
ASISs and perpendicular to the frontal plane. The sagittal plane was then defined as the plane
passing through the pubic symphysis and perpendicular to both frontal and transverse planes.
Following establishment of the pelvis coordinate system, the acetabulum face plane was
26

defined as the best fit plane of the selected bony points along acetabulum rim using a least square
plane fitting algorithm (FIGURE 4-1b). A vector perpendicular to the acetabulum face plane was
created to represent the acetabulum orientation vector (FIGURE 4-1b). Acetabular abduction was
measured as the acute angle between the sagittal plane of the pelvis and the projection line of the
acetabulum orientation vector onto the frontal plane of the pelvis (FIGURE 4-1c). Acetabular
anteversion was measured as the angle between the frontal plane of the pelvis and the projection
line of the acetabulum orientation vector onto the transverse plane of the pelvis (FIGURE 4-1d).
To evaluate proximal femur orientation, the femur coordinate system (FIGURE 4-2a) was
first defined. The frontal plane of the femur was defined by the plane tangent to the posterior
surface of the femur. Similarly, the transverse plane was defined by the plane that was tangent to
FIGURE 4-1. Pelvis morphology measurements. (a) pelvis coordinate system, (b) acetabulum
face plane and acetabulum orientation vector, (c) acetabular anteversion angle, (d) acetabular
abduction angle.
27

the inferior surface of the femur. The sagittal plane was then defined as the plane perpendicular to
both the frontal and transverse planes.
The femoral head-neck axis was created using a consecutive ellipse fitting algorithm (55)
(FIGURE 4-2b). The femoral shaft axis was determined by the first principal component vector
determined from all cloud points of the shaft surface below lesser trochanter and above femoral
condyles (FIGURE 4-2b). The femoral neck-shaft angle was then measured as the obtuse angle
between the projected lines of femur head-neck axis and femur shaft axis onto the frontal plane of
femur (FIGURE 4-2c). Femoral anteversion angle measured as the angle between the frontal plane
of femur and the projection line of the femoral head-neck axis onto the transverse plane of femur
(FIGURE 4-2d).
As an indicator of pelvis width, the linear distance between the left and right hip joint
centers was measured. The femoral head centers were defined as the center of a sphere fit to the
respective femoral head surface models. The bilateral hip width distance was normalized to femur
length of the dominant leg, which was defined as the linear distance between the femoral head
FIGURE 4-2. Proximal femur morphology measurements. (a) femoral coordinate system, (b)
femoral head-neck and shaft axes, (c) femoral neck-shaft angle, (d) femoral anteversion angle.
28

center and the midpoint between medial and lateral epicondyles.
EMG Activity
Raw EMG signals obtained from gluteus medius during the strength testing and running
trials were processed in MATLAB software (MathWorks Inc., MA, USA). EMG data were band-
pass filtered at 20 to 450 Hz and smoothed using a 100 ms window (56). Maximum activation of
gluteus medius was determined as the peak value from the 3 strength trials. EMG data during
running were then normalized to the peak value and expressed as a percentage of maximum
voluntary isometric contraction (% MVIC).
Muscle activity of gluteus medius was quantified during 2 phases of the running cycle: late
swing and stance. Late swing phase muscle activation was quantified as the average activity 100
ms prior to peak hip adduction. Stance phase muscle activation was quantified as the average EMG
activity from initial contact and peak hip adduction. Data obtained from the 3 successful trials were
averaged for statistical analysis.
Hip Kinematics
Visual 3D software (C-Motion, Rockville, MD) was used to quantify 3D kinematics of the
dominant leg during running. Motion trajectory data were low pass filtered at 20 Hz using a 4th
order Butterworth filter (49). Joint kinematics were calculated using Cardan angles with a rotation
sequence of flexion/extension, abduction/adduction, and internal/external rotation (50). Hip joint
angles were calculated as the motion of the femur relative to pelvis. Peak hip adduction during the
late swing phase and stance phase was identified for each trial. Data obtained from the 3 successful
trials were averaged for statistical analysis.
Statistical Analysis
Independent sample t-tests were used to compare sex differences in hip abductor strength,
29

gluteus medius activation (stance and swing), and the morphology variables of interest. Multiple
linear regression was used to evaluate the ability of each independent variable of interest to predict
peak hip adduction during late swing and stance. For each analysis, we first tested for the presence
of a sex interaction, if no interaction was present, males and females were combined in the final
regression model. Otherwise, potential relationship in each sex group was analyzed separately. All
statistical analyses were performed using R Statistical Software (version 3.6.3; R Foundation for
Statistical Computing, Vienna, Austria), using significance level α=0.05.
RESULTS
Time series data for frontal plane hip motions during running are presented in FIGURE 4-
3. Compared to males, females exhibited significantly greater peak hip adduction during both the
late swing phase (8.5 ± 2.6° vs 6.2 ± 2.8°, p = 0.04) and stance phase of running (13.4 ± 4.2° vs
10.0 ± 3.2°, p = 0.02).
Descriptive data for each of the predictor variables of interest for both males and females
FIGURE 4-3. Sex difference in frontal plane hip motion during running. Data are mean ± SD.

30

are presented in TABLE 4-2. Compared to males, females exhibited significantly lower hip
abductor torque production than males (p=0.04, TABLE 4-2). Of the morphology variables
examined, femoral neck-shaft angle (p=0.01, TABLE 4-2) and the bilateral hip width to femur
length ratio (p=0.03, TABLE 4-2) were significantly larger in females than in males. There were
no significant sex differences in hip abductor muscle activation or any other morphological
variables of interest.
TABLE 4-2. Sex differences in hip abductor muscle strength, muscle activation, and pelvis and femur
morphology variables.
Variables
Females
(n=15)
Males
(n=14)
p-value
effect
size
Hip Abductor Strength (Nm/kg) 1.8 ± 0.3 2.0 ± 0.3 0.04* 0.70
Late Swing Gluteus Medius Activation (%MVIC) 16.5 ± 6.6 15.7 ± 8.1 0.33 0.11
Stance Phase Gluteus Medius Activation (%MVIC) 50.8 ± 17.8 46.8 ± 15.7 0.39 0.24
Acetabulum Abduction Angle (°) 50.6 ± 3.1 49.9 ± 5.0 0.34 0.16
Acetabulum Anteversion Angle (°) 21.6 ± 6.4 19.8 ± 5.5 0.21 0.30
Femoral Neck Shaft Angle (°) 134.1 ± 5.0 129.9 ± 4.1 0.01* 0.91
Femoral Anteversion Angle (°) 14.7 ± 10.6 11.9 ± 8.2 0.21 0.30
Bilateral Hip Width/Femur Length 0.44 ± 0.02 0.42 ± 0.03 0.03* 0.77
Values are mean ± SD. *: significant sex difference, p<0.05
TABLE 4-3. Predictions of hip abductor muscle strength, muscle activation, and pelvis & femur
morphology variables to peak hip adduction angles during running.
Variables
Prediction to Late Swing
Peak Hip Adduction
Prediction to Stance
Peak Hip Adduction
r R
2
p-value r R
2
p-value
Hip Abductor Strength -0.27 0.07 0.16 -0.08 0.01 0.69
Gluteus Medius Activation 0.05 0.03 0.80
F: 0.50
M: -0.43
0.25
0.19
F: 0.06
M: 0.14
Acetabulum Abduction Angle -0.09 0.01 0.33 -0.16 0.03 0.20
Acetabulum Anteversion Angle -0.08 0.01 0.35 0.02 <0.01 0.46
Femoral Neck Shaft Angle 0.24 0.06 0.10 0.03 <0.01 0.43
Femoral Anteversion Angle 0.36 0.13 0.03* 0.41 0.16 0.01*
Bilateral Hip Width/Femur Length 0.12 0.02 0.26 0.26 0.07 0.09
F: Females, M: Males. *: significant correlation, p<0.05
31

Predictors of Late Swing Phase Peak Hip Adduction
No sex interactions were identified for any of the regression models related to the
prediction of late swing phase peak hip adduction angles (p>0.05). As such, data for males and
females were combined in each model. Femoral anteversion was found to predict late swing phase
hip adduction (p=0.03, TABLE 4-3, FIGURE 4-4a). No other variables of interest were found to
e predictive of peak hip adduction during this phase.
Predictors of Stance Phase Peak Hip Adduction
A significant sex interaction effect was identified for gluteus medius activation (p<0.05).
As such, males and females were considered separately. Gluteus medius activation was not
predictive of stance phase peak hip adduction in either sex group (p>0.05, TABLE 4-3). There
were no significant sex interactions identified in the remaining linear regression models (p>0.05).
For males and females combined, only the femoral anteversion angle was predictive of peak stance
phase hip adduction (p =0.01, TABLE 4-3, FIGURE 4-4b). No other variables of interest were
r=0.36, p=0.03 r=0.41, p=0.01
FIGURE 4-4. Correlations of femoral anteversion angle to peak hip adduction during the late swing
phase (a) and stance phase (b) of running. Red dots represent females, and green dots represent males.

32

predictive of peak hip adduction during the stance.
DISCUSSION
The purpose of the current study was to comprehensively evaluate the influence of pelvis
and femur morphology, hip abductor strength, and muscle activation on hip adduction during the
late swing and stance phases of running. For all variables examined, only femoral anteversion was
found to be predictive of peak hip adduction during the late swing and stance phases of running.
Our findings highlight the difficulty in predicting movement behavior during complex motor tasks
using simple measures of muscle performance and skeletal structure.
Of the 5 morphology variables examined, sex differences only were found for the bilateral
hip width to femur length ratio and femoral neck-shaft angle. With respect to our finding of higher
hip width to femur length ratio, there is evidence that females have a greater bilateral acetabulum
distance and a shorter femur length compared to males (57). The finding of a greater femoral neck-
shaft angle in females is consistent with previous studies (58-61) but not all (57,61-63).
Inconsistent findings among studies can be attributed to differences in sample sizes and
populations evaluated, as the femoral neck-shaft angle has been reported to decrease with age (58).
As for measures of acetabulum abduction, acetabulum anteversion, and femoral
anteversion, we did not find statistical differences between males and females. Our findings are in
agreement with previous studies (57,64,65) but in contrast to others that have reported significant
sex differences in acetabulum abduction (66-68), acetabulum anteversion (62,65-68), and femoral
anteversion (60,62). Inconsistent findings in literature could be a result of different methodologies
used to obtain morphological measurements (2D vs 3D). In addition, the age of populations
examined may be a contributing factor as it has been shown that pelvis morphology differs between
females and males post-puberty and pre-menopause, but are similar otherwise (69).
33

Regarding the ability of the morphology variables of interest to predict peak hip adduction
during running, only femoral anteversion emerged as a predictor during late swing and stance. It
is important to note however that only 13% and 16% of the variance in peak hip adduction was
explained by femoral anteversion for the late swing and stance phases, respectively. The fact that
a transverse plane morphology variable was found to be predictive of frontal plane motion
highlights the potential of an inter-planar influence of femur morphology on joint kinematics. This
could be due in part to the crosstalk that can occur between the frontal and transverse plane hip
joint angles when calculated using Cardan angles (70). This premise of inter-planar influence of
femur morphology measurements on lower limb kinematics is supported by a previous study that
reported persons who exhibit greater femoral anteversion also exhibit greater degrees of frontal
plane knee valgus during single-leg landing tasks (53).
Theoretically, a smaller femoral neck shaft angle would have the potential to bias the hip
into greater adduction. In contrast to this assumption, females exhibited greater femoral neck-shaft
angles despite displaying greater degrees of hip adduction than males. The finding that the femoral
neck-shaft angle was not predictive of peak hip adduction is in agreement with Baggeley et al. (5)
who also did not find a relationship between the femoral neck-shaft angle and hip adduction in
female runners.
Apart from the morphology measurements, we also examined the potential influence of hip
abductor muscle strength on sex differences in hip adduction during running. Our finding of lower
hip abductor strength in females compared to males is consistent with previous literature (8,30,31).
The lack of predictive ability of hip abductor strength to explain hip adduction during running also
agrees with previous studies that involved both female and male participants (9,10,12,13,34).
Taken together, the findings of the current study and previous work in this area call into question
34

the influence of hip abductor strength on hip adduction during running.
Lastly, our results related to gluteus medius neuromuscular recruitment revealed similar
levels of activation between males and females during the late swing and stance phases. This
finding is consistent with previous studies in this area (9,27). In addition, gluteus medius activation
was not predictive of stance phase hip adduction, which is in agreement with results of Chumanov
et al. (10). Similar to hip abduction strength, our results call into question the role of hip abductor
activation in contributing to hip adduction during running.
Several limitations need to be considered when interpreting the results of this study. First,
the participants in our study were young, healthy, recreational runners. Caution is needed when
generalizing results of this study to various clinical populations in which greater degrees of hip
adduction may be expected (i.e., runners with patellofemoral pain or iliotibial band syndrome). In
addition, a larger sample size would be required to detect sex differences in several of the
morphology variables examined. Last, we only examined a few of the more common pelvis and
femur morphological variables in this study. Further exploration of other morphology variables in
relation to hip adduction during running is indicated.
CONCLUSION
Hip abductor muscle strength and gluteus medius activation were not found to be predictive
of frontal plane hip kinematics during running. In contrast, femoral anteversion was found to be
predictive of peak hip adduction during late swing and stance phase of running. However, only a
small amount of the variance in peak hip adduction could be explained by this variable. Our
findings highlight the complexity of predicting joint kinematics based on traditional measures of
muscle performance and morphology. Future studies in this area should consider the inter-planar
interaction effects of morphologic measures on sex differences in hip adduction during running.
35

Chapter 5. Classification of Runners with High versus Low Hip
Adduction using Pelvis and Femur Morphology Measures
ABSTRACT
Purpose: To determine the most relevant pelvis and femur morphological characteristics for
differentiating runners with high versus low hip adduction during running.
Methods: Fifteen female and 14 male runners underwent instrumented kinematics analysis of
overground running and CT scanning of pelvis and femur. Peak hip adduction during the stance
phase of running was identified for each participant. Using the cohort average peak hip adduction
as the classifying threshold, participants were categorized into high or low hip adduction groups.
To determine the most relevant morphologic features for discriminating high and low hip
adduction runners, a feature selection-based support vector machine classification analysis was
performed. The relationships between the best discriminant variables and classification outcomes
were then assessed using sensitivity analyses.
Results: Out of 15 morphology variables examined, femoral head anteversion and femur length
were shown to be the best discriminant variables for group classification. Together, these variables
achieved a prediction accuracy of 0.93, sensitivity of 1, and specificity of 0.88. Sensitivity analyses
revealed a greater importance of femoral head anteversion than femur length in contributing to hip
adduction group classification.
Conclusion: Our results highlight the importance of femur morphology in contributing to
increased hip adduction during running. Further study is needed to fully understand the underlying
mechanism(s) of how femur morphology influences hip adduction during running.
36

INTRODUCTION
Excessive hip adduction is a common movement impairment that has been proposed to
underlie various running injuries, such as patellofemoral pain (3,14,15) and iliotibial band
syndrome (16,17). Given that females have been reported to exhibit greater degrees of hip
adduction during running compared to males (4,19,26,40,52), there has been interest in
understanding potential causes of sex differences in frontal plane hip kinematics. Since females
and males differ in various aspects of pelvis and femur morphology features, it has been proposed
that bony morphology may play a role (5,27,36,71).
To date, 4 studies have investigated the potential influence of pelvis and/or femur
morphology on hip adduction during running. However, non-significant or weak relationships
have been reported. For example, Brindle et al (36) examined the relationship between bilateral
greater trochanteric distance (normalized to femur length) and peak stance phase hip adduction in
female runners and reported a significant yet very low association (R
2
=0.05). Similarly, Chumanov
et al (27) revealed that bi-trochanteric width to limb length ratio was significantly associated with
hip adduction excursion during walking (R
2
=0.14) but not during running. With respect to angular
measurements, Baggaley et al. (5) reported that femoral neck-shaft angle was not associated with
the magnitude of hip adduction in females during running. In addition, the findings from Chapter
IV examined the relationships between 5 pelvis and femur morphology measures and peak hip
adduction during stance of running and found that femoral head-neck anteversion was the only
predictor of hip adduction during running (R
2
=0.16).
A common limitation in the aforementioned studies is that a linear one-to-one relationship
between hip adduction and the various morphology variables examined has been assumed.
Statistical approaches that utilize linear regression or simple correlations may be susceptible to
37

noise and/or outliers, and consequently, often lack sufficient sensitivity to identify relevant
relationships. Another method to determine the most relevant morphological factors associated
with hip adduction during running is the hybrid approach of classification analysis and feature
selection techniques. In contrast to regression analysis, classification relaxes the one-to-one
assumption between potential predictors and the dependent variable of interest and instead seeks
a boundary that best separates groups of interests. In particular, the support vector machine (SVM)
has been shown as the most robust classification algorithm as it is highly effective when working
with a high dimensional dataset and capable of performing both linear and non-linear
classifications (72-76). However, traditional SVM classification analysis considers all independent
variables entered, and thus is incapable to select the best performing factors. With respect to
solving redundancy, feature selection techniques (e.g., best subset, forward, or backward selection)
have been commonly employed to identify the subset of variables or features that best distinguish
among groups. While feature selection-based SVM analysis has not been typically used in the
running literature, it has shown promise in various applications, such as gene selection for cancer
detection (76) and biomechanics feature identification for sports brain injury classification (75).
Using a feature selection-based SVM algorithm, the purpose of this current study was to
determine the most relevant pelvis and femur morphological variables for discriminating runners
with high and low hip adduction during the stance of running. Following identification of
discriminant variables, a secondary sensitivity analysis was performed to examine the order of
importance of each discriminant variable with respect to group classification. Information of this
study will extend our understanding of the influence of pelvis and femur morphology on hip
adduction during running.

38

METHODS
Participants
The current study represents a secondary analysis of data collected in Chapter IV. Twenty-
nine recreational runners participated in this study (14 males and 15 females; TABLE 4-1). To be
eligible for this study, participants must have been between 18-45 years of age, and currently
running at least 16 km per week. All participants were natural heel strikers (e.g., runners who
contacted the ground with the rear third of their foot), which was verified using sagittal plane
images from high-speed video (120 Hz). Only heel strike runners were recruited owing to the
known biomechanical differences among runners with varying foot-strike patterns (48).
Potential participants were excluded if they reported any of the following: (1) women who
were pregnant or breastfeeding; (2) current lower extremity pain; (3) previous history of lower
extremity surgery, fracture, osteoarthritis, hip dysplasia, or (4) any lower extremity condition that
resulted in pain or discomfort during running within 6 months prior to participation.
Procedure
As reported in Chapter IV, each participant completed 2 phases of data collection: 1) CT
scanning of the pelvis and femur, and 2) instrumented gait analysis. Prior to data collection,
participants were informed as to the objectives, procedures, and potential risks of participation in
the study and provided informed consent as approved by the Health Science Institutional Review
Board of the University of Southern California.
CT Scanning
Pelvis and femur morphology data were obtained using a 320 detector Toshiba Aquilion
One CT scanner. Participants were positioned supine in the CT system with the knees extended.
39

The lower limbs were stabilized in a neutral position to avoid hip external rotation and abduction.
Axial plane images of the pelvis and both femurs were obtained for each participant (0.5 mm slice
thickness, zero tilt, 100 mA, and 80 KVP).
Instrumented Running Analysis
Three-dimensional lower extremity kinematics during running were collected using an 11-
camera motion capture system (Qualisys, Gothenburg, Sweden) at a sampling rate of 250 Hz.
Ground reaction force data were obtained at a rate of 1500 Hz using a single force plate (AMTI,
Newton, MA).
Prior to data collection, 21 anatomical markers (reflective 14-mm diameter) were placed
on the following bony landmarks: distal phalanges of second toes, first and fifth metatarsal heads,
medial and lateral malleoli, medial and lateral epicondyles of femurs, greater trochanters, iliac
crests, L5-S1 junction, and anterior superior iliac spines (ASISs). In addition, tracking marker
clusters mounted on semi-rigid plastic plates were placed on the back of the sacrum and the lateral
surfaces of the participant's thighs, shanks, and heel counters of the shoes. A standing calibration
trial was first obtained to define the segmental coordinate systems and joint axes. After the
calibration trial, anatomical markers were removed, except for those at the ASISs, iliac crests, and
L5-S1 junction.
Following marker placement, participants were instructed to run overground along a 14-
meter runway at a controlled speed of 4 m/s. A successful trial was defined when the running speed
was within ± 5% of the target speed and the foot of the dominant leg fell within the borders of the
force plate from initial contact to toe-off. A total of 3 successful trials were collected from each
participant’s dominant leg. Kinematic and kinetic data were collected and synchronized using a
motion capturing software (Qualisys Track Manager version 2.12).
40

Data Analysis
Hip Kinematics
Visual 3D software (C-Motion, Rockville, MD) was used to quantify 3-D kinematics of
the dominant leg during running. Motion trajectory data were low pass filtered at 20 Hz using a
4th order Butterworth filter (49). Joint kinematics were calculated using Cardan angles with a
rotation sequence of flexion/extension, abduction/adduction, and internal/external rotation (50).
Hip joint angles were calculated as the motion of the femur relative to pelvis. Peak hip adduction
during the stance phase was identified for each trial. Data obtained from the 3 successful trials
were averaged.
Morphology Measurements
Using a commercial software package (Avizo, FEI Visualization Sciences Group, USA),
CT images of pelvis and femur were separately segmented, reconstructed, and smoothed to derive
3D subject-specific pelvis and femur geometry models. Manually selected landmarks and surface
regions were identified on the pelvis (anterior superior iliac spines, pubic symphysis, 32 evenly
spaced points along the acetabulum rim, and acetabulum fossa cloud points) and femur (femoral
head surface cloud points, femoral neck surface cloud points, femoral head-neck surface cloud
points, medial and lateral femoral epicondyles, the point immediately inferior to lesser trochanter)
models. Next, the models and selected landmarks were then imported into MATLAB software
(MathWorks Inc., MA, USA) to obtain a total of 15 structural measurements of interest.
Pelvis Morphology. To obtain pelvis morphology measurements, the pelvis coordinate
system (FIGURE 4-1a) was first defined. The frontal plane was determined as the plane passing
through the bilateral ASISs and pubic symphysis. The transverse plane was defined as the plane
containing bilateral ASISs and perpendicular to the frontal plane. The sagittal plane was then
41

defined as the plane passing through pubic symphysis and perpendicular to both frontal and
transverse planes. To eliminate the influence of subject positioning during CT scanning on
subsequent pelvis measurements, the pelvis geometry model and all associated landmarks/regions
were neutralized using the calculated pelvis coordinate system transformation matrix, prior to
obtaining pelvis measurements.
Following establishment of the pelvis coordinate system, measurements of acetabulum
abduction and anteversion were obtained. First, the acetabulum face plane was defined, which is
best fit plane of the selected bony points along acetabulum rim using a least square plane fitting
algorithm (FIGURE 4-1b). A vector perpendicular to the acetabulum face plane was created to
represent the acetabulum orientation vector (FIGURE 4-1b). The acetabular abduction angle was
then measured as the acute angle between the sagittal plane of the pelvis and the projection line of
the acetabulum orientation vector onto the frontal plane of the pelvis (FIGURE 4-1c). The
acetabular anteversion angle was measured as the angle between the frontal plane of the pelvis and
the projection line of the acetabulum orientation vector onto the transverse plane of the pelvis
(FIGURE 4-1d).
In addition to the angular measurements described above, 4 linear measurements of the
pelvis were obtained. These measurements included bilateral iliac crest width, bilateral ASIS width,
bilateral ischial tuberosity width, and acetabulum depth. TABLE 5-1 outlines the landmarks used
for each linear measurement.
Femur Morphology. Similar to the pelvis measurements, the femur coordinate system
(FIGURE 5-1a) was first defined. Specifically, the frontal plane of the femur was defined by the
plane tangent to the posterior surface of the femur. The transverse plane was defined by the plane
that was tangent to the inferior surface of the femur. The sagittal plane was then defined as the
42

plane perpendicular to both the frontal and transverse planes. Also, the femur model and all
associated landmarks/regions were neutralized using the femur coordinate system transformation
matrix before subsequent femur measurements, to avoid confounding influences of subject
positioning during CT scanning.
Following establishment of the femur coordinate system, 5 angular measurements were
obtained. First, the mechanical axes of the femoral head, neck, head-neck (combined), and shaft
were defined. The femoral head axis was determined by the line through the center of the femoral
head, which was determined using the center of a sphere fitting the cloud points of the femoral
head surface, and the center of the femoral fovea area, which was the mean point of the femoral
fovea region. The fovea region was determined by finding the concave area on the convex femur
head surface model using a customized MATLAB script. The femoral neck axis was determined
by the line best fit the femoral neck region using a consecutive ellipse fitting algorithm (55).
Similarly, the femoral head-neck axis was created using the head-neck region with the consecutive
ellipse fitting algorithm (55). The femoral shaft axis was determined by the first principal
component vector determined from all cloud points of the shaft surface between lesser trochanter
and the midpoint of medial and lateral epicondyles. The femoral mechanical axis was determined
by the line through the femoral head center and the midpoint between medial and lateral
epicondyles.
The femoral neck-shaft angle was then measured as the obtuse angle between the projected
lines of femur neck axis and femur shaft axis onto the frontal plane of femur (FIGURE 5-1b). The
femoral valgus angle was determined as the acute angle between projected lines of femoral
mechanical axis and femur shaft axis onto the frontal plane of femur (FIGURE 5-1c). The femoral
head-neck anteversion angle was measured as the angle between the frontal plane of femur and the
43

projection line of the femoral head-neck axis vector onto the transverse plane of femur (FIGURE
5-1d). The femoral neck anteversion angle was measured as the angle between the frontal plane of
femur and the projection line of the femoral neck axis onto the transverse plane of femur (FIGURE
5-1e). Finally, the femoral head anteversion was calculated as the acute angle between the frontal
plane of femur and the projection line of femoral head orientation axis onto the transverse plane
of femur (FIGURE 5-1f).
Apart from the angular measurements of the femur, 2 linear measurements were obtained
(femur head radius and femur length) TABLE 5-1. Additionally, bilateral measurements were
carried out using femur landmarks to obtain bilateral hip joint width, bilateral greater trochanter
FIGURE 5-1. Femoral orientation measurements. (a) femoral coordinate system, (b) femoral
neck-shaft angle, (c) femoral valgus angle, (d) femoral head-neck anteversion angle, (e) femoral
neck anteversion angle, (f) femoral head anteversion angle.
44

width (TABLE 5-1). Except for bilateral linear measurements, all other morphologic variables
from the dominant side of each participant were used for statistical analyses.
TABLE 5-1. Description of linear morphology measurements of pelvis and femur.
Variables Description
Bilateral Iliac Crest Width
Linear distance between the most lateral points on right and left
ilium
Bilateral ASIS Width Linear distance between the right and left ASIS
Bilateral Ischial Tuberosity Width Linear distance between the right and left ischial tuberosities
Bilateral Hip Joint Width Linear distance between the right and left hip joint centers
Bilateral Greater Trochanter Width
Linear distance between the most laterally prominent points on
right and left greater trochanters
Acetabulum Depth Longest distance from acetabulum fossa to acetabulum face plane
Femoral Head Radius
The radius of the sphere that were used to fit the femoral head
surface model
Femur Length Vertical distance from the proximal to distal femur

Statistical Analysis
SVM Classification and Feature Evaluation
Participant Classification. Participants whose peak hip adduction were higher than the
mean hip adduction value were classified into the high adduction group, while those with peak hip
adduction below the median were classified into the low adduction group.
Feature Selection-based SVM Classification. Owing to the fact that different
measurement units (e.g., degrees or meters) were involved each variable was standardized to
distributed as N(0,1) prior to classification as recommended by Wu et al. (76). To determine the
best subset of discriminant variables, all possible subsets of the 14 morphology variables were
established. Next, using each subset of variables, a two-class SVM classifier with a radial basis
function kernel was trained. The performance of the classifier trained based on each subset of
variables was evaluated using the prediction accuracy (equation 1) measure with leave-one-out
45

cross validation, to avoid overfitting of the training samples.
𝑎𝑐𝑐𝑢𝑟𝑎𝑐𝑦 =
𝑛𝑢𝑚𝑏𝑒𝑟 𝑜𝑓 𝑡𝑟𝑢𝑒 𝑝𝑜𝑠𝑖𝑡𝑖𝑣𝑒 𝑎𝑛𝑑 𝑡𝑟𝑢𝑒 𝑛𝑒𝑔𝑎𝑡𝑖𝑣𝑒
𝑡𝑜𝑡𝑎𝑙 𝑛𝑢𝑚𝑏𝑒𝑟 𝑜𝑓 𝑐𝑙𝑎𝑠𝑠𝑖𝑓𝑖𝑐𝑎𝑡𝑖𝑜𝑛 𝑙𝑎 𝑏 𝑒𝑙𝑠 (Eq. 1)
For example, using one subset of variables, a SVM classifier was trained with the dataset
that included all but one participant. Next, the trained classifier was used to predict whether the
remaining participant belonged to high or low hip adduction group. This train-and-predict process
was repeated until each participant had served as the testing subject (i.e., leave-one-out cross-
validation). To evaluate the classification performance of this particular subset of variables, the
prediction accuracy was calculated by comparing the ground true classification labels (based on
kinematics) and the predicted labels (based on the SVM classifier) for all participants. As such,
the leave-one-out cross-validation and subsequent performance evaluation were carried out for
every possible subset of variables.
Following examination of all possible subsets, the variable subset that yielded the highest
prediction accuracy was then selected as the best performing variables for discriminating the high
and low hip adduction runners. If the highest prediction accuracy was found in more than one
subset of variables, the subset that consisted of the smallest number of variables were selected as
the best combination of performing variables.
While accuracy was the sole evaluation criterion for choosing the best discriminant
variables, the prediction sensitivity and specificity (equation 2-3 respectively) also were calculated
with the cross-validated dataset to provide additional information about the performance of each
classifier.
sensitivity =
𝑛𝑢𝑚𝑏𝑒𝑟 𝑜𝑓 𝑡𝑟𝑢𝑒 𝑝𝑜𝑠𝑖𝑡𝑖𝑣𝑒
𝑡𝑜𝑡𝑎𝑙 𝑛𝑢𝑚𝑏𝑒𝑟 𝑜𝑓 𝑡𝑟𝑢𝑒 𝑝𝑜𝑠𝑖𝑡𝑖𝑣𝑒 𝑎𝑛𝑑 𝑓𝑎𝑙𝑠𝑒 𝑛𝑒𝑔𝑎𝑡 𝑖𝑣𝑒 (Eq. 2)
46

specificity =
𝑛𝑢𝑚𝑏𝑒𝑟 𝑜𝑓 𝑡𝑟𝑢𝑒 𝑛𝑒𝑔𝑎𝑡𝑖𝑣𝑒
𝑡𝑜𝑡𝑎𝑙 𝑛𝑢𝑚𝑏𝑒𝑟 𝑜𝑓 𝑡𝑟𝑢𝑒 𝑛𝑒𝑔𝑎𝑡𝑖𝑣𝑒 𝑎𝑛𝑑 𝑓𝑎𝑙𝑠𝑒 𝑝 𝑜 𝑠𝑖𝑡𝑖𝑣𝑒 (Eq. 3)
Importance Order of Each Best Performing Variable. While feature selection-based
SVM classification can eliminate variable redundancy and help identify the best performing
variables, this approach provides limited information as to the order of importance. Therefore, we
adopted the method of Sanz et al (77) and created pseudo-samples to iteratively probe the
importance of each of the best performing variables in group classification outcomes as described
below.
The best classification model was established by training a SVM classifier with the best
discriminant variables only (determined above). The training process used the dataset from all 29
participants involved in this study. Next, a few pseudo-datasets that included 50 new samples were
created. The number of pseudo-datasets created equaled the number of best discriminant variables
determined above. Within each pseudo-dataset, the number of independent variables also is the
same as the number of best discriminant variables. For example, if the best subset of discriminant
variables included k (1 ≤ k ≤14) morphology measurements, there would be k pseudo-datasets of
50 new samples. In each pseudo-dataset, there also would be k independent variables.) In this way,
each discriminant variable could be uniquely varied in magnitude from -2 to +2 standard deviations
(SDs) around mean (i.e., 0 for a standardized dataset) with equidistance, whereas the magnitudes
of the rest discriminant variables were set constant as the mean (i.e., 0 for a standardized dataset)
of respective variables.
Following the establishment of the best classification model and pseudo-datasets, we
applied the best classification model to each pseudo-dataset to classify each of the 50 samples. The
output of the classifier included a decision value for each sample, based on the positive or negative
47

sign of which the respective sample was assigned to the high or low hip adduction group,
respectively. This classification process was implemented to each pseudo-dataset so that the
importance of each discriminant variable to response classification outcomes could be assessed.
To determine the order of importance between the best discriminant variables, the
variability of the decision values among 50 samples of each pseudo-dataset was calculated. The
rationale being that by varying the magnitudes of one variable while maintaining the rest of the
variables constant, the morphology variable that possesses a large contribution to response
classification would result in a large variability in the classification decision values. (77). The
variability was examined using the univariate robust metric median absolute deviation (MAD) (77).
Group Differences in Morphology Variables
We first assessed the normality within each group using Shapiro-Wilk tests. If normality
was present in both groups, the difference in each morphology variable of interest was compared
between high and low adduction runners using independent t-tests; otherwise, the Wilcoxon Rank-
Sum test was used. All statistical analyses were performed using MATLAB software (MathWorks,
Natick, MA) with an alpha value of 0.05.
RESULTS
Classification Evaluation
Group Classification. The mean value of peak stance phase hip adduction of all 29 runners
was 11.7°. Thirteen and 16 participants were categorized into the high and low adduction groups,
respectively. Compared to the low group, the high adduction group exhibited significantly greater
peak hip adduction during the stance phase of running (15.5 ± 2.6° vs 8.7 ± 1.8°, p < 0.001).
Best Subset Variables for Classification. Using the cross-validated dataset, a
48

combination of 2 morphology variables was shown to best discriminate the high and low adduction
groups: femoral head anteversion and femur length. Together, these variables achieved a
prediction accuracy of 0.93, sensitivity of 1, and specificity of 0.88. TABLE 5-2 presents the
performance comparison of classifiers that used femoral head anteversion and/or femur length.
TABLE 5-2. Classifier prediction performance using cross-validated datasets
Performance
Femoral Head Anteversion
and Femur Length
(Combined)
Femoral Head Anteversion
Only
Femur Length
Only
True Positive 13 10 0
True Negative 14 15 9
False Positive 2 1 7
False Negative 0 3 13
Misclassified cases 2 4 20
Prediction Accuracy 0.93 0.86 0.31
Prediction Sensitivity 1 0.77 0
Prediction Specificity 0.88 0.94 0.56

Order of Importance. The best classifier that was built with femoral head anteversion and
femur length variables using the dataset from all 29 participants only misclassified one participant
(FIGURE 5-2). Based on the best classifier, results from sensitivity analysis using pseudo-samples
showed that femoral head anteversion obtained higher importance than femur length (0.45 vs 0.08,
respectively, FIGURE 5-3).
Group Differences in Morphology Variables
Group differences in pelvis and femur morphology variables are presented in TABLE 5-3.
Compared to the low hip adduction group, the high adduction group exhibited significantly larger
femoral head anteversion (p<0.01, TABLE 5-3), femoral neck anteversion (p<0.01, TABLE 5-3),
and femoral head-neck anteversion (p=0.01, TABLE 5-3). No other variables of interest were
49

found to be significantly different between groups (p>0.05, TABLE 5-3).

FIGURE 5-3. Importance of femoral head anteversion and femur length in the best SVM classifier.

FIGURE 5-2. Classification Outcome using the dataset from all 29 participants. Red region
represents the SVM-classified high adduction group, and the dark blue region represents the SVM-
classified low adduction. Black dots represent runners with kinematics-classified low hip adduction
and yellow dots represent runners with kinematics-classified high hip adduction. Only 1 participant
was false-positively classified as a high adduction runner (outlined by a white rectangular box).

50

TABLE 5-3. Group differences in hip abductor strength and pelvis & femur morphology variables.
Variables
All
Runners
(n=29)
High
Adduction
Group
(n=13)
Low
Adduction
Group
(n=16)
p
value
effect
size
Acetabulum Abduction Angle (°) 50.3 ± 4.1 48.9 ± 3.9 51.4 ± 4.0 0.09 0.66
Acetabulum Anteversion Angle (°) 20.7 ± 5.9 21.6 ± 4.4 20.0 ± 7.0 0.49 0.26
Femoral Head Anteversion Angle (°) 3.2 ± 8.3 8.6 ± 4.1 -1.1 ± 8.5 <0.01* 1.40
Femoral Neck Anteversion Angle (°) 17.0 ± 9.0 22.3 ± 5.7 12.8 ± 9.0 <0.01* 1.23
Femoral Head-Neck Anteversion Angle (°) 13.55 ± 9.4 18.4 ± 5.6 9.6 ± 10.1 0.01* 1.05
Femoral Neck Shaft Angle (°) 129.0 ± 6.0 130 ± 5.2 128.2 ± 6.7 0.43 0.3
Femoral Valgus Angle (°) 4.6 ± 0.9 4.5 ± 0.9 4.7 ± 0.9 0.70 0.14
Bilateral Iliac Crest Width (cm) 26.7 ± 1.6 27.0 ± 1.9 26.5 ± 1.3 0.42 0.31
Bilateral ASIS Width (cm) 22.4 ± 1.9 22.4 ± 2.4 22.3 ± 1.6 0.85 0.07
Bilateral Ischial Tuberosity Width (cm) 10.3 ± 1.3 10.6 ± 1.0 10.1 ± 1.5 0.26 0.43
Bilateral Hip Joint Width (cm) 17.8 ± 0.9 18.0 ± 0.9 17.6 ± 1.0 0.35 0.36
Bilateral Greater Trochanter Width (cm) 29.7 ± 2.0 30.0 ± 2.1 30.0 ± 2.0 0.65 0.17
Acetabulum Depth (cm) 2.4 ± 0.2 2.4 ± 0.2 2.4 ± 0.2 0.55 0.23
Femoral Head Radius (cm) 2.3 ± 0.2 2.2 ± 0.2 2.3 ± 0.2 1.00 0.03
Femur Length (cm) 41.1 ± 2.8 45.8 ± 3.5 46.0 ± 3.0 0.94 0.03
Values are mean ± SD. *: significant group difference, p<0.05
DISCUSSION
The purpose of the current study was to determine the most relevant pelvis and femur
morphology variables for differentiating runners with high and low hip adduction. Of the 15
variables examined, the combination of femoral head anteversion angle and femur length was
found to best discriminate between runners with high and low hip adduction. Together, these two
variables achieved high levels of prediction accuracy, specificity, and sensitivity.
Based on the classification of participants using peak hip adduction, those assigned to the
high and low groups exhibited 15.5° and 8.7° of peak hip adduction during the stance phase of
running, respectively. The average group difference of 6.8° is considered clinically meaningful, as
prospective studies involving female runners have shown that differences as small as 4 degrees
can differentiate runners who develop iliotibial band syndrome (2) and patellofemoral pain (3)
51

from those who do not.
Results from feature selection-based SVM classification identified a combination of 2
discriminant morphology measures (i.e., femoral head anteversion and femur length). Femoral
head anteversion was shown to be the most important variable and alone exhibited a prediction
accuracy of 0.86 for group classification. It is interesting to note that while all 3 femoral
anteversion measurements were significantly different between groups, femoral head anteversion
emerged as the best discriminant factor. This suggests there may exist a ‘localized’ influence
(within the femoral head) on hip adduction. Further, our findings revealed that except for one
participant, runners with higher femoral head anteversion angles (greater than 5°) were classified
into the high adduction group.
The influence of femoral head anteversion on peak hip adduction during running may be
related to 2 potential mechanisms. First, given the fact that the femoral head anteversion in the
current study was measured using the axis that went through the centers of femoral fovea and
femoral head, it is reasonable to speculate a contributory role of ligamentum teres in affecting
frontal plane hip kinematics. With fibers oriented primarily in the superior-inferior direction,
ligamentum teres has been reported to be an important structure in limiting hip adduction motion
(78-81) and possesses comparative mechanical properties to anterior cruciate ligament (82) and
the medial patellofemoral ligament (83). Anatomically speaking, a larger angle of femoral head
anteversion would re-orient the ligamentum teres and consequently reduce its effectiveness in
resisting hip adduction motion.
Another potential explanation for the influence of femoral head anteversion on frontal
plane hip kinematics may be related to interaction between frontal and transverse plane hip
kinematics in contributing to optimizing hip joint congruency. Within the hip joint, femoral fovea
52

is the only region of the femoral head that is not covered by articular cartilage. A larger femoral
head anteversion angle would shift the location of the fovea more anteriorly, and as a consequence,
decrease the contact area of the cartilage on the anterior surface of femoral head. Given that the
anterior-superior-medial surface on femoral head is important for loading distribution during
weight-bearing (84,85), increased hip internal rotation and hip adduction may be compensation
strategies to restore optimal joint congruency. This premise is supported by results from post-hoc
analyses in which participants assigned to the high adduction group also demonstrated greater hip
internal rotation (3.5 ± 3.5 vs 0.14 ± 3.4, p=0.014) at the time of peak hip adduction. Additionally,
femoral head anteversion was found to be significantly correlated with both hip adduction (r=0.54,
p=0.002) and hip internal rotation angle (r=0.37, p=0.047).
Apart from femoral head anteversion, femur length also was identified as an important
variable for classifying high and low hip adduction during running. However, the prediction
accuracy of using femur length alone to discriminate groups was low (0.31) and worse than random
chance (i.e., 0.5). In addition, femur length was found similar between high and low hip adduction
groups of runners Collectively, these results reject the premise that femur length alone plays an
important role in affecting frontal plane hip adduction. Rather, it is likely that femur length may
act to accentuate the influence of femoral head anteversion. For example, FIGURE 5-2
demonstrated a clear interaction effect between femoral head anteversion and femur length,
especially when femoral head anteversion angle ranged between -5° to 5° or when the angle was
very large (i.e., 20°). By including femur length to the femoral head anteversion classification
model, 2 misclassified cases were corrected, and the prediction accuracy was improved from 0.86
to 0.93. Considering the small improvement of classification performance, the biomechanical
mechanisms behind the interactions of femoral head anteversion and femur length may be trivial.
53

It is likely that femur length functions as a limb length or body height indicator/scaler in the
classification model that defines the relationship between femoral head anteversion and hip
adduction during running.
Several limitations need to be considered when interpreting the results of this study. First,
the participants in our study were young, healthy, recreational runners. Caution is needed when
generalizing results of this study to various clinical and/or elder populations with known bony
morphology variations. In addition, while we examined the best subset of morphology variables
for discriminating runners with high and low hip adduction, additional work is needed to
understand the complex mechanisms behind the relationship between morphology variables and
hip kinematics. Further, a small sample size of runners was included in this study. Future studies
with larger sample size will be able to provide a clear distinction of high and low hip adduction
groups from the baseline, and thus help identify the most reliable relationships between
morphology and kinematics.
CONCLUSION
The feature selection-based SVM method identified 2 morphology variables (i.e., femoral
head anteversion and femur length) for discriminating runner with high adduction and low
adduction. Together these variables achieved high prediction accuracy, sensitivity, and specificity.
The findings point to the importance of femur morphology as being contributory to hip adduction
in runners. However, a complex relationship between morphology and frontal plane hip kinematics
exists. While potential mechanisms were proposed, future studies are needed to better understand
the morphological influences on frontal plane hip motion.
54

Chapter 6. Summary and Conclusions
Excessive hip adduction has been shown to be related to various lower limb injuries during
running, especially in females. Identification of contributing factors of excessive hip adduction is
the first step for designing effective prevention and rehabilitation strategies for related injuries. As
such, the purpose of this dissertation was to comprehensively examine the roles of hip abductor
strength, neuromuscular activation, and pelvis and femur morphology in contributing to frontal
plane hip kinematics during running. To achieve this objective, 3 studies were conducted. The
purpose of Chapter III was to compare sex differences in frontal plane hip kinematics during
running and determine the relationship between late swing and stance phase kinematics. The
purposes of Chapter IV were to 1) compare sex differences in hip abductor strength, activation,
and pelvis and femur bony morphology, and 2) to determine the best combination of predictor(s)
of peak hip adduction during running. The purpose of Chapter V was to determine the most
relevant pelvis and femur morphological characteristics for differentiating runners with high
versus low hip adduction during running.
The results of Chapter III revealed that compared to males, females exhibited significantly
greater peak hip adduction during both the late swing (8.5 ± 2.6 vs 6.1 ± 2.8°, p = 0.019) and stance
(13.3 ± 4.2 vs 9.6 ± 3.4°, p = 0.011) phases of running. In addition, late swing peak hip adduction
was predictive of subsequent stance phase peak hip adduction (r = 0.63, p < 0.001). The findings
of Chapter IV revealed that compared to males, females exhibited significantly lower hip abductor
strength (1.8 ± 0.3 vs 2.0 ± 0.3 Nm/kg, p=0.04), greater femoral neck-shaft angles (134.1 ± 5.0°
vs 129.9 ± 4.1°, p=0.01), and greater hip width to femur length ratios (0.44 ± 0.02 vs 0.42 ± 0.03,
p=0.03). Further, of all variables examined, femoral anteversion was found to be the only
significant predictor of peak hip adduction during late swing (r=0.36, p=0.05) and stance (r=0.41,
55

p=0.03). The results of Chapter V identified the best combination of morphological variables (i.e.,
femoral head anteversion and femur length) for distinguishing runners with high versus low hip
adduction, which achieved a prediction accuracy of 0.93, sensitivity of 1 and specificity of 0.88.
Also, the results of a sensitivity analysis revealed that femoral head anteversion was more
important than femur length in contributing to hip adduction group classification.
The results of Chapter III confirmed previous research findings that females exhibit greater
peak hip adduction during the stance phase of running compared to males. However, the novel
finding of this study was that females also demonstrated significantly greater hip adduction during
the late swing phase of running compared to males. Importantly 40% of the variance in stance
phase peak hip adduction could be explained by late swing values for both males and females
combined. Given the fact that a much lower neuromuscular demand would be required during late
swing, this finding was suggestive of a potential influence of pelvis and femur morphology in
contributing to frontal plane hip kinematics, With the known sex differences in pelvis and femur
morphology reported in literature, a primary purpose of Chapter IV was to further explore the
potential influence of morphology on hip adduction during running.
Results from Chapter IV confirmed the significant influence of femur morphology (i.e.,
femoral head-neck anteversion) on hip adduction during both late swing and stance of running.
This finding contrasted with results of hip abductor strength and activation which were not found
to be predictive of peak hip adduction. Despite being a significant predictor, femoral head-neck
anteversion only explained a small amount of the variance in peak hip adduction during stance and
swing. This finding may be due in part to the fact that only 5 commonly measured morphology
variables were examined. Another reason for the weak and or non-significant relationships found
in Chapter IV may have been related to the statistical approach involved. Specifically, we assumed
56

a linear one-to-one relationship between morphology and hip kinematics and utilized a linear
regression analysis which is susceptible to noise and/or outliers, and consequently, often lacks
sufficient sensitivity to identify relevant relationships.
Based on the statistical limitations identified in Chapter IV, Chapter V employed a feature
selection-based classification analysis with an expanded list of morphology variables, to detect the
best subset of morphologic measures for discriminating runners with high and low adduction
during running. Given that the most predictive bony structural variable revealed in Chapter IV was
the femoral head-neck anteversion, additional related measurements were examined, specifically
femoral head anteversion and femoral neck anteversion. Results of Chapter V revealed femoral
head anteversion was the most important discriminant factor and significantly different between
runners with high and low hip adduction. Specifically, for most of the runners with a femoral head
anteversion angle greater than 5° were classified into the high adduction group and vice versa.
While the underlying mechanism by which femoral head anteversion influences frontal
plane hip motion is not entirely clear, potential causes were hypothesized in Chapter V. Given the
fact that femoral head anteversion in was measured using the axis that went through the centers of
femoral fovea and femoral head, 2 potential mechanisms may be contributory: 1) the function of
ligamentum teres in resisting hip adduction motions, and/or 2) the interaction between frontal and
transverse plane hip motions in contributing to hip joint congruency during running. In regard to
ligamentum teres, it is conceivable that femoral head anteversion could affect hip adduction
through re-orienting the ligamentum teres within the hip joint. If true, it suggests that hip adduction
may be resisted passively as opposed to through active neuromuscular function of the hip
abductors. This premise may explain previous findings that strengthening of hip abductor muscle
strength showed no change in hip adduction motion during running (86).
57

The second proposed mechanism by which femoral head anteversion could influence hip
adduction during running was related to joint congruency. Given the fact that femoral fovea is the
only region of the femoral head that is not covered by articular cartilage, a larger femoral head
anteversion angle would shift the location of the fovea more anteriorly, and as a consequence,
decrease the contact area of the cartilage on the anterior surface of femoral head. To restore optimal
acetabulum-femoral head contact congruency during running, increased hip adduction and internal
rotation may be compensatory strategies to make use of the anterior-superior-medial surface on
the femoral head which is important for loading distribution during weight-bearing (84,85). This
premise was supported by results of a post-hoc analysis presented in Chapter V.
Taken together, the innovation of this dissertation lies in a series of systematic
investigations of the potential influences of hip abductor strength, muscle activation, and pelvis
and femur morphology on frontal plane hip kinematics during running. The results revealed a
contributory influence of bony morphology as opposed to muscle strength and activation. It is
likely habitual physical activities such as running are energetically efficient for healthy individulas
and therefore do not require a high-level output from the neuromuscular system. In addition, while
neuromuscular deficits have been reported in injured population (32,87,88), such variables do not
appear to be relevant in the development of running injuries (4,9,10,13,27). Accordingly, the
findings of this dissertation called into question the contributory roles of muscle strength and
activation in contributing to hip adduction during running.
With respect to morphological influences, previous studies (5,27,36,71) have assumed a
simple, linear relationship between morphology and kinematics. However, the findings of this
dissertation suggest there exists a complex interaction exists between structure and function.
Specifically, various morphology features may interact to affect hip adduction during running.
58

However, despite the complex relationship indicated, the best discriminant morphology measure
was found to be a localized shape variation (i.e., femoral head anteversion). This result contrasts
with previous hypotheses suggesting that broader morphology characteristics such as pelvis with
and femoral neck-shaft angle would be more contributory to hip adduction during running.
Future Directions
As indicated above, the findings of this dissertation provide an important first step in
understanding the influence of pelvis and femur morphology on running kinematics, , However,
future studies are still warranted to understand the underlying mechanisms. Specifically, in-vitro
or finite element studies should be conducted in order to test the tension and contribution of
ligamentum teres in affecting frontal plane hip motions. In addition, joint congruency analysis
should be pursued in a larger sample size to fully explore the relationship between frontal plane
and transverse plane hip motions and their contributions to optimizing joint congruency during
running.
Furthermore, considering that bony morphology is not modifiable in conservative
treatment settings, the interactive relationship among muscle strength, neuromuscular recruitment,
and bony morphology needs to be established to inform injury prevention rehabilitation protocols.
Such information will not only be beneficial for designing effective, subject-specific injury
prevention and rehabilitation protocols but also provide in-depth knowledge as to the underlying
mechanisms of excessive hip adduction.
As this dissertation only recruited healthy young runners, future studies should consider
investigating the muscle strength, activation, and bony morphology influences on hip adduction in
other populations, such as injured patients and/or people with known morphology anomalies such
as femoroacetabular impingement syndrome or hip dysplasia. While muscle strength and
59

activation may not be contributory to excessive hip adduction in a healthy population, these
variables may play important roles after the development of injuries.

60

References
1. Utting MR, Davies G, Newman JH. Is anterior knee pain a predisposing factor to patellofemoral
osteoarthritis? The Knee. 2005;12(5):362-5.
2. Noehren B, Davis I, Hamill J. ASB clinical biomechanics award winner 2006 prospective study of
the biomechanical factors associated with iliotibial band syndrome. Clin. Biomech. (Bristol, Avon).
2007;22(9):951-6.
3. Noehren B, Hamill J, Davis I. Prospective evidence for a hip etiology in patellofemoral pain. Med.
Sci. Sports Exerc. 2013;45(6):1120-4.
4. Russell KA, Palmieri RM, Zinder SM, Ingersoll CD. Sex differences in valgus knee angle during
a single-leg drop jump. Journal of athletic training. 2006;41(2):166-71.
5. Baggaley M, Noehren B, Clasey JL, Shapiro R, Pohl MB. Frontal plane kinematics of the hip during
running: Are they related to hip anatomy and strength? Gait Posture. 2015;42(4):505-10.
6. Taunton JE, Ryan MB, Clement DB, McKenzie DC, Lloyd-Smith DR, Zumbo BD. A retrospective
case-control analysis of 2002 running injuries. Br. J. Sports Med. 2002;36(2):95-101.
7. Malinzak RA, Colby SM, Kirkendall DT, Yu B, Garrett WE. A comparison of knee joint motion
patterns between men and women in selected athletic tasks. Clin. Biomech. (Bristol, Avon).
2001;16(5):438-45.
8. Jacobs CA, Uhl TL, Mattacola CG, Shapiro R, Rayens WS. Hip abductor function and lower
extremity landing kinematics: sex differences. Journal of athletic training. 2007;42(1):76-83.
9. Willson JD, Petrowitz I, Butler RJ, Kernozek TW. Male and female gluteal muscle activity and
lower extremity kinematics during running. Clin. Biomech. (Bristol, Avon). 2012;27(10):1052-7.
10. Zeller BL, McCrory JL, Kibler WB, Uhl TL. Differences in kinematics and electromyographic
activity between men and women during the single-legged squat. Am. J. Sports Med.
2003;31(3):449-56.
61

11. Hollman JH, Galardi CM, Lin IH, Voth BC, Whitmarsh CL. Frontal and transverse plane hip
kinematics and gluteus maximus recruitment correlate with frontal plane knee kinematics during
single-leg squat tests in women. Clin. Biomech. (Bristol, Avon). 2014;29(4):468-74.
12. Souza RB, Powers CM. Differences in hip kinematics, muscle strength, and muscle activation
between subjects with and without patellofemoral pain. J. Orthop. Sports Phys. Ther.
2009;39(1):12-9.
13. Brophy RH, Backus S, Kraszewski AP et al. Differences between sexes in lower extremity
alignment and muscle activation during soccer kick. J. Bone Joint Surg. Am. 2010;92(11):2050-8.
14. de Oliveira Silva D, Briani RV, Pazzinatto MF et al. Q-angle static or dynamic measurements,
which is the best choice for patellofemoral pain? Clinical Biomechanics.
15. Powers CM. The influence of altered lower-extremity kinematics on patellofemoral joint
dysfunction: a theoretical perspective. J. Orthop. Sports Phys. Ther. 2003;33(11):639-46.
16. Tateuchi H, Shiratori S, Ichihashi N. The effect of angle and moment of the hip and knee joint on
iliotibial band hardness. Gait Posture. 2015;41(2):522-8.
17. Ferber R, Noehren B, Hamill J, Davis IS. Competitive female runners with a history of iliotibial
band syndrome demonstrate atypical hip and knee kinematics. J. Orthop. Sports Phys. Ther.
2010;40(2):52-8.
18. Mall NA, Chalmers PN, Moric M et al. Incidence and trends of anterior cruciate ligament
reconstruction in the United States. Am. J. Sports Med. 2014;42(10):2363-70.
19. Glaviano NR, Kew M, Hart JM, Saliba S. Demographic and epidemiological trends in
patellofemoral pain. Int. J. Sports Phys. Ther. 2015;10(3):281-90.
20. Markolf KL, Burchfield DM, Shapiro MM, Shepard MF, Finerman GA, Slauterbeck JL. Combined
knee loading states that generate high anterior cruciate ligament forces. J. Orthop. Res.
1995;13(6):930-5.
21. Clement DB, Taunton JE, Smart GW, McNicol KL. A survey of overuse running injuries. Phys.
Sportsmed. 1981;9(5):47-58.
62

22. Macera CA, Pate RR, Powell KE, Jackson KL, Kendrick JS, Craven TE. Predicting lower-
extremity injuries among habitual runners. Arch. Intern. Med. 1989;149(11):2565-8.
23. Wills AK, Ewins DJ, Ramasamy A, Etherington J. A prospective study of lower extremity
kinematics during gait in persons with patellofemoral pain syndrome. Med. Sci. Sports Exerc.
2005;37(Supplement):S54-S5.
24. Boling MC, Padua DA, Marshall SW, Guskiewicz K, Pyne S, Beutler A. A prospective
investigation of biomechanical risk factors for patellofemoral pain syndrome. The American
Journal of Sports Medicine. 2009;37(11):2108-16.
25. Boling M, Padua D, Marshall S, Guskiewicz K, Pyne S, Beutler A. Gender differences in the
incidence and prevalence of patellofemoral pain syndrome. Scand. J. Med. Sci. Sports.
2010;20(5):725-30.
26. Gehring D, Mornieux G, Fleischmann J, Gollhofer A. Knee and hip joint biomechanics are gender-
specific in runners with high running mileage. Int. J. Sports Med. 2014;35(2):153-8.
27. Chumanov ES, Wall-Scheffler C, Heiderscheit BC. Gender differences in walking and running on
level and inclined surfaces. Clin. Biomech. (Bristol, Avon). 2008;23(10):1260-8.
28. Ferber R, McClay Davis I, Williams Iii DS. Gender differences in lower extremity mechanics
during running. Clinical Biomechanics. 2003;18(4):350-7.
29. Willson JD, Ireland ML, Davis I. Core strength and lower extremity alignment during single leg
squats. Med. Sci. Sports Exerc. 2006;38(5):945-52.
30. Sugimoto D, Mattacola CG, Mullineaux DR, Palmer TG, Hewett TE. Comparison of isokinetic hip
abduction and adduction peak torques and ratio between sexes. Clin. J. Sport Med. 2014;24(5):422-
8.
31. Claiborne TL, Armstrong CW, Gandhi V, Pincivero DM. Relationship between hip and knee
strength and knee valgus during a single leg squat. J. Appl. Biomech. 2006;22(1):41-50.
63

32. Nakagawa TH, Moriya ET, Maciel CD, Serrao FV. Trunk, pelvis, hip, and knee kinematics, hip
strength, and gluteal muscle activation during a single-leg squat in males and females with and
without patellofemoral pain syndrome. J. Orthop. Sports Phys. Ther. 2012;42(6):491-501.
33. Allison KF, Keenan KA, Sell TC et al. Musculoskeletal, biomechanical, and physiological gender
differences in the US military. US Army Med. Dep. J. 2015:22-32.
34. Hannigan JJ, Osternig LR, Chou LS. Sex-Specific Relationships between Hip Strength and Hip,
Pelvis, and Trunk Kinematics in Healthy Runners. J. Appl. Biomech. 2017:1-22.
35. Pohl MB, Kendall KD, Patel C, Wiley JP, Emery C, Ferber R. Experimentally reduced hip-abductor
muscle strength and frontal-plane biomechanics during walking. Journal of athletic training.
2015;50(4):385-91.
36. Brindle RA, Ebaugh DD, Willson JD, Finley MA, Shewokis PA, Milner CE. Relationships of hip
abductor strength, neuromuscular control, and hip width to femoral length ratio with peak hip
adduction angle in healthy female runners. J. Sports Sci. 2020:1-7.
37. Sangeux M, Pascoe J, Graham K, Ramanauskas F, Cain T. Three-dimensional measurement of
femoral neck anteversion and neck shaft angle. J. Comput. Assist. Tomogr. 2014;39.
38. Fiorentino NM, Atkins PR, Kutschke MJ, Foreman KB, Anderson AE. In-vivo quantification of
dynamic hip joint center errors and soft tissue artifact. Gait Posture. 2016;50:246-51.
39. Boyer KA, Freedman Silvernail J, Hamill J. Age and sex influences on running mechanics and
coordination variability. J. Sports Sci. 2017;35(22):2225-31.
40. Sakaguchi M, Ogawa H, Shimizu N, Kanehisa H, Yanai T, Kawakami Y. Gender differences in
hip and ankle joint kinematics on knee abduction during running. European journal of sport science.
2014;14 Suppl 1:S302-9.
41. Almonroeder TG, Benson LC. Sex differences in lower extremity kinematics and patellofemoral
kinetics during running. J. Sports Sci. 2016:1-7.
64

42. Peterson JB, Doan J, Bomar JD, Wenger DR, Pennock AT, Upasani VV. Sex differences in
cartilage topography and orientation of the developing acetabulum: Implications for hip
preservation surgery. Clin. Orthop. Relat. Res. 2015;473(8):2489-94.
43. Brown KM. Selective pressures in the human bony pelvis: Decoupling sexual dimorphism in the
anterior and posterior spaces. Am. J. Phys. Anthropol. 2015;157(3):428-40.
44. Coleman WH. Sex differences in the growth of the human bony pelvis. Am. J. Phys. Anthropol.
1969;31(2):125-51.
45. Kurki HK. Pelvic dimorphism in relation to body size and body size dimorphism in humans. J.
Hum. Evol. 2011;61(6):631-43.
46. Malloy PJ, Morgan AM, Meinerz CM, Geiser CF, Kipp K. Hip external rotator strength is
associated with better dynamic control of the lower extremity during landing tasks. J. Strength
Cond. Res. 2016;30(1):282-91.
47. Hollman JH, Ginos BE, Kozuchowski J, Vaughn AS, Krause DA, Youdas JW. Relationships
between knee valgus, hip-muscle strength, and hip-muscle recruitment during a single-limb step-
down. Journal of sport rehabilitation. 2009;18(1):104-17.
48. Almeida MO, Davis IS, Lopes AD. Biomechanical Differences of Foot-Strike Patterns During
Running: A Systematic Review With Meta-analysis. J. Orthop. Sports Phys. Ther.
2015;45(10):738-55.
49. Sinclair J, Taylor PJ, Hobbs SJ. Digital filtering of three-dimensional lower extremity kinematics:
an assessment. J Hum Kinet. 2013;39:25-36.
50. Wu G, Siegler S, Allard P et al. 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.
2002;35(4):543-8.
51. Schache AG, Blanch P, Rath D, Wrigley T, Bennell K. Differences between the sexes in the three-
dimensional angular rotations of the lumbo-pelvic-hip complex during treadmill running. J. Sports
Sci. 2003;21(2):105-18.
65

52. Kernozek TW, Torry MR, H VANH, Cowley H, Tanner S. Gender differences in frontal and
sagittal plane biomechanics during drop landings. Med. Sci. Sports Exerc. 2005;37(6):1003-12;
discussion 13.
53. Kaneko M, Sakuraba K. Association between femoral anteversion and lower extremity posture
upon single-leg landing: Implications for anterior cruciate ligament injury. Journal of physical
therapy science. 2013;25(10):1213-7.
54. Rainoldi A, Melchiorri G, Caruso I. A method for positioning electrodes during surface EMG
recordings in lower limb muscles. J. Neurosci. Methods. 2004;134(1):37-43.
55. Bonneau N, Libourel PA, Simonis C et al. A three-dimensional axis for the study of femoral neck
orientation. J. Anat. 2012;221(5):465-76.
56. Merletti R. Surface Electromyography: Physiology, Engineering and Applications. Hoboken:
Hoboken : John Wiley & Sons, Incorporated; 2016.
57. Ma H, Han Y, Yang Q et al. Three-dimensional computed tomography reconstruction
measurements of acetabulum in Chinese adults. Anatomical record (Hoboken, N.J. : 2007).
2014;297(4):643-9.
58. Elbuken F, Baykara M, Ozturk C. Standardisation of the neck-shaft angle and measurement of age-,
gender- and BMI-related changes in the femoral neck using DXA. Singapore Med. J.
2012;53(9):587-90.
59. Rickels T, Kreuzer S, Lovell T, Nogler M, Puri L. Age and gender related differences in femoral
neck - shaft angles ORS 2011 Annual Meeting. 2011.
60. Hartel MJ, Petersik A, Schmidt A et al. Determination of femoral neck angle and torsion angle
utilizing a aovel three-dimensional modeling and analytical technology based on CT datasets. PLoS
One. 2016;11(3).
61. Anderson JY, Trinkaus E. Patterns of sexual, bilateral and interpopulational variation in human
femoral neck-shaft angles. J. Anat. 1998;192(2):279-85.
66

62. Jiang N, Peng L, Al-Qwbani M et al. Femoral version, neck-shaft angle, and acetabular anteversion
in Chinese han population: A retrospective analysis of 466 healthy adults. Medicine.
2015;94(21):e891.
63. Nakahara I, Takao M, Sakai T, Nishii T, Yoshikawa H, Sugano N. Gender differences in 3D
morphology and bony impingement of human hips. J. Orthop. Res. 2011;29(3):333-9.
64. Patro BP, Behera S, Das SS, Das G, Patra SK, Prabhat V. Estimation of femoral neck anteversion
in adults: A comparison between clinical method, radiography, and Computed Tomography at a
tertiary-care center in Eastern India. Cureus. 2019;11(4):e4469-e.
65. Zhang H, Wang Y, Ai S, Chen X, Wang L, Dai K. Three-dimensional acetabular orientation
measurement in a reliable coordinate system among one hundred Chinese. PLoS One.
2017;12(2):e0172297.
66. Miyasaka D, Sakai Y, Ibuchi S, Suzuki H, Imai N, Endo N. Sex- and age-specific differences in
femoral head coverage and acetabular morphology among healthy subjects-derivation of normal
ranges and thresholds for abnormality. Skeletal Radiol. 2017;46(4):523-31.
67. Thelen T, Thelen P, Demezon H, Aunoble S, Le Huec JC. Normative 3D acetabular orientation
measurements by the low-dose EOS imaging system in 102 asymptomatic subjects in standing
position: Analyses by side, gender, pelvic incidence and reproducibility. Orthop. Traumatol. Surg.
Res. 2017;103(2):209-15.
68. Higgins SW, Spratley EM, Boe RA, Hayes CW, Jiranek WA, Wayne JS. A novel approach for
determining three-dimensional acetabular orientation: results from two hundred subjects. J. Bone
Joint Surg. Am. 2014;96(21):1776-84.
69. Huseynov A, Zollikofer CPE, Coudyzer W et al. Developmental evidence for obstetric adaptation
of the human female pelvis. Proc. Natl. Acad. Sci. U. S. A. 2016;113(19):5227-32.
70. Sinclair J, Taylor PJ, Bottoms L. The appropriateness of the helical axis technique and six available
cardan sequences for the representation of 3-d lead leg kinematics during the fencing lunge. J Hum
Kinet. 2013;37:7-15.
67

71. Liu J, Lewton KL, Colletti PM, Powers CM. Sex difference in hip adduction during running:
influence of hip abductor strength, activation, and pelvis & femur morphology. 2021.
72. Avidan S. Subset selection for efficient SVM tracking. In: Proceedings of the 2003 IEEE Computer
Society Conference on Computer Vision and Pattern Recognition, 2003. Proceedings. 2003. p. I-I.
73. Brown LE, Tsamardinos I, Hardin DP. To feature space and back: Identifying top-weighted
features in polynomial Support Vector Machine models. Intelligent Data Analysis. 2012;16:551-
79.
74. Salimi A, Ziaii M, Amiri A, Hosseinjani Zadeh M, Karimpouli S, Moradkhani M. Using a feature
subset selection method and support vector machine to address curse of dimensionality and
redundancy in Hyperion hyperspectral data classification. The Egyptian Journal of Remote Sensing
and Space Science. 2018;21(1):27-36.
75. Wu LC, Kuo C, Loza J et al. Detection of American football head impacts using biomechanical
features and support vector machine classification. Sci. Rep. 2017;8(1):855.
76. Guyon I, Weston J, Barnhill S, Vapnik V. Gene selection for cancer classification using support
vector machines. Machine Learning. 2002;46(1):389-422.
77. Sanz H, Valim C, Vegas E, Oller JM, Reverter F. SVM-RFE: selection and visualization of the
most relevant features through non-linear kernels. BMC Bioinformatics. 2018;19(1):432.
78. van Arkel RJ, Amis AA, Cobb JP, Jeffers JRT. The capsular ligaments provide more hip rotational
restraint than the acetabular labrum and the ligamentum teres : an experimental study. Bone Joint
J. 2015;97-B(4):484-91.
79. O'Donnell JM, Devitt BM, Arora M. The role of the ligamentum teres in the adult hip: redundant
or relevant? A review. J Hip Preserv Surg. 2018;5(1):15-22.
80. Field RE, Rajakulendran K. The labro-acetabular complex. J. Bone Joint Surg. Am. 2011;93 Suppl
2:22-7.
81. Demange K, Kakuda C, Pereira C, Sakaki M, Albuquerque R. Influence of the femoral head
ligament on hip mechanical function. Acta ortop. bras. 2007;15.
68

82. Wenger D, Miyanji F, Mahar A, Oka R. The Mechanical Properties of the Ligamentum Teres: A
Pilot Study to Assess Its Potential for Improving Stability in Children's Hip Surgery. Journal of
Pediatric Orthopaedics. 2007;27(4):408-10.
83. Kader DF, Rajeev A. A review of functional anatomy and surgical reconstruction of medial
patellofemoral ligament. Journal of Arthroscopy and Joint Surgery. 2014;1(1):5-10.
84. Chen L, Hong G, Fang B et al. Predicting the collapse of the femoral head due to osteonecrosis:
From basic methods to application prospects. Journal of Orthopaedic Translation. 2017;11:62-72.
85. Neumann DA. Kinesiology of the musculoskeletal system : foundations for physical rehabilitation.
First edition. St. Louis : Mosby, [2002] ©2002; 2002.
86. Willy RW, Davis IS. The effect of a hip-strengthening program on mechanics during running and
during a single-leg squat. J. Orthop. Sports Phys. Ther. 2011;41(9):625-32.
87. Fredericson M, Cookingham C, Chaudhari A, Dowdell B, Oestreicher N, Sahrmann S. Hip
Abductor Weakness in Distance Runners with Iliotibial Band Syndrome. Clinical journal of sport
medicine : official journal of the Canadian Academy of Sport Medicine. 2000;10:169-75.
88. Brown AM, Zifchock RA, Lenhoff M, Song J, Hillstrom HJ. Hip muscle response to a fatiguing
run in females with iliotibial band syndrome. Hum Movement Sci. 2019;64:181-90.

Abstract (if available)

Abstract Excessive hip adduction is a common movement impairment that has been proposed to underlie various lower extremity injuries (1-3). Compared to males, females are more likely to exhibit excessive hip adduction during weight-bearing activities such as running (4-7). Despite decades of research investigating the potential roles of hip abductor muscle strength and activation as being potential causes of excessive hip adduction (8-13), studies have reported that these variables have little influence on frontal plane hip motion (5,8,11). Another factor that has not been systematically explored in the context of sex differences in frontal plane hip kinematics is the potential influence of pelvis and femur bony morphology. The premise that bony morphology may play a role in hip joint kinematics is supported by 2 underlying assertions: 1) males and females differ in various aspects of femur and pelvis morphology and 2) sex differences in frontal plane hip kinematics have been reported at initial contact and the stance phase of running. The latter suggests that hip kinematics during late swing (a time point at which there is little muscular demand) may have an influence on hip adduction during stance. Given the limited work in this area, the objective of this dissertation was to comprehensively examine the roles of hip abductor strength, neuromuscular activation, and pelvis and femur bony morphology in contributing to frontal plane hip kinematics during running. To achieve this objective, 3 studies were undertaken. ❧ The purpose of Chapter III was to compare sex differences in frontal plane hip kinematics during both late swing and stance phases of running and determine the relationship between late swing and stance phase kinematics. Fifteen female and 16 male runners (all heel strikers) ran overground at a speed of 4 m/s. Hip joint kinematics during running were quantified using a 3D motion capture system. Sex differences in peak hip adduction during the late swing and stance phases were compared using independent sample t-tests. Linear regression analysis was used to determine the relationship between late swing and stance phase peak hip adduction. Compared to males, females exhibited significantly greater peak hip adduction during both the late swing (8.5 ± 2.6 vs 6.1 ± 2.8°, p = 0.019) and stance phases of running (13.3 ± 4.2 vs 9.6 ± 3.4°, p = 0.011). Furthermore, late swing peak hip adduction was predictive of subsequent stance phase peak hip adduction (r = 0.63, p < 0.001). ❧ The purpose of Chapter IV was twofold: 1) to compare sex differences in hip abductor strength, activation, and pelvis and femur bony morphology, and 2) to determine the best combination of predictor(s) of peak hip adduction during late swing and stance phases of running. Fifteen female and 14 male runners underwent strength testing, instrumented analysis of overground running (e.g., kinematics and muscle activation), and computed tomography (CT) scanning of pelvis and femur. Morphologic measurements included the bilateral hip width to femur length ratio, acetabulum abduction, acetabulum anteversion, femoral anteversion, and femoral neck-shaft angles. Sex differences for all variables (e.g., kinematics, strength, activation, and morphology) were examined using independent t tests. Linear regression was used to assess the ability of each independent variable of interest to predict peak hip adduction during the late swing and stance phase of running. Compared to males, females exhibited significantly lower hip abductor strength (1.8 ± 0.3 vs 2.0 ± 0.3 Nm/kg, p=0.04), greater femoral neck-shaft angles (134.1 ± 5.0° vs 129.9 ± 4.1°, p=0.01), and a greater hip width to femur length ratio (0.44 ± 0.02 vs 0.42 ± 0.03, p=0.03). However, femoral anteversion was found to be the only significant predictor of peak hip adduction during late swing (r=0.36, p=0.05) and stance (r=0.41, p=0.03). ❧ The purpose of Chapter V was to determine the most relevant pelvis and femur morphological characteristics for differentiating runners with high versus low hip adduction during the stance phase of running. Fifteen female and 14 male runners underwent instrumented kinematics analysis of overground running and CT scanning of pelvis and femur. Using a hybrid approach of support vector machine classification and best subset feature selection techniques, the combination of femoral head anteversion and femur length was shown to be the best performing variables for distinguishing runners with high versus low peak hip adduction during running. Together, these two variables achieved a prediction accuracy of 0.93, sensitivity of 1 and specificity of 0.88. Also, the results of a sensitivity analysis revealed that femoral head anteversion was more important than femur length in contributing to hip adduction group classification. ❧ Taken together, the results of this dissertation highlight the role of bony morphology as opposed to muscle strength and activation as being contributory to hip adduction during running. With respect to morphology, there appears to be a ‘localized’ influence on frontal plane hip kinematics related to femoral head orientation. While it was beyond the scope of this dissertation to determine the mechanisms underlying this ‘localized’ morphology influence, 2 potential mechanisms may be contributory: 1) the function of ligamentum teres in limiting excessive hip adduction, and/or 2) interaction between frontal and transverse plane hip kinematics in contributing to joint congruency during running. Further investigation is needed to fully understand the complex relationship between proximal femur morphology and frontal plane hip kinematics during dynamic tasks.

Linked assets

University of Southern California Dissertations and Theses

Conceptually similar

PDF

Pathomechanics of femoroacetabular impingement syndrome: utilizing subject-specific modeling approaches to investigate the influence of hip joint morphology and neuromuscular control

PDF

Corticomotor excitability of gluteus maximus: influence on hip extensor strength and hip mechanics

PDF

Hip and pelvis kinematics and kinetics in persons with femoroacetabular impingement

PDF

Influence of sagittal plane trunk posture on lower extremity biomechanics during running

PDF

The influence of hip muscle performance on postural stability & ankle biomechanics: implications for ankle injury

PDF

Multifidus morphology, fatigability and activation in persons with chronic unilateral low back pain

PDF

The influence of tibiofemoral kinematics and knee extensor mechanics on patellar tendon stress: a comparison of persons with and without patellar tendinopathy

PDF

The influence of hip muscle performance on lower extremity biomechanics and neuromuscular activation: implications for anterior cruciate ligament injury

PDF

The influence of patellofemoral joint loading on patella strain and patella water content in females with patellofemoral pain

PDF

Trunk neuromechanics during turning: a window into recurrent low back pain

PDF

Development of a movement performance assessment to predict ACL re-injury

PDF

The footprint of pain: investigating persistence of altered trunk control in recurrent low back pain

Asset Metadata

Creator Liu, Jia (author)

Core Title Sex differences in hip adduction during running: influence of hip abductor strength, muscle activation, and pelvis & femur morphology

School School of Dentistry

Degree Doctor of Philosophy

Degree Program Biokinesiology

Publication Date 02/27/2021

Defense Date 12/11/2020

Publisher University of Southern California (original), University of Southern California. Libraries (digital)

Tag activation,hip adduction,morphology,OAI-PMH Harvest,Running,sex differences,Strength

Language English

Contributor Electronically uploaded by the author (provenance)

Advisor Powers, Christopher M. (committee chair), Baker, Lucinda (committee member), Colletti, Patrick M. (committee member), Lewton, Kristi L. (committee member), Salem, George J. (committee member)

Creator Email liu598@usc.edu

Permanent Link (DOI) https://doi.org/10.25549/usctheses-c89-424645

Unique identifier UC11667575

Identifier etd-LiuJia-9297.pdf (filename),usctheses-c89-424645 (legacy record id)

Legacy Identifier etd-LiuJia-9297.pdf

Dmrecord 424645

Document Type Dissertation

Rights Liu, Jia

Type texts

Source University of Southern California (contributing entity), University of Southern California Dissertations and Theses (collection)

Access Conditions The author retains rights to his/her dissertation, thesis or other graduate work according to U.S. copyright law. Electronic access is being provided by the USC Libraries in agreement with the a...

Repository Name University of Southern California Digital Library

Repository Location USC Digital Library, University of Southern California, University Park Campus MC 2810, 3434 South Grand Avenue, 2nd Floor, Los Angeles, California 90089-2810, USA